NEWS revision 1.1.1.6
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 GCC 6 Release Series 10 11 October 26, 2018 12 13 The [1]GNU project and the GCC developers are pleased to announce the 14 release of GCC 6.5. 15 16 This release is a bug-fix release, containing fixes for regressions in 17 GCC 6.4 relative to previous releases of GCC. 18 19 Release History 20 21 GCC 6.5 22 October 26, 2018 ([2]changes, [3]documentation) 23 24 GCC 6.4 25 July 4, 2017 ([4]changes, [5]documentation) 26 27 GCC 6.3 28 December 21, 2016 ([6]changes, [7]documentation) 29 30 GCC 6.2 31 August 22, 2016 ([8]changes, [9]documentation) 32 33 GCC 6.1 34 April 27, 2016 ([10]changes, [11]documentation) 35 36 References and Acknowledgements 37 38 GCC used to stand for the GNU C Compiler, but since the compiler 39 supports several other languages aside from C, it now stands for the 40 GNU Compiler Collection. 41 42 A list of [12]successful builds is updated as new information becomes 43 available. 44 45 The GCC developers would like to thank the numerous people that have 46 contributed new features, improvements, bug fixes, and other changes as 47 well as test results to GCC. This [13]amazing group of volunteers is 48 what makes GCC successful. 49 50 For additional information about GCC please refer to the [14]GCC 51 project web site or contact the [15]GCC development mailing list. 52 53 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 54 55 56 For questions related to the use of GCC, please consult these web 57 pages and the [18]GCC manuals. If that fails, the 58 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 59 web pages and the development of GCC are welcome on our developer 60 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 61 archives. 62 63 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 64 distribution of this entire article is permitted in any medium, 65 provided this notice is preserved. 66 67 These pages are [23]maintained by the GCC team. Last modified 68 2018-10-26[24]. 69 70 References 71 72 1. http://www.gnu.org/ 73 2. http://gcc.gnu.org/gcc-6/changes.html 74 3. http://gcc.gnu.org/onlinedocs/6.5.0/ 75 4. http://gcc.gnu.org/gcc-6/changes.html 76 5. http://gcc.gnu.org/onlinedocs/6.4.0/ 77 6. http://gcc.gnu.org/gcc-6/changes.html 78 7. http://gcc.gnu.org/onlinedocs/6.3.0/ 79 8. http://gcc.gnu.org/gcc-6/changes.html 80 9. http://gcc.gnu.org/onlinedocs/6.2.0/ 81 10. http://gcc.gnu.org/gcc-6/changes.html 82 11. http://gcc.gnu.org/onlinedocs/6.1.0/ 83 12. http://gcc.gnu.org/gcc-6/buildstat.html 84 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 85 14. http://gcc.gnu.org/index.html 86 15. mailto:gcc (a] gcc.gnu.org 87 16. http://gcc.gnu.org/mirrors.html 88 17. http://gcc.gnu.org/svn.html 89 18. https://gcc.gnu.org/onlinedocs/ 90 19. mailto:gcc-help (a] gcc.gnu.org 91 20. mailto:gcc (a] gcc.gnu.org 92 21. https://gcc.gnu.org/lists.html 93 22. https://www.fsf.org/ 94 23. https://gcc.gnu.org/about.html 95 24. http://validator.w3.org/check/referer 96 ====================================================================== 97 http://gcc.gnu.org/gcc-6/changes.html 98 GCC 6 Release Series 99 Changes, New Features, and Fixes 100 101 This page is a brief summary of some of the huge number of improvements 102 in GCC 6. For more information, see the [1]Porting to GCC 6 page and 103 the [2]full GCC documentation. 104 105 Caveats 106 107 * The default mode for C++ is now -std=gnu++14 instead of 108 -std=gnu++98. 109 * Support for a number of older systems and recently unmaintained or 110 untested target ports of GCC has been declared obsolete in GCC 6. 111 Unless there is activity to revive them, the next release of GCC 112 will have their sources permanently removed. 113 The following ports for individual systems on particular 114 architectures have been obsoleted: 115 + SH5 / SH64 (sh64-*-*) as announced [3]here. 116 * The AVR port requires binutils version 2.26.1 or later for the fix 117 for [4]PR71151 to work. 118 119 General Optimizer Improvements 120 121 * UndefinedBehaviorSanitizer gained a new sanitization option, 122 -fsanitize=bounds-strict, which enables strict checking of array 123 bounds. In particular, it enables -fsanitize=bounds as well as 124 instrumentation of flexible array member-like arrays. 125 * Type-based alias analysis now disambiguates accesses to different 126 pointers. This improves precision of the alias oracle by about 127 20-30% on higher-level C++ programs. Programs doing invalid type 128 punning of pointer types may now need -fno-strict-aliasing to work 129 correctly. 130 * Alias analysis now correctly supports the weakref and alias 131 attributes. This allows accessing both a variable and its alias in 132 one translation unit which is common with link-time optimization. 133 * Value range propagation now assumes that the this pointer in C++ 134 member functions is non-null. This eliminates common null pointer 135 checks but also breaks some non-conforming code-bases (such as 136 Qt-5, Chromium, KDevelop). As a temporary work-around 137 -fno-delete-null-pointer-checks can be used. Wrong code can be 138 identified by using -fsanitize=undefined. 139 * Link-time optimization improvements: 140 + warning and error attributes are now correctly preserved by 141 declaration linking and thus -D_FORTIFY_SOURCE=2 is now 142 supported with -flto. 143 + Type merging was fixed to handle C and Fortran 144 interoperability rules as defined by the Fortran 2008 language 145 standard. 146 As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable 147 with char in all cases because it is an array while char is 148 scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In 149 general, this inter-operability cannot be implemented, for 150 example on targets where the argument passing convention for 151 arrays differs from scalars. 152 + More type information is now preserved at link time, reducing 153 the loss of accuracy of the type-based alias analysis compared 154 to builds without link-time optimization. 155 + Invalid type punning on global variables and declarations is 156 now reported with -Wodr-type-mismatch. 157 + The size of LTO object files was reduced by about 11% 158 (measured by compiling Firefox 46.0). 159 + Link-time parallelization (enabled using -flto=n) was 160 significantly improved by decreasing the size of streamed data 161 when partitioning programs. The size of streamed IL while 162 compiling Firefox 46.0 was reduced by 66%. 163 + The linker plugin was extended to pass information about the 164 type of binary produced to the GCC back end. (That can also be 165 controlled manually by -flinker-output.) This makes it 166 possible to properly configure the code generator and support 167 incremental linking. Incremental linking of LTO objects by gcc 168 -r is now supported for plugin-enabled setups. 169 There are two ways to perform incremental linking: 170 1. Linking by ld -r will result in an object file with all 171 sections from individual object files mechanically 172 merged. This delays the actual link-time optimization to 173 the final linking step and thus permits whole program 174 optimization. Linking the final binary with such object 175 files is however slower. 176 2. Linking by gcc -r will lead to link-time optimization and 177 emit the final binary into the object file. Linking such 178 an object file is fast but avoids any benefits from whole 179 program optimization. 180 GCC 7 will support incremental link-time optimization with gcc 181 -r. 182 * Inter-procedural optimization improvements: 183 + Basic jump threading is now performed before profile 184 construction and inline analysis, resulting in more realistic 185 size and time estimates that drive the heuristics of the 186 inliner and function cloning passes. 187 + Function cloning now more aggressively eliminates unused 188 function parameters. 189 190 New Languages and Language specific improvements 191 192 Compared to GCC 5, the GCC 6 release series includes a much improved 193 implementation of the [5]OpenACC 2.0a specification. Highlights are: 194 * In addition to single-threaded host-fallback execution, offloading 195 is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit 196 little-endian GNU/Linux host systems. For nvptx offloading, with 197 the OpenACC parallel construct, the execution model allows for an 198 arbitrary number of gangs, up to 32 workers, and 32 vectors. 199 * Initial support for parallelized execution of OpenACC kernels 200 constructs: 201 + Parallelization of a kernels region is switched on by 202 -fopenacc combined with -O2 or higher. 203 + Code is offloaded onto multiple gangs, but executes with just 204 one worker, and a vector length of 1. 205 + Directives inside a kernels region are not supported. 206 + Loops with reductions can be parallelized. 207 + Only kernels regions with one loop nest are parallelized. 208 + Only the outer-most loop of a loop nest can be parallelized. 209 + Loop nests containing sibling loops are not parallelized. 210 Typically, using the OpenACC parallel construct gives much better 211 performance, compared to the initial support of the OpenACC kernels 212 construct. 213 * The device_type clause is not supported. The bind and nohost 214 clauses are not supported. The host_data directive is not supported 215 in Fortran. 216 * Nested parallelism (cf. CUDA dynamic parallelism) is not supported. 217 * Usage of OpenACC constructs inside multithreaded contexts (such as 218 created by OpenMP, or pthread programming) is not supported. 219 * If a call to the acc_on_device function has a compile-time constant 220 argument, the function call evaluates to a compile-time constant 221 value only for C and C++ but not for Fortran. 222 223 See the [6]OpenACC and [7]Offloading wiki pages for further 224 information. 225 226 C family 227 228 * Version 4.5 of the [8]OpenMP specification is now supported in the 229 C and C++ compilers. 230 * The C and C++ compilers now support attributes on enumerators. For 231 instance, it is now possible to mark enumerators as deprecated: 232 233 enum { 234 newval, 235 oldval __attribute__ ((deprecated ("too old"))) 236 }; 237 238 * Source locations for the C and C++ compilers are now tracked as 239 ranges, rather than just points, making it easier to identify the 240 subexpression of interest within a complicated expression. For 241 example: 242 243 test.cc: In function 'int test(int, int, foo, int, int)': 244 test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo' 245 ) 246 return p + q * r * s + t; 247 ~~^~~ 248 249 In addition, there is now initial support for precise diagnostic 250 locations within strings: 251 252 format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in 253 t' argument [-Wformat=] 254 printf("%*d"); 255 ^ 256 257 * Diagnostics can now contain "fix-it hints", which are displayed in 258 context underneath the relevant source code. For example: 259 260 fixits.c: In function 'bad_deref': 261 fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'? 262 return ptr.x; 263 ^ 264 -> 265 266 * The C and C++ compilers now offer suggestions for misspelled field 267 names: 268 269 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 270 you mean 'color'? 271 return ptr->colour; 272 ^~~~~~ 273 274 * New command-line options have been added for the C and C++ 275 compilers: 276 + -Wshift-negative-value warns about left shifting a negative 277 value. 278 + -Wshift-overflow warns about left shift overflows. This 279 warning is enabled by default. -Wshift-overflow=2 also warns 280 about left-shifting 1 into the sign bit. 281 + -Wtautological-compare warns if a self-comparison always 282 evaluates to true or false. This warning is enabled by -Wall. 283 + -Wnull-dereference warns if the compiler detects paths that 284 trigger erroneous or undefined behavior due to dereferencing a 285 null pointer. This option is only active when 286 -fdelete-null-pointer-checks is active, which is enabled by 287 optimizations in most targets. The precision of the warnings 288 depends on the optimization options used. 289 + -Wduplicated-cond warns about duplicated conditions in an 290 if-else-if chain. 291 + -Wmisleading-indentation warns about places where the 292 indentation of the code gives a misleading idea of the block 293 structure of the code to a human reader. For example, given 294 [9]CVE-2014-1266: 295 296 sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange': 297 sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin 298 g-indentation] 299 if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0) 300 ^~ 301 sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly 302 indented as if it is guarded by the 'if' 303 goto fail; 304 ^~~~ 305 306 This warning is enabled by -Wall. 307 * The C and C++ compilers now emit saner error messages if 308 merge-conflict markers are present in a source file. 309 310 test.c:3:1: error: version control conflict marker in file 311 <<<<<<< HEAD 312 ^~~~~~~ 313 314 C 315 316 * It is possible to disable warnings when an initialized field of a 317 structure or a union with side effects is being overridden when 318 using designated initializers via a new warning option 319 -Woverride-init-side-effects. 320 * A new type attribute scalar_storage_order applying to structures 321 and unions has been introduced. It specifies the storage order (aka 322 endianness) in memory of scalar fields in structures or unions. 323 324 C++ 325 326 * The default mode has been changed to -std=gnu++14. 327 * [10]C++ Concepts are now supported when compiling with -fconcepts. 328 * -flifetime-dse is more aggressive in dead-store elimination in 329 situations where a memory store to a location precedes a 330 constructor to that memory location. 331 * G++ now supports [11]C++17 fold expressions, u8 character literals, 332 extended static_assert, and nested namespace definitions. 333 * G++ now allows constant evaluation for all non-type template 334 arguments. 335 * G++ now supports C++ Transactional Memory when compiling with 336 -fgnu-tm. 337 338 Runtime Library (libstdc++) 339 340 * Extensions to the C++ Library to support mathematical special 341 functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland. 342 * Experimental support for C++17, including the following new 343 features: 344 + std::uncaught_exceptions function (this is also available for 345 -std=gnu++NN modes); 346 + new member functions try_emplace and insert_or_assign for 347 unique_key maps; 348 + non-member functions std::size, std::empty, and std::data for 349 accessing containers and arrays; 350 + std::invoke; 351 + std::shared_mutex; 352 + std::void_t and std::bool_constant metaprogramming utilities. 353 Thanks to Ville Voutilainen for contributing many of the C++17 354 features. 355 * An experimental implementation of the File System TS. 356 * Experimental support for most features of the second version of the 357 Library Fundamentals TS. This includes polymorphic memory resources 358 and array support in shared_ptr, thanks to Fan You. 359 * Some assertions checked by Debug Mode can now also be enabled by 360 _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro 361 have less run-time overhead than the full _GLIBCXX_DEBUG checks and 362 don't affect the library ABI, so can be enabled per-translation 363 unit. 364 * Timed mutex types are supported on more targets, including Darwin. 365 * Improved std::locale support for DragonFly and FreeBSD, thanks to 366 John Marino and Andreas Tobler. 367 368 Fortran 369 370 * Fortran 2008 SUBMODULE support. 371 * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY 372 support. 373 * Improved support for Fortran 2003 deferred-length character 374 variables. 375 * Improved support for OpenMP and OpenACC. 376 * The MATMUL intrinsic is now inlined for straightforward cases if 377 front-end optimization is active. The maximum size for inlining can 378 be set to n with the -finline-matmul-limit=n option and turned off 379 with -finline-matmul-limit=0. 380 * The -Wconversion-extra option will warn about REAL constants which 381 have excess precision for their kind. 382 * The -Winteger-division option has been added, which warns about 383 divisions of integer constants which are truncated. This option is 384 included in -Wall by default. 385 386 libgccjit 387 388 * The driver code is now run in-process within libgccjit, providing a 389 small speed-up of the compilation process. 390 * The API has gained entrypoints for 391 + [12]timing how long was spent in different parts of code, 392 + [13]creating switch statements, 393 + [14]allowing unreachable basic blocks in a function, and 394 + [15]adding arbitrary command-line options to a compilation. 395 396 New Targets and Target Specific Improvements 397 398 AArch64 399 400 * A number of AArch64-specific options have been added. The most 401 important ones are summarised in this section; for more detailed 402 information please refer to the documentation. 403 * The command-line options -march=native, -mcpu=native and 404 -mtune=native are now available on native AArch64 GNU/Linux 405 systems. Specifying these options causes GCC to auto-detect the 406 host CPU and choose the optimal setting for that system. 407 * -fpic is now supported when generating code for the small code 408 model (-mcmodel=small). The size of the global offset table (GOT) 409 is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the 410 ILP32 SysV ABI. 411 * The AArch64 port now supports target attributes and pragmas. Please 412 refer to the [16]documentation for details of available attributes 413 and pragmas as well as usage instructions. 414 * Link-time optimization across translation units with different 415 target-specific options is now supported. 416 * The option -mtls-size= is now supported. It can be used to specify 417 the bit size of TLS offsets, allowing GCC to generate better TLS 418 instruction sequences. 419 * The option -fno-plt is now fully functional. 420 * The ARMv8.1-A architecture and the Large System Extensions are now 421 supported. They can be used by specifying the -march=armv8.1-a 422 option. Additionally, the +lse option extension can be used in a 423 similar fashion to other option extensions. The Large System 424 Extensions introduce new instructions that are used in the 425 implementation of atomic operations. 426 * The ACLE half-precision floating-point type __fp16 is now supported 427 in the C and C++ languages. 428 * The ARM Cortex-A35 processor is now supported via the 429 -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the 430 equivalent target attributes and pragmas. 431 * The Qualcomm QDF24xx processor is now supported via the 432 -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent 433 target attributes and pragmas. 434 * Code generation for the ARM Cortex-A57 processor is improved. Among 435 general code generation improvements, a better algorithm is added 436 for allocating registers to floating-point multiply-accumulate 437 instructions offering increased performance when compiling with 438 -mcpu=cortex-a57 or -mtune=cortex-a57. 439 * Code generation for the ARM Cortex-A53 processor is improved. A 440 more accurate instruction scheduling model for the processor is now 441 used, and a number of compiler tuning parameters have been set to 442 offer increased performance when compiling with -mcpu=cortex-a53 or 443 -mtune=cortex-a53. 444 * Code generation for the Samsung Exynos M1 processor is improved. A 445 more accurate instruction scheduling model for the processor is now 446 used, and a number of compiler tuning parameters have been set to 447 offer increased performance when compiling with -mcpu=exynos-m1 or 448 -mtune=exynos-m1. 449 * Improvements in the generation of conditional branches and literal 450 pools allow the compiler to compile functions of a large size. 451 Constant pools are now placed into separate rodata sections. The 452 new option -mpc-relative-literal-loads generates per-function 453 literal pools, limiting the maximum size of functions to 1MiB. 454 * Several correctness issues generating Advanced SIMD instructions 455 for big-endian targets have been fixed resulting in improved code 456 generation for ACLE intrinsics with -mbig-endian. 457 458 ARM 459 460 * Support for revisions of the ARM architecture prior to ARMv4t has 461 been deprecated and will be removed in a future GCC release. The 462 -mcpu and -mtune values that are deprecated are: arm2, arm250, 463 arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, 464 arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, 465 arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm, 466 strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value 467 arm7tdmi is still supported. The values of -march that are 468 deprecated are: armv2,armv2a,armv3,armv3m,armv4. 469 * The ARM port now supports target attributes and pragmas. Please 470 refer to the [17]documentation for details of available attributes 471 and pragmas as well as usage instructions. 472 * Support has been added for the following processors (GCC 473 identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM 474 Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC 475 identifiers can be used as arguments to the -mcpu or -mtune 476 options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35. 477 478 Heterogeneous Systems Architecture 479 480 * GCC can now generate HSAIL (Heterogeneous System Architecture 481 Intermediate Language) for simple OpenMP device constructs if 482 configured with --enable-offload-targets=hsa. A new libgomp plugin 483 then runs the HSA GPU kernels implementing these constructs on HSA 484 capable GPUs via a standard HSA run time. 485 If the HSA compilation back end determines it cannot output HSAIL 486 for a particular input, it gives a warning by default. These 487 warnings can be suppressed with -Wno-hsa. To give a few examples, 488 the HSA back end does not implement compilation of code using 489 function pointers, automatic allocation of variable sized arrays, 490 functions with variadic arguments as well as a number of other less 491 common programming constructs. 492 When compilation for HSA is enabled, the compiler attempts to 493 compile composite OpenMP constructs 494 495 #pragma omp target teams distribute parallel for 496 497 into parallel HSA GPU kernels. 498 499 IA-32/x86-64 500 501 * GCC now supports the Intel CPU named Skylake with AVX-512 502 extensions through -march=skylake-avx512. The switch enables the 503 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 504 AVX-512DQ. 505 * Support for new AMD instructions monitorx and mwaitx has been 506 added. This includes new intrinsic and built-in support. It is 507 enabled through option -mmwaitx. The instructions monitorx and 508 mwaitx implement the same functionality as the old monitor and 509 mwait instructions. In addition mwaitx adds a configurable timer. 510 The timer value is received as third argument and stored in 511 register %ebx. 512 * x86-64 targets now allow stack realignment from a word-aligned 513 stack pointer using the command-line option -mstackrealign or 514 __attribute__ ((force_align_arg_pointer)). This allows functions 515 compiled with a vector-aligned stack to be invoked from objects 516 that keep only word-alignment. 517 * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These 518 can be used to access data via the %fs and %gs segments without 519 having to resort to inline assembly. Please refer to the 520 [18]documentation for usage instructions. 521 * Support for AMD Zen (family 17h) processors is now available 522 through the -march=znver1 and -mtune=znver1 options. 523 524 MeP 525 526 * Support for the MeP (mep-elf) architecture has been deprecated and 527 will be removed in a future GCC release. 528 529 MSP430 530 531 * The MSP430 compiler now has the ability to automatically distribute 532 code and data between low memory (addresses below 64K) and high 533 memory. This only applies to parts that actually have both memory 534 regions and only if the linker script for the part has been 535 specifically set up to support this feature. 536 A new attribute of either can be applied to both functions and 537 data, and this tells the compiler to place the object into low 538 memory if there is room and into high memory otherwise. Two other 539 new attributes - lower and upper - can be used to explicitly state 540 that an object should be placed in the specified memory region. If 541 there is not enough left in that region the compilation will fail. 542 Two new command-line options - -mcode-region=[lower|upper|either] 543 and -mdata-region=[lower|upper|either] - can be used to tell the 544 compiler what to do with objects that do not have one of these new 545 attributes. 546 547 PowerPC / PowerPC64 / RS6000 548 549 * PowerPC64 now supports IEEE 128-bit floating-point using the 550 __float128 data type. In GCC 6, this is not enabled by default, but 551 you can enable it with -mfloat128. The IEEE 128-bit floating-point 552 support requires the use of the VSX instruction set. IEEE 128-bit 553 floating-point values are passed and returned as a single vector 554 value. The software emulator for IEEE 128-bit floating-point 555 support is only built on PowerPC GNU/Linux systems where the 556 default CPU is at least power7. On future ISA 3.0 systems (POWER 9 557 and later), you will be able to use the -mfloat128-hardware option 558 to use the ISA 3.0 instructions that support IEEE 128-bit 559 floating-point. An additional type (__ibm128) has been added to 560 refer to the IBM extended double type that normally implements long 561 double. This will allow for a future transition to implementing 562 long double with IEEE 128-bit floating-point. 563 * Basic support has been added for POWER9 hardware that will use the 564 recently published OpenPOWER ISA 3.0 instructions. The following 565 new switches are available: 566 + -mcpu=power9: Implement all of the ISA 3.0 instructions 567 supported by the compiler. 568 + -mtune=power9: In the future, apply tuning for POWER9 systems. 569 Currently, POWER8 tunings are used. 570 + -mmodulo: Generate code using the ISA 3.0 integer instructions 571 (modulus, count trailing zeros, array index support, integer 572 multiply/add). 573 + -mpower9-fusion: Generate code to suitably fuse instruction 574 sequences for a POWER9 system. 575 + -mpower9-dform: Generate code to use the new D-form 576 (register+offset) memory instructions for the vector 577 registers. 578 + -mpower9-vector: Generate code using the new ISA 3.0 vector 579 (VSX or Altivec) instructions. 580 + -mpower9-minmax: Reserved for future development. 581 + -mtoc-fusion: Keep TOC entries together to provide more fusion 582 opportunities. 583 * New constraints have been added to support IEEE 128-bit 584 floating-point and ISA 3.0 instructions: 585 + wb: Altivec register if -mpower9-dform is enabled. 586 + we: VSX register if -mpower9-vector is enabled for 64-bit code 587 generation. 588 + wo: VSX register if -mpower9-vector is enabled. 589 + wp: Reserved for future use if long double is implemented with 590 IEEE 128-bit floating-point instead of IBM extended double. 591 + wq: VSX register if -mfloat128 is enabled. 592 + wF: Memory operand suitable for POWER9 fusion load/store. 593 + wG: Memory operand suitable for TOC fusion memory references. 594 + wL: Integer constant identifying the element number mfvsrld 595 accesses within a vector. 596 * Support has been added for __builtin_cpu_is() and 597 __builtin_cpu_supports(), allowing for very fast access to 598 AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of 599 glibc 2.23 or later. 600 * All hardware transactional memory builtins now correctly behave as 601 memory barriers. Programmers can use #ifdef __TM_FENCE__ to 602 determine whether their "old" compiler treats the builtins as 603 barriers. 604 * Split-stack support has been added for gccgo on PowerPC64 for both 605 big- and little-endian (but not for 32-bit). The gold linker from 606 at least binutils 2.25.1 must be available in the PATH when 607 configuring and building gccgo to enable split stack. (The 608 requirement for binutils 2.25.1 applies to PowerPC64 only.) The 609 split-stack feature allows a small initial stack size to be 610 allocated for each goroutine, which increases as needed. 611 * GCC on PowerPC now supports the standard lround function. 612 * A new configuration option ---with-advance-toolchain=at was added 613 for PowerPC 64-bit GNU/Linux systems to use the header files, 614 library files, and the dynamic linker from a specific Advance 615 Toolchain release instead of the default versions that are provided 616 by the GNU/Linux distribution. In general, this option is intended 617 for the developers of GCC, and it is not intended for general use. 618 * The "q", "S", "T", and "t" asm-constraints have been removed. 619 * The "b", "B", "m", "M", and "W" format modifiers have been removed. 620 621 S/390, System z, IBM z Systems 622 623 * Support for the IBM z13 processor has been added. When using the 624 -march=z13 option, the compiler will generate code making use of 625 the new instructions and registers introduced with the vector 626 extension facility. The -mtune=z13 option enables z13 specific 627 instruction scheduling without making use of new instructions. 628 Compiling code with -march=z13 reduces the default alignment of 629 vector types bigger than 8 bytes to 8. This is an ABI change and 630 care must be taken when linking modules compiled with different 631 arch levels which interchange variables containing vector type 632 values. For newly compiled code the GNU linker will emit a warning. 633 * The -mzvector option enables a C/C++ language extension. This 634 extension provides a new keyword vector which can be used to define 635 vector type variables. (Note: This is not available when enforcing 636 strict standard compliance e.g. with -std=c99. Either enable GNU 637 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 638 Additionally a set of overloaded builtins is provided which is 639 partially compatible to the PowerPC Altivec builtins. In order to 640 make use of these builtins the vecintrin.h header file needs to be 641 included. 642 * The new command-line options -march=native, and -mtune=native are 643 now available on native IBM z Systems. Specifying these options 644 causes GCC to auto-detect the host CPU and choose the optimal 645 setting for that system. If GCC is unable to detect the host CPU 646 these options have no effect. 647 * The IBM z Systems port now supports target attributes and pragmas. 648 Please refer to the [19]documentation for details of available 649 attributes and pragmas as well as usage instructions. 650 * -fsplit-stack is now supported as part of the IBM z Systems port. 651 This feature requires a recent gold linker to be used. 652 * Support for the g5 and g6 -march=/-mtune= CPU level switches has 653 been deprecated and will be removed in a future GCC release. -m31 654 from now on defaults to -march=z900 if not specified otherwise. 655 -march=native on a g5/g6 machine will default to -march=z900. 656 657 SH 658 659 * Support for SH5 / SH64 has been declared obsolete and will be 660 removed in future releases. 661 * Support for the FDPIC ABI has been added. It can be enabled using 662 the new -mfdpic target option and --enable-fdpic configure option. 663 664 SPARC 665 666 * An ABI bug has been fixed in 64-bit mode. Unfortunately, this 667 change will break binary compatibility with earlier releases for 668 code it affects, but this should be pretty rare in practice. The 669 conditions are: a 16-byte structure containing a double or a 8-byte 670 vector in the second half is passed to a subprogram in slot #15, 671 for example as 16th parameter if the first 15 ones have at most 8 672 bytes. The double or vector was wrongly passed in floating-point 673 register %d32 in lieu of on the stack as per the SPARC calling 674 conventions. 675 676 Operating Systems 677 678 AIX 679 680 * DWARF debugging support for AIX 7.1 has been enabled as an optional 681 debugging format. A more recent Technology Level (TL) and GCC built 682 with that level are required for full exploitation of DWARF 683 debugging capabilities. 684 685 Linux 686 687 * Support for the [20]musl C library was added for the AArch64, ARM, 688 MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and 689 x86_64 targets. It can be selected using the new -mmusl option in 690 case musl is not the default libc. GCC defaults to musl libc if it 691 is built with a target triplet matching the *-linux-musl* pattern. 692 693 RTEMS 694 695 * The RTEMS thread model implementation changed. Mutexes now use 696 self-contained objects defined in Newlib <sys/lock.h> instead of 697 Classic API semaphores. The keys for thread specific data and the 698 once function are directly defined via <pthread.h>. Self-contained 699 condition variables are provided via Newlib <sys/lock.h>. The RTEMS 700 thread model also supports C++11 threads. 701 * OpenMP support now uses self-contained objects provided by Newlib 702 <sys/lock.h> and offers a significantly better performance compared 703 to the POSIX configuration of libgomp. It is possible to configure 704 thread pools for each scheduler instance via the environment 705 variable GOMP_RTEMS_THREAD_POOLS. 706 707 Solaris 708 709 * Solaris 12 is now fully supported. Minimal support had already been 710 present in GCC 5.3. 711 * Solaris 12 provides a full set of startup files (crt1.o, crti.o, 712 crtn.o), which GCC now prefers over its own ones. 713 * Position independent executables (PIE) are now supported on Solaris 714 12. 715 * Constructor priority is now supported on Solaris 12 with the system 716 linker. 717 * libvtv has been ported to Solaris 11 and up. 718 719 Windows 720 721 * The option -mstackrealign is now automatically activated in 32-bit 722 mode whenever the use of SSE instructions is requested. 723 724 Other significant improvements 725 726 * The gcc and g++ driver programs will now provide suggestions for 727 misspelled command-line options. 728 729 $ gcc -static-libfortran test.f95 730 gcc: error: unrecognized command line option '-static-libfortran'; did you mean 731 '-static-libgfortran'? 732 733 * The --enable-default-pie configure option enables generation of PIE 734 by default. 735 736 GCC 6.2 737 738 This is the [21]list of problem reports (PRs) from GCC's bug tracking 739 system that are known to be fixed in the 6.2 release. This list might 740 not be complete (that is, it is possible that some PRs that have been 741 fixed are not listed here). 742 743 Target Specific Changes 744 745 SPARC 746 747 * Support for --with-cpu-32 and --with-cpu-64 configure options has 748 been added on bi-architecture platforms. 749 * Support for the SPARC M7 (Niagara 7) processor has been added. 750 * Support for the VIS 4.0 instruction set has been added. 751 752 GCC 6.3 753 754 This is the [22]list of problem reports (PRs) from GCC's bug tracking 755 system that are known to be fixed in the 6.3 release. This list might 756 not be complete (that is, it is possible that some PRs that have been 757 fixed are not listed here). 758 759 Target Specific Changes 760 761 IA-32/x86-64 762 763 * Support for the [23]deprecated pcommit instruction has been 764 removed. 765 766 GCC 6.4 767 768 This is the [24]list of problem reports (PRs) from GCC's bug tracking 769 system that are known to be fixed in the 6.4 release. This list might 770 not be complete (that is, it is possible that some PRs that have been 771 fixed are not listed here). 772 773 Operating Systems 774 775 RTEMS 776 777 * The ABI changes on ARM so that no short enums are used by default. 778 779 GCC 6.5 780 781 This is the [25]list of problem reports (PRs) from GCC's bug tracking 782 system that are known to be fixed in the 6.5 release. This list might 783 not be complete (that is, it is possible that some PRs that have been 784 fixed are not listed here). 785 786 787 For questions related to the use of GCC, please consult these web 788 pages and the [26]GCC manuals. If that fails, the 789 [27]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 790 web pages and the development of GCC are welcome on our developer 791 list at [28]gcc (a] gcc.gnu.org. All of [29]our lists have public 792 archives. 793 794 Copyright (C) [30]Free Software Foundation, Inc. Verbatim copying and 795 distribution of this entire article is permitted in any medium, 796 provided this notice is preserved. 797 798 These pages are [31]maintained by the GCC team. Last modified 799 2018-10-26[32]. 800 801 References 802 803 1. http://gcc.gnu.org/gcc-6/porting_to.html 804 2. http://gcc.gnu.org/onlinedocs/index.html#current 805 3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html 806 4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151 807 5. https://www.openacc.org/ 808 6. https://gcc.gnu.org/wiki/OpenACC 809 7. https://gcc.gnu.org/wiki/Offloading 810 8. https://www.openmp.org/specifications/ 811 9. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266 812 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf 813 11. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 814 12. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html 815 13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch 816 14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks 817 15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option 818 16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes 819 17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes 820 18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces 821 19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes 822 20. http://www.musl-libc.org/ 823 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2 824 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3 825 23. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction 826 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4 827 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5 828 26. https://gcc.gnu.org/onlinedocs/ 829 27. mailto:gcc-help (a] gcc.gnu.org 830 28. mailto:gcc (a] gcc.gnu.org 831 29. https://gcc.gnu.org/lists.html 832 30. https://www.fsf.org/ 833 31. https://gcc.gnu.org/about.html 834 32. http://validator.w3.org/check/referer 835 ====================================================================== 836 http://gcc.gnu.org/gcc-5/index.html 837 GCC 5 Release Series 838 839 October 10, 2017 840 841 The [1]GNU project and the GCC developers are pleased to announce the 842 release of GCC 5.5. 843 844 This release is a bug-fix release, containing fixes for regressions in 845 GCC 5.4 relative to previous releases of GCC. 846 847 Release History 848 849 GCC 5.5 850 October 10, 2017 ([2]changes, [3]documentation) 851 852 GCC 5.4 853 June 3, 2016 ([4]changes, [5]documentation) 854 855 GCC 5.3 856 December 4, 2015 ([6]changes, [7]documentation) 857 858 GCC 5.2 859 July 16, 2015 ([8]changes, [9]documentation) 860 861 GCC 5.1 862 April 22, 2015 ([10]changes, [11]documentation) 863 864 References and Acknowledgements 865 866 GCC used to stand for the GNU C Compiler, but since the compiler 867 supports several other languages aside from C, it now stands for the 868 GNU Compiler Collection. 869 870 A list of [12]successful builds is updated as new information becomes 871 available. 872 873 The GCC developers would like to thank the numerous people that have 874 contributed new features, improvements, bug fixes, and other changes as 875 well as test results to GCC. This [13]amazing group of volunteers is 876 what makes GCC successful. 877 878 For additional information about GCC please refer to the [14]GCC 879 project web site or contact the [15]GCC development mailing list. 880 881 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 882 883 884 For questions related to the use of GCC, please consult these web 885 pages and the [18]GCC manuals. If that fails, the 886 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 887 web pages and the development of GCC are welcome on our developer 888 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 889 archives. 890 891 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 892 distribution of this entire article is permitted in any medium, 893 provided this notice is preserved. 894 895 These pages are [23]maintained by the GCC team. Last modified 896 2018-09-30[24]. 897 898 References 899 900 1. http://www.gnu.org/ 901 2. http://gcc.gnu.org/gcc-5/changes.html 902 3. http://gcc.gnu.org/onlinedocs/5.5.0/ 903 4. http://gcc.gnu.org/gcc-5/changes.html 904 5. http://gcc.gnu.org/onlinedocs/5.4.0/ 905 6. http://gcc.gnu.org/gcc-5/changes.html 906 7. http://gcc.gnu.org/onlinedocs/5.3.0/ 907 8. http://gcc.gnu.org/gcc-5/changes.html 908 9. http://gcc.gnu.org/onlinedocs/5.2.0/ 909 10. http://gcc.gnu.org/gcc-5/changes.html 910 11. http://gcc.gnu.org/onlinedocs/5.1.0/ 911 12. http://gcc.gnu.org/gcc-5/buildstat.html 912 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 913 14. http://gcc.gnu.org/index.html 914 15. mailto:gcc (a] gcc.gnu.org 915 16. http://gcc.gnu.org/mirrors.html 916 17. http://gcc.gnu.org/svn.html 917 18. https://gcc.gnu.org/onlinedocs/ 918 19. mailto:gcc-help (a] gcc.gnu.org 919 20. mailto:gcc (a] gcc.gnu.org 920 21. https://gcc.gnu.org/lists.html 921 22. https://www.fsf.org/ 922 23. https://gcc.gnu.org/about.html 923 24. http://validator.w3.org/check/referer 924 ====================================================================== 925 http://gcc.gnu.org/gcc-5/changes.html 926 GCC 5 Release Series 927 Changes, New Features, and Fixes 928 929 Caveats 930 931 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 932 * The C++ runtime library (libstdc++) uses a new ABI by default (see 933 [1]below). 934 * The Graphite framework for loop optimizations no longer requires 935 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 936 The installation manual contains more information about 937 requirements to build GCC. 938 * The non-standard C++0x type traits has_trivial_default_constructor, 939 has_trivial_copy_constructor and has_trivial_copy_assign have been 940 deprecated and will be removed in a future version. The standard 941 C++11 traits is_trivially_default_constructible, 942 is_trivially_copy_constructible and is_trivially_copy_assignable 943 should be used instead. 944 * On AVR, support has been added for the devices 945 ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer. 946 * The AVR port uses a new scheme to describe supported devices: For 947 each supported device the compiler provides a device-specific 948 [2]spec file. If the compiler is used together with AVR-LibC, this 949 requires at least GCC 5.2 and a version of AVR-LibC which 950 implements [3]feature #44574. 951 952 General Optimizer Improvements 953 954 * Inter-procedural optimization improvements: 955 + An Identical Code Folding (ICF) pass (controlled via 956 -fipa-icf) has been added. Compared to the identical code 957 folding performed by the Gold linker this pass does not 958 require function sections. It also performs merging before 959 inlining, so inter-procedural optimizations are aware of the 960 code re-use. On the other hand not all unifications performed 961 by a linker are doable by GCC which must honor aliasing 962 information. During link-time optimization of Firefox, this 963 pass unifies about 31000 functions, that is 14% overall. 964 + The devirtualization pass was significantly improved by adding 965 better support for speculative devirtualization and dynamic 966 type detection. About 50% of virtual calls in Firefox are now 967 speculatively devirtualized during link-time optimization. 968 + A new comdat localization pass allows the linker to eliminate 969 more dead code in presence of C++ inline functions. 970 + Virtual tables are now optimized. Local aliases are used to 971 reduce dynamic linking time of C++ virtual tables on ELF 972 targets and data alignment has been reduced to limit data 973 segment bloat. 974 + A new -fno-semantic-interposition option can be used to 975 improve code quality of shared libraries where interposition 976 of exported symbols is not allowed. 977 + Write-only variables are now detected and optimized out. 978 + With profile feedback the function inliner can now bypass 979 --param inline-insns-auto and --param inline-insns-single 980 limits for hot calls. 981 + The IPA reference pass was significantly sped up making it 982 feasible to enable -fipa-reference with -fprofile-generate. 983 This also solves a bottleneck seen when building Chromium with 984 link-time optimization. 985 + The symbol table and call-graph API was reworked to C++ and 986 simplified. 987 + The interprocedural propagation of constants now also 988 propagates alignments of pointer parameters. This for example 989 means that the vectorizer often does not need to generate loop 990 prologues and epilogues to make up for potential 991 misalignments. 992 * Link-time optimization improvements: 993 + One Definition Rule based merging of C++ types has been 994 implemented. Type merging enables better devirtualization and 995 alias analysis. Streaming extra information needed to merge 996 types adds about 2-6% of memory size and object size increase. 997 This can be controlled by -flto-odr-type-merging. 998 + Command-line optimization and target options are now streamed 999 on a per-function basis and honored by the link-time 1000 optimizer. This change makes link-time optimization a more 1001 transparent replacement of per-file optimizations. It is now 1002 possible to build projects that require different optimization 1003 settings for different translation units (such as -ffast-math, 1004 -mavx, or -finline). Contrary to earlier GCC releases, the 1005 optimization and target options passed on the link command 1006 line are ignored. 1007 Note that this applies only to those command-line options that 1008 can be passed to optimize and target attributes. Command-line 1009 options affecting global code generation (such as -fpic), 1010 warnings (such as -Wodr), optimizations affecting the way 1011 static variables are optimized (such as -fcommon), debug 1012 output (such as -g), and --param parameters can be applied 1013 only to the whole link-time optimization unit. In these cases, 1014 it is recommended to consistently use the same options at both 1015 compile time and link time. 1016 + GCC bootstrap now uses slim LTO object files. 1017 + Memory usage and link times were improved. Tree merging was 1018 sped up, memory usage of GIMPLE declarations and types was 1019 reduced, and, support for on-demand streaming of variable 1020 constructors was added. 1021 * Feedback directed optimization improvements: 1022 + A new auto-FDO mode uses profiles collected by low overhead 1023 profiling tools (perf) instead of more expensive program 1024 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 1025 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 1026 traditional feedback directed optimization. 1027 + Profile precision was improved in presence of C++ inline and 1028 extern inline functions. 1029 + The new gcov-tool utility allows manipulating profiles. 1030 + Profiles are now more tolerant to source file changes (this 1031 can be controlled by --param profile-func-internal-id). 1032 * Register allocation improvements: 1033 + A new local register allocator (LRA) sub-pass, controlled by 1034 -flra-remat, implements control-flow sensitive global register 1035 rematerialization. Instead of spilling and restoring a 1036 register value, it is recalculated if it is profitable. The 1037 sub-pass improved SPEC2000 generated code by 1% and 0.5% 1038 correspondingly on ARM and x86-64. 1039 + Reuse of the PIC hard register, instead of using a fixed 1040 register, was implemented on x86/x86-64 targets. This improves 1041 generated PIC code performance as more hard registers can be 1042 used. Shared libraries can significantly benefit from this 1043 optimization. Currently it is switched on only for x86/x86-64 1044 targets. As RA infrastructure is already implemented for PIC 1045 register reuse, other targets might follow this in the future. 1046 + A simple form of inter-procedural RA was implemented. When it 1047 is known that a called function does not use caller-saved 1048 registers, save/restore code is not generated around the call 1049 for such registers. This optimization can be controlled by 1050 -fipa-ra 1051 + LRA is now much more effective at generating spills of general 1052 registers into vector registers instead of memory on 1053 architectures (e.g., modern Intel processors) where this is 1054 profitable. 1055 * UndefinedBehaviorSanitizer gained a few new sanitization options: 1056 + -fsanitize=float-divide-by-zero: detect floating-point 1057 division by zero; 1058 + -fsanitize=float-cast-overflow: check that the result of 1059 floating-point type to integer conversions do not overflow; 1060 + -fsanitize=bounds: enable instrumentation of array bounds and 1061 detect out-of-bounds accesses; 1062 + -fsanitize=alignment: enable alignment checking, detect 1063 various misaligned objects; 1064 + -fsanitize=object-size: enable object size checking, detect 1065 various out-of-bounds accesses. 1066 + -fsanitize=vptr: enable checking of C++ member function calls, 1067 member accesses and some conversions between pointers to base 1068 and derived classes, detect if the referenced object does not 1069 have the correct dynamic type. 1070 * Pointer Bounds Checker, a bounds violation detector, has been added 1071 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 1072 instrumented with run-time checks of used pointers against their 1073 bounds to detect pointer bounds violations (overflows). The Pointer 1074 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 1075 new ISA extension Intel MPX support. See the Pointer Bounds Checker 1076 [4]Wiki page for more details. 1077 1078 New Languages and Language specific improvements 1079 1080 * [5]OpenMP 4.0 specification offloading features are now supported 1081 by the C, C++, and Fortran compilers. Generic changes: 1082 + Infrastructure (suitable for any vendor). 1083 + Testsuite which covers offloading from the [6]OpenMP 4.0 1084 Examples document. 1085 Specific for upcoming Intel Xeon Phi products: 1086 + Run-time library. 1087 + Card emulator. 1088 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 1089 specification. OpenACC is intended for programming accelerator 1090 devices such as GPUs. See [7]the OpenACC wiki page for more 1091 information. 1092 1093 C family 1094 1095 * The default setting of the -fdiagnostics-color= command-line option 1096 is now [8]configurable when building GCC using configuration option 1097 --with-diagnostics-color=. The possible values are: never, always, 1098 auto and auto-if-env. The new default auto uses color only when the 1099 standard error is a terminal. The default in GCC 4.9 was 1100 auto-if-env, which is equivalent to auto if there is a non-empty 1101 GCC_COLORS environment variable, and never otherwise. As in GCC 1102 4.9, an empty GCC_COLORS variable in the environment will always 1103 disable colors, no matter what the default is or what command-line 1104 options are used. 1105 * A new command-line option -Wswitch-bool has been added for the C 1106 and C++ compilers, which warns whenever a switch statement has an 1107 index of boolean type. 1108 * A new command-line option -Wlogical-not-parentheses has been added 1109 for the C and C++ compilers, which warns about "logical not" used 1110 on the left hand side operand of a comparison. 1111 * A new command-line option -Wsizeof-array-argument has been added 1112 for the C and C++ compilers, which warns when the sizeof operator 1113 is applied to a parameter that has been declared as an array in a 1114 function definition. 1115 * A new command-line option -Wbool-compare has been added for the C 1116 and C++ compilers, which warns about boolean expressions compared 1117 with an integer value different from true/false. 1118 * Full support for [9]Cilk Plus has been added to the GCC compiler. 1119 Cilk Plus is an extension to the C and C++ languages to support 1120 data and task parallelism. 1121 * A new attribute no_reorder prevents reordering of selected symbols 1122 against other such symbols or inline assembler. This enables to 1123 link-time optimize the Linux kernel without having to resort to 1124 -fno-toplevel-reorder that disables several optimizations. 1125 * New preprocessor constructs, __has_include and __has_include_next, 1126 to test the availability of headers have been added. 1127 This demonstrates a way to include the header <optional> only if it 1128 is available: 1129 1130 #ifdef __has_include 1131 # if __has_include(<optional>) 1132 # include <optional> 1133 # define have_optional 1 1134 # elif __has_include(<experimental/optional>) 1135 # include <experimental/optional> 1136 # define have_optional 1 1137 # define experimental_optional 1138 # else 1139 # define have_optional 0 1140 # endif 1141 #endif 1142 1143 The header search paths for __has_include and __has_include_next 1144 are equivalent to those of the standard directive #include and the 1145 extension #include_next respectively. 1146 * A new built-in function-like macro to determine the existence of an 1147 attribute, __has_attribute, has been added. The equivalent built-in 1148 macro __has_cpp_attribute was added to C++ to support 1149 [10]Feature-testing recommendations for C++. The macro 1150 __has_attribute is added to all C-like languages as an extension: 1151 1152 int 1153 #ifdef __has_attribute 1154 # if __has_attribute(__noinline__) 1155 __attribute__((__noinline__)) 1156 # endif 1157 #endif 1158 foo(int x); 1159 1160 If an attribute exists, a nonzero constant integer is returned. For 1161 standardized C++ attributes a date is returned, otherwise the 1162 constant returned is 1. Both __has_attribute and 1163 __has_cpp_attribute will add underscores to an attribute name if 1164 necessary to resolve the name. For C++11 and onwards the attribute 1165 may be scoped. 1166 * A new set of built-in functions for arithmetics with overflow 1167 checking has been added: __builtin_add_overflow, 1168 __builtin_sub_overflow and __builtin_mul_overflow and for 1169 compatibility with clang also other variants. These builtins have 1170 two integral arguments (which don't need to have the same type), 1171 the arguments are extended to infinite precision signed type, +, - 1172 or * is performed on those, and the result is stored in an integer 1173 variable pointed to by the last argument. If the stored value is 1174 equal to the infinite precision result, the built-in functions 1175 return false, otherwise true. The type of the integer variable that 1176 will hold the result can be different from the types of the first 1177 two arguments. The following snippet demonstrates how this can be 1178 used in computing the size for the calloc function: 1179 1180 void * 1181 calloc (size_t x, size_t y) 1182 { 1183 size_t sz; 1184 if (__builtin_mul_overflow (x, y, &sz)) 1185 return NULL; 1186 void *ret = malloc (sz); 1187 if (ret) memset (res, 0, sz); 1188 return ret; 1189 } 1190 1191 On e.g. i?86 or x86-64 the above will result in a mul instruction 1192 followed by a jump on overflow. 1193 * The option -fextended-identifiers is now enabled by default for 1194 C++, and for C99 and later C versions. Various bugs in the 1195 implementation of extended identifiers have been fixed. 1196 1197 C 1198 1199 * The default mode has been changed to -std=gnu11. 1200 * A new command-line option -Wc90-c99-compat has been added to warn 1201 about features not present in ISO C90, but present in ISO C99. 1202 * A new command-line option -Wc99-c11-compat has been added to warn 1203 about features not present in ISO C99, but present in ISO C11. 1204 * It is possible to disable warnings about conversions between 1205 pointers that have incompatible types via a new warning option 1206 -Wno-incompatible-pointer-types; warnings about implicit 1207 incompatible integer to pointer and pointer to integer conversions 1208 via a new warning option -Wno-int-conversion; and warnings about 1209 qualifiers on pointers being discarded via a new warning option 1210 -Wno-discarded-qualifiers. 1211 * To allow proper use of const qualifiers with multidimensional 1212 arrays, GCC will not warn about incompatible pointer types anymore 1213 for conversions between pointers to arrays with and without const 1214 qualifier (except when using -pedantic). Instead, a new warning is 1215 emitted only if the const qualifier is lost. This can be controlled 1216 with a new warning option -Wno-discarded-array-qualifiers. 1217 * The C front end now generates more precise caret diagnostics. 1218 * The -pg command-line option now only affects the current file in an 1219 LTO build. 1220 1221 C++ 1222 1223 * G++ now supports [11]C++14 variable templates. 1224 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 1225 * Excessive template instantiation depth is now a fatal error. This 1226 prevents excessive diagnostics that usually do not help to identify 1227 the problem. 1228 * G++ and libstdc++ now implement the feature-testing macros from 1229 [12]Feature-testing recommendations for C++. 1230 * G++ now allows typename in a template template parameter. 1231 1232 template<template<typename> typename X> struct D; // OK 1233 1234 * G++ now supports [13]C++14 aggregates with non-static data member 1235 initializers. 1236 1237 struct A { int i, j = i; }; 1238 A a = { 42 }; // a.j is also 42 1239 1240 * G++ now supports [14]C++14 extended constexpr. 1241 1242 constexpr int f (int i) 1243 { 1244 int j = 0; 1245 for (; i > 0; --i) 1246 ++j; 1247 return j; 1248 } 1249 1250 constexpr int i = f(42); // i is 42 1251 1252 * G++ now supports the [15]C++14 sized deallocation functions. 1253 1254 void operator delete (void *, std::size_t) noexcept; 1255 void operator delete[] (void *, std::size_t) noexcept; 1256 1257 * A new One Definition Rule violation warning (controlled by -Wodr) 1258 detects mismatches in type definitions and virtual table contents 1259 during link-time optimization. 1260 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 1261 developers to annotate programs with final specifiers (or anonymous 1262 namespaces) to improve code generation. These warnings can be used 1263 at compile time, but they are more useful in combination with 1264 link-time optimization. 1265 * G++ no longer supports [16]N3639 variable length arrays, as they 1266 were removed from the C++14 working paper prior to ratification. 1267 GNU VLAs are still supported, so VLA support is now the same in 1268 C++14 mode as in C++98 and C++11 modes. 1269 * G++ now allows passing a non-trivially-copyable class via C 1270 varargs, which is conditionally-supported with 1271 implementation-defined semantics in the standard. This uses the 1272 same calling convention as a normal value parameter. 1273 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 1274 various mangling bugs are fixed, but G++ will still emit aliases 1275 with the old, wrong mangling where feasible. -Wabi=2 will warn 1276 about differences between ABI version 2 and the current setting. 1277 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 1278 to be unaffected, but -Wabi=8 will warn about a non-static data 1279 member with type std::nullptr_t which changes position due to this 1280 change. 1281 1282 Runtime Library (libstdc++) 1283 1284 * A [17]Dual ABI is provided by the library. A new ABI is enabled by 1285 default. The old ABI is still supported and can be used by defining 1286 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 1287 standard library headers. 1288 * A new implementation of std::string is enabled by default, using 1289 the small string optimization instead of copy-on-write reference 1290 counting. 1291 * A new implementation of std::list is enabled by default, with an 1292 O(1) size() function; 1293 * [18]Full support for C++11, including the following new features: 1294 + std::deque and std::vector<bool> meet the allocator-aware 1295 container requirements; 1296 + movable and swappable iostream classes; 1297 + support for std::align and std::aligned_union; 1298 + type traits std::is_trivially_copyable, 1299 std::is_trivially_constructible, std::is_trivially_assignable 1300 etc.; 1301 + I/O manipulators std::put_time, std::get_time, std::hexfloat 1302 and std::defaultfloat; 1303 + generic locale-aware std::isblank; 1304 + locale facets for Unicode conversion; 1305 + atomic operations for std::shared_ptr; 1306 + std::notify_all_at_thread_exit() and functions for making 1307 futures ready at thread exit. 1308 * Support for the C++11 hexfloat manipulator changes how the num_put 1309 facet formats floating point types when 1310 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 1311 This change affects all language modes, even though the C++98 1312 standard gave no special meaning to that combination of flags. To 1313 prevent the use of hexadecimal notation for floating point types 1314 use str.unsetf(std::ios_base::floatfield) to clear the relevant 1315 bits in str.flags(). 1316 * [19]Full experimental support for C++14, including the following 1317 new features: 1318 + std::is_final type trait; 1319 + heterogeneous comparison lookup in associative containers. 1320 + global functions cbegin, cend, rbegin, rend, crbegin, and 1321 crend for range access to containers, arrays and initializer 1322 lists. 1323 * [20]Improved experimental support for the Library Fundamentals TS, 1324 including: 1325 + class std::experimental::any; 1326 + function template std::experimental::apply; 1327 + function template std::experimental::sample; 1328 + function template std::experimental::search and related 1329 searcher types; 1330 + variable templates for type traits; 1331 + function template std::experimental::not_fn. 1332 * New random number distributions logistic_distribution and 1333 uniform_on_sphere_distribution as extensions. 1334 * [21]GDB Xmethods for containers and std::unique_ptr. 1335 1336 Fortran 1337 1338 * Compatibility notice: 1339 + The version of the module files (.mod) has been incremented. 1340 + For free-form source files [22]-Werror=line-truncation is now 1341 enabled by default. Note that comments exceeding the line 1342 length are not diagnosed. (For fixed-form source code, the 1343 same warning is available but turned off by default, such that 1344 excess characters are ignored. -ffree-line-length-n and 1345 -ffixed-line-length-n can be used to modify the default line 1346 lengths of 132 and 72 columns, respectively.) 1347 + The -Wtabs option is now more sensible: with -Wtabs the 1348 compiler warns if it encounters tabs and with -Wno-tabs this 1349 warning is turned off. Before, -Wno-tabs warned and -Wtabs 1350 disabled the warning. As before, this warning is also enabled 1351 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 1352 options of -std=. 1353 * Incomplete support for colorizing diagnostics emitted by gfortran 1354 has been added. The option [23]-fdiagnostics-color controls when 1355 color is used in diagnostics. The default value of this option can 1356 be [24]configured when building GCC. The GCC_COLORS environment 1357 variable can be used to customize the colors or disable coloring 1358 completely. Sample diagnostics output: 1359 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 1360 test.f90:6:1: 1361 1362 0 continue 1363 1 1364 Error: Zero is not a valid statement label at (1) 1365 test.f90:9:6: 1366 1367 USE foo 1368 1 1369 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 1370 1371 * The -Wuse-without-only option has been added to warn when a USE 1372 statement has no ONLY qualifier and thus implicitly imports all 1373 public entities of the used module. 1374 * Formatted READ and WRITE statements now work correctly in 1375 locale-aware programs. For more information and potential caveats, 1376 see [25]Section 5.3 Thread-safety of the runtime library in the 1377 manual. 1378 * [26]Fortran 2003: 1379 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 1380 IEEE_ARITHMETIC) are now supported. 1381 * [27]Fortran 2008: 1382 + [28]Coarrays: Full experimental support of Fortran 2008's 1383 coarrays with -fcoarray=lib except for allocatable/pointer 1384 components of derived-type coarrays. GCC currently only ships 1385 with a single-image library (libcaf_single), but multi-image 1386 support based on MPI and GASNet is provided by the libraries 1387 of the [29]OpenCoarrays project. 1388 * TS18508 Additional Parallel Features in Fortran: 1389 + Support for the collective intrinsic subroutines CO_MAX, 1390 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 1391 including -fcoarray=lib support. 1392 + Support for the new atomic intrinsics has been added, 1393 including -fcoarray=lib support. 1394 * Fortran 2015: 1395 + Support for IMPLICIT NONE (external, type). 1396 + ERROR STOP is now permitted in pure procedures. 1397 1398 Go 1399 1400 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 1401 * Building GCC 5 with Go enabled will install two new programs: 1402 [30]go and [31]gofmt. 1403 1404 libgccjit 1405 1406 New in GCC 5 is the ability to build GCC as a shared library for 1407 embedding in other processes (such as interpreters), suitable for 1408 Just-In-Time compilation to machine code. 1409 1410 The shared library has a [32]C API and a [33]C++ wrapper API providing 1411 some "syntactic sugar". There are also bindings available from 3rd 1412 parties for [34]Python and for [35]D. 1413 1414 For example, this library can be used by interpreters for [36]compiling 1415 functions from bytecode to machine code. 1416 1417 The library can also be used for ahead-of-time compilation, enabling 1418 GCC to be plugged into a pre-existing front end. An example of using 1419 this to build a compiler for an esoteric language we'll refer to as 1420 "brainf" can be seen [37]here. 1421 1422 libgccjit is licensed under the GPLv3 (or at your option, any later 1423 version) 1424 1425 It should be regarded as experimental at this time. 1426 1427 New Targets and Target Specific Improvements 1428 1429 Reporting stack usage 1430 1431 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 1432 -fstack-usage option. 1433 1434 AArch64 1435 1436 * Code generation for the ARM Cortex-A57 processor has been improved. 1437 A more accurate instruction scheduling model for the processor is 1438 now used, and a number of compiler tuning parameters have been set 1439 to offer increased performance when compiling with -mcpu=cortex-a57 1440 or -mtune=cortex-a57. 1441 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 1442 and can be enabled by giving the -mfix-cortex-a53-835769 option. 1443 Alternatively it can be enabled by default by configuring GCC with 1444 the --enable-fix-cortex-a53-835769 option. 1445 * The optional cryptographic extensions to the ARMv8-A architecture 1446 are no longer enabled by default when specifying the 1447 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 1448 options. To enable these extensions add +crypto to the value of 1449 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 1450 * Support has been added for the following processors (GCC 1451 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 1452 initial support for its big.LITTLE combination with the ARM 1453 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 1454 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 1455 The GCC identifiers can be used as arguments to the -mcpu or -mtune 1456 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 1457 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 1458 support for the Cortex-A72. 1459 * The transitional options -mlra and -mno-lra have been removed. The 1460 AArch64 backend now uses the local register allocator (LRA) only. 1461 1462 ARM 1463 1464 * Thumb-1 assembly code is now generated in unified syntax. The new 1465 option -masm-syntax-unified specifies whether inline assembly code 1466 is using unified syntax. By default the option is off which means 1467 non-unified syntax is used. However this is subject to change in 1468 future releases. Eventually the non-unified syntax will be 1469 deprecated. 1470 * It is now a configure-time error to use the --with-cpu configure 1471 option with either of --with-tune or --with-arch. 1472 * Code generation for the ARM Cortex-A57 processor has been improved. 1473 A more accurate instruction scheduling model for the processor is 1474 now used, and a number of compiler tuning parameters have been set 1475 to offer increased performance when compiling with -mcpu=cortex-a57 1476 or -mtune=cortex-a57. 1477 * Support has been added for the following processors (GCC 1478 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 1479 initial support for its big.LITTLE combination with the ARM 1480 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 1481 initial support for its big.LITTLE combination with the ARM 1482 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 1483 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 1484 The GCC identifiers can be used as arguments to the -mcpu or -mtune 1485 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 1486 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 1487 support for the Cortex-A72. 1488 * The deprecated option -mwords-little-endian has been removed. 1489 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 1490 -mtpcs-leaf-frame which are only applicable to the old ABI have 1491 been deprecated. 1492 * The transitional options -mlra and -mno-lra have been removed. The 1493 ARM backend now uses the local register allocator (LRA) only. 1494 1495 AVR 1496 1497 * The compiler no more supports individual devices like ATmega8. 1498 Specifying, say, -mmcu=atmega8 triggers the usage of the 1499 device-specific [38]spec file specs-atmega8 which is part of the 1500 installation and describes options for the sub-processes like 1501 compiler proper, assembler and linker. You can add support for a 1502 new device -mmcu=mydevice as follows: 1503 1. In an empty directory /someplace, create a new directory 1504 device-specs. 1505 2. Copy a device spec file from the installed device-specs 1506 folder, follow the comments in that file and then save it as 1507 /someplace/device-specs/specs-mydevice. 1508 3. Add -B /someplace -mmcu=mydevice to the compiler's 1509 command-line options. Notice that /someplace must specify an 1510 absolute path and that mydevice must not start with "avr". 1511 4. Provided you have a device-specific library libmydevice.a 1512 available, you can put it at /someplace, dito for a 1513 device-specific startup file crtmydevice.o. 1514 The contents of the device spec files depend on the compiler's 1515 configuration, in particular on --with-avrlibc=no and whether or 1516 not it is configured for RTEMS. 1517 * A new command-line option -nodevicelib has been added. It prevents 1518 the compiler from linking against AVR-LibC's device-specific 1519 library libdevice.a. 1520 * The following three command-line options have been added: 1521 1522 -mrmw 1523 Set if the device supports the read-modify-write 1524 instructions LAC, LAS, LAT and XCH. 1525 1526 -mn-flash=size 1527 Specify the flash size of the device in units of 64 KiB, 1528 rounded up to the next integer as needed. This option 1529 affects the availability of the [39]AVR address-spaces. 1530 1531 -mskip-bug 1532 Set if the device is affected by the respective silicon 1533 bug. 1534 1535 In general, you don't need to set these options by hand. The new 1536 device-specific spec file will set them as needed. 1537 1538 IA-32/x86-64 1539 1540 * New [40]ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of 1541 Intel's CPU codenamed Skylake Server was added to GCC. That 1542 includes inline assembly support, new intrinsics, and basic 1543 autovectorization. These new AVX-512 extensions are available via 1544 the following GCC switches: AVX-512 Vector Length EVEX feature: 1545 -mavx512vl, AVX-512 Byte and Word instructions: -mavx512bw, AVX-512 1546 Dword and Qword instructions: -mavx512dq, AVX-512 FMA-52 1547 instructions: -mavx512ifma and for AVX-512 Vector Bit Manipulation 1548 Instructions: -mavx512vbmi. 1549 * New ISA extensions support Intel MPX was added to GCC. This new 1550 extension is available via the -mmpx compiler switch. Intel MPX is 1551 a set of processor features which, with compiler, run-time library 1552 and OS support, brings increased robustness to software by run-time 1553 checking pointer references against their bounds. In GCC Intel MPX 1554 is supported by Pointer Bounds Checker and libmpx run-time 1555 libraries. 1556 * The new -mrecord-mcount option for -pg generates a Linux kernel 1557 style table of pointers to mcount or __fentry__ calls at the 1558 beginning of functions. The new -mnop-mcount option in addition 1559 also generates nops in place of the __fentry__ or mcount call, so 1560 that a call per function can be later patched in. This can be used 1561 for low overhead tracing or hot code patching. 1562 * The new -malign-data option controls how GCC aligns variables. 1563 -malign-data=compat uses increased alignment compatible with GCC 1564 4.8 and earlier, -malign-data=abi uses alignment as specified by 1565 the psABI, and -malign-data=cacheline uses increased alignment to 1566 match the cache line size. -malign-data=compat is the default. 1567 * The new -mskip-rax-setup option skips setting up the RAX register 1568 when SSE is disabled and there are no variable arguments passed in 1569 vector registers. This can be used to optimize the Linux kernel. 1570 1571 MIPS 1572 1573 * MIPS Releases 3 and 5 are now directly supported. Use the 1574 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 1575 to enable code-generation for these processors. 1576 * The Imagination P5600 processor is now supported using the 1577 -march=p5600 command-line option. 1578 * The Cavium Octeon3 processor is now supported using the 1579 -march=octeon3 command-line option. 1580 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 1581 command-line options. 1582 * The o32 ABI has been modified and extended. The o32 64-bit 1583 floating-point register support is now obsolete and has been 1584 removed. It has been replaced by three ABI extensions FPXX, FP64A, 1585 and FP64. The meaning of the -mfp64 command-line option has 1586 changed. It is now used to enable the FP64A and FP64 ABI 1587 extensions. 1588 + The FPXX extension requires that code generated to access 1589 double-precision values use even-numbered registers. Code that 1590 adheres to this extension is link-compatible with all other 1591 o32 double-precision ABI variants and will execute correctly 1592 in all hardware FPU modes. The command-line options -mabi=32 1593 -mfpxx can be used to enable this extension. MIPS II is the 1594 minimum processor required. 1595 + The o32 FP64A extension requires that floating-point registers 1596 be 64-bit and odd-numbered single-precision registers are not 1597 allowed. Code that adheres to the o32 FP64A variant is 1598 link-compatible with all other o32 double-precision ABI 1599 variants. The command-line options -mabi=32 -mfp64 1600 -mno-odd-spreg can be used to enable this extension. MIPS32R2 1601 is the minimum processor required. 1602 + The o32 FP64 extension also requires that floating-point 1603 registers be 64-bit, but permits the use of single-precision 1604 registers. Code that adheres to the o32 FP64 variant is 1605 link-compatible with o32 FPXX and o32 FP64A variants only, 1606 i.e. it is not compatible with the original o32 1607 double-precision ABI. The command-line options -mabi=32 -mfp64 1608 -modd-spreg can be used to enable this extension. MIPS32R2 is 1609 the minimum processor required. 1610 The new ABI variants can be enabled by default using the configure 1611 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 1612 It is strongly recommended that all vendors begin to set o32 FPXX 1613 as the default ABI. This will be required to run the generated code 1614 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 1615 MIPSR6 cores. 1616 * GCC will now pass all floating-point options to the assembler if 1617 GNU binutils 2.25 is used. As a result, any inline assembly code 1618 that uses hard-float instructions should be amended to include a 1619 .set directive to override the global assembler options when 1620 compiling for soft-float targets. 1621 1622 NDS32 1623 1624 * The variadic function ABI implementation is now compatible with 1625 past Andes toolchains where the caller uses registers to pass 1626 arguments and the callee is in charge of pushing them on stack. 1627 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 1628 been removed since they are not yet available in the nds32 port of 1629 GNU binutils. 1630 * A new option -mcmodel=[small|medium|large] supports varied code 1631 models on code generation. The -mgp-direct option became 1632 meaningless and can be discarded. 1633 1634 RX 1635 1636 * A new command line option -mno-allow-string-insns can be used to 1637 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 1638 SWHILE and RMPA instructions. An erratum released by Renesas shows 1639 that it is unsafe to use these instructions on addresses within the 1640 I/O space of the processor. The new option can be used when the 1641 programmer is concerned that the I/O space might be accessed. The 1642 default is still to enable these instructions. 1643 1644 SH 1645 1646 * The compiler will now pass the appropriate --isa= option to the 1647 assembler. 1648 * The default handling for the GBR has been changed from call 1649 clobbered to call preserved. The old behavior can be reinstated by 1650 specifying the option -fcall-used-gbr. 1651 * Support for the SH4A fpchg instruction has been added which will be 1652 utilized when switching between single and double precision FPU 1653 modes. 1654 * The compiler no longer uses the __fpscr_values array for switching 1655 between single and double FPU precision modes on non-SH4A targets. 1656 Instead mode switching will now be performed by storing, modifying 1657 and reloading the FPSCR, so that other FPSCR bits are preserved 1658 across mode switches. The __fpscr_values array that is defined in 1659 libgcc is still present for backwards compatibility, but it will 1660 not be referenced by compiler generated code anymore. 1661 * New builtin functions __builtin_sh_get_fpscr and 1662 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 1663 function will mask the specified bits in such a way that the SZ, PR 1664 and FR mode bits will be preserved, while changing the other bits. 1665 These new functions do not reference the __fpscr_values array. The 1666 old functions __set_fpscr and __get_fpscr in libgcc which access 1667 the __fpscr_values array are still present for backwards 1668 compatibility, but their usage is highly discouraged. 1669 * Some improvements to code generated for __atomic built-in 1670 functions. 1671 * When compiling for SH2E the compiler will no longer force the usage 1672 of delay slots for conditional branch instructions bt and bf. The 1673 old behavior can be reinstated (e.g. to work around a hardware bug 1674 in the original SH7055) by specifying the new option 1675 -mcbranch-force-delay-slot. 1676 1677 Operating Systems 1678 1679 AIX 1680 1681 * GCC now supports stabs debugging continuation lines to allow long 1682 stabs debug information without overflow that generates AIX linker 1683 errors. 1684 1685 DragonFly BSD 1686 1687 * GCC now supports the DragonFly BSD operating system. 1688 1689 FreeBSD 1690 1691 * GCC now supports the FreeBSD operating system for the arm port 1692 through the arm*-*-freebsd* target triplets. 1693 1694 VxWorks MILS 1695 1696 * GCC now supports the MILS (Multiple Independent Levels of Security) 1697 variant of WindRiver's VxWorks operating system for PowerPC 1698 targets. 1699 1700 Other significant improvements 1701 1702 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 1703 to set the compiler to use. 1704 1705 * When the new command-line option -freport-bug is used, GCC 1706 automatically generates a developer-friendly reproducer whenever an 1707 internal compiler error is encountered. 1708 1709 GCC 5.2 1710 1711 This is the [41]list of problem reports (PRs) from GCC's bug tracking 1712 system that are known to be fixed in the 5.2 release. This list might 1713 not be complete (that is, it is possible that some PRs that have been 1714 fixed are not listed here). 1715 1716 Target Specific Changes 1717 1718 IA-32/x86-64 1719 1720 * Support for new AMD instructions monitorx and mwaitx has been 1721 added. This includes new intrinsic and built-in support. It is 1722 enabled through option -mmwaitx. The instructions monitorx and 1723 mwaitx implement the same functionality as the old monitor and 1724 mwait instructions. In addition mwaitx adds a configurable timer. 1725 The timer value is received as third argument and stored in 1726 register %ebx. 1727 1728 S/390, System z, IBM z Systems 1729 1730 * Support for the IBM z13 processor has been added. When using the 1731 -march=z13 option, the compiler will generate code making use of 1732 the new instructions and registers introduced with the vector 1733 extension facility. The -mtune=z13 option enables z13 specific 1734 instruction scheduling without making use of new instructions. 1735 Compiling code with -march=z13 reduces the default alignment of 1736 vector types bigger than 8 bytes to 8. This is an ABI change and 1737 care must be taken when linking modules compiled with different 1738 arch levels which interchange variables containing vector type 1739 values. For newly compiled code the GNU linker will emit a warning. 1740 * The -mzvector option enables a C/C++ language extension. This 1741 extension provides a new keyword vector which can be used to define 1742 vector type variables. (Note: This is not available when enforcing 1743 strict standard compliance e.g. with -std=c99. Either enable GNU 1744 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 1745 Additionally a set of overloaded builtins is provided which is 1746 partially compatible to the PowerPC Altivec builtins. In order to 1747 make use of these builtins the vecintrin.h header file needs to be 1748 included. 1749 1750 GCC 5.3 1751 1752 This is the [42]list of problem reports (PRs) from GCC's bug tracking 1753 system that are known to be fixed in the 5.3 release. This list might 1754 not be complete (that is, it is possible that some PRs that have been 1755 fixed are not listed here). 1756 1757 Target Specific Changes 1758 1759 IA-32/x86-64 1760 1761 * GCC now supports the Intel CPU named Skylake with AVX-512 1762 extensions through -march=skylake-avx512. The switch enables the 1763 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 1764 AVX-512DQ. 1765 1766 S/390, System z, IBM z Systems 1767 1768 * With this version of GCC IBM z Systems support has been added to 1769 the GO runtime environment. GCC 5.3 has proven to be able to 1770 compile larger GO applications on IBM z Systems. 1771 1772 GCC 5.4 1773 1774 This is the [43]list of problem reports (PRs) from GCC's bug tracking 1775 system that are known to be fixed in the 5.4 release. This list might 1776 not be complete (that is, it is possible that some PRs that have been 1777 fixed are not listed here). 1778 1779 GCC 5.5 1780 1781 This is the [44]list of problem reports (PRs) from GCC's bug tracking 1782 system that are known to be fixed in the 5.5 release. This list might 1783 not be complete (that is, it is possible that some PRs that have been 1784 fixed are not listed here). 1785 1786 Target Specific Changes 1787 1788 IA-32/x86-64 1789 1790 * Support for the [45]deprecated pcommit instruction has been 1791 removed. 1792 1793 1794 For questions related to the use of GCC, please consult these web 1795 pages and the [46]GCC manuals. If that fails, the 1796 [47]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1797 web pages and the development of GCC are welcome on our developer 1798 list at [48]gcc (a] gcc.gnu.org. All of [49]our lists have public 1799 archives. 1800 1801 Copyright (C) [50]Free Software Foundation, Inc. Verbatim copying and 1802 distribution of this entire article is permitted in any medium, 1803 provided this notice is preserved. 1804 1805 These pages are [51]maintained by the GCC team. Last modified 1806 2018-09-30[52]. 1807 1808 References 1809 1810 1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx 1811 2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 1812 3. https://savannah.nongnu.org/bugs/?44574 1813 4. https://gcc.gnu.org/wiki/Intel MPX support in the GCC compiler 1814 5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf 1815 6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf 1816 7. https://gcc.gnu.org/wiki/OpenACC 1817 8. https://gcc.gnu.org/install/configure.html 1818 9. https://www.cilkplus.org/ 1819 10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 1820 11. http://gcc.gnu.org/projects/cxx1y.html 1821 12. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 1822 13. http://gcc.gnu.org/projects/cxx1y.html 1823 14. http://gcc.gnu.org/projects/cxx1y.html 1824 15. http://gcc.gnu.org/projects/cxx1y.html 1825 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 1826 17. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 1827 18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 1828 19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 1829 20. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 1830 21. https://sourceware.org/gdb/current/onlinedocs/gdb/Xmethods-In-Python.html 1831 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 1832 23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 1833 24. https://gcc.gnu.org/install/configure.html 1834 25. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 1835 26. https://gcc.gnu.org/wiki/Fortran2003Status 1836 27. https://gcc.gnu.org/wiki/Fortran2008Status 1837 28. https://gcc.gnu.org/wiki/Coarray 1838 29. http://www.opencoarrays.org/ 1839 30. https://golang.org/cmd/go/ 1840 31. https://golang.org/cmd/gofmt/ 1841 32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 1842 33. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 1843 34. https://github.com/davidmalcolm/pygccjit 1844 35. https://github.com/ibuclaw/gccjitd 1845 36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 1846 37. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 1847 38. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 1848 39. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 1849 40. https://software.intel.com/sites/default/files/managed/39/c5/325462-sdm-vol-1-2abcd-3abcd.pdf 1850 41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 1851 42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 1852 43. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 1853 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 1854 45. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction 1855 46. https://gcc.gnu.org/onlinedocs/ 1856 47. mailto:gcc-help (a] gcc.gnu.org 1857 48. mailto:gcc (a] gcc.gnu.org 1858 49. https://gcc.gnu.org/lists.html 1859 50. https://www.fsf.org/ 1860 51. https://gcc.gnu.org/about.html 1861 52. http://validator.w3.org/check/referer 1862 ====================================================================== 1863 http://gcc.gnu.org/gcc-4.9/index.html 1864 GCC 4.9 Release Series 1865 1866 Aug 3, 2016 1867 1868 The [1]GNU project and the GCC developers are pleased to announce the 1869 release of GCC 4.9.4. 1870 1871 This release is a bug-fix release, containing fixes for regressions in 1872 GCC 4.9.3 relative to previous releases of GCC. 1873 1874 This release series is no longer maintained. 1875 1876 Release History 1877 1878 GCC 4.9.4 1879 Aug 3, 2016 ([2]changes, [3]documentation) 1880 1881 GCC 4.9.3 1882 June 26, 2015 ([4]changes, [5]documentation) 1883 1884 GCC 4.9.2 1885 October 30, 2014 ([6]changes, [7]documentation) 1886 1887 GCC 4.9.1 1888 July 16, 2014 ([8]changes, [9]documentation) 1889 1890 GCC 4.9.0 1891 April 22, 2014 ([10]changes, [11]documentation) 1892 1893 References and Acknowledgements 1894 1895 GCC used to stand for the GNU C Compiler, but since the compiler 1896 supports several other languages aside from C, it now stands for the 1897 GNU Compiler Collection. 1898 1899 A list of [12]successful builds is updated as new information becomes 1900 available. 1901 1902 The GCC developers would like to thank the numerous people that have 1903 contributed new features, improvements, bug fixes, and other changes as 1904 well as test results to GCC. This [13]amazing group of volunteers is 1905 what makes GCC successful. 1906 1907 For additional information about GCC please refer to the [14]GCC 1908 project web site or contact the [15]GCC development mailing list. 1909 1910 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 1911 1912 1913 For questions related to the use of GCC, please consult these web 1914 pages and the [18]GCC manuals. If that fails, the 1915 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1916 web pages and the development of GCC are welcome on our developer 1917 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 1918 archives. 1919 1920 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 1921 distribution of this entire article is permitted in any medium, 1922 provided this notice is preserved. 1923 1924 These pages are [23]maintained by the GCC team. Last modified 1925 2018-09-30[24]. 1926 1927 References 1928 1929 1. http://www.gnu.org/ 1930 2. http://gcc.gnu.org/gcc-4.9/changes.html 1931 3. http://gcc.gnu.org/onlinedocs/4.9.4/ 1932 4. http://gcc.gnu.org/gcc-4.9/changes.html 1933 5. http://gcc.gnu.org/onlinedocs/4.9.3/ 1934 6. http://gcc.gnu.org/gcc-4.9/changes.html 1935 7. http://gcc.gnu.org/onlinedocs/4.9.2/ 1936 8. http://gcc.gnu.org/gcc-4.9/changes.html 1937 9. http://gcc.gnu.org/onlinedocs/4.9.1/ 1938 10. http://gcc.gnu.org/gcc-4.9/changes.html 1939 11. https://gcc.gnu.org/onlinedocs/4.9.0/ 1940 12. http://gcc.gnu.org/gcc-4.9/buildstat.html 1941 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 1942 14. http://gcc.gnu.org/index.html 1943 15. mailto:gcc (a] gcc.gnu.org 1944 16. http://gcc.gnu.org/mirrors.html 1945 17. http://gcc.gnu.org/svn.html 1946 18. https://gcc.gnu.org/onlinedocs/ 1947 19. mailto:gcc-help (a] gcc.gnu.org 1948 20. mailto:gcc (a] gcc.gnu.org 1949 21. https://gcc.gnu.org/lists.html 1950 22. https://www.fsf.org/ 1951 23. https://gcc.gnu.org/about.html 1952 24. http://validator.w3.org/check/referer 1953 ====================================================================== 1954 http://gcc.gnu.org/gcc-4.9/changes.html 1955 GCC 4.9 Release Series 1956 Changes, New Features, and Fixes 1957 1958 Caveats 1959 1960 * The mudflap run time checker has been removed. The mudflap options 1961 remain, but do nothing. 1962 * Support for a number of older systems and recently unmaintained or 1963 untested target ports of GCC has been declared obsolete in GCC 4.9. 1964 Unless there is activity to revive them, the next release of GCC 1965 will have their sources permanently removed. 1966 The following ports for individual systems on particular 1967 architectures have been obsoleted: 1968 + Solaris 9 (*-*-solaris2.9). Details can be found in the 1969 [1]announcement. 1970 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 1971 float64x1_t exported by arm_neon.h are defined to be the same as 1972 their base types. This results in incorrect application of 1973 parameter passing rules to arguments of types int64x1_t and 1974 uint64x1_t, with respect to the AAPCS64 ABI specification. In 1975 addition, names of C++ functions with parameters of these types 1976 (including float64x1_t) are not mangled correctly. The current 1977 typedef declarations also unintentionally allow implicit casting 1978 between singleton vector types and their base types. These issues 1979 will be resolved in a near future release. See [2]PR60825 for more 1980 information. 1981 1982 More information on porting to GCC 4.9 from previous versions of GCC 1983 can be found in the [3]porting guide for this release. 1984 1985 General Optimizer Improvements 1986 1987 * AddressSanitizer, a fast memory error detector, is now available on 1988 ARM. 1989 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 1990 detector, has been added and can be enabled via 1991 -fsanitize=undefined. Various computations will be instrumented to 1992 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 1993 currently available for the C and C++ languages. 1994 * Link-time optimization (LTO) improvements: 1995 + Type merging was rewritten. The new implementation is 1996 significantly faster and uses less memory. 1997 + Better partitioning algorithm resulting in less streaming 1998 during link time. 1999 + Early removal of virtual methods reduces the size of object 2000 files and improves link-time memory usage and compile time. 2001 + Function bodies are now loaded on-demand and released early 2002 improving overall memory usage at link time. 2003 + C++ hidden keyed methods can now be optimized out. 2004 + When using a linker plugin, compiling with the -flto option 2005 now generates slim object files (.o) which only contain 2006 intermediate language representation for LTO. Use 2007 -ffat-lto-objects to create files which contain additionally 2008 the object code. To generate static libraries suitable for LTO 2009 processing, use gcc-ar and gcc-ranlib; to list symbols from a 2010 slim object file use gcc-nm. (This requires that ar, ranlib 2011 and nm have been compiled with plugin support.) 2012 Memory usage building Firefox with debug enabled was reduced from 2013 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 2014 * Inter-procedural optimization improvements: 2015 + New type inheritance analysis module improving 2016 devirtualization. Devirtualization now takes into account 2017 anonymous name-spaces and the C++11 final keyword. 2018 + New speculative devirtualization pass (controlled by 2019 -fdevirtualize-speculatively. 2020 + Calls that were speculatively made direct are turned back to 2021 indirect where direct call is not cheaper. 2022 + Local aliases are introduced for symbols that are known to be 2023 semantically equivalent across shared libraries improving 2024 dynamic linking times. 2025 * Feedback directed optimization improvements: 2026 + Profiling of programs using C++ inline functions is now more 2027 reliable. 2028 + New time profiling determines typical order in which functions 2029 are executed. 2030 + A new function reordering pass (controlled by 2031 -freorder-functions) significantly reduces startup time of 2032 large applications. Until binutils support is completed, it is 2033 effective only with link-time optimization. 2034 + Feedback driven indirect call removal and devirtualization now 2035 handle cross-module calls when link-time optimization is 2036 enabled. 2037 2038 New Languages and Language specific improvements 2039 2040 * Version 4.0 of the [4]OpenMP specification is now supported in the 2041 C and C++ compilers and starting with the 4.9.1 release also in the 2042 Fortran compiler. The new -fopenmp-simd option can be used to 2043 enable OpenMP's SIMD directives while ignoring other OpenMP 2044 directives. The new [5]-fsimd-cost-model= option permits to tune 2045 the vectorization cost model for loops annotated with OpenMP and 2046 Cilk Plus simd directives. -Wopenmp-simd warns when the current 2047 cost model overrides simd directives set by the user. 2048 * The -Wdate-time option has been added for the C, C++ and Fortran 2049 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 2050 macros are used. Those macros might prevent bit-wise-identical 2051 reproducible compilations. 2052 2053 Ada 2054 2055 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 2056 2057 C family 2058 2059 * Support for colorizing diagnostics emitted by GCC has been added. 2060 The [6]-fdiagnostics-color=auto will enable it when outputting to 2061 terminals, -fdiagnostics-color=always unconditionally. The 2062 GCC_COLORS environment variable can be used to customize the colors 2063 or disable coloring. If GCC_COLORS variable is present in the 2064 environment, the default is -fdiagnostics-color=auto, otherwise 2065 -fdiagnostics-color=never. 2066 Sample diagnostics output: 2067 $ g++ -fdiagnostics-color=always -S -Wall test.C 2068 test.C: In function int foo(): 2069 test.C:1:14: warning: no return statement in function returning non-void [-W 2070 return-type] 2071 int foo () { } 2072 ^ 2073 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 2074 -ftemplate-depth= to increase the maximum) instantiating struct X<100> 2075 template <int N> struct X { static const int value = X<N-1>::value; }; temp 2076 late struct X<1000>; 2077 ^ 2078 test.C:2:46: recursively required from const int X<999>::value 2079 test.C:2:46: required from const int X<1000>::value 2080 test.C:2:88: required from here 2081 2082 test.C:2:46: error: incomplete type X<100> used in nested name specifier 2083 2084 * With the new [7]#pragma GCC ivdep, the user can assert that there 2085 are no loop-carried dependencies which would prevent concurrent 2086 execution of consecutive iterations using SIMD (single instruction 2087 multiple data) instructions. 2088 * Support for [8]Cilk Plus has been added and can be enabled with the 2089 -fcilkplus option. Cilk Plus is an extension to the C and C++ 2090 languages to support data and task parallelism. The present 2091 implementation follows ABI version 1.2; all features but _Cilk_for 2092 have been implemented. 2093 2094 C 2095 2096 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 2097 <stdatomic.h> header) are now supported. 2098 * ISO C11 generic selections (_Generic keyword) are now supported. 2099 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 2100 __thread) is now supported. 2101 * ISO C11 support is now at a similar level of completeness to ISO 2102 C99 support: substantially complete modulo bugs, extended 2103 identifiers (supported except for corner cases when 2104 -fextended-identifiers is used), floating-point issues (mainly but 2105 not entirely relating to optional C99 features from Annexes F and 2106 G) and the optional Annexes K (Bounds-checking interfaces) and L 2107 (Analyzability). 2108 * A new C extension __auto_type provides a subset of the 2109 functionality of C++11 auto in GNU C. 2110 2111 C++ 2112 2113 * The G++ implementation of [9]C++1y return type deduction for normal 2114 functions has been updated to conform to [10]N3638, the proposal 2115 accepted into the working paper. Most notably, it adds 2116 decltype(auto) for getting decltype semantics rather than the 2117 template argument deduction semantics of plain auto: 2118 2119 int& f(); 2120 auto i1 = f(); // int 2121 decltype(auto) i2 = f(); // int& 2122 2123 * G++ supports [11]C++1y lambda capture initializers: 2124 2125 [x = 42]{ ... }; 2126 2127 Actually, they have been accepted since GCC 4.5, but now the 2128 compiler doesn't warn about them with -std=c++1y, and supports 2129 parenthesized and brace-enclosed initializers as well. 2130 * G++ supports [12]C++1y variable length arrays. G++ has supported 2131 GNU/C99-style VLAs for a long time, but now additionally supports 2132 initializers and lambda capture by reference. In C++1y mode G++ 2133 will complain about VLA uses that are not permitted by the draft 2134 standard, such as forming a pointer to VLA type or applying sizeof 2135 to a VLA variable. Note that it now appears that VLAs will not be 2136 part of C++14, but will be part of a separate document and then 2137 perhaps C++17. 2138 2139 void f(int n) { 2140 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 2141 [&a]{ for (int i : a) { cout << i << endl; } }(); 2142 &a; // error, taking address of VLA 2143 } 2144 2145 * G++ supports the [13]C++1y [[deprecated]] attribute modulo bugs in 2146 the underlying [[gnu::deprecated]] attribute. Classes and functions 2147 can be marked deprecated and a diagnostic message added: 2148 2149 class A; 2150 int bar(int n); 2151 #if __cplusplus > 201103 2152 class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 2153 [[deprecated("bar is unsafe; use foo() instead")]] 2154 int bar(int n); 2155 2156 int foo(int n); 2157 class B; 2158 #endif 2159 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 2160 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 2161 o() instead 2162 2163 * G++ supports [14]C++1y digit separators. Long numeric literals can 2164 be subdivided with a single quote ' to enhance readability: 2165 2166 int i = 1048576; 2167 int j = 1'048'576; 2168 int k = 0x10'0000; 2169 int m = 0'004'000'000; 2170 int n = 0b0001'0000'0000'0000'0000'0000; 2171 2172 double x = 1.602'176'565e-19; 2173 double y = 1.602'176'565e-1'9; 2174 2175 * G++ supports [15]C++1y generic (polymorphic) lambdas. 2176 2177 // a functional object that will increment any type 2178 auto incr = [](auto x) { return x++; }; 2179 2180 * As a GNU extension, G++ supports explicit template parameter syntax 2181 for generic lambdas. This can be combined in the expected way with 2182 the standard auto syntax. 2183 2184 // a functional object that will add two like-type objects 2185 auto add = [] <typename T> (T a, T b) { return a + b; }; 2186 2187 * G++ supports unconstrained generic functions as specified by 4.1.2 2188 and 5.1.1 of [16]N3889: Concepts Lite Specification. Briefly, auto 2189 may be used as a type-specifier in a parameter declaration of any 2190 function declarator in order to introduce an implicit function 2191 template parameter, akin to generic lambdas. 2192 2193 // the following two function declarations are equivalent 2194 auto incr(auto x) { return x++; } 2195 template <typename T> 2196 auto incr(T x) { return x++; } 2197 2198 Runtime Library (libstdc++) 2199 2200 * [17]Improved support for C++11, including: 2201 + support for <regex>; 2202 + The associative containers in <map> and <set> and the 2203 unordered associative containers in <unordered_map> and 2204 <unordered_set> meet the allocator-aware container 2205 requirements; 2206 * [18]Improved experimental support for the upcoming ISO C++ 2207 standard, C++14, including: 2208 + fixing constexpr member functions without const; 2209 + implementation of the std::exchange() utility function; 2210 + addressing tuples by type; 2211 + implemention of std::make_unique; 2212 + implemention of std::shared_lock; 2213 + making std::result_of SFINAE-friendly; 2214 + adding operator() to std::integral_constant; 2215 + adding user-defined literals for standard library types 2216 std::basic_string, std::chrono::duration, and std::complex; 2217 + adding two range overloads to non-modifying sequence oprations 2218 std::equal and std::mismatch; 2219 + adding IO manipulators for quoted strings; 2220 + adding constexpr members to <utility>, <complex>, <chrono>, 2221 and some containers; 2222 + adding compile-time std::integer_sequence; 2223 + adding cleaner transformation traits; 2224 + making <functional>s operator functors easier to use and more 2225 generic; 2226 * An implementation of std::experimental::optional. 2227 * An implementation of std::experimental::string_view. 2228 * The non-standard function std::copy_exception has been deprecated 2229 and will be removed in a future version. std::make_exception_ptr 2230 should be used instead. 2231 2232 Fortran 2233 2234 * Compatibility notice: 2235 + Module files: The version of the module files (.mod) has been 2236 incremented; additionally, module files are now compressed. 2237 Fortran MODULEs compiled by earlier GCC versions have to be 2238 recompiled, when they are USEd by files compiled with GCC 4.9. 2239 GCC 4.9 is not able to read .mod files of earlier GCC 2240 versions; attempting to do so gives an error message. Note: 2241 The ABI of the produced assembler data itself has not changed: 2242 object files and libraries are fully compatible with older 2243 versions (except as stated below). 2244 + ABI changes: 2245 o The [19]argument passing ABI has changed for scalar dummy 2246 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 2247 which have both the VALUE and the OPTIONAL attributes. 2248 o To support finalization the virtual table associated with 2249 polymorphic variables has changed. Code containing CLASS 2250 should be recompiled, including all files which define 2251 derived types involved in the type definition used by 2252 polymorphic variables. (Note: Due to the incremented 2253 module version, trying to mix old code with new code will 2254 usually give an error message.) 2255 + GNU Fortran no longer deallocates allocatable variables or 2256 allocatable components of variables declared in the main 2257 program. Since Fortran 2008, the standard explicitly states 2258 that variables declared in the Fortran main program 2259 automatically have the SAVE attribute. 2260 + When opening files, the close-on-exec flag is set if the 2261 system supports such a feature. This is generally considered 2262 good practice these days, but if there is a need to pass file 2263 descriptors to child processes the parent process must now 2264 remember to clear the close-on-exec flag by calling fcntl(), 2265 e.g. via ISO_C_BINDING, before executing the child process. 2266 * The deprecated command-line option -fno-whole-file has been 2267 removed. (-fwhole-file is the default since GCC 4.6.) 2268 -fwhole-file/-fno-whole-file continue to be accepted but do not 2269 influence the code generation. 2270 * The compiler no longer unconditionally warns about DO loops with 2271 zero iterations. This warning is now controlled by the -Wzerotrip 2272 option, which is implied by -Wall. 2273 * The new NO_ARG_CHECK attribute of the [20]!GCC$ directive can be 2274 used to disable the type-kind-rank (TKR) argument check for a dummy 2275 argument. The feature is similar to ISO/IEC TS 29133:2012's 2276 TYPE(*), except that it additionally also disables the rank check. 2277 Variables with NO_ARG_CHECK have to be dummy arguments and may only 2278 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 2279 to another NO_ARG_CHECK dummy argument; also the other constraints 2280 of TYPE(*) apply. The dummy arguments should be declared as scalar 2281 or assumed-size variable of type type(*) (recommended) or of type 2282 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 2283 the data without further type or shape information is passed, 2284 similar to C's void*. Note that also TS 29113's 2285 type(*),dimension(..) accepts arguments of any type and rank; 2286 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 2287 descriptor which contains the array shape and stride of the 2288 argument. 2289 * [21]Fortran 2003: 2290 + Finalization is now supported. It is currently only done for a 2291 subset of those situations in which it should occur. 2292 + Experimental support for scalar character components with 2293 deferred length (i.e. allocatable string length) in derived 2294 types has been added. (Deferred-length character variables are 2295 supported since GCC 4.6.) 2296 * [22]Fortran 2008: 2297 + When STOP or ERROR STOP are used to terminate the execution 2298 and any exception (but inexact) is signaling, a warning is 2299 printed to ERROR_UNIT, indicating which exceptions are 2300 signaling. The [23]-ffpe-summary= command-line option can be 2301 used to fine-tune for which exceptions the warning should be 2302 shown. 2303 + Rounding on input (READ) is now handled on systems where 2304 strtod honours the rounding mode. (For output, rounding is 2305 supported since GCC 4.5.) Note that for input, the compatible 2306 rounding mode is handled as nearest (i.e., rounding to an even 2307 least significant [cf. IEC 60559:1989] for a tie, while 2308 compatible rounds away from zero in that case). 2309 2310 Go 2311 2312 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 2313 2314 New Targets and Target Specific Improvements 2315 2316 AArch64 2317 2318 * The ARMv8-A crypto and CRC instructions are now supported through 2319 intrinsics. These are enabled when the architecture supports these 2320 and are available through the -march=armv8-a+crc and 2321 -march=armv8-a+crypto options. 2322 * Initial support for ILP32 has now been added to the compiler. This 2323 is now available through the command-line option -mabi=ilp32. 2324 Support for ILP32 is considered experimental as the ABI 2325 specification is still beta. 2326 * Coverage of more of the ISA including the SIMD extensions has been 2327 added. The Advanced SIMD intrinsics have also been improved. 2328 * The new local register allocator (LRA) is now on by default for the 2329 AArch64 backend. 2330 * The REE (Redundant extension elimination) pass has now been enabled 2331 by default for the AArch64 backend. 2332 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 2333 * Initial big.LITTLE tuning support for the combination of Cortex-A57 2334 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 2335 option. 2336 * A number of structural changes have been made to both the ARM and 2337 AArch64 backends to facilitate improved code-generation. 2338 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 2339 has been added and can be enabled by giving the 2340 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 2341 default by configuring GCC with the --enable-fix-cortex-a53-835769 2342 option. 2343 2344 ARC 2345 2346 * A port for Synopsys Designware ARC has been contributed by Embecosm 2347 and Synopsys Inc. 2348 2349 ARM 2350 2351 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 2352 disabled by default. This was found to generate better code in only 2353 a small number of cases. It can be turned back on with the 2354 -mneon-for-64bits option. 2355 * Further support for the ARMv8-A architecture, notably implementing 2356 the restriction around IT blocks in the Thumb32 instruction set has 2357 been added. The -mrestrict-it option can be used with 2358 -march=armv7-a or the -march=armv7ve options to make code 2359 generation fully compatible with the deprecated instructions in 2360 ARMv8-A. 2361 * Support has now been added for the ARMv7ve variant of the 2362 architecture. This can be used by the -march=armv7ve option. 2363 * The ARMv8-A crypto and CRC instructions are now supported through 2364 intrinsics and are available through the -march=armv8-a+crc and 2365 mfpu=crypto-neon-fp-armv8 options. 2366 * LRA is now on by default for the ARM target. This can be turned off 2367 using the -mno-lra option. This option is a purely transitionary 2368 command-line option and will be removed in a future release. We are 2369 interested in any bug reports regarding functional and performance 2370 regressions with LRA. 2371 * A new option -mslow-flash-data to improve performance of programs 2372 fetching data on slow flash memory has now been introduced for the 2373 ARMv7-M profile cores. 2374 * A new option -mpic-data-is-text-relative for targets that allows 2375 data segments to be relative to text segments has been added. This 2376 is on by default for all targets except VxWorks RTP. 2377 * A number of infrastructural changes have been made to both the ARM 2378 and AArch64 backends to facilitate improved code-generation. 2379 * GCC now supports Cortex-A12 and the Cortex-R7 through the 2380 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 2381 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 2382 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 2383 * Initial big.LITTLE tuning support for the combination of Cortex-A57 2384 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 2385 option. Similar support was added for the combination of Cortex-A15 2386 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 2387 * Further performance optimizations for the Cortex-A15 and the 2388 Cortex-M4 have been added. 2389 * A number of code generation improvements for Thumb2 to reduce code 2390 size when compiling for the M-profile processors. 2391 2392 AVR 2393 2394 * A new command-line option -mfract-convert-truncate has been added. 2395 It allows compiler to use truncation instead of rounding towards 2396 zero for fractional fixed-point types. 2397 2398 IA-32/x86-64 2399 2400 * -mfpmath=sse is now implied by -ffast-math on all targets where 2401 SSE2 is supported. 2402 * Intel AVX-512 support was added to GCC. That includes inline 2403 assembly support, new registers and extending existing ones, new 2404 intrinsics (covered by corresponding testsuite), and basic 2405 autovectorization. AVX-512 instructions are available via the 2406 following GCC switches: AVX-512 foundation instructions: -mavx512f, 2407 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 2408 reciprocal instructions: -mavx512er, AVX-512 conflict detection 2409 instructions: -mavx512cd. 2410 * It is now possible to call x86 intrinsics from select functions in 2411 a file that are tagged with the corresponding target attribute 2412 without having to compile the entire file with the -mxxx option. 2413 This improves the usability of x86 intrinsics and is particularly 2414 useful when doing [24]Function Multiversioning. 2415 * GCC now supports the new Intel microarchitecture named Silvermont 2416 through -march=silvermont. 2417 * GCC now supports the new Intel microarchitecture named Broadwell 2418 through -march=broadwell. 2419 * Optimizing for other Intel microarchitectures have been renamed to 2420 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 2421 * -march=generic has been retuned for better support of Intel core 2422 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 2423 Pentium-M, and Pentium4 based CPUs is no longer considered 2424 important for generic. 2425 * -mtune=intel can now be used to generate code running well on the 2426 most current Intel processors, which are Haswell and Silvermont for 2427 GCC 4.9. 2428 * Support to encode 32-bit assembly instructions in 16-bit format is 2429 now available through the -m16 command-line option. 2430 * Better inlining of memcpy and memset that is aware of value ranges 2431 and produces shorter alignment prologues. 2432 * -mno-accumulate-outgoing-args is now honored when unwind 2433 information is output. Argument accumulation is also now turned off 2434 for portions of programs optimized for size. 2435 * Support for new AMD family 15h processors (Excavator core) is now 2436 available through the -march=bdver4 and -mtune=bdver4 options. 2437 2438 MSP430 2439 2440 * A new command-line option -mcpu= has been added to the MSP430 2441 backend. This option is used to specify the ISA to be used. 2442 Accepted values are msp430 (the default), msp430x and msp430xv2. 2443 The ISA is no longer deduced from the -mmcu= option as there are 2444 far too many different MCU names. The -mmcu= option is still 2445 supported, and this is still used to select linker scripts and 2446 generate a C preprocessor symbol that will be recognised by the 2447 msp430.h header file. 2448 2449 NDS32 2450 2451 * A new nds32 port supports the 32-bit architecture from Andes 2452 Technology Corporation. 2453 * The port provides initial support for the V2, V3, V3m instruction 2454 set architectures. 2455 2456 Nios II 2457 2458 * A port for the Altera Nios II has been contributed by Mentor 2459 Graphics. 2460 2461 PowerPC / PowerPC64 / RS6000 2462 2463 * GCC now supports Power ISA 2.07, which includes support for 2464 Hardware Transactional Memory (HTM), Quadword atomics and several 2465 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 2466 integer and decimal integer operations. 2467 * Support for the POWER8 processor is now available through the 2468 -mcpu=power8 and -mtune=power8 options. 2469 * The libitm library has been modified to add a HTM fastpath that 2470 automatically uses POWER's HTM hardware instructions when it is 2471 executing on a HTM enabled processor. 2472 * Support for the new powerpc64le-linux platform has been added. It 2473 defaults to generating code that conforms to the ELFV2 ABI. 2474 2475 S/390, System z 2476 2477 * Support for the Transactional Execution Facility included with the 2478 IBM zEnterprise zEC12 processor has been added. A set of GCC style 2479 builtins as well as XLC style builtins are provided. The builtins 2480 are enabled by default when using the -march=zEC12 option but can 2481 explicitly be disabled with -mno-htm. Using the GCC builtins also 2482 libitm supports hardware transactions on S/390. 2483 * The hotpatch features allows to prepare functions for hotpatching. 2484 A certain amount of bytes is reserved before the function entry 2485 label plus a NOP is inserted at its very beginning to implement a 2486 backward jump when applying a patch. The feature can either be 2487 enabled per compilation unit via the command-line option -mhotpatch 2488 or per function using the hotpatch attribute. 2489 * The shrink wrap optimization is now supported on S/390 and enabled 2490 by default. 2491 * A major rework of the routines to determine which registers need to 2492 be saved and restored in function prologue/epilogue now allow to 2493 use floating point registers as save slots. This will happen for 2494 certain leaf function with -march=z10 or higher. 2495 * The LRA rtl pass replaces reload by default on S/390. 2496 2497 RX 2498 2499 * The port now allows to specify the RX100, RX200, and RX600 2500 processors with the command-line options -mcpu=rx100, -mcpu=rx200 2501 and -mcpu=rx600. 2502 2503 SH 2504 2505 * Minor improvements to code generated for integer arithmetic and 2506 code that involves the T bit. 2507 * Added support for the SH2A clips and clipu instructions. The 2508 compiler will now try to utilize them for min/max expressions such 2509 as max (-128, min (127, x)). 2510 * Added support for the cmp/str instruction through built-in 2511 functions such as __builtin_strlen. When not optimizing for size, 2512 the compiler will now expand calls to e.g. strlen as an inlined 2513 sequences which utilize the cmp/str instruction. 2514 * Improved code generated around volatile memory loads and stores. 2515 * The option -mcbranchdi has been deprecated. Specifying it will 2516 result in a warning and will not influence code generation. 2517 * The option -mcmpeqdi has been deprecated. Specifying it will result 2518 in a warning and will not influence code generation. 2519 2520 GCC 4.9.1 2521 2522 This is the [25]list of problem reports (PRs) from GCC's bug tracking 2523 system that are known to be fixed in the 4.9.1 release. This list might 2524 not be complete (that is, it is possible that some PRs that have been 2525 fixed are not listed here). 2526 2527 Version 4.0 of the OpenMP specification is supported even in Fortran, 2528 not just C and C++. 2529 2530 GCC 4.9.2 2531 2532 This is the [26]list of problem reports (PRs) from GCC's bug tracking 2533 system that are known to be fixed in the 4.9.2 release. This list might 2534 not be complete (that is, it is possible that some PRs that have been 2535 fixed are not listed here). 2536 2537 GCC 4.9.3 2538 2539 This is the [27]list of problem reports (PRs) from GCC's bug tracking 2540 system that are known to be fixed in the 4.9.3 release. This list might 2541 not be complete (that is, it is possible that some PRs that have been 2542 fixed are not listed here). 2543 2544 GCC 4.9.4 2545 2546 This is the [28]list of problem reports (PRs) from GCC's bug tracking 2547 system that are known to be fixed in the 4.9.4 release. This list might 2548 not be complete (that is, it is possible that some PRs that have been 2549 fixed are not listed here). 2550 2551 2552 For questions related to the use of GCC, please consult these web 2553 pages and the [29]GCC manuals. If that fails, the 2554 [30]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2555 web pages and the development of GCC are welcome on our developer 2556 list at [31]gcc (a] gcc.gnu.org. All of [32]our lists have public 2557 archives. 2558 2559 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 2560 distribution of this entire article is permitted in any medium, 2561 provided this notice is preserved. 2562 2563 These pages are [34]maintained by the GCC team. Last modified 2564 2018-09-30[35]. 2565 2566 References 2567 2568 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 2569 2. https://gcc.gnu.org/PR60825 2570 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 2571 4. http://www.openmp.org/specifications/ 2572 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 2573 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 2574 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 2575 8. https://www.cilkplus.org/ 2576 9. http://gcc.gnu.org/projects/cxx1y.html 2577 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 2578 11. http://gcc.gnu.org/projects/cxx1y.html 2579 12. http://gcc.gnu.org/projects/cxx1y.html 2580 13. http://gcc.gnu.org/projects/cxx1y.html 2581 14. http://gcc.gnu.org/projects/cxx1y.html 2582 15. http://gcc.gnu.org/projects/cxx1y.html 2583 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 2584 17. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 2585 18. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 2586 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 2587 20. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 2588 21. https://gcc.gnu.org/wiki/Fortran2003Status 2589 22. https://gcc.gnu.org/wiki/Fortran2008Status 2590 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 2591 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 2592 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 2593 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 2594 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 2595 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4 2596 29. https://gcc.gnu.org/onlinedocs/ 2597 30. mailto:gcc-help (a] gcc.gnu.org 2598 31. mailto:gcc (a] gcc.gnu.org 2599 32. https://gcc.gnu.org/lists.html 2600 33. https://www.fsf.org/ 2601 34. https://gcc.gnu.org/about.html 2602 35. http://validator.w3.org/check/referer 2603 ====================================================================== 2604 http://gcc.gnu.org/gcc-4.8/index.html 2605 GCC 4.8 Release Series 2606 2607 June 23, 2015 2608 2609 The [1]GNU project and the GCC developers are pleased to announce the 2610 release of GCC 4.8.5. 2611 2612 This release is a bug-fix release, containing fixes for regressions in 2613 GCC 4.8.4 relative to previous releases of GCC. 2614 2615 This release series is no longer maintained. 2616 2617 Release History 2618 2619 GCC 4.8.5 2620 June 23, 2015 ([2]changes, [3]documentation) 2621 2622 GCC 4.8.4 2623 December 19, 2014 ([4]changes, [5]documentation) 2624 2625 GCC 4.8.3 2626 May 22, 2014 ([6]changes, [7]documentation) 2627 2628 GCC 4.8.2 2629 October 16, 2013 ([8]changes, [9]documentation) 2630 2631 GCC 4.8.1 2632 May 31, 2013 ([10]changes, [11]documentation) 2633 2634 GCC 4.8.0 2635 March 22, 2013 ([12]changes, [13]documentation) 2636 2637 References and Acknowledgements 2638 2639 GCC used to stand for the GNU C Compiler, but since the compiler 2640 supports several other languages aside from C, it now stands for the 2641 GNU Compiler Collection. 2642 2643 A list of [14]successful builds is updated as new information becomes 2644 available. 2645 2646 The GCC developers would like to thank the numerous people that have 2647 contributed new features, improvements, bug fixes, and other changes as 2648 well as test results to GCC. This [15]amazing group of volunteers is 2649 what makes GCC successful. 2650 2651 For additional information about GCC please refer to the [16]GCC 2652 project web site or contact the [17]GCC development mailing list. 2653 2654 To obtain GCC please use [18]our mirror sites or [19]our SVN server. 2655 2656 2657 For questions related to the use of GCC, please consult these web 2658 pages and the [20]GCC manuals. If that fails, the 2659 [21]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2660 web pages and the development of GCC are welcome on our developer 2661 list at [22]gcc (a] gcc.gnu.org. All of [23]our lists have public 2662 archives. 2663 2664 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 2665 distribution of this entire article is permitted in any medium, 2666 provided this notice is preserved. 2667 2668 These pages are [25]maintained by the GCC team. Last modified 2669 2018-09-30[26]. 2670 2671 References 2672 2673 1. http://www.gnu.org/ 2674 2. http://gcc.gnu.org/gcc-4.8/changes.html 2675 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 2676 4. http://gcc.gnu.org/gcc-4.8/changes.html 2677 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 2678 6. http://gcc.gnu.org/gcc-4.8/changes.html 2679 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 2680 8. http://gcc.gnu.org/gcc-4.8/changes.html 2681 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 2682 10. http://gcc.gnu.org/gcc-4.8/changes.html 2683 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 2684 12. http://gcc.gnu.org/gcc-4.8/changes.html 2685 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 2686 14. http://gcc.gnu.org/gcc-4.8/buildstat.html 2687 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 2688 16. http://gcc.gnu.org/index.html 2689 17. mailto:gcc (a] gcc.gnu.org 2690 18. http://gcc.gnu.org/mirrors.html 2691 19. http://gcc.gnu.org/svn.html 2692 20. https://gcc.gnu.org/onlinedocs/ 2693 21. mailto:gcc-help (a] gcc.gnu.org 2694 22. mailto:gcc (a] gcc.gnu.org 2695 23. https://gcc.gnu.org/lists.html 2696 24. https://www.fsf.org/ 2697 25. https://gcc.gnu.org/about.html 2698 26. http://validator.w3.org/check/referer 2699 ====================================================================== 2700 http://gcc.gnu.org/gcc-4.8/changes.html 2701 GCC 4.8 Release Series 2702 Changes, New Features, and Fixes 2703 2704 Caveats 2705 2706 GCC now uses C++ as its implementation language. This means that to 2707 build GCC from sources, you will need a C++ compiler that understands 2708 C++ 2003. For more details on the rationale and specific changes, 2709 please refer to the [1]C++ conversion page. 2710 2711 To enable the Graphite framework for loop optimizations you now need 2712 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 2713 the [2]GCC infrastructure directory. The installation manual contains 2714 more information about requirements to build GCC. 2715 2716 GCC now uses a more aggressive analysis to derive an upper bound for 2717 the number of iterations of loops using constraints imposed by language 2718 standards. This may cause non-conforming programs to no longer work as 2719 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 2720 option, -fno-aggressive-loop-optimizations, was added to disable this 2721 aggressive analysis. In some loops that have known constant number of 2722 iterations, but undefined behavior is known to occur in the loop before 2723 reaching or during the last iteration, GCC will warn about the 2724 undefined behavior in the loop instead of deriving lower upper bound of 2725 the number of iterations for the loop. The warning can be disabled with 2726 -Wno-aggressive-loop-optimizations. 2727 2728 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 2729 for the layout of vectors that could lead to wrong code being 2730 generated. Vectors larger than 8 bytes in size are now by default 2731 aligned to an 8-byte boundary. This is an ABI change: code that makes 2732 explicit use of vector types may be incompatible with binary objects 2733 built with older versions of GCC. Auto-vectorized code is not affected 2734 by this change. 2735 2736 On AVR, support has been removed for the command-line option 2737 -mshort-calls deprecated in GCC 4.7. 2738 2739 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2 2740 is turned on per default for all non-RTEMS configurations. This option 2741 arranges for a better integration of [3]AVR Libc with avr-gcc. For 2742 technical details, see [4]PR54461. To turn off the option in non-RTEMS 2743 configurations, use --with-avrlibc=no. If the compiler is configured 2744 for RTEMS, the option is always turned off. 2745 2746 More information on porting to GCC 4.8 from previous versions of GCC 2747 can be found in the [5]porting guide for this release. 2748 2749 General Optimizer Improvements (and Changes) 2750 2751 * DWARF4 is now the default when generating DWARF debug information. 2752 When -g is used on a platform that uses DWARF debugging 2753 information, GCC will now default to -gdwarf-4 2754 -fno-debug-types-section. 2755 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 2756 consumers support DWARF4 by default. Before GCC 4.8 the default 2757 version used was DWARF2. To make GCC 4.8 generate an older DWARF 2758 version use -g together with -gdwarf-2 or -gdwarf-3. The default 2759 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 2760 * A new general optimization level, -Og, has been introduced. It 2761 addresses the need for fast compilation and a superior debugging 2762 experience while providing a reasonable level of run-time 2763 performance. Overall experience for development should be better 2764 than the default optimization level -O0. 2765 * A new option -ftree-partial-pre was added to control the partial 2766 redundancy elimination (PRE) optimization. This option is enabled 2767 by default at the -O3 optimization level, and it makes PRE more 2768 aggressive. 2769 * The option -fconserve-space has been removed; it was no longer 2770 useful on most targets since GCC supports putting variables into 2771 BSS without making them common. 2772 * The struct reorg and matrix reorg optimizations (command-line 2773 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 2774 removed. They did not always work correctly, nor did they work with 2775 link-time optimization (LTO), hence were only applicable to 2776 programs consisting of a single translation unit. 2777 * Several scalability bottle-necks have been removed from GCC's 2778 optimization passes. Compilation of extremely large functions, e.g. 2779 due to the use of the flatten attribute in the "Eigen" C++ linear 2780 algebra templates library, is significantly faster than previous 2781 releases of GCC. 2782 * Link-time optimization (LTO) improvements: 2783 + LTO partitioning has been rewritten for better reliability and 2784 maintanibility. Several important bugs leading to link 2785 failures have been fixed. 2786 * Interprocedural optimization improvements: 2787 + A new symbol table has been implemented. It builds on existing 2788 callgraph and varpool modules and provide a new API. Unusual 2789 symbol visibilities and aliases are handled more consistently 2790 leading to, for example, more aggressive unreachable code 2791 removal with LTO. 2792 + The inline heuristic can now bypass limits on the size of of 2793 inlined functions when the inlining is particularly 2794 profitable. This happens, for example, when loop bounds or 2795 array strides get propagated. 2796 + Values passed through aggregates (either by value or 2797 reference) are now propagated at the inter-procedural level 2798 leading to better inlining decisions (for example in the case 2799 of Fortran array descriptors) and devirtualization. 2800 * [6]AddressSanitizer , a fast memory error detector, has been added 2801 and can be enabled via -fsanitize=address. Memory access 2802 instructions will be instrumented to detect heap-, stack-, and 2803 global-buffer overflow as well as use-after-free bugs. To get nicer 2804 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 2805 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 2806 x86-64 Darwin. 2807 * [7]ThreadSanitizer has been added and can be enabled via 2808 -fsanitize=thread. Instructions will be instrumented to detect data 2809 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 2810 * A new local register allocator (LRA) has been implemented, which 2811 replaces the 26 year old reload pass and improves generated code 2812 quality. For now it is active on the IA-32 and x86-64 targets. 2813 * Support for transactional memory has been implemented on the 2814 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 2815 Alpha. 2816 2817 New Languages and Language specific improvements 2818 2819 C family 2820 2821 * Each diagnostic emitted now includes the original source line and a 2822 caret '^' indicating the column. The option 2823 -fno-diagnostics-show-caret suppresses this information. 2824 * The option -ftrack-macro-expansion=2 is now enabled by default. 2825 This allows the compiler to display the macro expansion stack in 2826 diagnostics. Combined with the caret information, an example 2827 diagnostic showing these two features is: 2828 2829 t.c:1:94: error: invalid operands to binary < (have struct mystruct and float 2830 ) 2831 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 2832 _b = (B); __a < __b ? __b : __a; }) 2833 2834 ^ 2835 t.c:7:7: note: in expansion of macro 'MYMAX' 2836 X = MYMAX(P, F); 2837 ^ 2838 2839 * A new -Wsizeof-pointer-memaccess warning has been added (also 2840 enabled by -Wall) to warn about suspicious length parameters to 2841 certain string and memory built-in functions if the argument uses 2842 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 2843 (ptr)); if ptr is not an array, but a pointer, and suggests a 2844 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 2845 * The new option -Wpedantic is an alias for -pedantic, which is now 2846 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 2847 -Wno-error=pedantic work in the same way as for any other -W 2848 option. One caveat is that -Werror=pedantic is not equivalent to 2849 -pedantic-errors, since the latter makes into errors some warnings 2850 that are not controlled by -Wpedantic, and the former only affects 2851 diagnostics that are disabled when using -Wno-pedantic. 2852 * The option -Wshadow no longer warns if a declaration shadows a 2853 function declaration, unless the former declares a function or 2854 pointer to function, because this is [8]a common and valid case in 2855 real-world code. 2856 2857 C++ 2858 2859 * G++ now implements the [9]C++11 thread_local keyword; this differs 2860 from the GNU __thread keyword primarily in that it allows dynamic 2861 initialization and destruction semantics. Unfortunately, this 2862 support requires a run-time penalty for references to 2863 non-function-local thread_local variables defined in a different 2864 translation unit even if they don't need dynamic initialization, so 2865 users may want to continue to use __thread for TLS variables with 2866 static initialization semantics. 2867 If the programmer can be sure that no use of the variable in a 2868 non-defining TU needs to trigger dynamic initialization (either 2869 because the variable is statically initialized, or a use of the 2870 variable in the defining TU will be executed before any uses in 2871 another TU), they can avoid this overhead with the 2872 -fno-extern-tls-init option. 2873 OpenMP threadprivate variables now also support dynamic 2874 initialization and destruction by the same mechanism. 2875 * G++ now implements the [10]C++11 attribute syntax, e.g. 2876 2877 [[noreturn]] void f(); 2878 2879 and also the alignment specifier, e.g. 2880 2881 alignas(double) int i; 2882 2883 * G++ now implements [11]C++11 inheriting constructors, e.g. 2884 2885 struct A { A(int); }; 2886 struct B: A { using A::A; }; // defines B::B(int) 2887 B b(42); // OK 2888 2889 * As of GCC 4.8.1, G++ implements the change to decltype semantics 2890 from [12]N3276. 2891 2892 struct A f(); 2893 decltype(f()) g(); // OK, return type of f() is not required to be complete. 2894 2895 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g. 2896 2897 struct A { int f() &; }; 2898 int i = A().f(); // error, f() requires an lvalue object 2899 2900 * G++ now supports a -std=c++1y option for experimentation with 2901 features proposed for the next revision of the standard, expected 2902 around 2014. Currently the only difference from -std=c++11 is 2903 support for return type deduction in normal functions, as proposed 2904 in [14]N3386. Status of C++1y features in GCC 4.8 can be found 2905 [15]here. 2906 * The G++ namespace association extension, __attribute ((strong)), 2907 has been deprecated. Inline namespaces should be used instead. 2908 * G++ now supports a -fext-numeric-literal option to control whether 2909 GNU numeric literal suffixes are accepted as extensions or 2910 processed as C++11 user-defined numeric literal suffixes. The flag 2911 is on (use suffixes for GNU literals) by default for -std=gnu++*, 2912 and -std=c++98. The flag is off (use suffixes for user-defined 2913 literals) by default for -std=c++11 and later. 2914 2915 Runtime Library (libstdc++) 2916 2917 * [16]Improved experimental support for the new ISO C++ standard, 2918 C++11, including: 2919 + forward_list meets the allocator-aware container requirements; 2920 + this_thread::sleep_for(), this_thread::sleep_until() and 2921 this_thread::yield() are defined without requiring the 2922 configure option --enable-libstdcxx-time; 2923 * Improvements to <random>: 2924 + SSE optimized normal_distribution. 2925 + Use of hardware RNG instruction for random_device on new x86 2926 processors (requires the assembler to support the 2927 instruction.) 2928 and <ext/random>: 2929 + New random number engine simd_fast_mersenne_twister_engine 2930 with an optimized SSE implementation. 2931 + New random number distributions beta_distribution, 2932 normal_mv_distribution, rice_distribution, 2933 nakagami_distribution, pareto_distribution, k_distribution, 2934 arcsine_distribution, hoyt_distribution. 2935 * Added --disable-libstdcxx-verbose configure option to disable 2936 diagnostic messages issued when a process terminates abnormally. 2937 This may be useful for embedded systems to reduce the size of 2938 executables that link statically to the library. 2939 2940 Fortran 2941 2942 * Compatibility notice: 2943 + Module files: The version of module files (.mod) has been 2944 incremented. Fortran MODULEs compiled by earlier GCC versions 2945 have to be recompiled, when they are USEd by files compiled 2946 with GCC 4.8. GCC 4.8 is not able to read .mod files created 2947 by earlier versions; attempting to do so gives an error 2948 message. 2949 Note: The ABI of the produced assembler data itself has not 2950 changed; object files and libraries are fully compatible with 2951 older versions except as noted below. 2952 + ABI: Some internal names (used in the assembler/object file) 2953 have changed for symbols declared in the specification part of 2954 a module. If an affected module or a file using it via use 2955 association is recompiled, the module and all files which 2956 directly use such symbols have to be recompiled as well. This 2957 change only affects the following kind of module symbols: 2958 o Procedure pointers. Note: C-interoperable function 2959 pointers (type(c_funptr)) are not affected nor are 2960 procedure-pointer components. 2961 o Deferred-length character strings. 2962 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 2963 backtrace at an arbitrary place in user code; program execution 2964 continues normally afterwards. 2965 * The [18]-Wc-binding-type warning option has been added (disabled by 2966 default). It warns if the a variable might not be C interoperable; 2967 in particular, if the variable has been declared using an intrinsic 2968 type with default kind instead of using a kind parameter defined 2969 for C interoperability in the intrinsic ISO_C_Binding module. 2970 Before, this warning was always printed. The -Wc-binding-type 2971 option is enabled by -Wall. 2972 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 2973 options have been added, which diagnose when code is inserted for 2974 automatic (re)allocation of a variable during assignment. This 2975 option can be used to decide whether it is safe to use 2976 [20]-fno-realloc-lhs. Additionally, it can be used to find 2977 automatic (re)allocation in hot loops. (For arrays, replacing 2978 "var=" by "var(:)=" disables the automatic reallocation.) 2979 * The [21]-Wcompare-reals command-line option has been added. When 2980 this is set, warnings are issued when comparing REAL or COMPLEX 2981 types for equality and inequality; consider replacing a == b by 2982 abs(ab) < eps with a suitable eps. -Wcompare-reals is enabled by 2983 -Wextra. 2984 * The [22]-Wtarget-lifetime command-line option has been added 2985 (enabled with -Wall), which warns if the pointer in a pointer 2986 assignment might outlive its target. 2987 * Reading floating point numbers which use "q" for the exponential 2988 (such as 4.0q0) is now supported as vendor extension for better 2989 compatibility with old data files. It is strongly recommended to 2990 use for I/O the equivalent but standard conforming "e" (such as 2991 4.0e0). 2992 (For Fortran source code, consider replacing the "q" in 2993 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 2994 suitable qp). Note that in Fortran source code replacing "q" by 2995 a simple "e" is not equivalent.) 2996 * The GFORTRAN_TMPDIR environment variable for specifying a 2997 non-default directory for files opened with STATUS="SCRATCH", is 2998 not used anymore. Instead gfortran checks the POSIX/GNU standard 2999 TMPDIR environment variable. If TMPDIR is not defined, gfortran 3000 falls back to other methods to determine the directory for 3001 temporary files as documented in the [23]user manual. 3002 * [24]Fortran 2003: 3003 + Support for unlimited polymorphic variables (CLASS(*)) has 3004 been added. Nonconstant character lengths are not yet 3005 supported. 3006 * [25]TS 29113: 3007 + Assumed types (TYPE(*)) are now supported. 3008 + Experimental support for assumed-rank arrays (dimension(..)) 3009 has been added. Note that currently gfortran's own array 3010 descriptor is used, which is different from the one defined in 3011 TS29113, see [26]gfortran's header file or use the [27]Chasm 3012 Language Interoperability Tools. 3013 3014 Go 3015 3016 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 3017 release. 3018 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 3019 release. The library support is not quite complete. 3020 * Go has been tested on GNU/Linux and Solaris platforms for various 3021 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 3022 work on other platforms as well. 3023 3024 New Targets and Target Specific Improvements 3025 3026 AArch64 3027 3028 * A new port has been added to support AArch64, the new 64-bit 3029 architecture from ARM. Note that this is a separate port from the 3030 existing 32-bit ARM port. 3031 * The port provides initial support for the Cortex-A53 and the 3032 Cortex-A57 processors with the command line options 3033 -mcpu=cortex-a53 and -mcpu=cortex-a57. 3034 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 3035 has been added and can be enabled by giving the 3036 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 3037 default by configuring GCC with the --enable-fix-cortex-a53-835769 3038 option. 3039 3040 ARM 3041 3042 * Initial support has been added for the AArch32 extensions defined 3043 in the ARMv8 architecture. 3044 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 3045 * A new option, -mcpu=marvell-pj4, has been added to generate code 3046 for the Marvell PJ4 processor. 3047 * The compiler can now automatically generate the VFMA, VFMS, REVSH 3048 and REV16 instructions. 3049 * A new vectorizer cost model for Advanced SIMD configurations to 3050 improve the auto-vectorization strategies used. 3051 * The scheduler now takes into account the number of live registers 3052 to reduce the amount of spilling that can occur. This should 3053 improve code performance in large functions. The limit can be 3054 removed by using the option -fno-sched-pressure. 3055 * Improvements have been made to the Marvell iWMMX code generation 3056 and support for the iWMMX2 SIMD unit has been added. The option 3057 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 3058 * A number of code generation improvements for Thumb2 to reduce code 3059 size when compiling for the M-profile processors. 3060 * The RTEMS (arm-rtems) port has been updated to use the EABI. 3061 * Code generation support for the old FPA and Maverick floating-point 3062 architectures has been removed. Ports that previously relied on 3063 these features have also been removed. This includes the targets: 3064 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 3065 + arm*-*-elf (use arm*-*-eabi) 3066 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 3067 + arm*-*-ecos-elf (no alternative) 3068 + arm*-*-freebsd (no alternative) 3069 + arm*-wince-pe* (no alternative). 3070 3071 AVR 3072 3073 * Support for the "Embedded C" fixed-point has been added. For 3074 details, see the [28]GCC wiki and the [29]user manual. The support 3075 is not complete. 3076 * A new print modifier %r for register operands in inline assembler 3077 is supported. It will print the raw register number without the 3078 register prefix 'r': 3079 /* Return the most significant byte of 'val', a 64-bit value. */ 3080 3081 unsigned char msb (long long val) 3082 { 3083 unsigned char c; 3084 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 3085 return c; 3086 } 3087 The inline assembler in this example will generate code like 3088 mov r24, 8+7 3089 provided c is allocated to R24 and val is allocated to R8R15. This 3090 works because the GNU assembler accepts plain register numbers 3091 without register prefix. 3092 * Static initializers with 3-byte symbols are supported now: 3093 extern const __memx char foo; 3094 const __memx void *pfoo = &foo; 3095 This requires at least Binutils 2.23. 3096 3097 IA-32/x86-64 3098 3099 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 3100 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 3101 stack alignment, this is ABI incompatible and intended to be used 3102 in controlled environments where stack space is an important 3103 limitation. This option will lead to wrong code when functions 3104 compiled with 16 byte stack alignment (such as functions from a 3105 standard library) are called with misaligned stack. In this case, 3106 SSE instructions may lead to misaligned memory access traps. In 3107 addition, variable arguments will be handled incorrectly for 16 3108 byte aligned objects (including x87 long double and __int128), 3109 leading to wrong results. You must build all modules with 3110 -mpreferred-stack-boundary=3, including any libraries. This 3111 includes the system libraries and startup modules. 3112 * Support for the new Intel processor codename Broadwell with RDSEED, 3113 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 3114 -mrdseed command-line options. 3115 * Support for the Intel RTM and HLE intrinsics, built-in functions 3116 and code generation is available via -mrtm and -mhle. 3117 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 3118 Intrinsics and built-in functions are available via -mfxsr, -mxsave 3119 and -mxsaveopt respectively. 3120 * New -maddress-mode=[short|long] options for x32. 3121 -maddress-mode=short overrides default 64-bit addresses to 32-bit 3122 by emitting the 0x67 address-size override prefix. This is the 3123 default address mode for x32. 3124 * New built-in functions to detect run-time CPU type and ISA: 3125 + A built-in function __builtin_cpu_is has been added to detect 3126 if the run-time CPU is of a particular type. It returns a 3127 positive integer on a match and zero otherwise. It accepts one 3128 string literal argument, the CPU name. For example, 3129 __builtin_cpu_is("westmere") returns a positive integer if the 3130 run-time CPU is an Intel Core i7 Westmere processor. Please 3131 refer to the [30]user manual for the list of valid CPU names 3132 recognized. 3133 + A built-in function __builtin_cpu_supports has been added to 3134 detect if the run-time CPU supports a particular ISA feature. 3135 It returns a positive integer on a match and zero otherwise. 3136 It accepts one string literal argument, the ISA feature. For 3137 example, __builtin_cpu_supports("ssse3") returns a positive 3138 integer if the run-time CPU supports SSSE3 instructions. 3139 Please refer to the [31]user manual for the list of valid ISA 3140 names recognized. 3141 Caveat: If these built-in functions are called before any static 3142 constructors are invoked, like during IFUNC initialization, then 3143 the CPU detection initialization must be explicitly run using this 3144 newly provided built-in function, __builtin_cpu_init. The 3145 initialization needs to be done only once. For example, this is how 3146 the invocation would look like inside an IFUNC initializer: 3147 static void (*some_ifunc_resolver(void))(void) 3148 { 3149 __builtin_cpu_init(); 3150 if (__builtin_cpu_is("amdfam10h") ... 3151 if (__builtin_cpu_supports("popcnt") ... 3152 } 3153 3154 * Function Multiversioning Support with G++: 3155 It is now possible to create multiple function versions each 3156 targeting a specific processor and/or ISA. Function versions have 3157 the same signature but different target attributes. For example, 3158 here is a program with function versions: 3159 __attribute__ ((target ("default"))) 3160 int foo(void) 3161 { 3162 return 1; 3163 } 3164 3165 __attribute__ ((target ("sse4.2"))) 3166 int foo(void) 3167 { 3168 return 2; 3169 } 3170 3171 int main (void) 3172 { 3173 int (*p) = &foo; 3174 assert ((*p)() == foo()); 3175 return 0; 3176 } 3177 3178 Please refer to this [32]wiki for more information. 3179 * The x86 back end has been improved to allow option -fschedule-insns 3180 to work reliably. This option can be used to schedule instructions 3181 better and leads to improved performace in certain cases. 3182 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 3183 from the Mingw-w64 trunk. 3184 * Support for new AMD family 15h processors (Steamroller core) is now 3185 available through the -march=bdver3 and -mtune=bdver3 options. 3186 * Support for new AMD family 16h processors (Jaguar core) is now 3187 available through the -march=btver2 and -mtune=btver2 options. 3188 3189 FRV 3190 3191 * This target now supports the -fstack-usage command-line option. 3192 3193 MIPS 3194 3195 * GCC can now generate code specifically for the R4700, Broadcom XLP 3196 and MIPS 34kn processors. The associated -march options are 3197 -march=r4700, -march=xlp and -march=34kn respectively. 3198 * GCC now generates better DSP code for MIPS 74k cores thanks to 3199 further scheduling optimizations. 3200 * The MIPS port now supports the -fstack-check option. 3201 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 3202 * Previous versions of GCC would silently accept -fpic and -fPIC for 3203 -mno-abicalls targets like mips*-elf. This combination was not 3204 intended or supported, and did not generate position-independent 3205 code. GCC 4.8 now reports an error when this combination is used. 3206 3207 PowerPC / PowerPC64 / RS6000 3208 3209 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 3210 restore or update the VRSAVE register by default. The respective 3211 operating systems manage the VRSAVE register directly. 3212 * Large TOC support has been added for AIX through the command line 3213 option -mcmodel=large. 3214 * Native Thread-Local Storage support has been added for AIX. 3215 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 3216 when targetting processors that support those hardware features on 3217 AIX 6.1 and above. 3218 3219 RX 3220 3221 * This target will now issue a warning message whenever multiple fast 3222 interrupt handlers are found in the same compilation unit. This 3223 feature can be turned off by the new 3224 -mno-warn-multiple-fast-interrupts command-line option. 3225 3226 S/390, System z 3227 3228 * Support for the IBM zEnterprise zEC12 processor has been added. 3229 When using the -march=zEC12 option, the compiler will generate code 3230 making use of the following new instructions: 3231 + load and trap instructions 3232 + 2 new compare and trap instructions 3233 + rotate and insert selected bits - without CC clobber 3234 The -mtune=zEC12 option enables zEC12 specific instruction 3235 scheduling without making use of new instructions. 3236 * Register pressure sensitive instruction scheduling is enabled by 3237 default. 3238 * The ifunc function attribute is enabled by default. 3239 * memcpy and memcmp invokations on big memory chunks or with run time 3240 lengths are not generated inline anymore when tuning for z10 or 3241 higher. The purpose is to make use of the IFUNC optimized versions 3242 in Glibc. 3243 3244 SH 3245 3246 * The default alignment settings have been reduced to be less 3247 aggressive. This results in more compact code for optimization 3248 levels other than -Os. 3249 * Improved support for the __atomic built-in functions: 3250 + A new option -matomic-model=model selects the model for the 3251 generated atomic sequences. The following models are 3252 supported: 3253 3254 soft-gusa 3255 Software gUSA sequences (SH3* and SH4* only). On 3256 SH4A targets this will now also partially utilize 3257 the movco.l and movli.l instructions. This is the 3258 default when the target is sh3*-*-linux* or 3259 sh4*-*-linux*. 3260 3261 hard-llcs 3262 Hardware movco.l / movli.l sequences (SH4A only). 3263 3264 soft-tcb 3265 Software thread control block sequences. 3266 3267 soft-imask 3268 Software interrupt flipping sequences (privileged 3269 mode only). This is the default when the target is 3270 sh1*-*-linux* or sh2*-*-linux*. 3271 3272 none 3273 Generates function calls to the respective __atomic 3274 built-in functions. This is the default for SH64 3275 targets or when the target is not sh*-*-linux*. 3276 3277 + The option -msoft-atomic has been deprecated. It is now an 3278 alias for -matomic-model=soft-gusa. 3279 + A new option -mtas makes the compiler generate the tas.b 3280 instruction for the __atomic_test_and_set built-in function 3281 regardless of the selected atomic model. 3282 + The __sync functions in libgcc now reflect the selected atomic 3283 model when building the toolchain. 3284 * Added support for the mov.b and mov.w instructions with 3285 displacement addressing. 3286 * Added support for the SH2A instructions movu.b and movu.w. 3287 * Various improvements to code generated for integer arithmetic. 3288 * Improvements to conditional branches and code that involves the T 3289 bit. A new option -mzdcbranch tells the compiler to favor 3290 zero-displacement branches. This is enabled by default for SH4* 3291 targets. 3292 * The pref instruction will now be emitted by the __builtin_prefetch 3293 built-in function for SH3* targets. 3294 * The fmac instruction will now be emitted by the fmaf standard 3295 function and the __builtin_fmaf built-in function. 3296 * The -mfused-madd option has been deprecated in favor of the 3297 machine-independent -ffp-contract option. Notice that the fmac 3298 instruction will now be generated by default for expressions like a 3299 * b + c. This is due to the compiler default setting 3300 -ffp-contract=fast. 3301 * Added new options -mfsrra and -mfsca to allow the compiler using 3302 the fsrra and fsca instructions on targets other than SH4A (where 3303 they are already enabled by default). 3304 * Added support for the __builtin_bswap32 built-in function. It is 3305 now expanded as a sequence of swap.b and swap.w instructions 3306 instead of a library function call. 3307 * The behavior of the -mieee option has been fixed and the negative 3308 form -mno-ieee has been added to control the IEEE conformance of 3309 floating point comparisons. By default -mieee is now enabled and 3310 the option -ffinite-math-only implicitly sets -mno-ieee. 3311 * Added support for the built-in functions __builtin_thread_pointer 3312 and __builtin_set_thread_pointer. This assumes that GBR is used to 3313 hold the thread pointer of the current thread. Memory loads and 3314 stores relative to the address returned by __builtin_thread_pointer 3315 will now also utilize GBR based displacement address modes. 3316 * The -mdiv= option for targets other than SHmedia has been fixed and 3317 documented. 3318 3319 SPARC 3320 3321 * Added optimized instruction scheduling for Niagara4. 3322 3323 TILE-Gx 3324 3325 * Added support for the -mcmodel=MODEL command-line option. The 3326 models supported are small and large. 3327 3328 V850 3329 3330 * This target now supports the E3V5 architecture via the use of the 3331 new -mv850e3v5 command-line option. It also has experimental 3332 support for the e3v5 LOOP instruction which can be enabled via the 3333 new -mloop command-line option. 3334 3335 XStormy16 3336 3337 * This target now supports the -fstack-usage command-line option. 3338 3339 Operating Systems 3340 3341 OpenBSD 3342 3343 * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and 3344 support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated. 3345 3346 Windows (Cygwin) 3347 3348 * Executables are now linked against shared libgcc by default. The 3349 previous default was to link statically, which can still be done by 3350 explicitly specifying -static or static-libgcc on the command line. 3351 However it is strongly advised against, as it will cause problems 3352 for any application that makes use of DLLs compiled by GCC. It 3353 should be alright for a monolithic stand-alone application that 3354 only links against the Windows DLLs, but offers little or no 3355 benefit. 3356 3357 GCC 4.8.1 3358 3359 This is the [33]list of problem reports (PRs) from GCC's bug tracking 3360 system that are known to be fixed in the 4.8.1 release. This list might 3361 not be complete (that is, it is possible that some PRs that have been 3362 fixed are not listed here). 3363 3364 The C++11 <chrono> std::chrono::system_clock and 3365 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 3366 both are now separate (never typedefs of each other), both use 3367 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 3368 std::chrono::steady_clock is now finally monotonic, and both classes 3369 are mangled differently than in the previous GCC releases. 3370 std::chrono::system_clock::now() with std::chrono::microseconds resp. 3371 std::chrono::seconds resolution is still exported for backwards 3372 compatibility with default configured libstdc++. Note that libstdc++ 3373 configured with --enable-libstdcxx-time= used to be ABI incompatible 3374 with default configured libstdc++ for those two classes and no ABI 3375 compatibility can be offered for those configurations, so any C++11 3376 code that uses those classes and has been compiled and linked against 3377 libstdc++ configured with the non-default --enable-libstdcxx-time= 3378 configuration option needs to be recompiled. 3379 3380 GCC 4.8.2 3381 3382 This is the [34]list of problem reports (PRs) from GCC's bug tracking 3383 system that are known to be fixed in the 4.8.2 release. This list might 3384 not be complete (that is, it is possible that some PRs that have been 3385 fixed are not listed here). 3386 3387 GCC 4.8.3 3388 3389 This is the [35]list of problem reports (PRs) from GCC's bug tracking 3390 system that are known to be fixed in the 4.8.3 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 Support for the new powerpc64le-linux platform has been added. It 3395 defaults to generating code that conforms to the ELFV2 ABI. 3396 3397 GCC 4.8.4 3398 3399 This is the [36]list of problem reports (PRs) from GCC's bug tracking 3400 system that are known to be fixed in the 4.8.4 release. This list might 3401 not be complete (that is, it is possible that some PRs that have been 3402 fixed are not listed here). 3403 3404 GCC 4.8.5 3405 3406 This is the [37]list of problem reports (PRs) from GCC's bug tracking 3407 system that are known to be fixed in the 4.8.5 release. This list might 3408 not be complete (that is, it is possible that some PRs that have been 3409 fixed are not listed here). 3410 3411 3412 For questions related to the use of GCC, please consult these web 3413 pages and the [38]GCC manuals. If that fails, the 3414 [39]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3415 web pages and the development of GCC are welcome on our developer 3416 list at [40]gcc (a] gcc.gnu.org. All of [41]our lists have public 3417 archives. 3418 3419 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and 3420 distribution of this entire article is permitted in any medium, 3421 provided this notice is preserved. 3422 3423 These pages are [43]maintained by the GCC team. Last modified 3424 2018-09-30[44]. 3425 3426 References 3427 3428 1. https://gcc.gnu.org/wiki/cxx-conversion 3429 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 3430 3. http://www.nongnu.org/avr-libc/ 3431 4. https://gcc.gnu.org/PR54461 3432 5. https://gcc.gnu.org/gcc-4.8/porting_to.html 3433 6. https://github.com/google/sanitizers 3434 7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 3435 8. https://lkml.org/lkml/2006/11/28/239 3436 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 3437 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 3438 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 3439 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 3440 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 3441 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 3442 15. http://gcc.gnu.org/projects/cxx1y.html 3443 16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 3444 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 3445 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 3446 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 3447 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 3448 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 3449 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 3450 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 3451 24. https://gcc.gnu.org/wiki/Fortran2003Status 3452 25. https://gcc.gnu.org/wiki/TS29113Status 3453 26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text/plain&view=co 3454 27. http://chasm-interop.sourceforge.net/ 3455 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 3456 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 3457 30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 3458 31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 3459 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning 3460 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 3461 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 3462 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 3463 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 3464 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 3465 38. https://gcc.gnu.org/onlinedocs/ 3466 39. mailto:gcc-help (a] gcc.gnu.org 3467 40. mailto:gcc (a] gcc.gnu.org 3468 41. https://gcc.gnu.org/lists.html 3469 42. https://www.fsf.org/ 3470 43. https://gcc.gnu.org/about.html 3471 44. http://validator.w3.org/check/referer 3472 ====================================================================== 3473 http://gcc.gnu.org/gcc-4.7/index.html 3474 GCC 4.7 Release Series 3475 3476 June 12, 2014 3477 3478 The [1]GNU project and the GCC developers are pleased to announce the 3479 release of GCC 4.7.4. 3480 3481 This release is a bug-fix release, containing fixes for regressions in 3482 GCC 4.7.3 relative to previous releases of GCC. 3483 3484 This release series is no longer maintained. 3485 3486 Release History 3487 3488 GCC 4.7.4 3489 June 12, 2014 ([2]changes, [3]documentation) 3490 3491 GCC 4.7.3 3492 April 11, 2013 ([4]changes, [5]documentation) 3493 3494 GCC 4.7.2 3495 September 20, 2012 ([6]changes, [7]documentation) 3496 3497 GCC 4.7.1 3498 June 14, 2012 ([8]changes, [9]documentation) 3499 3500 GCC 4.7.0 3501 March 22, 2012 ([10]changes, [11]documentation) 3502 3503 References and Acknowledgements 3504 3505 GCC used to stand for the GNU C Compiler, but since the compiler 3506 supports several other languages aside from C, it now stands for the 3507 GNU Compiler Collection. 3508 3509 A list of [12]successful builds is updated as new information becomes 3510 available. 3511 3512 The GCC developers would like to thank the numerous people that have 3513 contributed new features, improvements, bug fixes, and other changes as 3514 well as test results to GCC. This [13]amazing group of volunteers is 3515 what makes GCC successful. 3516 3517 For additional information about GCC please refer to the [14]GCC 3518 project web site or contact the [15]GCC development mailing list. 3519 3520 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 3521 3522 3523 For questions related to the use of GCC, please consult these web 3524 pages and the [18]GCC manuals. If that fails, the 3525 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3526 web pages and the development of GCC are welcome on our developer 3527 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 3528 archives. 3529 3530 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 3531 distribution of this entire article is permitted in any medium, 3532 provided this notice is preserved. 3533 3534 These pages are [23]maintained by the GCC team. Last modified 3535 2018-09-30[24]. 3536 3537 References 3538 3539 1. http://www.gnu.org/ 3540 2. http://gcc.gnu.org/gcc-4.7/changes.html 3541 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 3542 4. http://gcc.gnu.org/gcc-4.7/changes.html 3543 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 3544 6. http://gcc.gnu.org/gcc-4.7/changes.html 3545 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 3546 8. http://gcc.gnu.org/gcc-4.7/changes.html 3547 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 3548 10. http://gcc.gnu.org/gcc-4.7/changes.html 3549 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 3550 12. http://gcc.gnu.org/gcc-4.7/buildstat.html 3551 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 3552 14. http://gcc.gnu.org/index.html 3553 15. mailto:gcc (a] gcc.gnu.org 3554 16. http://gcc.gnu.org/mirrors.html 3555 17. http://gcc.gnu.org/svn.html 3556 18. https://gcc.gnu.org/onlinedocs/ 3557 19. mailto:gcc-help (a] gcc.gnu.org 3558 20. mailto:gcc (a] gcc.gnu.org 3559 21. https://gcc.gnu.org/lists.html 3560 22. https://www.fsf.org/ 3561 23. https://gcc.gnu.org/about.html 3562 24. http://validator.w3.org/check/referer 3563 ====================================================================== 3564 http://gcc.gnu.org/gcc-4.7/changes.html 3565 GCC 4.7 Release Series 3566 Changes, New Features, and Fixes 3567 3568 Caveats 3569 3570 * The -fconserve-space flag has been deprecated. The flag had no 3571 effect for most targets: only targets without a global .bss section 3572 and without support for switchable sections. Furthermore, the flag 3573 only had an effect for G++, where it could result in wrong 3574 semantics (please refer to the GCC manual for further details). The 3575 flag will be removed in GCC 4.8 3576 * Support for a number of older systems and recently unmaintained or 3577 untested target ports of GCC has been declared obsolete in GCC 4.7. 3578 Unless there is activity to revive them, the next release of GCC 3579 will have their sources permanently removed. 3580 All GCC ports for the following processor architectures have been 3581 declared obsolete: 3582 + picoChip (picochip-*) 3583 The following ports for individual systems on particular 3584 architectures have been obsoleted: 3585 + IRIX 6.5 (mips-sgi-irix6.5) 3586 + MIPS OpenBSD (mips*-*-openbsd*) 3587 + Solaris 8 (*-*-solaris2.8). Details can be found in the 3588 [1]announcement. 3589 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 3590 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 3591 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 3592 default, which for some sources generates code that accesses memory 3593 on unaligned addresses. This requires the kernel of those systems 3594 to enable such accesses (controlled by CP15 register c1, refer to 3595 ARM documentation). Alternatively, or for compatibility with 3596 kernels where unaligned accesses are not supported, all code has to 3597 be compiled with -mno-unaligned-access. Upstream Linux kernel 3598 releases have automatically and unconditionally supported unaligned 3599 accesses as emitted by GCC due to this option being active since 3600 version 2.6.28. 3601 * Support on ARM for the legacy floating-point accelerator (FPA) and 3602 the mixed-endian floating-point format that it used has been 3603 obsoleted. The ports that still use this format have been obsoleted 3604 as well. Many legacy ARM ports already provide an alternative that 3605 uses the VFP floating-point format. The obsolete ports will be 3606 deleted in the next release. 3607 The obsolete ports with alternatives are: 3608 + arm*-*-rtems (use arm*-*-rtemseabi) 3609 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 3610 + arm*-*-elf (use arm*-*-eabi) 3611 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 3612 Note, however, that these alternatives are not binary compatible 3613 with their legacy counterparts (although some can support running 3614 legacy applications). 3615 The obsolete ports that currently lack a modern alternative are: 3616 + arm*-*-ecos-elf 3617 + arm*-*-freebsd 3618 + arm*-wince-pe* 3619 New ports that support more recent versions of the architecture are 3620 welcome. 3621 * Support for the Maverick co-processor on ARM has been obsoleted. 3622 Code to support it will be deleted in the next release. 3623 * Support has been removed for Unix International threads on Solaris 3624 2, so the --enable-threads=solaris configure option and the 3625 -threads compiler option don't work any longer. 3626 * Support has been removed for the Solaris BSD Compatibility Package, 3627 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 3628 from Solaris 11, and was only intended as a migration aid from 3629 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 3630 recognized any longer. 3631 * The AVR port's libgcc has been improved and its multilib structure 3632 has been enhanced. As a result, all objects contributing to an 3633 application must either be compiled with GCC versions up to 4.6.x 3634 or with GCC versions 4.7.1 or later. If the compiler is used with 3635 AVR Libc, you need a version that supports the new layout, i.e. 3636 implements [2]#35407. 3637 * The AVR port's -mshort-calls command-line option has been 3638 deprecated. It will be removed in the GCC 4.8 release. See -mrelax 3639 for a replacement. 3640 * The AVR port only references startup code that clears .bss and the 3641 common section resp. initializes the .data and .rodata section 3642 provided respective sections (or subsections thereof) are not 3643 empty, see [3]PR18145. Applications that put all static storage 3644 objects into non-standard sections and / or define all static 3645 storage objects in assembler modules, must reference __do_clear_bss 3646 resp. __do_copy_data by hand or undefine the symbol(s) by means of 3647 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 3648 * The ARM port's -mwords-little-endian option has been deprecated. It 3649 will be removed in a future release. 3650 * Support has been removed for the NetWare x86 configuration 3651 obsoleted in GCC 4.6. 3652 * It is no longer possible to use the "l" constraint in MIPS16 asm 3653 statements. 3654 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 3655 library which affected the ABI in C++11 mode: a data member was 3656 added to std::list changing its size and altering the definitions 3657 of some member functions, and std::pair's move constructor was 3658 non-trivial which altered the calling convention for functions with 3659 std::pair arguments or return types. The ABI incompatibilities have 3660 been fixed for GCC version 4.7.2 but as a result C++11 code 3661 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 3662 code compiled with different GCC versions and with C++98/C++03 code 3663 compiled with any version. 3664 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 3665 rules for the layout of vectors that could lead to wrong code being 3666 generated. Vectors larger than 8 bytes in size are now by default 3667 aligned to an 8-byte boundary. This is an ABI change: code that 3668 makes explicit use of vector types may be incompatible with binary 3669 objects built with older versions of GCC. Auto-vectorized code is 3670 not affected by this change. (This change affects GCC versions 3671 4.7.2 and later.) 3672 * More information on porting to GCC 4.7 from previous versions of 3673 GCC can be found in the [4]porting guide for this release. 3674 3675 General Optimizer Improvements 3676 3677 * Support for a new parameter --param case-values-threshold=n was 3678 added to allow users to control the cutoff between doing switch 3679 statements as a series of if statements and using a jump table. 3680 * Link-time optimization (LTO) improvements: 3681 + Improved scalability and reduced memory usage. Link time 3682 optimization of Firefox now requires 3GB of RAM on a 64-bit 3683 system, while over 8GB was needed previously. Linking time has 3684 been improved, too. The serial stage of linking Firefox has 3685 been sped up by about a factor of 10. 3686 + Reduced size of object files and temporary storage used during 3687 linking. 3688 + Streaming performance (both outbound and inbound) has been 3689 improved. 3690 + ld -r is now supported with LTO. 3691 + Several bug fixes, especially in symbol table handling and 3692 merging. 3693 * Interprocedural optimization improvements: 3694 + Heuristics now take into account that after inlining code will 3695 be optimized out because of known values (or properties) of 3696 function parameters. For example: 3697 void foo(int a) 3698 { 3699 if (a > 10) 3700 ... huge code ... 3701 } 3702 void bar (void) 3703 { 3704 foo (0); 3705 } 3706 3707 The call of foo will be inlined into bar even when optimizing 3708 for code size. Constructs based on __builtin_constant_p are 3709 now understood by the inliner and code size estimates are 3710 evaluated a lot more realistically. 3711 + The representation of C++ virtual thunks and aliases (both 3712 implicit and defined via the alias attribute) has been 3713 re-engineered. Aliases no longer pose optimization barriers 3714 and calls to an alias can be inlined and otherwise optimized. 3715 + The inter-procedural constant propagation pass has been 3716 rewritten. It now performs generic function specialization. 3717 For example when compiling the following: 3718 void foo(bool flag) 3719 { 3720 if (flag) 3721 ... do something ... 3722 else 3723 ... do something else ... 3724 } 3725 void bar (void) 3726 { 3727 foo (false); 3728 foo (true); 3729 foo (false); 3730 foo (true); 3731 foo (false); 3732 foo (true); 3733 } 3734 3735 GCC will now produce two copies of foo. One with flag being 3736 true, while other with flag being false. This leads to 3737 performance improvements previously possible only by inlining 3738 all calls. Cloning causes a lot less code size growth. 3739 * A string length optimization pass has been added. It attempts to 3740 track string lengths and optimize various standard C string 3741 functions like strlen, strchr, strcpy, strcat, stpcpy and their 3742 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 3743 enabled by default at -O2 or above, unless optimizing for size, and 3744 can be disabled by the -fno-optimize-strlen option. The pass can 3745 e.g. optimize 3746 char *bar (const char *a) 3747 { 3748 size_t l = strlen (a) + 2; 3749 char *p = malloc (l); if (p == NULL) return p; 3750 strcpy (p, a); strcat (p, "/"); return p; 3751 } 3752 3753 into: 3754 char *bar (const char *a) 3755 { 3756 size_t tmp = strlen (a); 3757 char *p = malloc (tmp + 2); if (p == NULL) return p; 3758 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 3759 } 3760 3761 or for hosted compilations where stpcpy is available in the runtime 3762 and headers provide its prototype, e.g. 3763 void foo (char *a, const char *b, const char *c, const char *d) 3764 { 3765 strcpy (a, b); strcat (a, c); strcat (a, d); 3766 } 3767 3768 can be optimized into: 3769 void foo (char *a, const char *b, const char *c, const char *d) 3770 { 3771 strcpy (stpcpy (stpcpy (a, b), c), d); 3772 } 3773 3774 New Languages and Language specific improvements 3775 3776 * Version 3.1 of the [5]OpenMP specification is now supported for the 3777 C, C++, and Fortran compilers. 3778 3779 Ada 3780 3781 * The command-line option -feliminate-unused-debug-types has been 3782 re-enabled by default, as it is for the other languages, leading to 3783 a reduction in debug info size of 12.5% and more for relevant 3784 cases, as well as to a small compilation speedup. 3785 3786 C family 3787 3788 * A new built-in, __builtin_assume_aligned, has been added, through 3789 which the compiler can be hinted about pointer alignment and can 3790 use it to improve generated code. 3791 * A new warning option -Wunused-local-typedefs was added for C, C++, 3792 Objective-C and Objective-C++. This warning diagnoses typedefs 3793 locally defined in a function, and otherwise not used. 3794 * A new experimental command-line option -ftrack-macro-expansion was 3795 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 3796 the compiler to emit diagnostic about the current macro expansion 3797 stack when a compilation error occurs in a macro expansion. 3798 * Experimental support for transactional memory has been added. It 3799 includes support in the compiler, as well as a supporting runtime 3800 library called libitm. To compile code with transactional memory 3801 constructs, use the -fgnu-tm option. 3802 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 3803 and 32-bit/64-bit x86 platforms. 3804 For more details on transactional memory see [6]the GCC WiKi. 3805 * Support for atomic operations specifying the C++11/C11 memory model 3806 has been added. These new __atomic routines replace the existing 3807 __sync built-in routines. 3808 Atomic support is also available for memory blocks. Lock-free 3809 instructions will be used if a memory block is the same size and 3810 alignment as a supported integer type. Atomic operations which do 3811 not have lock-free support are left as function calls. A set of 3812 library functions is available on the GCC atomic wiki in the 3813 "External Atomics Library" section. 3814 For more details on the memory models and features, see the 3815 [7]atomic wiki. 3816 * When a binary operation is performed on vector types and one of the 3817 operands is a uniform vector, it is possible to replace the vector 3818 with the generating element. For example: 3819 typedef int v4si __attribute__ ((vector_size (16))); 3820 v4si res, a = {1,2,3,4}; 3821 int x; 3822 3823 res = 2 + a; /* means {2,2,2,2} + a */ 3824 res = a - x; /* means a - {x,x,x,x} */ 3825 3826 C 3827 3828 * There is support for some more features from the C11 revision of 3829 the ISO C standard. GCC now accepts the options -std=c11 and 3830 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 3831 + Unicode strings (previously supported only with options such 3832 as -std=gnu11, now supported with -std=c11), and the 3833 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 3834 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 3835 + Alignment support (_Alignas, _Alignof, max_align_t, 3836 <stdalign.h>). 3837 + A built-in function __builtin_complex is provided to support C 3838 library implementation of the CMPLX family of macros. 3839 3840 C++ 3841 3842 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 3843 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 3844 -Wc++0x-compat, respectively. 3845 * G++ now implements [8]C++11 extended friend syntax: 3846 3847 template<class W> 3848 class Q 3849 { 3850 static const int I = 2; 3851 public: 3852 friend W; 3853 }; 3854 3855 struct B 3856 { 3857 int ar[Q<B>::I]; 3858 }; 3859 3860 * Thanks to Ville Voutilainen, G++ now implements [9]C++11 explicit 3861 override control. 3862 3863 struct B { 3864 virtual void f() const final; 3865 virtual void f(int); 3866 }; 3867 3868 struct D : B { 3869 void f() const; // error: D::f attempts to override final B::f 3870 void f(long) override; // error: doesn't override anything 3871 void f(int) override; // ok 3872 }; 3873 3874 struct E final { }; 3875 struct F: E { }; // error: deriving from final class 3876 3877 * G++ now implements [10]C++11 non-static data member initializers. 3878 3879 struct A { 3880 int i = 42; 3881 } a; // initializes a.i to 42 3882 3883 * Thanks to Ed Smith-Rowland, G++ now implements [11]C++11 3884 user-defined literals. 3885 3886 // Not actually a good approximation. :) 3887 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 3888 long double pi = 180.0_degrees; 3889 3890 * G++ now implements [12]C++11 alias-declarations. 3891 3892 template <class T> using Ptr = T*; 3893 Ptr<int> ip; // decltype(ip) is int* 3894 3895 * Thanks to Ville Voutilainen and Pedro Lamaro, G++ now implements 3896 [13]C++11 delegating constructors. 3897 3898 struct A { 3899 A(int); 3900 A(): A(42) { } // delegate to the A(int) constructor 3901 }; 3902 3903 * G++ now fully implements C++11 atomic classes rather than just 3904 integer derived classes. 3905 3906 class POD { 3907 int a; 3908 int b; 3909 }; 3910 std::atomic<POD> my_atomic_POD; 3911 3912 * G++ now sets the predefined macro __cplusplus to the correct value, 3913 199711L for C++98/03, and 201103L for C++11. 3914 * G++ now correctly implements the two-phase lookup rules such that 3915 an unqualified name used in a template must have an appropriate 3916 declaration found either in scope at the point of definition of the 3917 template or by argument-dependent lookup at the point of 3918 instantiation. As a result, code that relies on a second 3919 unqualified lookup at the point of instantiation to find functions 3920 declared after the template or in dependent bases will be rejected. 3921 The compiler will suggest ways to fix affected code, and using the 3922 -fpermissive compiler flag will allow the code to compile with a 3923 warning. 3924 3925 template <class T> 3926 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 3927 void g(int) { } // fix by moving this declaration before the declaration of f 3928 3929 template <class T> 3930 struct A: T { 3931 // error, B::g(B) not found by argument-dependent lookup 3932 void f() { g(T()); } // fix by using this->g or A::g 3933 }; 3934 3935 struct B { void g(B); }; 3936 3937 int main() 3938 { 3939 f<int>(); 3940 A<B>().f(); 3941 } 3942 3943 * G++ now properly re-uses stack space allocated for temporary 3944 objects when their lifetime ends, which can significantly lower 3945 stack consumption for some C++ functions. As a result of this, some 3946 code with undefined behavior will now break: 3947 3948 const int &f(const int &i) { return i; } 3949 .... 3950 const int &x = f(1); 3951 const int &y = f(2); 3952 3953 Here, x refers to the temporary allocated to hold the 1 argument, 3954 which only lives until the end of the initialization; it 3955 immediately becomes a dangling reference. So the next statement 3956 re-uses the stack slot to hold the 2 argument, and users of x get 3957 that value instead. 3958 Note that this should not cause any change of behavior for 3959 temporaries of types with non-trivial destructors, as they are 3960 already destroyed at end of full-expression; the change is that now 3961 the storage is released as well. 3962 * A new command-line option -Wdelete-non-virtual-dtor has been added 3963 to warn when delete is used to destroy an instance of a class which 3964 has virtual functions and non-virtual destructor. It is unsafe to 3965 delete an instance of a derived class through a pointer to a base 3966 class if the base class does not have a virtual destructor. This 3967 warning is enabled by -Wall. 3968 * A new command-line option -Wzero-as-null-pointer-constant has been 3969 added to warn when a literal '0' is used as null pointer constant. 3970 It can be useful to facilitate the conversion to nullptr in C++11. 3971 * As per C++98, access-declarations are now deprecated by G++. 3972 Using-declarations are to be used instead. Furthermore, some 3973 efforts have been made to improve the support of class scope 3974 using-declarations. In particular, using-declarations referring to 3975 a dependent type now work as expected ([14]bug c++/14258). 3976 * The ELF symbol visibility of a template instantiation is now 3977 properly constrained by the visibility of its template arguments 3978 ([15]bug c++/35688). 3979 3980 Runtime Library (libstdc++) 3981 3982 * [16]Improved experimental support for the new ISO C++ standard, 3983 C++11, including: 3984 + using noexcept in most of the library; 3985 + implementations of pointer_traits, allocator_traits and 3986 scoped_allocator_adaptor; 3987 + uses-allocator construction for tuple; 3988 + vector meets the allocator-aware container requirements; 3989 + replacing monotonic_clock with steady_clock; 3990 + enabling the thread support library on most POSIX targets; 3991 + many small improvements to conform to the FDIS. 3992 * Added --enable-clocale=newlib configure option. 3993 * Debug Mode iterators for unordered associative containers. 3994 * Avoid polluting the global namespace and do not include <unistd.h>. 3995 3996 Fortran 3997 3998 * The compile flag [17]-fstack-arrays has been added, which causes 3999 all local arrays to be put on stack memory. For some programs this 4000 will improve the performance significantly. If your program uses 4001 very large local arrays, it is possible that you will have to 4002 extend your runtime limits for stack memory. 4003 * The [18]-Ofast flag now also implies [19]-fno-protect-parens and 4004 [20]-fstack-arrays. 4005 * Front-end optimizations can now be selected by the 4006 [21]-ffrontend-optimize option and deselected by the 4007 -fno-frontend-optimize option. 4008 * When front-end optimization removes a function call, 4009 [22]-Wfunction-elimination warns about that. 4010 * When performing front-end-optimization, the 4011 [23]-faggressive-function-elimination option allows the removal of 4012 duplicate function calls even for impure functions. 4013 * The flag [24]-Wreal-q-constant has been added, which warns if 4014 floating-point literals have been specified using q (such as 4015 1.0q0); the q marker is now supported as a vendor extension to 4016 denote quad precision (REAL(16) or, if not available, REAL(10)). 4017 Consider using a kind parameter (such as in 1.0_qp) instead, which 4018 can be obtained via [25]SELECTED_REAL_KIND. 4019 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 4020 Fortran now always prints error messages to standard error. If you 4021 wish to redirect standard error, please consult the manual for your 4022 OS, shell, batch environment etc. as appropriate. 4023 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 4024 variable have been removed. When encountering a serious error, 4025 gfortran will now always abort the program. Whether a core dump is 4026 generated depends on the user environment settings; see the ulimit 4027 -c setting for POSIX shells, limit coredumpsize for C shells, and 4028 the [26]WER user-mode dumps settings on Windows. 4029 * The [27]-fbacktrace option is now enabled by default. When 4030 encountering a fatal error, gfortran will attempt to print a 4031 backtrace to standard error before aborting. It can be disabled 4032 with -fno-backtrace. Note: On POSIX targets with the addr2line 4033 utility from GNU binutils, GNU Fortran can print a backtrace with 4034 function name, file name, line number information in addition to 4035 the addresses; otherwise only the addresses are printed. 4036 * [28]Fortran 2003: 4037 + Generic interface names which have the same name as derived 4038 types are now supported, which allows to write constructor 4039 functions. Note that Fortran does not support static 4040 constructor functions; only default initialization or an 4041 explicit structure-constructor initialization are available. 4042 + [29]Polymorphic (class) arrays are now supported. 4043 * [30]Fortran 2008: 4044 + Support for the DO CONCURRENT construct has been added, which 4045 allows the user to specify that individual loop iterations 4046 have no interdependencies. 4047 + [31]Coarrays: Full single-image support except for polymorphic 4048 coarrays. Additionally, preliminary support for multiple 4049 images via an MPI-based [32]coarray communication library has 4050 been added. Note: The library version is not yet usable as 4051 remote coarray access is not yet possible. 4052 * [33]TS 29113: 4053 + New flag [34]-std=f2008ts permits programs that are expected 4054 to conform to the Fortran 2008 standard and the draft 4055 Technical Specification (TS) 29113 on Further Interoperability 4056 of Fortran with C. 4057 + The OPTIONAL attribute is now allowed for dummy arguments of 4058 BIND(C) procedures. 4059 + The RANK intrinsic has been added. 4060 + The implementation of the ASYNCHRONOUS attribute in GCC is 4061 compatible with the candidate draft of TS 29113 (since GCC 4062 4.6). 4063 4064 Go 4065 4066 * GCC 4.7 implements the [35]Go 1 language standard. The library 4067 support in 4.7.0 is not quite complete, due to release timing. 4068 Release 4.7.1 includes complete support for Go 1. The Go library is 4069 from the Go 1.0.1 release. 4070 * Go has been tested on GNU/Linux and Solaris platforms. It may work 4071 on other platforms as well. 4072 4073 New Targets and Target Specific Improvements 4074 4075 ARM 4076 4077 * GCC now supports the Cortex-A7 processor implementing the v7-a 4078 version of the architecture using the option -mcpu=cortex-a7. 4079 * The default vector size in auto-vectorization for NEON is now 128 4080 bits. If vectorization fails thusly, the vectorizer tries again 4081 with 64-bit vectors. 4082 * A new option -mvectorize-with-neon-double was added to allow users 4083 to change the vector size to 64 bits. 4084 4085 AVR 4086 4087 * GCC now supports the XMEGA architecture. This requires GNU binutils 4088 2.22 or later. 4089 * Support for the [36]named address spaces __flash, __flash1, , 4090 __flash5 and __memx has been added. These address spaces locate 4091 read-only data in flash memory and allow reading from flash memory 4092 by means of ordinary C code, i.e. without the need of (inline) 4093 assembler code: 4094 4095 const __flash int values[] = { 42, 31 }; 4096 4097 int add_values (const __flash int *p, int i) 4098 { 4099 return values[i] + *p; 4100 } 4101 4102 * Support has been added for the AVR-specific configure option 4103 --with-avrlibc=yes in order to arrange for better integration of 4104 [37]AVR-Libc. This configure option is supported in avr-gcc 4.7.2 4105 and newer and will only take effect in non-RTEMS configurations. If 4106 avr-gcc is configured for RTEMS, the option will be ignored which 4107 is the same as specifying --with-avrlibc=no. See [38]PR54461 for 4108 more technical details. 4109 * Support for AVR-specific [39]built-in functions has been added. 4110 * Support has been added for the signed and unsigned 24-bit scalar 4111 integer types __int24 and __uint24. 4112 * New command-line options -maccumulate-args, -mbranch-cost=cost and 4113 -mstrict-X were added to allow better fine-tuning of code 4114 optimization. 4115 * The command-line option -fdata-sections now also takes affect on 4116 the section names of variables with the progmem attribute. 4117 * A new inline assembler print modifier %i to print a RAM address as 4118 I/O address has been added: 4119 4120 #include <avr/io.h> /* Port Definitions from AVR-LibC */ 4121 4122 void set_portb (uint8_t value) 4123 { 4124 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 4125 } 4126 4127 The offset between an I/O address and the RAM address for that I/O 4128 location is device-specific. This offset is taken into account when 4129 printing a RAM address with the %i modifier so that the address is 4130 suitable to be used as operand in an I/O command. The address must 4131 be a constant integer known at compile time. 4132 * The inline assembler constraint "R" to represent integers in the 4133 range 6 5 has been removed without replacement. 4134 * Many optimizations to: 4135 + 64-bit integer arithmetic 4136 + Widening multiplication 4137 + Integer division by a constant 4138 + Avoid constant reloading in multi-byte instructions. 4139 + Micro-optimizations for special instruction sequences. 4140 + Generic built-in functions like __builtin_ffs*, 4141 __builtin_clz*, etc. 4142 + If-else decision trees generated by switch instructions 4143 + Merging of data located in flash memory 4144 + New libgcc variants for devices with 8-bit wide stack pointer 4145 + 4146 * Better documentation: 4147 + Handling of EIND and indirect jumps on devices with more than 4148 128 KiB of program memory. 4149 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 4150 registers. 4151 + Function attributes OS_main and OS_task. 4152 + AVR-specific built-in macros. 4153 4154 C6X 4155 4156 * Support has been added for the Texas Instruments C6X family of 4157 processors. 4158 4159 CR16 4160 4161 * Support has been added for National Semiconductor's CR16 4162 architecture. 4163 4164 Epiphany 4165 4166 * Support has been added for Adapteva's Epiphany architecture. 4167 4168 IA-32/x86-64 4169 4170 * Support for Intel AVX2 intrinsics, built-in functions and code 4171 generation is available via -mavx2. 4172 * Support for Intel BMI2 intrinsics, built-in functions and code 4173 generation is available via -mbmi2. 4174 * Implementation and automatic generation of __builtin_clz* using the 4175 lzcnt instruction is available via -mlzcnt. 4176 * Support for Intel FMA3 intrinsics and code generation is available 4177 via -mfma. 4178 * A new -mfsgsbase command-line option is available that makes GCC 4179 generate new segment register read/write instructions through 4180 dedicated built-ins. 4181 * Support for the new Intel rdrnd instruction is available via 4182 -mrdrnd. 4183 * Two additional AVX vector conversion instructions are available via 4184 -mf16c. 4185 * Support for new Intel processor codename IvyBridge with RDRND, 4186 FSGSBASE and F16C is available through -march=core-avx-i. 4187 * Support for the new Intel processor codename Haswell with AVX2, 4188 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 4189 * Support for new AMD family 15h processors (Piledriver core) is now 4190 available through -march=bdver2 and -mtune=bdver2 options. 4191 * Support for [40]the x32 psABI is now available through the -mx32 4192 option. 4193 * Windows mingw targets are using the -mms-bitfields option by 4194 default. 4195 * Windows x86 targets are using the __thiscall calling convention for 4196 C++ class-member functions. 4197 * Support for the configure option --with-threads=posix for Windows 4198 mingw targets. 4199 4200 MIPS 4201 4202 * GCC now supports thread-local storage (TLS) for MIPS16. This 4203 requires GNU binutils 2.22 or later. 4204 * GCC can now generate code specifically for the Cavium Octeon+ and 4205 Octeon2 processors. The associated command-line options are 4206 -march=octeon+ and -march=octeon2 respectively. Both options 4207 require GNU binutils 2.22 or later. 4208 * GCC can now work around certain 24k errata, under the control of 4209 the command-line option -mfix-24k. These workarounds require GNU 4210 binutils 2.20 or later. 4211 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 4212 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 4213 toolchain that generates 32-bit code by default. Use the 4214 configure-time option --enable-targets=all to select these extra 4215 multilibs. 4216 * Passing -fno-delayed-branch now also stops the assembler from 4217 automatically filling delay slots. 4218 4219 PowerPC/PowerPC64 4220 4221 * Vectors of type vector long long or vector long are passed and 4222 returned using the same method as other vectors with the VSX 4223 instruction set. Previously GCC did not adhere to the ABI for 4224 128-bit vectors with 64-bit integer base types (PR 48857). This 4225 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 4226 * A new option -mno-pointers-to-nested-functions was added to allow 4227 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 4228 that the compiler should not load up the chain register (r11) 4229 before calling a function through a pointer. If you use this 4230 option, you cannot call nested functions through a pointer, or call 4231 other languages that might use the static chain. 4232 * A new option msave-toc-indirect was added to allow AIX 4233 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 4234 save the TOC in the prologue for indirect calls or generate the 4235 save inline. This can speed up some programs that call through a 4236 function pointer a lot, but it can slow down other functions that 4237 only call through a function pointer in exceptional cases. 4238 * The PowerPC port will now enable machine-specific built-in 4239 functions when the user switches the target machine using the 4240 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 4241 sequences. In addition, the target macros are updated. However, due 4242 to the way the -save-temps switch is implemented, you won't see the 4243 effect of these additional macros being defined in preprocessor 4244 output. 4245 4246 SH 4247 4248 * A new option -msoft-atomic has been added. When it is specified, 4249 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 4250 the new __atomic routines. 4251 * Since it is neither supported by GAS nor officially documented, 4252 code generation for little endian SH2A has been disabled. 4253 Specifying -ml with -m2a* will now result in a compiler error. 4254 * The defunct -mbranch-cost option has been fixed. 4255 * Some improvements to the generated code of: 4256 + Utilization of the tst #imm,R0 instruction. 4257 + Dynamic shift instructions on SH2A. 4258 + Integer absolute value calculations. 4259 * The -mdiv= option for targets other than SHmedia has been fixed and 4260 documented. 4261 4262 SPARC 4263 4264 * The option -mflat has been reinstated. When it is specified, the 4265 compiler will generate code for a single register window model. 4266 This is essentially a new implementation and the corresponding 4267 debugger support has been added to GDB 7.4. 4268 * Support for the options -mtune=native and -mcpu=native has been 4269 added on selected native platforms (GNU/Linux and Solaris). 4270 * Support for the SPARC T3 (Niagara 3) processor has been added. 4271 * VIS: 4272 + An intrinsics header visintrin.h has been added. 4273 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 4274 compare instructions have been added. 4275 + The little-endian version of alignaddr is now supported. 4276 + When possible, VIS builtins are marked const, which should 4277 increase the compiler's ability to optimize VIS operations. 4278 + The compiler now properly tracks the %gsr register and how it 4279 behaves as an input for various VIS instructions. 4280 + Akin to fzero, the compiler can now generate fone instructions 4281 in order to set all of the bits of a floating-point register 4282 to 1. 4283 + The documentation for the VIS intrinsics in the GCC manual has 4284 been brought up to date and many inaccuracies were fixed. 4285 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 4286 non-condition-code setting edge instructions have been added. 4287 Their availability is controlled by the new -mvis2 and 4288 -mno-vis2 options. They are enabled by default on 4289 UltraSPARC-III and later CPUs. 4290 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 4291 has been added. These instructions are enabled by default on SPARC 4292 T3 (Niagara 3) and later CPUs. 4293 4294 TILE-Gx/TILEPro 4295 4296 * Support has been added for the Tilera TILE-Gx and TILEPro families 4297 of processors. 4298 4299 Other significant improvements 4300 4301 * A new option (-grecord-gcc-switches) was added that appends 4302 compiler command-line options that might affect code generation to 4303 the DW_AT_producer attribute string in the DWARF debugging 4304 information. 4305 * GCC now supports various new GNU extensions to the DWARF debugging 4306 information format, like [41]entry value and [42]call site 4307 information, [43]typed DWARF stack or [44]a more compact macro 4308 representation. Support for these extensions has been added to GDB 4309 7.4. They can be disabled through the -gstrict-dwarf command-line 4310 option. 4311 4312 GCC 4.7.1 4313 4314 This is the [45]list of problem reports (PRs) from GCC's bug tracking 4315 system that are known to be fixed in the 4.7.1 release. This list might 4316 not be complete (that is, it is possible that some PRs that have been 4317 fixed are not listed here). 4318 4319 The Go front end in the 4.7.1 release fully supports the [46]Go 1 4320 language standard. 4321 4322 GCC 4.7.2 4323 4324 This is the [47]list of problem reports (PRs) from GCC's bug tracking 4325 system that are known to be fixed in the 4.7.2 release. This list might 4326 not be complete (that is, it is possible that some PRs that have been 4327 fixed are not listed here). 4328 4329 GCC 4.7.3 4330 4331 This is the [48]list of problem reports (PRs) from GCC's bug tracking 4332 system that are known to be fixed in the 4.7.3 release. This list might 4333 not be complete (that is, it is possible that some PRs that have been 4334 fixed are not listed here). 4335 4336 GCC 4.7.4 4337 4338 This is the [49]list of problem reports (PRs) from GCC's bug tracking 4339 system that are known to be fixed in the 4.7.4 release. This list might 4340 not be complete (that is, it is possible that some PRs that have been 4341 fixed are not listed here). 4342 4343 4344 For questions related to the use of GCC, please consult these web 4345 pages and the [50]GCC manuals. If that fails, the 4346 [51]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4347 web pages and the development of GCC are welcome on our developer 4348 list at [52]gcc (a] gcc.gnu.org. All of [53]our lists have public 4349 archives. 4350 4351 Copyright (C) [54]Free Software Foundation, Inc. Verbatim copying and 4352 distribution of this entire article is permitted in any medium, 4353 provided this notice is preserved. 4354 4355 These pages are [55]maintained by the GCC team. Last modified 4356 2018-09-30[56]. 4357 4358 References 4359 4360 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 4361 2. http://savannah.nongnu.org/bugs/?35407 4362 3. https://gcc.gnu.org/PR18145 4363 4. https://gcc.gnu.org/gcc-4.7/porting_to.html 4364 5. http://www.openmp.org/specifications/ 4365 6. https://gcc.gnu.org/wiki/TransactionalMemory 4366 7. https://gcc.gnu.org/wiki/Atomic/GCCMM 4367 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4368 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4369 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4370 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4371 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4372 13. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 4373 14. https://gcc.gnu.org/PR14258 4374 15. https://gcc.gnu.org/PR35688 4375 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 4376 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 4377 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 4378 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 4379 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 4380 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 4381 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 4382 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 4383 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 4384 25. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 4385 26. https://docs.microsoft.com/en-us/windows/desktop/wer/collecting-user-mode-dumps 4386 27. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 4387 28. https://gcc.gnu.org/wiki/Fortran2003Status 4388 29. https://gcc.gnu.org/wiki/OOP 4389 30. https://gcc.gnu.org/wiki/Fortran2008Status 4390 31. https://gcc.gnu.org/wiki/Coarray 4391 32. https://gcc.gnu.org/wiki/CoarrayLib 4392 33. https://gcc.gnu.org/wiki/TS29113Status 4393 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 4394 35. https://golang.org/doc/go1 4395 36. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 4396 37. http://nongnu.org/avr-libc/ 4397 38. https://gcc.gnu.org/PR54461 4398 39. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built_002din-Functions.html 4399 40. https://sites.google.com/site/x32abi/ 4400 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1 4401 42. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2 4402 43. http://www.dwarfstd.org/ShowIssue.php?issue=140425.1 4403 44. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1 4404 45. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 4405 46. https://golang.org/doc/go1 4406 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 4407 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 4408 49. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 4409 50. https://gcc.gnu.org/onlinedocs/ 4410 51. mailto:gcc-help (a] gcc.gnu.org 4411 52. mailto:gcc (a] gcc.gnu.org 4412 53. https://gcc.gnu.org/lists.html 4413 54. https://www.fsf.org/ 4414 55. https://gcc.gnu.org/about.html 4415 56. http://validator.w3.org/check/referer 4416 ====================================================================== 4417 http://gcc.gnu.org/gcc-4.6/index.html 4418 GCC 4.6 Release Series 4419 4420 April 12, 2013 4421 4422 The [1]GNU project and the GCC developers are pleased to announce the 4423 release of GCC 4.6.4. 4424 4425 This release is a bug-fix release, containing fixes for regressions in 4426 GCC 4.6.3 relative to previous releases of GCC. 4427 4428 This release series is no longer maintained. 4429 4430 Release History 4431 4432 GCC 4.6.4 4433 April 12, 2013 ([2]changes, [3]documentation) 4434 4435 GCC 4.6.3 4436 March 1, 2012 ([4]changes, [5]documentation) 4437 4438 GCC 4.6.2 4439 October 26, 2011 ([6]changes, [7]documentation) 4440 4441 GCC 4.6.1 4442 June 27, 2011 ([8]changes, [9]documentation) 4443 4444 GCC 4.6.0 4445 March 25, 2011 ([10]changes, [11]documentation) 4446 4447 References and Acknowledgements 4448 4449 GCC used to stand for the GNU C Compiler, but since the compiler 4450 supports several other languages aside from C, it now stands for the 4451 GNU Compiler Collection. 4452 4453 A list of [12]successful builds is updated as new information becomes 4454 available. 4455 4456 The GCC developers would like to thank the numerous people that have 4457 contributed new features, improvements, bug fixes, and other changes as 4458 well as test results to GCC. This [13]amazing group of volunteers is 4459 what makes GCC successful. 4460 4461 For additional information about GCC please refer to the [14]GCC 4462 project web site or contact the [15]GCC development mailing list. 4463 4464 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 4465 4466 4467 For questions related to the use of GCC, please consult these web 4468 pages and the [18]GCC manuals. If that fails, the 4469 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4470 web pages and the development of GCC are welcome on our developer 4471 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 4472 archives. 4473 4474 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 4475 distribution of this entire article is permitted in any medium, 4476 provided this notice is preserved. 4477 4478 These pages are [23]maintained by the GCC team. Last modified 4479 2018-09-30[24]. 4480 4481 References 4482 4483 1. http://www.gnu.org/ 4484 2. http://gcc.gnu.org/gcc-4.6/changes.html 4485 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 4486 4. http://gcc.gnu.org/gcc-4.6/changes.html 4487 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 4488 6. http://gcc.gnu.org/gcc-4.6/changes.html 4489 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 4490 8. http://gcc.gnu.org/gcc-4.6/changes.html 4491 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 4492 10. http://gcc.gnu.org/gcc-4.6/changes.html 4493 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 4494 12. http://gcc.gnu.org/gcc-4.6/buildstat.html 4495 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 4496 14. http://gcc.gnu.org/index.html 4497 15. mailto:gcc (a] gcc.gnu.org 4498 16. http://gcc.gnu.org/mirrors.html 4499 17. http://gcc.gnu.org/svn.html 4500 18. https://gcc.gnu.org/onlinedocs/ 4501 19. mailto:gcc-help (a] gcc.gnu.org 4502 20. mailto:gcc (a] gcc.gnu.org 4503 21. https://gcc.gnu.org/lists.html 4504 22. https://www.fsf.org/ 4505 23. https://gcc.gnu.org/about.html 4506 24. http://validator.w3.org/check/referer 4507 ====================================================================== 4508 http://gcc.gnu.org/gcc-4.6/changes.html 4509 GCC 4.6 Release Series 4510 Changes, New Features, and Fixes 4511 4512 Caveats 4513 4514 * The options -b <machine> and -V <version> have been removed because 4515 they were unreliable. Instead, users should directly run 4516 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 4517 run a different version of gcc. 4518 * GCC now has stricter checks for invalid command-line options. In 4519 particular, when gcc was called to link object files rather than 4520 compile source code, it would previously accept and ignore all 4521 options starting with --, including linker options such as 4522 --as-needed and --export-dynamic, although such options would 4523 result in errors if any source code was compiled. Such options, if 4524 unknown to the compiler, are now rejected in all cases; if the 4525 intent was to pass them to the linker, options such as 4526 -Wl,--as-needed should be used. 4527 * Versions of the GNU C library up to and including 2.11.1 included 4528 an [1]incorrect implementation of the cproj function. GCC optimizes 4529 its builtin cproj according to the behavior specified and allowed 4530 by the ISO C99 standard. If you want to avoid discrepancies between 4531 the C library and GCC's builtin transformations when using cproj in 4532 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 4533 and actually rely on the incorrect behavior of cproj, then you can 4534 disable GCC's transformations using -fno-builtin-cproj. 4535 * The C-only intermodule optimization framework (IMA, enabled by 4536 -combine) has been removed in favor of the new generic link-time 4537 optimization framework (LTO) introduced in [2]GCC 4.5.0. 4538 * GCC now ships with the LGPL-licensed libquadmath library, which 4539 provides quad-precision mathematical functions for targets with a 4540 __float128 datatype. __float128 is available for targets on 32-bit 4541 x86, x86-64 and Itanium architectures. The libquadmath library is 4542 automatically built on such targets when building the Fortran 4543 compiler. 4544 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 4545 warnings were added for C, C++, Objective-C and Objective-C++. 4546 These warnings diagnose variables respective parameters which are 4547 only set in the code and never otherwise used. Usually such 4548 variables are useless and often even the value assigned to them is 4549 computed needlessly, sometimes expensively. The 4550 -Wunused-but-set-variable warning is enabled by default by -Wall 4551 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 4552 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 4553 rules for the layout of vectors that could lead to wrong code being 4554 generated. Vectors larger than 8 bytes in size are now by default 4555 aligned to an 8-byte boundary. This is an ABI change: code that 4556 makes explicit use of vector types may be incompatible with binary 4557 objects built with older versions of GCC. Auto-vectorized code is 4558 not affected by this change. (This change affects GCC versions 4559 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 4560 * On AVR, variables with the progmem attribute to locate data in 4561 flash memory must be qualified as const. 4562 * Support for a number of older systems and recently unmaintained or 4563 untested target ports of GCC has been declared obsolete in GCC 4.6. 4564 Unless there is activity to revive them, the next release of GCC 4565 will have their sources permanently removed. 4566 All GCC ports for the following processor architectures have been 4567 declared obsolete: 4568 + Argonaut ARC (arc-*) 4569 + National Semiconductor CRX (crx-*) 4570 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 4571 m68hc12-*-*, m6812-*-*) 4572 + Sunplus S+core (score-*) 4573 The following ports for individual systems on particular 4574 architectures have been obsoleted: 4575 + Interix (i[34567]86-*-interix3*) 4576 + NetWare x86 (i[3456x]86-*-netware*) 4577 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 4578 + MCore PE (mcore-*-pe*) 4579 + SH SymbianOS (sh*-*-symbianelf*) 4580 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 4581 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 4582 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 4583 vax-*-netbsd*, but not *-*-netbsdelf*) 4584 The i[34567]86-*-pe alias for Cygwin targets has also been 4585 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 4586 Certain configure options to control the set of libraries built 4587 with GCC on some targets have been obsoleted. On ARM targets, the 4588 options --disable-fpu, --disable-26bit, --disable-underscore, 4589 --disable-interwork, --disable-biendian and --disable-nofmult have 4590 been obsoleted. On MIPS targets, the options 4591 --disable-single-float, --disable-biendian and --disable-softfloat 4592 have been obsoleted. 4593 * Support has been removed for all the [3]configurations obsoleted in 4594 GCC 4.5. 4595 * More information on porting to GCC 4.6 from previous versions of 4596 GCC can be found in the [4]porting guide for this release. 4597 4598 General Optimizer Improvements 4599 4600 * A new general optimization level, -Ofast, has been introduced. It 4601 combines the existing optimization level -O3 with options that can 4602 affect standards compliance but result in better optimized code. 4603 For example, -Ofast enables -ffast-math. 4604 * Link-time optimization improvements: 4605 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 4606 stabilized to the point of being usable. It has become the 4607 default mode when using the LTO optimization model. Link time 4608 optimization can now split itself into multiple parallel 4609 compilations. Parallelism is controlled with -flto=n (where n 4610 specifies the number of compilations to execute in parallel). 4611 GCC can also cooperate with a GNU make job server by 4612 specifying the -flto=jobserver option and adding + to the 4613 beginning of the Makefile rule executing the linker. 4614 Classical LTO mode can be enforced by -flto-partition=none. 4615 This may result in small code quality improvements. 4616 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 4617 and other large applications can be built with LTO enabled. 4618 + The linker plugin support improvements 4619 o Linker plugin is now enabled by default when the linker 4620 is detected to have plugin support. This is the case for 4621 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 4622 the Gold linker on ELF targets. Plugin support of the 4623 Apple linker on Darwin is not compatible with GCC. The 4624 linker plugin can also be controlled by the 4625 -fuse-linker-plugin command-line option. 4626 o Resolution information from the linker plugin is used to 4627 drive whole program assumptions. Use of the linker plugin 4628 results in more aggressive optimization on binaries and 4629 on shared libraries that use the hidden visibility 4630 attribute. Consequently the use of -fwhole-program is not 4631 necessary in addition to LTO. 4632 + Hidden symbols used from non-LTO objects now have to be 4633 explicitly annotated with externally_visible when the linker 4634 plugin is not used. 4635 + C++ inline functions and virtual tables are now privatized 4636 more aggressively, leading to better inter-procedural 4637 optimization and faster dynamic linking. 4638 + Memory usage and intermediate language streaming performance 4639 have been improved. 4640 + Static constructors and destructors from individual units are 4641 inlined into a single function. This can significantly improve 4642 startup times of large C++ applications where static 4643 constructors are very common. For example, static constructors 4644 are used when including the iostream header. 4645 + Support for the Ada language has been added. 4646 * Interprocedural optimization improvements 4647 + The interprocedural framework was re-tuned for link time 4648 optimization. Several scalability issues were resolved. 4649 + Improved auto-detection of const and pure functions. Newly, 4650 noreturn functions are auto-detected. 4651 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 4652 available that informs users when adding attributes to headers 4653 might improve code generation. 4654 + A number of inlining heuristic improvements. In particular: 4655 o Partial inlining is now supported and enabled by default 4656 at -O2 and greater. The feature can be controlled via 4657 -fpartial-inlining. 4658 Partial inlining splits functions with short hot path to 4659 return. This allows more aggressive inlining of the hot 4660 path leading to better performance and often to code size 4661 reductions (because cold parts of functions are not 4662 duplicated). 4663 o Scalability for large compilation units was improved 4664 significantly. 4665 o Inlining of callbacks is now more aggressive. 4666 o Virtual methods are considered for inlining when the 4667 caller is inlined and devirtualization is then possible. 4668 o Inlining when optimizing for size (either in cold regions 4669 of a program or when compiling with -Os) was improved to 4670 better handle C++ programs with larger abstraction 4671 penalty, leading to smaller and faster code. 4672 + The IPA reference optimization pass detecting global variables 4673 used or modified by functions was strengthened and sped up. 4674 + Functions whose address was taken are now optimized out when 4675 all references to them are dead. 4676 + A new inter-procedural static profile estimation pass detects 4677 functions that are executed once or unlikely to be executed. 4678 Unlikely executed functions are optimized for size. Functions 4679 executed once are optimized for size except for the inner 4680 loops. 4681 + On most targets with named section support, functions used 4682 only at startup (static constructors and main), functions used 4683 only at exit and functions detected to be cold are placed into 4684 separate text segment subsections. This extends the 4685 -freorder-functions feature and is controlled by the same 4686 switch. The goal is to improve the startup time of large C++ 4687 programs. 4688 Proper function placement requires linker support. GNU ld 4689 2.21.51 on ELF targets was updated to place those functions 4690 together within the text section leading to better code 4691 locality and faster startup times of large C++ programs. The 4692 feature is also supported in the Apple linker. Support in the 4693 gold linker is planned. 4694 * A new switch -fstack-usage has been added. It makes the compiler 4695 output stack usage information for the program, on a per-function 4696 basis, in an auxiliary file. 4697 * A new switch -fcombine-stack-adjustments has been added. It can be 4698 used to enable or disable the compiler's stack-slot combining pass 4699 which before was enabled automatically at -O1 and above, but could 4700 not be controlled on its own. 4701 * A new switch -fstrict-volatile-bitfields has been added. Using it 4702 indicates that accesses to volatile bitfields should use a single 4703 access of the width of the field's type. This option can be useful 4704 for precisely defining and accessing memory-mapped peripheral 4705 registers from C or C++. 4706 4707 Compile time and memory usage improvements 4708 4709 * Datastructures used by the dataflow framework in GCC were 4710 reorganized for better memory usage and more cache locality. 4711 Compile time is improved especially on units with large functions 4712 (possibly resulting from a lot of inlining) not fitting into the 4713 processor cache. The compile time of the GCC C compiler binary with 4714 link-time optimization went down by over 10% (benchmarked on x86-64 4715 target). 4716 4717 New Languages and Language specific improvements 4718 4719 Ada 4720 4721 * Stack checking has been improved on selected architectures (Alpha, 4722 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 4723 overflows in all cases on these architectures. 4724 * Initial support for Ada 2012 has been added. 4725 4726 C family 4727 4728 * A new warning, enabled by -Wdouble-promotion, has been added that 4729 warns about cases where a value of type float is implicitly 4730 promoted to double. This is especially helpful for CPUs that handle 4731 the former in hardware, but emulate the latter in software. 4732 * A new function attribute leaf was introduced. This attribute allows 4733 better inter-procedural optimization across calls to functions that 4734 return to the current unit only via returning or exception 4735 handling. This is the case for most library functions that have no 4736 callbacks. 4737 * Support for a new data type __int128 for targets having wide enough 4738 machine-mode support. 4739 * The new function attribute callee_pop_aggregate allows to specify 4740 if the caller or callee is responsible for popping the aggregate 4741 return pointer value from the stack. 4742 * Support for selectively enabling and disabling warnings via #pragma 4743 GCC diagnostic has been added. For instance: 4744 #pragma GCC diagnostic error "-Wuninitialized" 4745 foo(a); /* error is given for this one */ 4746 #pragma GCC diagnostic push 4747 #pragma GCC diagnostic ignored "-Wuninitialized" 4748 foo(b); /* no diagnostic for this one */ 4749 #pragma GCC diagnostic pop 4750 foo(c); /* error is given for this one */ 4751 #pragma GCC diagnostic pop 4752 foo(d); /* depends on command-line options */ 4753 4754 * The -fmax-errors=N option is now supported. Using this option 4755 causes the compiler to exit after N errors have been issued. 4756 4757 C 4758 4759 * There is now experimental support for some features from the 4760 upcoming C1X revision of the ISO C standard. This support may be 4761 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 4762 Note that this support is experimental and may change incompatibly 4763 in future releases for consistency with changes to the C1X standard 4764 draft. The following features are newly supported as described in 4765 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 4766 meeting); some other features were already supported with no 4767 compiler changes being needed, or have some support but not in full 4768 accord with N1539 (as amended). 4769 + Static assertions (_Static_assert keyword) 4770 + Typedef redefinition 4771 + New macros in <float.h> 4772 + Anonymous structures and unions 4773 * The new -fplan9-extensions option directs the compiler to support 4774 some extensions for anonymous struct fields which are implemented 4775 by the Plan 9 compiler. A pointer to a struct may be automatically 4776 converted to a pointer to an anonymous field when calling a 4777 function, in order to make the types match. An anonymous struct 4778 field whose type is a typedef name may be referred to using the 4779 typedef name. 4780 4781 C++ 4782 4783 * Improved [7]experimental support for the upcoming C++0x ISO C++ 4784 standard, including support for constexpr (thanks to Gabriel Dos 4785 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 4786 noexcept, unrestricted unions, range-based for loops (thanks to 4787 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 4788 Rodrigo), implicitly deleted functions and implicit move 4789 constructors. 4790 * When an extern declaration within a function does not match a 4791 declaration in the enclosing context, G++ now properly declares the 4792 name within the namespace of the function rather than the namespace 4793 which was open just before the function definition ([8]c++/43145). 4794 * GCC now warns by default when casting integers to larger pointer 4795 types. These warnings can be disabled with the option 4796 -Wno-int-to-pointer-cast, which is now also available in C++. 4797 * G++ no longer optimizes using the assumption that a value of 4798 enumeration type will fall within the range specified by the 4799 standard, since that assumption is easily violated with a 4800 conversion from integer type ([9]c++/43680). The old behavior can 4801 be restored with -fstrict-enums. 4802 * The new -fnothrow-opt flag changes the semantics of a throw() 4803 exception specification to match the proposed semantics of the 4804 noexcept specification: just call terminate if an exception tries 4805 to propagate out of a function with such an exception 4806 specification. This dramatically reduces or eliminates the code 4807 size overhead from adding the exception specification. 4808 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 4809 a function that the compiler can tell doesn't throw if it would 4810 change the value of a noexcept expression. 4811 * The -Wshadow option now warns if a local variable or type 4812 declaration shadows another type in C++. Note that the compiler 4813 will not warn if a local variable shadows a struct/class/enum, but 4814 will warn if it shadows an explicit typedef. 4815 * When an identifier is not found in the current scope, G++ now 4816 offers suggestions about which identifier might have been intended. 4817 * G++ now issues clearer diagnostics for missing semicolons after 4818 class, struct, and union definitions. 4819 * G++ now issues clearer diagnostics for missing semicolons after 4820 class member declarations. 4821 * G++ now issues clearer diagnostics when a colon is used in a place 4822 where a double-colon was intended. 4823 * G++ no longer accepts mutable on reference members ([10]c++/33558). 4824 Use -fpermissive to allow the old, non-conforming behaviour. 4825 * A few mangling fixes have been made, to attribute const/volatile on 4826 function pointer types, decltype of a plain decl, and use of a 4827 function parameter in the declaration of another parameter. By 4828 default the compiler still uses the old mangling, but emits aliases 4829 with the new mangling on targets that support strong aliases. Users 4830 can switch over entirely to the new mangling with -fabi-version=5 4831 or -fabi-version=0. -Wabi will now warn about code that uses the 4832 old mangling. 4833 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 4834 type to be default initialized unless the type has a user-declared 4835 default constructor. In 4.6.2 G++ implements the proposed 4836 resolution of [11]DR 253, so default initialization is allowed if 4837 it initializes all subobjects. Code that fails to compile can be 4838 fixed by providing an initializer e.g. 4839 struct A { A(); }; 4840 struct B : A { int i; }; 4841 const B b = B(); 4842 Use -fpermissive to allow the old, non-conforming behaviour. 4843 4844 Runtime Library (libstdc++) 4845 4846 * [12]Improved experimental support for the upcoming ISO C++ 4847 standard, C++0x, including using constexpr and nullptr. 4848 * Performance improvements to the [13]Debug Mode, thanks to Franois 4849 Dumont. 4850 * Atomic operations used for reference-counting are annotated so that 4851 they can be understood by race detectors such as Helgrind, see 4852 [14]Data Race Hunting. 4853 * Most libstdc++ standard headers have been changed to no longer 4854 include the cstddef header as an implementation detail. Code that 4855 relied on that header being included as side-effect of including 4856 other standard headers will need to include cstddef explicitly. 4857 4858 Fortran 4859 4860 * On systems supporting the libquadmath library, GNU Fortran now also 4861 supports a quad-precision, kind=16 floating-point data type 4862 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 4863 hardware, calculations might be one to two orders of magnitude 4864 slower than with the 4, 8 or 10 bytes floating-point data types. 4865 This change does not affect systems which support REAL(16) in 4866 hardware nor those which do not support libquadmath. 4867 * Much improved compile time for large array constructors. 4868 * In order to reduce execution time and memory consumption, use of 4869 temporary arrays in assignment expressions is avoided for many 4870 cases. The compiler now reverses loops in order to avoid generating 4871 a temporary array where possible. 4872 * Improved diagnostics, especially with -fwhole-file. 4873 * The -fwhole-file flag is now enabled by default. This improves code 4874 generation and diagnostics. It can be disabled using the deprecated 4875 -fno-whole-file flag. 4876 * Support the generation of Makefile dependencies via the [15]-M... 4877 flags of GCC; you may need to specify the -cpp option in addition. 4878 The dependencies take modules, Fortran's include, and CPP's 4879 #include into account. Note: Using -M for the module path is no 4880 longer supported, use -J instead. 4881 * The flag -Wconversion has been modified to only issue warnings 4882 where a conversion leads to information loss. This drastically 4883 reduces the number of warnings; -Wconversion is thus now enabled 4884 with -Wall. The flag -Wconversion-extra has been added and also 4885 warns about other conversions; -Wconversion-extra typically issues 4886 a huge number of warnings, most of which can be ignored. 4887 * A new command-line option -Wunused-dummy-argument warns about 4888 unused dummy arguments and is included in -Wall. Before, 4889 -Wunused-variable also warned about unused dummy arguments. 4890 * Fortran 2003 support has been extended: 4891 + Improved support for polymorphism between libraries and 4892 programs and for complicated inheritance patterns (cf. 4893 [16]object-oriented programming). 4894 + Experimental support of the ASSOCIATE construct. 4895 + In pointer assignments it is now possible to specify the lower 4896 bounds of the pointer and, for a rank-1 or a simply contiguous 4897 data-target, to remap the bounds. 4898 + Automatic (re)allocation: In intrinsic assignments to 4899 allocatable variables the left-hand side will be automatically 4900 allocated (if unallocated) or reallocated (if the shape or 4901 type parameter is different). To avoid the small performance 4902 penalty, you can use a(:) = ... instead of a = ... for arrays 4903 and character strings or disable the feature using -std=f95 4904 or -fno-realloc-lhs. 4905 + Deferred type parameter: For scalar allocatable and pointer 4906 variables the character length can be deferred. 4907 + Namelist variables with allocatable and pointer attribute and 4908 nonconstant length type parameter are supported. 4909 * Fortran 2008 support has been extended: 4910 + Experimental [17]coarray support (for one image only, i.e. 4911 num_images() == 1); use the [18]-fcoarray=single flag to 4912 enable it. 4913 + The STOP and the new ERROR STOP statements now support all 4914 constant expressions. 4915 + Support for the CONTIGUOUS attribute. 4916 + Support for ALLOCATE with MOLD. 4917 + Support for the STORAGE_SIZE intrinsic inquiry function. 4918 + Support of the NORM2 and PARITY intrinsic functions. 4919 + The following bit intrinsics were added: POPCNT and POPPAR for 4920 counting the number of 1 bits and returning the parity; BGE, 4921 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 4922 for combined left and right shifts, MASKL and MASKR for simple 4923 left and right justified masks, MERGE_BITS for a bitwise merge 4924 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 4925 and the transformational bit intrinsics IALL, IANY and 4926 IPARITY. 4927 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 4928 + Support for the IMPURE attribute for procedures, which allows 4929 for ELEMENTAL procedures without the restrictions of PURE. 4930 + Null pointers (including NULL()) and not allocated variables 4931 can be used as actual argument to optional non-pointer, 4932 non-allocatable dummy arguments, denoting an absent argument. 4933 + Non-pointer variables with TARGET attribute can be used as 4934 actual argument to POINTER dummies with INTENT(IN) 4935 + Pointers including procedure pointers and those in a derived 4936 type (pointer components) can now be initialized by a target 4937 instead of only by NULL. 4938 + The EXIT statement (with construct-name) can now be used to 4939 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 4940 SELECT CASE and SELECT TYPE constructs. 4941 + Internal procedures can now be used as actual argument. 4942 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 4943 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 4944 have been added; these arrays contain the supported kind 4945 values for the respective types. 4946 + The module procedures C_SIZEOF of the intrinsic module 4947 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 4948 ISO_FORTRAN_ENV have been implemented. 4949 + Minor changes: obsolescence diagnostics for ENTRY was added 4950 for -std=f2008; a line may start with a semicolon; for 4951 internal and module procedures END can be used instead of END 4952 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 4953 a RADIX argument; intrinsic types are supported for 4954 TYPE(intrinsic-type-spec); multiple type-bound procedures can 4955 be declared in a single PROCEDURE statement; implied-shape 4956 arrays are supported for named constants (PARAMETER). The 4957 transformational, three argument versions of BESSEL_JN and 4958 BESSEL_YN were added the elemental, two-argument version had 4959 been added in GCC 4.4; note that the transformational 4960 functions use a recurrence algorithm. 4961 4962 Go 4963 4964 Support for the [19]Go programming language has been added to GCC. It 4965 is not enabled by default when you build GCC; use the 4966 --enable-languages configure option to build it. The driver program for 4967 compiling Go code is gccgo. 4968 4969 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 4970 is in progress. It may or may not work on other platforms. 4971 4972 Objective-C and Objective-C++ 4973 4974 * The -fobjc-exceptions flag is now required to enable Objective-C 4975 exception and synchronization syntax (introduced by the keywords 4976 @try, @catch, @finally and @synchronized). 4977 * A number of Objective-C 2.0 features and extensions are now 4978 supported by GCC. These features are enabled by default; you can 4979 disable them by using the new -fobjc-std=objc1 command-line option. 4980 * The Objective-C 2.0 dot-syntax is now supported. It is an 4981 alternative syntax for using getters and setters; object.count is 4982 automatically converted into [object count] or [object setCount: 4983 ...] depending on context; for example if (object.count > 0) is 4984 automatically compiled into the equivalent of if ([object count] > 4985 0) while object.count = 0; is automatically compiled into the 4986 equivalent ot [object setCount: 0];. The dot-syntax can be used 4987 with instance and class objects and with any setters or getters, no 4988 matter if they are part of a declared property or not. 4989 * Objective-C 2.0 declared properties are now supported. They are 4990 declared using the new @property keyword, and are most commonly 4991 used in conjunction with the new Objective-C 2.0 dot-syntax. The 4992 nonatomic, readonly, readwrite, assign, retain, copy, setter and 4993 getter attributes are all supported. Marking declared properties 4994 with __attribute__ ((deprecated)) is supported too. 4995 * The Objective-C 2.0 @synthesize and @dynamic keywords are 4996 supported. @synthesize causes the compiler to automatically 4997 synthesize a declared property, while @dynamic is used to disable 4998 all warnings for a declared property for which no implementation is 4999 provided at compile time. Synthesizing declared properties requires 5000 runtime support in most useful cases; to be able to use it with the 5001 GNU runtime, appropriate helper functions have been added to the 5002 GNU Objective-C runtime ABI, and are implemented by the GNU 5003 Objective-C runtime library shipped with GCC. 5004 * The Objective-C 2.0 fast enumeration syntax is supported in 5005 Objective-C. This is currently not yet available in Objective-C++. 5006 Fast enumeration requires support in the runtime, and such support 5007 has been added to the GNU Objective-C runtime library (shipped with 5008 GCC). 5009 * The Objective-C 2.0 @optional keyword is supported. It allows you 5010 to mark methods or properties in a protocol as optional as opposed 5011 to required. 5012 * The Objective-C 2.0 @package keyword is supported. It has currently 5013 the same effect as the @public keyword. 5014 * Objective-C 2.0 method attributes are supported. Currently the 5015 supported attributes are deprecated, sentinel, noreturn and format. 5016 * Objective-C 2.0 method argument attributes are supported. The most 5017 widely used attribute is unused, to mark an argument as unused in 5018 the implementation. 5019 * Objective-C 2.0 class and protocol attributes are supported. 5020 Currently the only supported attribute is deprecated. 5021 * Objective-C 2.0 class extensions are supported. A class extension 5022 has the same syntax as a category declaration with no category 5023 name, and the methods and properties declared in it are added 5024 directly to the main class. It is mostly used as an alternative to 5025 a category to add methods to a class without advertising them in 5026 the public headers, with the advantage that for class extensions 5027 the compiler checks that all the privately declared methods are 5028 actually implemented. 5029 * As a result of these enhancements, GCC can now be used to build 5030 Objective-C and Objective-C++ software that uses Foundation and 5031 other important system frameworks with the NeXT runtime on Darwin 9 5032 and Darwin 10 (OSX 10.5 and 10.6). 5033 * Many bugs in the compiler have been fixed in this release; in 5034 particular, LTO can now be used when compiling Objective-C and 5035 Objective-C++ and the parser is much more robust in dealing with 5036 invalid code. 5037 5038 Runtime Library (libobjc) 5039 5040 * The GNU Objective-C runtime library now defines the macro 5041 __GNU_LIBOBJC__ (with a value that is increased at every release 5042 where there is any change to the API) in objc/objc.h, making it 5043 easy to determine if the GNU Objective-C runtime library is being 5044 used, and if so, which version. Previous versions of the GNU 5045 Objective-C runtime library (and other Objective-C runtime 5046 libraries such as the Apple one) do not define this macro. 5047 * A new Objective-C 2.0 API, almost identical to the one implemented 5048 by the Apple Objective-C runtime, has been implemented in the GNU 5049 Objective-C runtime library. The new API hides the internals of 5050 most runtime structures but provides a more extensive set of 5051 functions to operate on them. It is much easier, for example, to 5052 create or modify classes at runtime. The new API also makes it 5053 easier to port software from Apple to GNU as almost no changes 5054 should be required. The old API is still supported for backwards 5055 compatibility; including the old objc/objc-api.h header file 5056 automatically selects the old API, while including the new 5057 objc/runtime.h header file automatically selects the new API. 5058 Support for the old API is being phased out and upgrading the 5059 software to use the new API is strongly recommended. To check for 5060 the availability of the new API, the __GNU_LIBOBJC__ macro can be 5061 used as older versions of the GNU Objective-C runtime library, 5062 which do not support the new API, do not define such a macro. 5063 * Runtime support for @synchronized has been added. 5064 * Runtime support for Objective-C 2.0 synthesized property accessors 5065 has been added. 5066 * Runtime support for Objective-C 2.0 fast enumeration has been 5067 added. 5068 5069 New Targets and Target Specific Improvements 5070 5071 ARM 5072 5073 * GCC now supports the Cortex-M4 processor implementing the v7-em 5074 version of the architecture using the option -mcpu=cortex-m4. 5075 * Scheduling descriptions for the Cortex-M4, the Neon and the 5076 floating point units of the Cortex-A9 and a pipeline description 5077 for the Cortex-A5 have been added. 5078 * Synchronization primitives such as __sync_fetch_and_add and friends 5079 are now inlined for supported architectures rather than calling 5080 into a kernel helper function. 5081 * SSA loop prefetching is enabled by default for the Cortex-A9 at 5082 -O3. 5083 * Several improvements were committed to improve code generation for 5084 the ARM architecture including a rewritten implementation for load 5085 and store multiples. 5086 * Several enhancements were committed to improve SIMD code generation 5087 for NEON by adding support for widening instructions, misaligned 5088 loads and stores, vector conditionals and support for 64 bit 5089 arithmetic. 5090 * Support was added for the Faraday cores fa526, fa606te, fa626te, 5091 fmp626te, fmp626 and fa726te and can be used with the respective 5092 names as parameters to the -mcpu= option. 5093 * Basic support was added for Cortex-A15 and is available through 5094 -mcpu=cortex-a15. 5095 * GCC for AAPCS configurations now more closely adheres to the AAPCS 5096 specification by enabling -fstrict-volatile-bitfields by default. 5097 5098 IA-32/x86-64 5099 5100 * The new -fsplit-stack option permits programs to use a 5101 discontiguous stack. This is useful for threaded programs, in that 5102 it is no longer necessary to specify the maximum stack size when 5103 creating a thread. This feature is currently only implemented for 5104 32-bit and 64-bit x86 GNU/Linux targets. 5105 * Support for emitting profiler counter calls before function 5106 prologues. This is enabled via a new command-line option -mfentry. 5107 * Optimization for the Intel Core 2 processors is now available 5108 through the -march=core2 and -mtune=core2 options. 5109 * Support for Intel Core i3/i5/i7 processors is now available through 5110 the -march=corei7 and -mtune=corei7 options. 5111 * Support for Intel Core i3/i5/i7 processors with AVX is now 5112 available through the -march=corei7-avx and -mtune=corei7-avx 5113 options. 5114 * Support for AMD Bobcat (family 14) processors is now available 5115 through the -march=btver1 and -mtune=btver1 options. 5116 * Support for AMD Bulldozer (family 15) processors is now available 5117 through the -march=bdver1 and -mtune=bdver1 options. 5118 * The default setting (when not optimizing for size) for 32-bit 5119 GNU/Linux and Darwin x86 targets has been changed to 5120 -fomit-frame-pointer. The default can be reverted to 5121 -fno-omit-frame-pointer by configuring GCC with the 5122 --enable-frame-pointer configure option. 5123 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 5124 __float128 on 32-bit and 64-bit x86 targets. 5125 * AVX floating-point arithmetic can now be enabled by default at 5126 configure time with the new --with-fpmath=avx option. 5127 * The SSA loop prefetching pass is enabled when using -O3 when 5128 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 5129 than K6). 5130 * Support for TBM (Trailing Bit Manipulation) built-in functions and 5131 code generation is available via -mtbm. 5132 * Support for AMD's BMI (Bit Manipulation) built-in functions and 5133 code generation is available via -mbmi. 5134 5135 MicroBlaze 5136 5137 * Support has been added for the Xilinx MicroBlaze softcore processor 5138 (microblaze-elf) embedded target. This configurable processor is 5139 supported on several Xilinx Spartan and Virtex FPGAs. 5140 5141 MIPS 5142 5143 * GCC now supports the Loongson 3A processor. Its canonical -march= 5144 and -mtune= name is loongson3a. 5145 5146 MN10300 / AM33 5147 5148 * The inline assembly register constraint "A" has been renamed "c". 5149 This constraint is used to select a floating-point register that 5150 can be used as the destination of a multiply-accumulate 5151 instruction. 5152 * New inline assembly register constraints "A" and "D" have been 5153 added. These constraint letters resolve to all general registers 5154 when compiling for AM33, and resolve to address registers only or 5155 data registers only when compiling for MN10300. 5156 * The MDR register is represented in the compiler. One can access the 5157 register via the "z" constraint in inline assembly. It can be 5158 marked as clobbered or used as a local register variable via the 5159 "mdr" name. The compiler uses the RETF instruction if the function 5160 does not modify the MDR register, so it is important that inline 5161 assembly properly annotate any usage of the register. 5162 5163 PowerPC/PowerPC64 5164 5165 * GCC now supports the Applied Micro Titan processor with 5166 -mcpu=titan. 5167 * The -mrecip option has been added, which indicates whether the 5168 reciprocal and reciprocal square root instructions should be used. 5169 * The -mveclibabi=mass option can be used to enable the compiler to 5170 autovectorize mathematical functions using the Mathematical 5171 Acceleration Subsystem library. 5172 * The -msingle-pic-base option has been added, which instructs the 5173 compiler to avoid loading the PIC base register in function 5174 prologues. The PIC base register must be initialized by the runtime 5175 system. 5176 * The -mblock-move-inline-limit option has been added, which enables 5177 the user to control the maximum size of inlined memcpy calls and 5178 similar. 5179 * PowerPC64 GNU/Linux support for applications requiring a large TOC 5180 section has been improved. A new command-line option, 5181 -mcmodel=MODEL, controls this feature; valid values for MODEL are 5182 small, medium, or large. 5183 * The Altivec builtin functions vec_ld and vec_st have been modified 5184 to generate the Altivec memory instructions LVX and STVX, even if 5185 the -mvsx option is used. In the initial GCC 4.5 release, these 5186 builtin functions were changed to generate VSX memory reference 5187 instructions instead of Altivec memory instructions, but there are 5188 differences between the two instructions. If the VSX instruction 5189 set is available, you can now use the new builtin functions 5190 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 5191 instructions. 5192 * The GCC compiler on AIX now defaults to a process layout with a 5193 larger data space allowing larger programs to be compiled. 5194 * The GCC long double type on AIX 6.1 and above has reverted to 64 5195 bit double precision, matching the AIX XL compiler default, because 5196 of missing C99 symbols required by the GCC runtime. 5197 * The default processor scheduling model and tuning for PowerPC64 5198 GNU/Linux and for AIX 6.1 and above now is POWER7. 5199 * Starting with GCC 4.6.1, vectors of type vector long long or vector 5200 long are passed and returned in the same method as other vectors 5201 with the VSX instruction set. Previously the GCC compiler did not 5202 adhere to the ABI for 128-bit vectors with 64-bit integer base 5203 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 5204 5205 S/390, zSeries and System z9/z10, IBM zEnterprise z196 5206 5207 * Support for the zEnterprise z196 processor has been added. When 5208 using the -march=z196 option, the compiler will generate code 5209 making use of the following instruction facilities: 5210 + Conditional load/store 5211 + Distinct-operands 5212 + Floating-point-extension 5213 + Interlocked-access 5214 + Population-count 5215 The -mtune=z196 option avoids the compare and branch instructions 5216 as well as the load address instruction with an index register as 5217 much as possible and performs instruction scheduling appropriate 5218 for the new out-of-order pipeline architecture. 5219 * When using the -m31 -mzarch options the generated code still 5220 conforms to the 32-bit ABI but uses the general purpose registers 5221 as 64-bit registers internally. This requires a Linux kernel saving 5222 the whole 64-bit registers when doing a context switch. Kernels 5223 providing that feature indicate that by the 'highgprs' string in 5224 /proc/cpuinfo. 5225 * The SSA loop prefetching pass is enabled when using -O3. 5226 5227 SPARC 5228 5229 * GCC now supports the LEON series of SPARC V8 processors. The code 5230 generated by the compiler can either be tuned to it by means of the 5231 --with-tune=leon configure option and -mtune=leon compilation 5232 option, or the compiler can be built for the sparc-leon-{elf,linux} 5233 and sparc-leon3-{elf,linux} targets directly. 5234 * GCC has stopped sign/zero-extending parameter registers in the 5235 callee for functions taking parameters with sub-word size in 32-bit 5236 mode, since this is redundant with the specification of the ABI. 5237 GCC has never done so in 64-bit mode since this is also redundant. 5238 * The command-line option -mfix-at697f has been added to enable the 5239 documented workaround for the single erratum of the Atmel AT697F 5240 processor. 5241 5242 Operating Systems 5243 5244 Android 5245 5246 * GCC now supports the Bionic C library and provides a convenient way 5247 of building native libraries and applications for the Android 5248 platform. Refer to the documentation of the -mandroid and -mbionic 5249 options for details on building native code. At the moment, Android 5250 support is enabled only for ARM. 5251 5252 Darwin/Mac OS X 5253 5254 * General 5255 + Initial support for CFString types has been added. 5256 This allows GCC to build projects including the system Core 5257 Foundation frameworks. The GCC Objective-C family supports 5258 CFString "toll-free bridged" as per the Mac OS X system tools. 5259 CFString is also recognized in the context of format 5260 attributes and arguments (see the documentation for format 5261 attributes for limitations). At present, 8-bit character types 5262 are supported. 5263 + Object file size reduction. 5264 The Darwin zeroed memory allocators have been re-written to 5265 make more use of .zerofill sections. For non-debug code, this 5266 can reduce object file size significantly. 5267 + Objective-C family 64-bit support (NeXT ABI 2). 5268 Initial support has been added to support 64-bit Objective-C 5269 code using the Darwin/OS X native (NeXT) runtime. ABI version 5270 2 will be selected automatically when 64-bit code is built. 5271 + Objective-C family 32-bit ABI 1. 5272 For 32-bit code ABI 1 is also now also allowed. At present it 5273 must be selected manually using -fobjc-abi-version=1 where 5274 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 5275 * x86 Architecture 5276 + The -mdynamic-no-pic option has been enabled. 5277 Code supporting -mdynamic-no-pic optimization has been added 5278 and is applicable to -m32 builds. The compiler bootstrap uses 5279 the option where appropriate. 5280 + The default value for -mtune= has been changed. 5281 Since Darwin systems are primarily Xeon, Core-2 or similar the 5282 default tuning has been changed to -mtune=core2. 5283 + Enable 128-bit long double (__float128) support on Darwin. 5284 * PPC Architecture 5285 + Darwin64 ABI. 5286 Several significant bugs have been fixed, such that GCC now 5287 produces code compatible with the Darwin64 PowerPC ABI. 5288 + libffi and boehm-gc. 5289 The Darwin ports of the libffi and boehm-gc libraries have 5290 been upgraded to include a Darwin64 implementation. This means 5291 that powerpc*-*-darwin9 platforms may now, for example, build 5292 Java applications with -m64 enabled. 5293 + Plug-in support has been enabled. 5294 + The -fsection-anchors option is now available although, 5295 presently, not heavily tested. 5296 5297 Solaris 2 5298 5299 New Features 5300 5301 * Support symbol versioning with the Sun linker. 5302 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 5303 * Support thread-local storage (TLS) with the Sun assembler on 5304 Solaris 2/x86. 5305 * Support TLS on Solaris 8/9 if prerequisites are met. 5306 * Support COMDAT group with the GNU assembler and recent Sun linker. 5307 * Support the Sun assembler visibility syntax. 5308 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 5309 -march=pentiumpro (Solaris 8/9). 5310 * Don't use SSE on Solaris 8/9 x86 by default. 5311 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 5312 5313 ABI Change 5314 5315 * Change the ABI for returning 8-byte vectors like __m64 in MMX 5316 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 5317 compilers. This is an incompatible change. If you use such types, 5318 you must either recompile all your code with the new compiler or 5319 use the new -mvect8-ret-in-mem option to remain compatible with 5320 previous versions of GCC and Sun Studio. 5321 5322 Windows x86/x86_64 5323 5324 * Initial support for decimal floating point. 5325 * Support for the __thiscall calling-convention. 5326 * Support for hot-patchable function prologues via the 5327 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 5328 * Improvements of stack-probing and stack-allocation mechanisms. 5329 * Support of push/pop-macro pragma as preprocessor command. 5330 With #pragma push_macro("macro-name") the current definition of 5331 macro-name is saved and can be restored with #pragma 5332 pop_macro("macro-name") to its saved definition. 5333 * Enable 128-bit long double (__float128) support on MinGW and 5334 Cygwin. 5335 5336 Other significant improvements 5337 5338 Installation changes 5339 5340 * An install-strip make target is provided that installs stripped 5341 executables, and may install libraries with unneeded or debugging 5342 sections stripped. 5343 * On Power7 systems, there is a potential problem if you build the 5344 GCC compiler with a host compiler using options that enable the VSX 5345 instruction set generation. If the host compiler has been patched 5346 so that the vec_ld and vec_st builtin functions generate Altivec 5347 memory instructions instead of VSX memory instructions, then you 5348 should be able to build the compiler with VSX instruction 5349 generation. 5350 5351 Changes for GCC Developers 5352 5353 Note: these changes concern developers that develop GCC itself or 5354 software that integrates with GCC, such as plugins, and not the general 5355 GCC users. 5356 * The gengtype utility, which previously was internal to the GCC 5357 build process, has been enchanced to provide GC root information 5358 for plugins as necessary. 5359 * The old GC allocation interface of ggc_alloc and friends was 5360 replaced with a type-safe alternative. 5361 5362 GCC 4.6.1 5363 5364 This is the [20]list of problem reports (PRs) from GCC's bug tracking 5365 system that are known to be fixed in the 4.6.1 release. This list might 5366 not be complete (that is, it is possible that some PRs that have been 5367 fixed are not listed here). 5368 5369 GCC 4.6.2 5370 5371 This is the [21]list of problem reports (PRs) from GCC's bug tracking 5372 system that are known to be fixed in the 4.6.2 release. This list might 5373 not be complete (that is, it is possible that some PRs that have been 5374 fixed are not listed here). 5375 5376 GCC 4.6.3 5377 5378 This is the [22]list of problem reports (PRs) from GCC's bug tracking 5379 system that are known to be fixed in the 4.6.3 release. This list might 5380 not be complete (that is, it is possible that some PRs that have been 5381 fixed are not listed here). 5382 5383 GCC 4.6.4 5384 5385 This is the [23]list of problem reports (PRs) from GCC's bug tracking 5386 system that are known to be fixed in the 4.6.4 release. This list might 5387 not be complete (that is, it is possible that some PRs that have been 5388 fixed are not listed here). 5389 5390 5391 For questions related to the use of GCC, please consult these web 5392 pages and the [24]GCC manuals. If that fails, the 5393 [25]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5394 web pages and the development of GCC are welcome on our developer 5395 list at [26]gcc (a] gcc.gnu.org. All of [27]our lists have public 5396 archives. 5397 5398 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and 5399 distribution of this entire article is permitted in any medium, 5400 provided this notice is preserved. 5401 5402 These pages are [29]maintained by the GCC team. Last modified 5403 2018-09-30[30]. 5404 5405 References 5406 5407 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 5408 2. http://gcc.gnu.org/gcc-4.5/changes.html 5409 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 5410 4. http://gcc.gnu.org/gcc-4.6/porting_to.html 5411 5. http://gcc.gnu.org/projects/lto/whopr.pdf 5412 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 5413 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html 5414 8. https://gcc.gnu.org/PR43145 5415 9. https://gcc.gnu.org/PR43680 5416 10. https://gcc.gnu.org/PR33558 5417 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 5418 12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 5419 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 5420 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 5421 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 5422 16. https://gcc.gnu.org/wiki/OOP 5423 17. https://gcc.gnu.org/wiki/Coarray 5424 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 5425 19. https://golang.org/ 5426 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 5427 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 5428 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 5429 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 5430 24. https://gcc.gnu.org/onlinedocs/ 5431 25. mailto:gcc-help (a] gcc.gnu.org 5432 26. mailto:gcc (a] gcc.gnu.org 5433 27. https://gcc.gnu.org/lists.html 5434 28. https://www.fsf.org/ 5435 29. https://gcc.gnu.org/about.html 5436 30. http://validator.w3.org/check/referer 5437 ====================================================================== 5438 http://gcc.gnu.org/gcc-4.5/index.html 5439 GCC 4.5 Release Series 5440 5441 Jul 2, 2012 5442 5443 The [1]GNU project and the GCC developers are pleased to announce the 5444 release of GCC 4.5.4. 5445 5446 This release is a bug-fix release, containing fixes for regressions in 5447 GCC 4.5.3 relative to previous releases of GCC. 5448 5449 This release series is no longer maintained. 5450 5451 Release History 5452 5453 GCC 4.5.4 5454 Jul 2, 2012 ([2]changes) 5455 5456 GCC 4.5.3 5457 Apr 28, 2011 ([3]changes) 5458 5459 GCC 4.5.2 5460 Dec 16, 2010 ([4]changes) 5461 5462 GCC 4.5.1 5463 Jul 31, 2010 ([5]changes) 5464 5465 GCC 4.5.0 5466 April 14, 2010 ([6]changes) 5467 5468 References and Acknowledgements 5469 5470 GCC used to stand for the GNU C Compiler, but since the compiler 5471 supports several other languages aside from C, it now stands for the 5472 GNU Compiler Collection. 5473 5474 A list of [7]successful builds is updated as new information becomes 5475 available. 5476 5477 The GCC developers would like to thank the numerous people that have 5478 contributed new features, improvements, bug fixes, and other changes as 5479 well as test results to GCC. This [8]amazing group of volunteers is 5480 what makes GCC successful. 5481 5482 For additional information about GCC please refer to the [9]GCC project 5483 web site or contact the [10]GCC development mailing list. 5484 5485 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 5486 5487 5488 For questions related to the use of GCC, please consult these web 5489 pages and the [13]GCC manuals. If that fails, the 5490 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5491 web pages and the development of GCC are welcome on our developer 5492 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 5493 archives. 5494 5495 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 5496 distribution of this entire article is permitted in any medium, 5497 provided this notice is preserved. 5498 5499 These pages are [18]maintained by the GCC team. Last modified 5500 2018-09-30[19]. 5501 5502 References 5503 5504 1. http://www.gnu.org/ 5505 2. http://gcc.gnu.org/gcc-4.5/changes.html 5506 3. http://gcc.gnu.org/gcc-4.5/changes.html 5507 4. http://gcc.gnu.org/gcc-4.5/changes.html 5508 5. http://gcc.gnu.org/gcc-4.5/changes.html 5509 6. http://gcc.gnu.org/gcc-4.5/changes.html 5510 7. http://gcc.gnu.org/gcc-4.5/buildstat.html 5511 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5512 9. http://gcc.gnu.org/index.html 5513 10. mailto:gcc (a] gcc.gnu.org 5514 11. http://gcc.gnu.org/mirrors.html 5515 12. http://gcc.gnu.org/svn.html 5516 13. https://gcc.gnu.org/onlinedocs/ 5517 14. mailto:gcc-help (a] gcc.gnu.org 5518 15. mailto:gcc (a] gcc.gnu.org 5519 16. https://gcc.gnu.org/lists.html 5520 17. https://www.fsf.org/ 5521 18. https://gcc.gnu.org/about.html 5522 19. http://validator.w3.org/check/referer 5523 ====================================================================== 5524 http://gcc.gnu.org/gcc-4.5/changes.html 5525 GCC 4.5 Release Series 5526 Changes, New Features, and Fixes 5527 5528 Caveats 5529 5530 * GCC now requires the [1]MPC library in order to build. See the 5531 [2]prerequisites page for version requirements. 5532 * Support for a number of older systems and recently unmaintained or 5533 untested target ports of GCC has been declared obsolete in GCC 4.5. 5534 Unless there is activity to revive them, the next release of GCC 5535 will have their sources permanently removed. 5536 The following ports for individual systems on particular 5537 architectures have been obsoleted: 5538 + IRIX releases before 6.5 (mips-sgi-irix5*, 5539 mips-sgi-irix6.[0-4]) 5540 + Solaris 7 (*-*-solaris2.7) 5541 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 5542 alpha-dec-osf5.0*) 5543 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 5544 can be found in the [3]announcement. 5545 Support for the classic POWER architecture implemented in the 5546 original RIOS and RIOS2 processors of the old IBM RS/6000 product 5547 line has been obsoleted in the rs6000 port. This does not affect 5548 the new generation Power and PowerPC architectures. 5549 * Support has been removed for all the [4]configurations obsoleted in 5550 GCC 4.4. 5551 * Support has been removed for the protoize and unprotoize utilities, 5552 obsoleted in GCC 4.4. 5553 * Support has been removed for tuning for Itanium1 (Merced) variants. 5554 Note that code tuned for Itanium2 should also run correctly on 5555 Itanium1. 5556 * GCC now generates unwind info also for epilogues. DWARF debuginfo 5557 generated by GCC now uses more features of DWARF3 than before, and 5558 also some DWARF4 features. GDB older than 7.0 is not able to handle 5559 either of these, so to debug GCC 4.5 generated binaries or 5560 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 5561 features with the -gdwarf-3 -gstrict-dwarf options, or use 5562 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 5563 epilogue unwind info is emitted unconditionally whenever unwind 5564 info is emitted. 5565 * On x86 targets, code containing floating-point calculations may run 5566 significantly more slowly when compiled with GCC 4.5 in strict C99 5567 conformance mode than they did with earlier GCC versions. This is 5568 due to stricter standard conformance of the compiler and can be 5569 avoided by using the option -fexcess-precision=fast; also see 5570 [5]below. 5571 * The function attribute noinline no longer prevents GCC from cloning 5572 the function. A new attribute noclone has been introduced for this 5573 purpose. Cloning a function means that it is duplicated and the new 5574 copy is specialized for certain contexts (for example when a 5575 parameter is a known constant). 5576 5577 General Optimizer Improvements 5578 5579 * The -save-temps now takes an optional argument. The -save-temps and 5580 -save-temps=cwd switches write the temporary files in the current 5581 working directory based on the original source file. The 5582 -save-temps=obj switch will write files into the directory 5583 specified with the -o option, and the intermediate filenames are 5584 based on the output file. This will allow the user to get the 5585 compiler intermediate files when doing parallel builds without two 5586 builds of the same filename located in different directories from 5587 interfering with each other. 5588 * Debugging dumps are now created in the same directory as the object 5589 file rather than in the current working directory. This allows the 5590 user to get debugging dumps when doing parallel builds without two 5591 builds of the same filename interfering with each other. 5592 * GCC has been integrated with the MPC library. This allows GCC to 5593 evaluate complex arithmetic at compile time [6]more accurately. It 5594 also allows GCC to evaluate calls to complex built-in math 5595 functions having constant arguments and replace them at compile 5596 time with their mathematically equivalent results. In doing so, GCC 5597 can generate correct results regardless of the math library 5598 implementation or floating point precision of the host platform. 5599 This also allows GCC to generate identical results regardless of 5600 whether one compiles in native or cross-compile configurations to a 5601 particular target. The following built-in functions take advantage 5602 of this new capability: cacos, cacosh, casin, casinh, catan, 5603 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 5604 and ctanh. The float and long double variants of these functions 5605 (e.g. csinf and csinl) are also handled. 5606 * A new link-time optimizer has been added ([7]-flto). When this 5607 option is used, GCC generates a bytecode representation of each 5608 input file and writes it to specially-named sections in each object 5609 file. When the object files are linked together, all the function 5610 bodies are read from these named sections and instantiated as if 5611 they had been part of the same translation unit. This enables 5612 interprocedural optimizations to work across different files (and 5613 even different languages), potentially improving the performance of 5614 the generated code. To use the link-timer optimizer, -flto needs to 5615 be specified at compile time and during the final link. If the 5616 program does not require any symbols to be exported, it is possible 5617 to combine -flto and the experimental [8]-fwhopr with 5618 [9]-fwhole-program to allow the interprocedural optimizers to use 5619 more aggressive assumptions. 5620 * The automatic parallelization pass was enhanced to support 5621 parallelization of outer loops. 5622 * Automatic parallelization can be enabled as part of Graphite. In 5623 addition to -ftree-parallelize-loops=, specify 5624 -floop-parallelize-all to enable the Graphite-based optimization. 5625 * The infrastructure for optimizing based on [10]restrict qualified 5626 pointers has been rewritten and should result in code generation 5627 improvements. Optimizations based on restrict qualified pointers 5628 are now also available when using -fno-strict-aliasing. 5629 * There is a new optimization pass that attempts to change prototype 5630 of functions to avoid unused parameters, pass only relevant parts 5631 of structures and turn arguments passed by reference to arguments 5632 passed by value when possible. It is enabled by -O2 and above as 5633 well as -Os and can be manually invoked using the new command-line 5634 switch -fipa-sra. 5635 * GCC now optimize exception handling code. In particular cleanup 5636 regions that are proved to not have any effect are optimized out. 5637 5638 New Languages and Language specific improvements 5639 5640 All languages 5641 5642 * The -fshow-column option is now on by default. This means error 5643 messages now have a column associated with them. 5644 5645 Ada 5646 5647 * Compilation of programs heavily using discriminated record types 5648 with variant parts has been sped up and generates more compact 5649 code. 5650 * Stack checking now works reasonably well on most plaforms. In some 5651 specific cases, stack overflows may still fail to be detected, but 5652 a compile-time warning will be issued for these cases. 5653 5654 C family 5655 5656 * If a header named in a #include directive is not found, the 5657 compiler exits immediately. This avoids a cascade of errors arising 5658 from declarations expected to be found in that header being 5659 missing. 5660 * A new built-in function __builtin_unreachable() has been added that 5661 tells the compiler that control will never reach that point. It may 5662 be used after asm statements that terminate by transferring control 5663 elsewhere, and in other places that are known to be unreachable. 5664 * The -Wlogical-op option now warns for logical expressions such as 5665 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 5666 mistakes. This option is disabled by default. 5667 * An asm goto feature has been added to allow asm statements that 5668 jump to C labels. 5669 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 5670 * The deprecated attribute now takes an optional string argument, for 5671 example, __attribute__((deprecated("text string"))), that will be 5672 printed together with the deprecation warning. 5673 5674 C 5675 5676 * The -Wenum-compare option, which warns when comparing values of 5677 different enum types, now works for C. It formerly only worked for 5678 C++. This warning is enabled by -Wall. It may be avoided by using a 5679 type cast. 5680 * The -Wcast-qual option now warns about casts which are unsafe in 5681 that they permit const-correctness to be violated without further 5682 warnings. Specifically, it warns about cases where a qualifier is 5683 added when all the lower types are not const. For example, it warns 5684 about a cast from char ** to const char **. 5685 * The -Wc++-compat option is significantly improved. It issues new 5686 warnings for: 5687 + Using C++ reserved operator names as identifiers. 5688 + Conversions to enum types without explicit casts. 5689 + Using va_arg with an enum type. 5690 + Using different enum types in the two branches of ?:. 5691 + Using ++ or -- on a variable of enum type. 5692 + Using the same name as both a struct, union or enum tag and a 5693 typedef, unless the typedef refers to the tagged type itself. 5694 + Using a struct, union, or enum which is defined within another 5695 struct or union. 5696 + A struct field defined using a typedef if there is a field in 5697 the struct, or an enclosing struct, whose name is the typedef 5698 name. 5699 + Duplicate definitions at file scope. 5700 + Uninitialized const variables. 5701 + A global variable with an anonymous struct, union, or enum 5702 type. 5703 + Using a string constant to initialize a char array whose size 5704 is the length of the string. 5705 * The new -Wjump-misses-init option warns about cases where a goto or 5706 switch skips the initialization of a variable. This sort of branch 5707 is an error in C++ but not in C. This warning is enabled by 5708 -Wc++-compat. 5709 * GCC now ensures that a C99-conforming <stdint.h> is present on most 5710 targets, and uses information about the types in this header to 5711 implement the Fortran bindings to those types. GCC does not ensure 5712 the presence of such a header, and does not implement the Fortran 5713 bindings, on the following targets: NetBSD, VxWorks, VMS, 5714 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 5715 * GCC now implements C90- and C99-conforming rules for constant 5716 expressions. This may cause warnings or errors for some code using 5717 expressions that can be folded to a constant but are not constant 5718 expressions as defined by ISO C. 5719 * All known target-independent C90 and C90 Amendment 1 conformance 5720 bugs, and all known target-independent C99 conformance bugs not 5721 related to floating point or extended identifiers, have been fixed. 5722 * The C decimal floating point support now includes support for the 5723 FLOAT_CONST_DECIMAL64 pragma. 5724 * The named address space feature from ISO/IEC TR 18037 is now 5725 supported. This is currently only implemented for the SPU 5726 processor. 5727 5728 C++ 5729 5730 * Improved [11]experimental support for the upcoming C++0x ISO C++ 5731 standard, including support for raw strings, lambda expressions and 5732 explicit type conversion operators. 5733 * When printing the name of a class template specialization, G++ will 5734 now omit any template arguments which come from default template 5735 arguments. This behavior (and the pretty-printing of function 5736 template specializations as template signature and arguments) can 5737 be disabled with the -fno-pretty-templates option. 5738 * Access control is now applied to typedef names used in a template, 5739 which may cause G++ to reject some ill-formed code that was 5740 accepted by earlier releases. The -fno-access-control option can be 5741 used as a temporary workaround until the code is corrected. 5742 * Compilation time for code that uses templates should now scale 5743 linearly with the number of instantiations rather than 5744 quadratically, as template instantiations are now looked up using 5745 hash tables. 5746 * Declarations of functions that look like builtin declarations of 5747 library functions are only considered to be redeclarations if they 5748 are declared with extern "C". This may cause problems with code 5749 that omits extern "C" on hand-written declarations of C library 5750 functions such as abort or memcpy. Such code is ill-formed, but was 5751 accepted by earlier releases. 5752 * Diagnostics that used to complain about passing non-POD types to 5753 ... or jumping past the declaration of a non-POD variable now check 5754 for triviality rather than PODness, as per C++0x. 5755 * In C++0x mode local and anonymous classes are now allowed as 5756 template arguments, and in declarations of variables and functions 5757 with linkage, so long as any such declaration that is used is also 5758 defined ([12]DR 757). 5759 * Labels may now have attributes, as has been permitted for a while 5760 in C. This is only permitted when the label definition and the 5761 attribute specifier is followed by a semicoloni.e., the label 5762 applies to an empty statement. The only useful attribute for a 5763 label is unused. 5764 * G++ now implements [13]DR 176. Previously G++ did not support using 5765 the injected-class-name of a template base class as a type name, 5766 and lookup of the name found the declaration of the template in the 5767 enclosing scope. Now lookup of the name finds the 5768 injected-class-name, which can be used either as a type or as a 5769 template, depending on whether or not the name is followed by a 5770 template argument list. As a result of this change, some code that 5771 was previously accepted may be ill-formed because 5772 1. The injected-class-name is not accessible because it's from a 5773 private base, or 5774 2. The injected-class-name cannot be used as an argument for a 5775 template template parameter. 5776 In either of these cases, the code can be fixed by adding a 5777 nested-name-specifier to explicitly name the template. The first 5778 can be worked around with -fno-access-control; the second is only 5779 rejected with -pedantic. 5780 * A new standard mangling for SIMD vector types has been added, to 5781 avoid name clashes on systems with vectors of varying length. By 5782 default the compiler still uses the old mangling, but emits aliases 5783 with the new mangling on targets that support strong aliases. Users 5784 can switch over entirely to the new mangling with -fabi-version=4 5785 or -fabi-version=0. -Wabi will now warn about code that uses the 5786 old mangling. 5787 * The command-line option -ftemplate-depth-N is now written as 5788 -ftemplate-depth=N and the old form is deprecated. 5789 * Conversions between NULL and non-pointer types are now warned by 5790 default. The new option -Wno-conversion-null disables these 5791 warnings. Previously these warnings were only available when using 5792 -Wconversion explicitly. 5793 5794 Runtime Library (libstdc++) 5795 5796 * Improved experimental support for the upcoming ISO C++ standard, 5797 C++0x, including: 5798 + Support for <future>, <functional>, and <random>. 5799 + Existing facilities now exploit explicit operators and the 5800 newly implemented core C++0x features. 5801 + The header <cstdatomic> has been renamed to <atomic>. 5802 * An experimental [14]profile mode has been added. This is an 5803 implementation of many C++ standard library constructs with an 5804 additional analysis layer that gives performance improvement advice 5805 based on recognition of suboptimal usage patterns. For example, 5806 #include <vector> 5807 int main() 5808 { 5809 std::vector<int> v; 5810 for (int k = 0; k < 1024; ++k) 5811 v.insert(v.begin(), k); 5812 } 5813 5814 When instrumented via the profile mode, can return suggestions 5815 about the initial size and choice of the container used as follows: 5816 vector-to-list: improvement = 5: call stack = 0x804842c ... 5817 : advice = change std::vector to std::list 5818 vector-size: improvement = 3: call stack = 0x804842c ... 5819 : advice = change initial container size from 0 to 1024 5820 5821 These constructs can be substituted for the normal libstdc++ 5822 constructs on a piecemeal basis, or all existing components can be 5823 transformed via the -D_GLIBCXX_PROFILE macro. 5824 * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR 5825 24733) has been added. This support is in header file 5826 <decimal/decimal>, uses namespace std::decimal, and includes 5827 classes decimal32, decimal64, and decimal128. 5828 * Sources have been audited for application of function attributes 5829 nothrow, const, pure, and noreturn. 5830 * Python pretty-printers have been added for many standard library 5831 components that simplify the internal representation and present a 5832 more intuitive view of components when used with 5833 appropriately-advanced versions of GDB. For more information, 5834 please consult the more [16]detailed description. 5835 * The default behavior for comparing typeinfo names has changed, so 5836 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 5837 * The new -static-libstdc++ option directs g++ to link the C++ 5838 library statically, even if the default would normally be to link 5839 it dynamically. 5840 5841 Fortran 5842 5843 * The COMMON default padding has been changed instead of adding the 5844 padding before a variable it is now added afterwards, which 5845 increases the compatibility with other vendors and helps to obtain 5846 the correct output in some cases. Cf. also the -falign-commons 5847 option ([17]added in 4.4). 5848 * The -finit-real= option now also supports the value snan for 5849 signaling not-a-number; to be effective, one additionally needs to 5850 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 5851 optimizations can turn a signaling NaN into a quiet one. 5852 * The new option -fcheck= has been added with the options bounds, 5853 array-temps, do, pointer, and recursive. The bounds and array-temps 5854 options are equivalent to -fbounds-check and 5855 -fcheck-array-temporaries. The do option checks for invalid 5856 modification of loop iteration variables, and the recursive option 5857 tests for recursive calls to subroutines/functions which are not 5858 marked as recursive. With pointer pointer association checks in 5859 calls are performed; however, neither undefined pointers nor 5860 pointers in expressions are handled. Using -fcheck=all enables all 5861 these run-time checks. 5862 * The run-time checking -fcheck=bounds now warns about invalid string 5863 lengths of character dummy arguments. Additionally, more 5864 compile-time checks have been added. 5865 * The new option [18]-fno-protect-parens has been added; if set, the 5866 compiler may reorder REAL and COMPLEX expressions without regard to 5867 parentheses. 5868 * GNU Fortran no longer links against libgfortranbegin. As before, 5869 MAIN__ (assembler symbol name) is the actual Fortran main program, 5870 which is invoked by the main function. However, main is now 5871 generated and put in the same object file as MAIN__. For the time 5872 being, libgfortranbegin still exists for backward compatibility. 5873 For details see the new [19]Mixed-Language Programming chapter in 5874 the manual. 5875 * The I/O library was restructured for performance and cleaner code. 5876 * Array assignments and WHERE are now run in parallel when OpenMP's 5877 WORKSHARE is used. 5878 * The experimental option -fwhole-file was added. The option allows 5879 whole-file checking of procedure arguments and allows for better 5880 optimizations. It can also be used with -fwhole-program, which is 5881 now also supported in gfortran. 5882 * More Fortran 2003 and Fortran 2008 mathematical functions can now 5883 be used as initialization expressions. 5884 * Some extended attributes such as STDCALL are now supported via the 5885 [20]GCC$ compiler directive. 5886 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 5887 intrinsic behaves now as if zero were always positive. 5888 * For legacy compatibiliy: On Cygwin and MinGW, the special files 5889 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 5890 supported. 5891 * Fortran 2003 support has been extended: 5892 + Procedure-pointer function results and procedure-pointer 5893 components (including PASS), 5894 + allocatable scalars (experimental), 5895 + DEFERRED type-bound procedures, 5896 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 5897 have been implemented. 5898 + The ALLOCATE statement supports type-specs and the SOURCE= 5899 argument. 5900 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 5901 type-bound procedure (i.e. as type-bound operators). 5902 + Rounding (ROUND=, RZ, ...) for output is now supported. 5903 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 5904 intrinsic module ISO_C_BINDING are now supported, except for 5905 the targets listed above as ones where GCC does not have 5906 <stdint.h> type information. 5907 + Extensible derived types with type-bound procedure or 5908 procedure pointer with PASS attribute now have to use CLASS in 5909 line with the Fortran 2003 standard; the workaround to use 5910 TYPE is no longer supported. 5911 + [21]Experimental, incomplete support for polymorphism, 5912 including CLASS, SELECT TYPE and dynamic dispatch of 5913 type-bound procedure calls. Some features do not work yet such 5914 as unlimited polymorphism (CLASS(*)). 5915 * Fortran 2008 support has been extended: 5916 + The OPEN statement now supports the NEWUNIT= option, which 5917 returns a unique file unit, thus preventing inadvertent use of 5918 the same unit in different parts of the program. 5919 + Support for unlimited format items has been added. 5920 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 5921 the intrinsic module ISO_FORTRAN_ENV are now supported. 5922 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 5923 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 5924 and ATANH have been added (for real and complex arguments) and 5925 ATAN(Y,X) is now an alias for ATAN2(Y,X). 5926 + The BLOCK construct has been implemented. 5927 5928 New Targets and Target Specific Improvements 5929 5930 AIX 5931 5932 * Full cross-toolchain support now available with GNU Binutils 5933 5934 ARM 5935 5936 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 5937 * GCC now supports the ARM v7E-M architecture. 5938 * GCC now supports VFPv4-based FPUs and FPUs with 5939 single-precision-only VFP. 5940 * GCC has many improvements to optimization for other ARM processors, 5941 including scheduling support for the integer pipeline on Cortex-A9. 5942 * GCC now supports the IEEE 754-2008 half-precision floating-point 5943 type, and a variant ARM-specific half-precision type. This type is 5944 specified using __fp16, with the layout determined by 5945 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 5946 VFPv4 half-precision instructions will be used. 5947 * GCC now supports the variant of AAPCS that uses VFP registers for 5948 parameter passing and return values. 5949 5950 AVR 5951 5952 * The -mno-tablejump option has been removed because it has the same 5953 effect as the -fno-jump-tables option. 5954 * Added support for these new AVR devices: 5955 + ATmega8U2 5956 + ATmega16U2 5957 + ATmega32U2 5958 5959 IA-32/x86-64 5960 5961 * GCC now will set the default for -march= based on the configure 5962 target. 5963 * GCC now supports handling floating-point excess precision arising 5964 from use of the x87 floating-point unit in a way that conforms to 5965 ISO C99. This is enabled with -fexcess-precision=standard and with 5966 standards conformance options such as -std=c99, and may be disabled 5967 using -fexcess-precision=fast. 5968 * Support for the Intel Atom processor is now available through the 5969 -march=atom and -mtune=atom options. 5970 * A new -mcrc32 option is now available to enable crc32 intrinsics. 5971 * A new -mmovbe option is now available to enable GCC to use the 5972 movbe instruction to implement __builtin_bswap32 and 5973 __builtin_bswap64. 5974 * SSE math now can be enabled by default at configure time with the 5975 new --with-fpmath=sse option. 5976 * There is a new intrinsic header file, <x86intrin.h>. It should be 5977 included before using any IA-32/x86-64 intrinsics. 5978 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 5979 Orochi processors are now available with the -mxop, -mfma4, and 5980 -mlwp options. 5981 * The -mabm option enables GCC to use the popcnt and lzcnt 5982 instructions on AMD processors. 5983 * The -mpopcnt option enables GCC to use the popcnt instructions on 5984 both AMD and Intel processors. 5985 5986 M68K/ColdFire 5987 5988 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 5989 and 5441x devices. 5990 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 5991 processors. 5992 5993 MeP 5994 5995 Support has been added for the Toshiba Media embedded Processor (MeP, 5996 or mep-elf) embedded target. 5997 5998 MIPS 5999 6000 * GCC now supports MIPS 1004K processors. 6001 * GCC can now be configured with options --with-arch-32, 6002 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 6003 default optimization separately for 32-bit and 64-bit modes. 6004 * MIPS targets now support an alternative _mcount interface, in which 6005 register $12 points to the function's save slot for register $31. 6006 This interface is selected by the -mcount-ra-address option; see 6007 the documentation for more details. 6008 * GNU/Linux targets can now generate read-only .eh_frame sections. 6009 This optimization requires GNU binutils 2.20 or above, and is only 6010 available if GCC is configured with a suitable version of binutils. 6011 * GNU/Linux targets can now attach special relocations to indirect 6012 calls, so that the linker can turn them into direct jumps or 6013 branches. This optimization requires GNU binutils 2.20 or later, 6014 and is automatically selected if GCC is configured with an 6015 appropriate version of binutils. It can be explicitly enabled or 6016 disabled using the -mrelax-pic-calls command-line option. 6017 * GCC now generates more heavily-optimized atomic operations on 6018 Octeon processors. 6019 * MIPS targets now support the -fstack-protector option. 6020 * GCC now supports an -msynci option, which specifies that synci is 6021 enough to flush the instruction cache, without help from the 6022 operating system. GCC uses this information to optimize 6023 automatically-generated cache flush operations, such as those used 6024 for nested functions in C. There is also a --with-synci 6025 configure-time option, which makes -msynci the default. 6026 * GCC supports four new function attributes for interrupt handlers: 6027 interrupt, use_shadow_register_set, keep_interrupts_masked and 6028 use_debug_exception_return. See the documentation for more details 6029 about these attributes. 6030 6031 RS/6000 (POWER/PowerPC) 6032 6033 * GCC now supports the Power ISA 2.06, which includes the VSX 6034 instructions that add vector 64-bit floating point support, new 6035 population count instructions, and conversions between floating 6036 point and unsigned types. 6037 * Support for the power7 processor is now available through the 6038 -mcpu=power7 and -mtune=power7. 6039 * GCC will now vectorize loops that contain simple math functions 6040 like copysign when generating code for altivec or VSX targets. 6041 * Support for the A2 processor is now available through the -mcpu=a2 6042 and -mtune=a2 options. 6043 * Support for the 476 processor is now available through the 6044 -mcpu={476,476fp} and -mtune={476,476fp} options. 6045 * Support for the e500mc64 processor is now available through the 6046 -mcpu=e500mc64 and -mtune=e500mc64 options. 6047 * GCC can now be configured with options --with-cpu-32, 6048 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 6049 default optimization separately for 32-bit and 64-bit modes. 6050 * Starting with GCC 4.5.4, vectors of type vector long long or vector 6051 long are passed and returned in the same method as other vectors 6052 with the VSX instruction set. Previously the GCC compiler did not 6053 adhere to the ABI for 128-bit vectors with 64-bit integer base 6054 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 6055 6056 RX 6057 6058 Support has been added for the Renesas RX Processor (rx-elf) target. 6059 6060 Operating Systems 6061 6062 Windows (Cygwin and MinGW) 6063 6064 * GCC now installs all the major language runtime libraries as DLLs 6065 when configured with the --enable-shared option. 6066 * GCC now makes use of the new support for aligned common variables 6067 in versions of binutils >= 2.20 to fix bugs in the support for SSE 6068 data types. 6069 * Improvements to the libffi support library increase the reliability 6070 of code generated by GCJ on all Windows platforms. Libgcj is 6071 enabled by default for the first time. 6072 * Libtool improvements simplify installation by placing the generated 6073 DLLs in the correct binaries directory. 6074 * Numerous other minor bugfixes and improvements, and substantial 6075 enhancements to the Fortran language support library. 6076 6077 > 6078 6079 Other significant improvements 6080 6081 Plugins 6082 6083 * It is now possible to extend the compiler without having to modify 6084 its source code. A new option -fplugin=file.so tells GCC to load 6085 the shared object file.so and execute it as part of the compiler. 6086 The internal documentation describes the details on how plugins can 6087 interact with the compiler. 6088 6089 Installation changes 6090 6091 * The move to newer autotools changed default installation 6092 directories and switches to control them: The --with-datarootdir, 6093 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 6094 used any more. Instead, you can now use --datarootdir, --docdir, 6095 --htmldir, and --pdfdir. The default installation directories have 6096 changed as follows according to the GNU Coding Standards: 6097 6098 datarootdir read-only architecture-independent data root [PREFIX/share] 6099 localedir locale-specific message catalogs [DATAROOTDIR/locale] 6100 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 6101 htmldir html documentation [DOCDIR] 6102 dvidir dvi documentation [DOCDIR] 6103 pdfdir pdf documentation [DOCDIR] 6104 psdir ps documentation [DOCDIR] 6105 The following variables have new default values: 6106 6107 datadir read-only architecture-independent data [DATAROOTDIR] 6108 infodir info documentation [DATAROOTDIR/info] 6109 mandir man documentation [DATAROOTDIR/man] 6110 6111 GCC 4.5.1 6112 6113 This is the [22]list of problem reports (PRs) from GCC's bug tracking 6114 system that are known to be fixed in the 4.5.1 release. This list might 6115 not be complete (that is, it is possible that some PRs that have been 6116 fixed are not listed here). 6117 6118 All languages 6119 6120 * GCC's new link-time optimizer ([23]-flto) now also works on a few 6121 non-ELF targets: 6122 + Cygwin (*-cygwin*) 6123 + MinGW (*-mingw*) 6124 + Darwin on x86-64 (x86_64-apple-darwin*) 6125 LTO is not enabled by default for these targets. To enable LTO, you 6126 should configure with the --enable-lto option. 6127 6128 GCC 4.5.2 6129 6130 This is the [24]list of problem reports (PRs) from GCC's bug tracking 6131 system that are known to be fixed in the 4.5.2 release. This list might 6132 not be complete (that is, it is possible that some PRs that have been 6133 fixed are not listed here). 6134 6135 GCC 4.5.3 6136 6137 This is the [25]list of problem reports (PRs) from GCC's bug tracking 6138 system that are known to be fixed in the 4.5.3 release. This list might 6139 not be complete (that is, it is possible that some PRs that have been 6140 fixed are not listed here). 6141 6142 On the PowerPC compiler, the Altivec builtin functions vec_ld and 6143 vec_st have been modified to generate the Altivec memory instructions 6144 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 6145 release, these builtin functions were changed to generate VSX memory 6146 reference instructions instead of Altivec memory instructions, but 6147 there are differences between the two instructions. If the VSX 6148 instruction set is available, you can now use the new builtin functions 6149 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 6150 instructions. 6151 6152 GCC 4.5.4 6153 6154 This is the [26]list of problem reports (PRs) from GCC's bug tracking 6155 system that are known to be fixed in the 4.5.4 release. This list might 6156 not be complete (that is, it is possible that some PRs that have been 6157 fixed are not listed here). 6158 6159 6160 For questions related to the use of GCC, please consult these web 6161 pages and the [27]GCC manuals. If that fails, the 6162 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6163 web pages and the development of GCC are welcome on our developer 6164 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 6165 archives. 6166 6167 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 6168 distribution of this entire article is permitted in any medium, 6169 provided this notice is preserved. 6170 6171 These pages are [32]maintained by the GCC team. Last modified 6172 2018-09-30[33]. 6173 6174 References 6175 6176 1. http://www.multiprecision.org/mpc/ 6177 2. https://gcc.gnu.org/install/prerequisites.html 6178 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 6179 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 6180 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86 6181 6. https://gcc.gnu.org/PR30789 6182 7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 6183 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 6184 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 6185 10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 6186 11. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html 6187 12. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 6188 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 6189 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 6190 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 6191 16. https://sourceware.org/gdb/wiki/STLSupport 6192 17. http://gcc.gnu.org/gcc-4.4/changes.html 6193 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 6194 19. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 6195 20. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 6196 21. https://gcc.gnu.org/wiki/OOP 6197 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 6198 23. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 6199 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 6200 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 6201 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 6202 27. https://gcc.gnu.org/onlinedocs/ 6203 28. mailto:gcc-help (a] gcc.gnu.org 6204 29. mailto:gcc (a] gcc.gnu.org 6205 30. https://gcc.gnu.org/lists.html 6206 31. https://www.fsf.org/ 6207 32. https://gcc.gnu.org/about.html 6208 33. http://validator.w3.org/check/referer 6209 ====================================================================== 6210 http://gcc.gnu.org/gcc-4.4/index.html 6211 GCC 4.4 Release Series 6212 6213 March 13, 2012 6214 6215 The [1]GNU project and the GCC developers are pleased to announce the 6216 release of GCC 4.4.7. 6217 6218 This release is a bug-fix release, containing fixes for regressions in 6219 GCC 4.4.6 relative to previous releases of GCC. 6220 6221 This release series is no longer maintained. 6222 6223 Release History 6224 6225 GCC 4.4.7 6226 March 13, 2012 ([2]changes) 6227 6228 GCC 4.4.6 6229 April 16, 2011 ([3]changes) 6230 6231 GCC 4.4.5 6232 October 1, 2010 ([4]changes) 6233 6234 GCC 4.4.4 6235 April 29, 2010 ([5]changes) 6236 6237 GCC 4.4.3 6238 January 21, 2010 ([6]changes) 6239 6240 GCC 4.4.2 6241 October 15, 2009 ([7]changes) 6242 6243 GCC 4.4.1 6244 July 22, 2009 ([8]changes) 6245 6246 GCC 4.4.0 6247 April 21, 2009 ([9]changes) 6248 6249 References and Acknowledgements 6250 6251 GCC used to stand for the GNU C Compiler, but since the compiler 6252 supports several other languages aside from C, it now stands for the 6253 GNU Compiler Collection. 6254 6255 A list of [10]successful builds is updated as new information becomes 6256 available. 6257 6258 The GCC developers would like to thank the numerous people that have 6259 contributed new features, improvements, bug fixes, and other changes as 6260 well as test results to GCC. This [11]amazing group of volunteers is 6261 what makes GCC successful. 6262 6263 For additional information about GCC please refer to the [12]GCC 6264 project web site or contact the [13]GCC development mailing list. 6265 6266 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 6267 6268 6269 For questions related to the use of GCC, please consult these web 6270 pages and the [16]GCC manuals. If that fails, the 6271 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6272 web pages and the development of GCC are welcome on our developer 6273 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 6274 archives. 6275 6276 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 6277 distribution of this entire article is permitted in any medium, 6278 provided this notice is preserved. 6279 6280 These pages are [21]maintained by the GCC team. Last modified 6281 2018-09-30[22]. 6282 6283 References 6284 6285 1. http://www.gnu.org/ 6286 2. http://gcc.gnu.org/gcc-4.4/changes.html 6287 3. http://gcc.gnu.org/gcc-4.4/changes.html 6288 4. http://gcc.gnu.org/gcc-4.4/changes.html 6289 5. http://gcc.gnu.org/gcc-4.4/changes.html 6290 6. http://gcc.gnu.org/gcc-4.4/changes.html 6291 7. http://gcc.gnu.org/gcc-4.4/changes.html 6292 8. http://gcc.gnu.org/gcc-4.4/changes.html 6293 9. http://gcc.gnu.org/gcc-4.4/changes.html 6294 10. http://gcc.gnu.org/gcc-4.4/buildstat.html 6295 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 6296 12. http://gcc.gnu.org/index.html 6297 13. mailto:gcc (a] gcc.gnu.org 6298 14. http://gcc.gnu.org/mirrors.html 6299 15. http://gcc.gnu.org/svn.html 6300 16. https://gcc.gnu.org/onlinedocs/ 6301 17. mailto:gcc-help (a] gcc.gnu.org 6302 18. mailto:gcc (a] gcc.gnu.org 6303 19. https://gcc.gnu.org/lists.html 6304 20. https://www.fsf.org/ 6305 21. https://gcc.gnu.org/about.html 6306 22. http://validator.w3.org/check/referer 6307 ====================================================================== 6308 http://gcc.gnu.org/gcc-4.4/changes.html 6309 GCC 4.4 Release Series 6310 Changes, New Features, and Fixes 6311 6312 The latest release in the 4.4 release series is [1]GCC 4.4.7. 6313 6314 Caveats 6315 6316 * __builtin_stdarg_start has been completely removed from GCC. 6317 Support for <varargs.h> had been deprecated since GCC 4.0. Use 6318 __builtin_va_start as a replacement. 6319 * Some of the errors issued by the C++ front end that could be 6320 downgraded to warnings in previous releases by using -fpermissive 6321 are now warnings by default. They can be converted into errors by 6322 using -pedantic-errors. 6323 * Use of the cpp assertion extension will now emit a warning when 6324 -Wdeprecated or -pedantic is used. This extension has been 6325 deprecated for many years, but never warned about. 6326 * Packed bit-fields of type char were not properly bit-packed on many 6327 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 6328 causes an ABI change. For example there is no longer a 4-bit 6329 padding between field a and b in this structure: 6330 struct foo 6331 { 6332 char a:4; 6333 char b:8; 6334 } __attribute__ ((packed)); 6335 There is a new warning to help identify fields that are affected: 6336 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 6337 The warning can be disabled with -Wno-packed-bitfield-compat. 6338 * On ARM EABI targets, the C++ mangling of the va_list type has been 6339 changed to conform to the current revision of the EABI. This does 6340 not affect the libstdc++ library included with GCC. 6341 * The SCOUNT and POS bits of the MIPS DSP control register are now 6342 treated as global. Previous versions of GCC treated these fields as 6343 call-clobbered instead. 6344 * The MIPS port no longer recognizes the h asm constraint. It was 6345 necessary to remove this constraint in order to avoid generating 6346 unpredictable code sequences. 6347 One of the main uses of the h constraint was to extract the high 6348 part of a multiplication on 64-bit targets. For example: 6349 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 6350 You can now achieve the same effect using 128-bit types: 6351 typedef unsigned int uint128_t __attribute__((mode(TI))); 6352 result = ((uint128_t) x * y) >> 64; 6353 The second sequence is better in many ways. For example, if x and y 6354 are constants, the compiler can perform the multiplication at 6355 compile time. If x and y are not constants, the compiler can 6356 schedule the runtime multiplication better than it can schedule an 6357 asm statement. 6358 * Support for a number of older systems and recently unmaintained or 6359 untested target ports of GCC has been declared obsolete in GCC 4.4. 6360 Unless there is activity to revive them, the next release of GCC 6361 will have their sources permanently removed. 6362 The following ports for individual systems on particular 6363 architectures have been obsoleted: 6364 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 6365 m68k-*-aout*) 6366 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 6367 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 6368 sh-*-*). This does not affect other more specific targets 6369 using the COFF object format on those architectures, or the 6370 more specific H8300 and SH targets (h8300-*-rtems*, 6371 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 6372 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 6373 + 2BSD on PDP-11 (pdp11-*-bsd) 6374 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 6375 powerpc-ibm-aix4.[12]*) 6376 + Tuning support for Itanium1 (Merced) variants. Note that code 6377 tuned for Itanium2 should also run correctly on Itanium1. 6378 * The protoize and unprotoize utilities have been obsoleted and will 6379 be removed in GCC 4.5. These utilities have not been installed by 6380 default since GCC 3.0. 6381 * Support has been removed for all the [2]configurations obsoleted in 6382 GCC 4.3. 6383 * Unknown -Wno-* options are now silently ignored by GCC if no other 6384 diagnostics are issued. If other diagnostics are issued, then GCC 6385 warns about the unknown options. 6386 * More information on porting to GCC 4.4 from previous versions of 6387 GCC can be found in the [3]porting guide for this release. 6388 6389 General Optimizer Improvements 6390 6391 * A new command-line switch -findirect-inlining has been added. When 6392 turned on it allows the inliner to also inline indirect calls that 6393 are discovered to have known targets at compile time thanks to 6394 previous inlining. 6395 * A new command-line switch -ftree-switch-conversion has been added. 6396 This new pass turns simple initializations of scalar variables in 6397 switch statements into initializations from a static array, given 6398 that all the values are known at compile time and the ratio between 6399 the new array size and the original switch branches does not exceed 6400 the parameter --param switch-conversion-max-branch-ratio (default 6401 is eight). 6402 * A new command-line switch -ftree-builtin-call-dce has been added. 6403 This optimization eliminates unnecessary calls to certain builtin 6404 functions when the return value is not used, in cases where the 6405 calls can not be eliminated entirely because the function may set 6406 errno. This optimization is on by default at -O2 and above. 6407 * A new command-line switch -fconserve-stack directs the compiler to 6408 minimize stack usage even if it makes the generated code slower. 6409 This affects inlining decisions. 6410 * When the assembler supports it, the compiler will now emit unwind 6411 information using assembler .cfi directives. This makes it possible 6412 to use such directives in inline assembler code. The new option 6413 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 6414 directives. 6415 * The [4]Graphite branch has been merged. This merge has brought in a 6416 new framework for loop optimizations based on a polyhedral 6417 intermediate representation. These optimizations apply to all the 6418 languages supported by GCC. The following new code transformations 6419 are available in GCC 4.4: 6420 + -floop-interchange performs loop interchange transformations 6421 on loops. Interchanging two nested loops switches the inner 6422 and outer loops. For example, given a loop like: 6423 DO J = 1, M 6424 DO I = 1, N 6425 A(J, I) = A(J, I) * C 6426 ENDDO 6427 ENDDO 6428 6429 loop interchange will transform the loop as if the user had 6430 written: 6431 DO I = 1, N 6432 DO J = 1, M 6433 A(J, I) = A(J, I) * C 6434 ENDDO 6435 ENDDO 6436 6437 which can be beneficial when N is larger than the caches, 6438 because in Fortran, the elements of an array are stored in 6439 memory contiguously by column, and the original loop iterates 6440 over rows, potentially creating at each access a cache miss. 6441 + -floop-strip-mine performs loop strip mining transformations 6442 on loops. Strip mining splits a loop into two nested loops. 6443 The outer loop has strides equal to the strip size and the 6444 inner loop has strides of the original loop within a strip. 6445 For example, given a loop like: 6446 DO I = 1, N 6447 A(I) = A(I) + C 6448 ENDDO 6449 6450 loop strip mining will transform the loop as if the user had 6451 written: 6452 DO II = 1, N, 4 6453 DO I = II, min (II + 3, N) 6454 A(I) = A(I) + C 6455 ENDDO 6456 ENDDO 6457 6458 + -floop-block performs loop blocking transformations on loops. 6459 Blocking strip mines each loop in the loop nest such that the 6460 memory accesses of the element loops fit inside caches. For 6461 example, given a loop like: 6462 DO I = 1, N 6463 DO J = 1, M 6464 A(J, I) = B(I) + C(J) 6465 ENDDO 6466 ENDDO 6467 6468 loop blocking will transform the loop as if the user had 6469 written: 6470 DO II = 1, N, 64 6471 DO JJ = 1, M, 64 6472 DO I = II, min (II + 63, N) 6473 DO J = JJ, min (JJ + 63, M) 6474 A(J, I) = B(I) + C(J) 6475 ENDDO 6476 ENDDO 6477 ENDDO 6478 ENDDO 6479 6480 which can be beneficial when M is larger than the caches, 6481 because the innermost loop will iterate over a smaller amount 6482 of data that can be kept in the caches. 6483 * A new register allocator has replaced the old one. It is called 6484 integrated register allocator (IRA) because coalescing, register 6485 live range splitting, and hard register preferencing are done 6486 on-the-fly during coloring. It also has better integration with the 6487 reload pass. IRA is a regional register allocator which uses modern 6488 Chaitin-Briggs coloring instead of Chow's priority coloring used in 6489 the old register allocator. More info about IRA internals and 6490 options can be found in the GCC manuals. 6491 * A new instruction scheduler and software pipeliner, based on the 6492 selective scheduling approach, has been added. The new pass 6493 performs instruction unification, register renaming, substitution 6494 through register copies, and speculation during scheduling. The 6495 software pipeliner is able to pipeline non-countable loops. The new 6496 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 6497 it is available for the Intel Itanium platform working by default 6498 as the second scheduling pass (after register allocation) at the 6499 -O3 optimization level. 6500 * When using -fprofile-generate with a multi-threaded program, the 6501 profile counts may be slightly wrong due to race conditions. The 6502 new -fprofile-correction option directs the compiler to apply 6503 heuristics to smooth out the inconsistencies. By default the 6504 compiler will give an error message when it finds an inconsistent 6505 profile. 6506 * The new -fprofile-dir=PATH option permits setting the directory 6507 where profile data files are stored when using -fprofile-generate 6508 and friends, and the directory used when reading profile data files 6509 using -fprofile-use and friends. 6510 6511 New warning options 6512 6513 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 6514 warning if any stack frame is larger than NUMBER bytes. This may be 6515 used to help ensure that code fits within a limited amount of stack 6516 space. 6517 * The command-line option -Wlarger-than-N is now written as 6518 -Wlarger-than=N and the old form is deprecated. 6519 * The new -Wno-mudflap option disables warnings about constructs 6520 which can not be instrumented when using -fmudflap. 6521 6522 New Languages and Language specific improvements 6523 6524 * Version 3.0 of the OpenMP specification is now supported for the C, 6525 C++, and Fortran compilers. 6526 * New character data types, per [5]TR 19769: New character types in 6527 C, are now supported for the C compiler in -std=gnu99 mode, as 6528 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 6529 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 6530 6531 C family 6532 6533 * A new optimize attribute was added to allow programmers to change 6534 the optimization level and particular optimization options for an 6535 individual function. You can also change the optimization options 6536 via the GCC optimize pragma for functions defined after the pragma. 6537 The GCC push_options pragma and the GCC pop_options pragma allow 6538 you temporarily save and restore the options used. The GCC 6539 reset_options pragma restores the options to what was specified on 6540 the command line. 6541 * Uninitialized warnings do not require enabling optimization 6542 anymore, that is, -Wuninitialized can be used together with -O0. 6543 Nonetheless, the warnings given by -Wuninitialized will probably be 6544 more accurate if optimization is enabled. 6545 * -Wparentheses now warns about expressions such as (!x | y) and (!x 6546 & y). Using explicit parentheses, such as in ((!x) | y), silences 6547 this warning. 6548 * -Wsequence-point now warns within if, while,do while and for 6549 conditions, and within for begin/end expressions. 6550 * A new option -dU is available to dump definitions of preprocessor 6551 macros that are tested or expanded. 6552 6553 C++ 6554 6555 * [6]Improved experimental support for the upcoming ISO C++ standard, 6556 C++0x. Including support for auto, inline namespaces, generalized 6557 initializer lists, defaulted and deleted functions, new character 6558 types, and scoped enums. 6559 * Those errors that may be downgraded to warnings to build legacy 6560 code now mention -fpermissive when -fdiagnostics-show-option is 6561 enabled. 6562 * -Wconversion now warns if the result of a static_cast to enumeral 6563 type is unspecified because the value is outside the range of the 6564 enumeral type. 6565 * -Wuninitialized now warns if a non-static reference or non-static 6566 const member appears in a class without constructors. 6567 * G++ now properly implements value-initialization, so objects with 6568 an initializer of () and an implicitly defined default constructor 6569 will be zero-initialized before the default constructor is called. 6570 6571 Runtime Library (libstdc++) 6572 6573 * Improved experimental support for the upcoming ISO C++ standard, 6574 C++0x, including: 6575 + Support for <chrono>, <condition_variable>, <cstdatomic>, 6576 <forward_list>, <initializer_list>, <mutex>, <ratio>, 6577 <system_error>, and <thread>. 6578 + unique_ptr, <algorithm> additions, exception propagation, and 6579 support for the new character types in <string> and <limits>. 6580 + Existing facilities now exploit initializer lists, defaulted 6581 and deleted functions, and the newly implemented core C++0x 6582 features. 6583 + Some standard containers are more efficient together with 6584 stateful allocators, i.e., no allocator is constructed on the 6585 fly at element construction time. 6586 * Experimental support for non-standard pointer types in containers. 6587 * The long standing libstdc++/30928 has been fixed for targets 6588 running glibc 2.10 or later. 6589 * As usual, many small and larger bug fixes, in particular quite a 6590 few corner cases in <locale>. 6591 6592 Fortran 6593 6594 * GNU Fortran now employs libcpp directly instead of using cc1 as an 6595 external preprocessor. The [7]-cpp option was added to allow manual 6596 invocation of the preprocessor without relying on filename 6597 extensions. 6598 * The [8]-Warray-temporaries option warns about array temporaries 6599 generated by the compiler, as an aid to optimization. 6600 * The [9]-fcheck-array-temporaries option has been added, printing a 6601 notification at run time, when an array temporary had to be created 6602 for an function argument. Contrary to -Warray-temporaries the 6603 warning is only printed if the array is noncontiguous. 6604 * Improved generation of DWARF debugging symbols 6605 * If using an intrinsic not part of the selected standard (via -std= 6606 and -fall-intrinsics) gfortran will now treat it as if this 6607 procedure were declared EXTERNAL and try to link to a user-supplied 6608 procedure. -Wintrinsics-std will warn whenever this happens. The 6609 now-useless option -Wnonstd-intrinsic was removed. 6610 * The flag -falign-commons has been added to control the alignment of 6611 variables in COMMON blocks, which is enabled by default in line 6612 with previous GCC version. Using -fno-align-commons one can force 6613 commons to be contiguous in memory as required by the Fortran 6614 standard, however, this slows down the memory access. The option 6615 -Walign-commons, which is enabled by default, warns when padding 6616 bytes were added for alignment. The proper solution is to sort the 6617 common objects by decreasing storage size, which avoids the 6618 alignment problems. 6619 * Fortran 2003 support has been extended: 6620 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 6621 now supported (except internal reads from/writes to wide 6622 strings). [10]-fbackslash now supports also \unnnn and 6623 \Unnnnnnnn to enter Unicode characters. 6624 + Asynchronous I/O (implemented as synchronous I/O) and the 6625 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 6626 are now supported in I/O statements. 6627 + Support for Fortran 2003 structure constructors and for array 6628 constructor with typespec has been added. 6629 + Procedure Pointers (but not yet as component in derived types 6630 and as function results) are now supported. 6631 + Abstract types, type extension, and type-bound procedures 6632 (both PROCEDURE and GENERIC but not as operators). Note: As 6633 CLASS/polymorphyic types are not implemented, type-bound 6634 procedures with PASS accept as non-standard extension TYPE 6635 arguments. 6636 * Fortran 2008 support has been added: 6637 + The -std=f2008 option and support for the file extensions 6638 .f2008 and .F2008 has been added. 6639 + The g0 format descriptor is now supported. 6640 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 6641 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 6642 are now available (some of them existed as GNU extension 6643 before). Note: The hyperbolic functions are not yet supporting 6644 complex arguments and the three- argument version of BESSEL_*N 6645 is not available. 6646 + The bit intrinsics LEADZ and TRAILZ have been added. 6647 6648 Java (GCJ) 6649 6650 Ada 6651 6652 * The Ada runtime now supports multilibs on many platforms including 6653 x86_64, SPARC and PowerPC. Their build is enabled by default. 6654 6655 New Targets and Target Specific Improvements 6656 6657 ARM 6658 6659 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 6660 Cortex-R4F processors and has many other improvements to 6661 optimization for ARM processors. 6662 * GCC now supports the VFPv3 variant with 16 double-precision 6663 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 6664 renamed to -mfpu=vfpv3. 6665 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 6666 erratum on Cortex-M3 processors. 6667 * GCC now supports the __sync_* atomic operations for ARM EABI 6668 GNU/Linux. 6669 * The section anchors optimization is now enabled by default when 6670 optimizing for ARM. 6671 * GCC now uses a new EABI-compatible profiling interface for EABI 6672 targets. This requires a function __gnu_mcount_nc, which is 6673 provided by GNU libc versions 2.8 and later. 6674 6675 AVR 6676 6677 * The -mno-tablejump option has been deprecated because it has the 6678 same effect as the -fno-jump-tables option. 6679 * Added support for these new AVR devices: 6680 + ATA6289 6681 + ATtiny13A 6682 + ATtiny87 6683 + ATtiny167 6684 + ATtiny327 6685 + ATmega8C1 6686 + ATmega16C1 6687 + ATmega32C1 6688 + ATmega8M1 6689 + ATmega16M1 6690 + ATmega32M1 6691 + ATmega32U4 6692 + ATmega16HVB 6693 + ATmega4HVD 6694 + ATmega8HVD 6695 + ATmega64C1 6696 + ATmega64M1 6697 + ATmega16U4 6698 + ATmega32U6 6699 + ATmega128RFA1 6700 + AT90PWM81 6701 + AT90SCR100 6702 + M3000F 6703 + M3000S 6704 + M3001B 6705 6706 IA-32/x86-64 6707 6708 * Support for Intel AES built-in functions and code generation is 6709 available via -maes. 6710 * Support for Intel PCLMUL built-in function and code generation is 6711 available via -mpclmul. 6712 * Support for Intel AVX built-in functions and code generation is 6713 available via -mavx. 6714 * Automatically align the stack for local variables with alignment 6715 requirement. 6716 * GCC can now utilize the SVML library for vectorizing calls to a set 6717 of C99 functions if -mveclibabi=svml is specified and you link to 6718 an SVML ABI compatible library. 6719 * On x86-64, the ABI has been changed in the following cases to 6720 conform to the x86-64 ABI: 6721 + Passing/returning structures with flexible array member: 6722 struct foo 6723 { 6724 int i; 6725 int flex[]; 6726 }; 6727 + Passing/returning structures with complex float member: 6728 struct foo 6729 { 6730 int i; 6731 __complex__ float f; 6732 }; 6733 + Passing/returning unions with long double member: 6734 union foo 6735 { 6736 int x; 6737 long double ld; 6738 }; 6739 Code built with previous versions of GCC that uses any of these is 6740 not compatible with code built with GCC 4.4.0 or later. 6741 * A new target attribute was added to allow programmers to change the 6742 target options like -msse2 or -march=k8 for an individual function. 6743 You can also change the target options via the GCC target pragma 6744 for functions defined after the pragma. 6745 * GCC can now be configured with options --with-arch-32, 6746 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 6747 --with-tune-64 to control the default optimization separately for 6748 32-bit and 64-bit modes. 6749 6750 IA-32/IA64 6751 6752 * Support for __float128 (TFmode) IEEE quad type and corresponding 6753 TCmode IEEE complex quad type is available via the soft-fp library 6754 on IA-32/IA64 targets. This includes basic arithmetic operations 6755 (addition, subtraction, negation, multiplication and division) on 6756 __float128 real and TCmode complex values, the full set of IEEE 6757 comparisons between __float128 values, conversions to and from 6758 float, double and long double floating point types, as well as 6759 conversions to and from signed or unsigned integer, signed or 6760 unsigned long integer and signed or unsigned quad (TImode, IA64 6761 only) integer types. Additionally, all operations generate the full 6762 set of IEEE exceptions and support the full set of IEEE rounding 6763 modes. 6764 6765 M68K/ColdFire 6766 6767 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 6768 processors. (Scheduling support for ColdFire V2 processors was 6769 added in GCC 4.3.) 6770 * GCC now supports the -mxgot option to support programs requiring 6771 many GOT entries on ColdFire. 6772 * The m68k-*-linux-gnu target now builds multilibs by default. 6773 6774 MIPS 6775 6776 * MIPS Technologies have extended the original MIPS SVR4 ABI to 6777 include support for procedure linkage tables (PLTs) and copy 6778 relocations. These extensions allow GNU/Linux executables to use a 6779 significantly more efficient code model than the one defined by the 6780 original ABI. 6781 GCC support for this code model is available via a new command-line 6782 option, -mplt. There is also a new configure-time option, 6783 --with-mips-plt, to make -mplt the default. 6784 The new code model requires support from the assembler, the linker, 6785 and the runtime C library. This support is available in binutils 6786 2.19 and GLIBC 2.9. 6787 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 6788 and 32-bit GNU/Linux shared libraries. This feature requires GNU 6789 binutils 2.19 or above. 6790 * Support for RMI's XLR processor is now available through the 6791 -march=xlr and -mtune=xlr options. 6792 * 64-bit targets can now perform 128-bit multiplications inline, 6793 instead of relying on a libgcc function. 6794 * Native GNU/Linux toolchains now support -march=native and 6795 -mtune=native, which select the host processor. 6796 * GCC now supports the R10K, R12K, R14K and R16K processors. The 6797 canonical -march= and -mtune= names for these processors are 6798 r10000, r12000, r14000 and r16000 respectively. 6799 * GCC can now work around the side effects of speculative execution 6800 on R10K processors. Please see the documentation of the 6801 -mr10k-cache-barrier option for details. 6802 * Support for the MIPS64 Release 2 instruction set has been added. 6803 The option -march=mips64r2 enables generation of these 6804 instructions. 6805 * GCC now supports Cavium Networks' Octeon processor. This support is 6806 available through the -march=octeon and -mtune=octeon options. 6807 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 6808 canonical -march= and -mtune= names for these processors are 6809 loongson2e and loongson2f. 6810 6811 picochip 6812 6813 Picochip is a 16-bit processor. A typical picoChip contains over 250 6814 small cores, each with small amounts of memory. There are three 6815 processor variants (STAN, MEM and CTRL) with different instruction sets 6816 and memory configurations and they can be chosen using the -mae option. 6817 6818 This port is intended to be a "C" only port. 6819 6820 Power Architecture and PowerPC 6821 6822 * GCC now supports the e300c2, e300c3 and e500mc processors. 6823 * GCC now supports Xilinx processors with a single-precision FPU. 6824 * Decimal floating point is now supported for e500 processors. 6825 6826 S/390, zSeries and System z9/z10 6827 6828 * Support for the IBM System z10 EC/BC processor has been added. When 6829 using the -march=z10 option, the compiler will generate code making 6830 use of instructions provided by the General-Instruction-Extension 6831 Facility and the Execute-Extension Facility. 6832 6833 VxWorks 6834 6835 * GCC now supports the thread-local storage mechanism used on 6836 VxWorks. 6837 6838 Xtensa 6839 6840 * GCC now supports thread-local storage (TLS) for Xtensa processor 6841 configurations that include the Thread Pointer option. TLS also 6842 requires support from the assembler and linker; this support is 6843 provided in the GNU binutils beginning with version 2.19. 6844 6845 Documentation improvements 6846 6847 Other significant improvements 6848 6849 GCC 4.4.1 6850 6851 This is the [11]list of problem reports (PRs) from GCC's bug tracking 6852 system that are known to be fixed in the 4.4.1 release. This list might 6853 not be complete (that is, it is possible that some PRs that have been 6854 fixed are not listed here). 6855 6856 GCC 4.4.2 6857 6858 This is the [12]list of problem reports (PRs) from GCC's bug tracking 6859 system that are known to be fixed in the 4.4.2 release. This list might 6860 not be complete (that is, it is possible that some PRs that have been 6861 fixed are not listed here). 6862 6863 GCC 4.4.3 6864 6865 This is the [13]list of problem reports (PRs) from GCC's bug tracking 6866 system that are known to be fixed in the 4.4.3 release. This list might 6867 not be complete (that is, it is possible that some PRs that have been 6868 fixed are not listed here). 6869 6870 GCC 4.4.4 6871 6872 This is the [14]list of problem reports (PRs) from GCC's bug tracking 6873 system that are known to be fixed in the 4.4.4 release. This list might 6874 not be complete (that is, it is possible that some PRs that have been 6875 fixed are not listed here). 6876 6877 GCC 4.4.5 6878 6879 This is the [15]list of problem reports (PRs) from GCC's bug tracking 6880 system that are known to be fixed in the 4.4.5 release. This list might 6881 not be complete (that is, it is possible that some PRs that have been 6882 fixed are not listed here). 6883 6884 GCC 4.4.6 6885 6886 This is the [16]list of problem reports (PRs) from GCC's bug tracking 6887 system that are known to be fixed in the 4.4.6 release. This list might 6888 not be complete (that is, it is possible that some PRs that have been 6889 fixed are not listed here). 6890 6891 GCC 4.4.7 6892 6893 This is the [17]list of problem reports (PRs) from GCC's bug tracking 6894 system that are known to be fixed in the 4.4.7 release. This list might 6895 not be complete (that is, it is possible that some PRs that have been 6896 fixed are not listed here). 6897 6898 6899 For questions related to the use of GCC, please consult these web 6900 pages and the [18]GCC manuals. If that fails, the 6901 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6902 web pages and the development of GCC are welcome on our developer 6903 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 6904 archives. 6905 6906 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 6907 distribution of this entire article is permitted in any medium, 6908 provided this notice is preserved. 6909 6910 These pages are [23]maintained by the GCC team. Last modified 6911 2018-09-30[24]. 6912 6913 References 6914 6915 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 6916 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 6917 3. http://gcc.gnu.org/gcc-4.4/porting_to.html 6918 4. https://gcc.gnu.org/wiki/Graphite 6919 5. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 6920 6. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html 6921 7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 6922 8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 6923 9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 6924 10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 6925 11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 6926 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 6927 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 6928 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 6929 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 6930 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 6931 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 6932 18. https://gcc.gnu.org/onlinedocs/ 6933 19. mailto:gcc-help (a] gcc.gnu.org 6934 20. mailto:gcc (a] gcc.gnu.org 6935 21. https://gcc.gnu.org/lists.html 6936 22. https://www.fsf.org/ 6937 23. https://gcc.gnu.org/about.html 6938 24. http://validator.w3.org/check/referer 6939 ====================================================================== 6940 http://gcc.gnu.org/gcc-4.3/index.html 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 2018-09-30[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. https://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 GCC 4.3 Release Series 7036 Changes, New Features, and Fixes 7037 7038 The latest release in the 4.3 release series is [1]GCC 4.3.5. 7039 7040 Caveats 7041 7042 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 7043 various front-end languages it supports. See the [4]prerequisites 7044 page for version requirements. 7045 * ColdFire targets now treat long double as having the same format as 7046 double. In earlier versions of GCC, they used the 68881 long double 7047 format instead. 7048 * The m68k-uclinux target now uses the same calling conventions as 7049 m68k-linux-gnu. You can select the original calling conventions by 7050 configuring for m68k-uclinuxoldabi instead. Note that 7051 m68k-uclinuxoldabi also retains the original 80-bit long double on 7052 ColdFire targets. 7053 * The -fforce-mem option has been removed because it has had no 7054 effect in the last few GCC releases. 7055 * The i386 -msvr3-shlib option has been removed since it is no longer 7056 used. 7057 * Fastcall for i386 has been changed not to pass aggregate arguments 7058 in registers, following Microsoft compilers. 7059 * Support for the AOF assembler has been removed from the ARM back 7060 end; this affects only the targets arm-semi-aof and armel-semi-aof, 7061 which are no longer recognized. We removed these targets without a 7062 deprecation period because we discovered that they have been 7063 unusable since GCC 4.0.0. 7064 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 7065 has been removed. This support had been deprecated since GCC 4.0.0. 7066 * Support for a number of older systems and recently unmaintained or 7067 untested target ports of GCC has been declared obsolete in GCC 4.3. 7068 Unless there is activity to revive them, the next release of GCC 7069 will have their sources permanently removed. 7070 All GCC ports for the following processor architectures have been 7071 declared obsolete: 7072 + Morpho MT (mt-*) 7073 The following aliases for processor architectures have been 7074 declared obsolete. Users should use the indicated generic target 7075 names instead, with compile-time options such as -mcpu or 7076 configure-time options such as --with-cpu to control the 7077 configuration more precisely. 7078 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 7079 instead). 7080 + parisc*-*-* (use hppa*-*-* instead). 7081 + m680[012]0-*-* (use m68k-*-* instead). 7082 All GCC ports for the following operating systems have been 7083 declared obsolete: 7084 + BeOS (*-*-beos*) 7085 + kaOS (*-*-kaos*) 7086 + GNU/Linux using the a.out object format (*-*-linux*aout*) 7087 + GNU/Linux using version 1 of the GNU C Library 7088 (*-*-linux*libc1*) 7089 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 7090 *-*-solaris2.[0-6].*) 7091 + Miscellaneous System V (*-*-sysv*) 7092 + WindISS (*-*-windiss*) 7093 Also, those for some individual systems on particular architectures 7094 have been obsoleted: 7095 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 7096 + CRIS with a.out object format (cris-*-aout) 7097 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 7098 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 7099 + PRO on PA-RISC (hppa1.1-*-pro*) 7100 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 7101 i[34567]86-sequent-sysv4*) 7102 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 7103 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 7104 was previously [5]removed in 2001, leaving only the support 7105 for UWIN as a target now being deprecated) 7106 + ChorusOS on PowerPC (powerpc-*-chorusos*) 7107 + All VAX configurations apart from NetBSD and OpenBSD 7108 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 7109 * The [6]-Wconversion option has been modified. Its purpose now is to 7110 warn for implicit conversions that may alter a value. This new 7111 behavior is available for both C and C++. Warnings about 7112 conversions between signed and unsigned integers can be disabled by 7113 using -Wno-sign-conversion. In C++, they are disabled by default 7114 unless -Wsign-conversion is explicitly requested. The old behavior 7115 of -Wconversion, that is, warn for prototypes causing a type 7116 conversion that is different from what would happen to the same 7117 argument in the absence of a prototype, has been moved to a new 7118 option -Wtraditional-conversion, which is only available for C. 7119 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 7120 been removed because they were deprecated for more than 3 GCC major 7121 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 7122 -mtune=pentiumpro as a replacement. 7123 * The -funsafe-math-optimizations option now automatically turns on 7124 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 7125 reassociation and thus may introduce or remove traps. 7126 * The -ftree-vectorize option is now on by default under -O3. In 7127 order to generate code for a SIMD extension, it has to be enabled 7128 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 7129 i?86 and x86_64. 7130 * More information on porting to GCC 4.3 from previous versions of 7131 GCC can be found in the [7]porting guide for this release. 7132 7133 General Optimizer Improvements 7134 7135 * The GCC middle-end has been integrated with the MPFR library. This 7136 allows GCC to evaluate and replace at compile-time calls to 7137 built-in math functions having constant arguments with their 7138 mathematically equivalent results. In making use of MPFR, GCC can 7139 generate correct results regardless of the math library 7140 implementation or floating point precision of the host platform. 7141 This also allows GCC to generate identical results regardless of 7142 whether one compiles in native or cross-compile configurations to a 7143 particular target. The following built-in functions take advantage 7144 of this new capability: acos, acosh, asin, asinh, atan2, atan, 7145 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 7146 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 7147 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 7148 tan, tanh, tgamma, y0, y1 and yn. The float and long double 7149 variants of these functions (e.g. sinf and sinl) are also handled. 7150 The sqrt and cabs functions with constant arguments were already 7151 optimized in prior GCC releases. Now they also use MPFR. 7152 * A new forward propagation pass on RTL was added. The new pass 7153 replaces several slower transformations, resulting in compile-time 7154 improvements as well as better code generation in some cases. 7155 * A new command-line switch -frecord-gcc-switches has been added to 7156 GCC, although it is only enabled for some targets. The switch 7157 causes the command line that was used to invoke the compiler to be 7158 recorded into the object file that is being created. The exact 7159 format of this recording is target and binary file format 7160 dependent, but it usually takes the form of a note section 7161 containing ASCII text. The switch is related to the -fverbose-asm 7162 switch, but that one only records the information in the assembler 7163 output file as comments, so the information never reaches the 7164 object file. 7165 * The inliner heuristic is now aware of stack frame consumption. New 7166 command-line parameters --param large-stack-frame and --param 7167 large-stack-frame-growth can be used to limit stack frame size 7168 growth caused by inlining. 7169 * During feedback directed optimizations, the expected block size the 7170 memcpy, memset and bzero functions operate on is discovered and for 7171 cases of commonly used small sizes, specialized inline code is 7172 generated. 7173 * __builtin_expect no longer requires its argument to be a compile 7174 time constant. 7175 * Interprocedural optimization was reorganized to work on functions 7176 in SSA form. This enables more precise and cheaper dataflow 7177 analysis and makes writing interprocedural optimizations easier. 7178 The following improvements have been implemented on top of this 7179 framework: 7180 + Pre-inline optimization: Selected local optimization passes 7181 are run before the inliner (and other interprocedural passes) 7182 are executed. This significantly improves the accuracy of code 7183 growth estimates used by the inliner and reduces the overall 7184 memory footprint for large compilation units. 7185 + Early inlining (a simple bottom-up inliner pass inlining only 7186 functions whose body is smaller than the expected call 7187 overhead) is now executed with the early optimization passes, 7188 thus inlining already optimized function bodies into an 7189 unoptimized function that is subsequently optimized by early 7190 optimizers. This enables the compiler to quickly eliminate 7191 abstraction penalty in C++ programs. 7192 + Interprocedural constant propagation now operate on SSA form 7193 increasing accuracy of the analysis. 7194 * A new internal representation for GIMPLE statements has been 7195 contributed, resulting in compile-time memory savings. 7196 * The vectorizer was enhanced to support vectorization of outer 7197 loops, intra-iteration parallelism (loop-aware SLP), vectorization 7198 of strided accesses and loops with multiple data-types. Run-time 7199 dependency testing using loop versioning was added. The cost model, 7200 turned on by -fvect-cost-model, was developed. 7201 7202 New Languages and Language specific improvements 7203 7204 * We have added new command-line options 7205 -finstrument-functions-exclude-function-list and 7206 -finstrument-functions-exclude-file-list. They provide more control 7207 over which functions are annotated by the -finstrument-functions 7208 option. 7209 7210 C family 7211 7212 * Implicit conversions between generic vector types are now only 7213 permitted when the two vectors in question have the same number of 7214 elements and compatible element types. (Note that the restriction 7215 involves compatible element types, not implicitly-convertible 7216 element types: thus, a vector type with element type int may not be 7217 implicitly converted to a vector type with element type unsigned 7218 int.) This restriction, which is in line with specifications for 7219 SIMD architectures such as AltiVec, may be relaxed using the flag 7220 -flax-vector-conversions. This flag is intended only as a 7221 compatibility measure and should not be used for new code. 7222 * -Warray-bounds has been added and is now enabled by default for 7223 -Wall . It produces warnings for array subscripts that can be 7224 determined at compile time to be always out of bounds. 7225 -Wno-array-bounds will disable the warning. 7226 * The constructor and destructor function attributes now accept 7227 optional priority arguments which control the order in which the 7228 constructor and destructor functions are run. 7229 * New [8]command-line options -Wtype-limits, -Wold-style-declaration, 7230 -Wmissing-parameter-type, -Wempty-body, -Wclobbered and 7231 -Wignored-qualifiers have been added for finer control of the 7232 diverse warnings enabled by -Wextra. 7233 * A new function attribute alloc_size has been added to mark up 7234 malloc style functions. For constant sized allocations this can be 7235 used to find out the size of the returned pointer using the 7236 __builtin_object_size() function for buffer overflow checking and 7237 similar. This supplements the already built-in malloc and calloc 7238 constant size handling. 7239 * Integer constants written in binary are now supported as a GCC 7240 extension. They consist of a prefix 0b or 0B, followed by a 7241 sequence of 0 and 1 digits. 7242 * A new predefined macro __COUNTER__ has been added. It expands to 7243 sequential integral values starting from 0. In conjunction with the 7244 ## operator, this provides a convenient means to generate unique 7245 identifiers. 7246 * A new command-line option -fdirectives-only has been added. It 7247 enables a special preprocessing mode which improves the performance 7248 of applications like distcc and ccache. 7249 * Fixed-point data types and operators have been added. They are 7250 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 7251 Currently, only MIPS targets are supported. 7252 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 7253 N1241, is now supported as a GCC extension to C for targets 7254 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 7255 and x86_64-*-linux-gnu. The feature introduces new data types 7256 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 7257 DD, and DL. 7258 7259 C++ 7260 7261 * [9]Experimental support for the upcoming ISO C++ standard, C++0x. 7262 * -Wc++0x-compat has been added and is now enabled by default for 7263 -Wall. It produces warnings for constructs whose meaning differs 7264 between ISO C++ 1998 and C++0x. 7265 * The -Wparentheses option now works for C++ as it does for C. It 7266 warns if parentheses are omitted when operators with confusing 7267 precedence are nested. It also warns about ambiguous else 7268 statements. Since -Wparentheses is enabled by -Wall, this may cause 7269 additional warnings with existing C++ code which uses -Wall. These 7270 new warnings may be disabled by using -Wall -Wno-parentheses. 7271 * The -Wmissing-declarations now works for C++ as it does for C. 7272 * The -fvisibility-ms-compat flag was added, to make it easier to 7273 port larger projects using shared libraries from Microsoft's Visual 7274 Studio to ELF and Mach-O systems. 7275 * C++ attribute handling has been overhauled for template arguments 7276 (ie dependent types). In particular, __attribute__((aligned(T))); 7277 works for C++ types. 7278 7279 Runtime Library (libstdc++) 7280 7281 * [10]Experimental support for the upcoming ISO C++ standard, C++0x. 7282 * Support for TR1 mathematical special functions and regular 7283 expressions. 7284 * Default what implementations give more elaborate exception strings 7285 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 7286 * Header dependencies have been streamlined, reducing unnecessary 7287 includes and pre-processed bloat. 7288 * Variadic template implementations of items in <tuple> and 7289 <functional>. 7290 * An experimental [11]parallel mode has been added. This is a 7291 parallel implementation of many C++ Standard library algorithms, 7292 like std::accumulate, std::for_each, std::transform, or std::sort, 7293 to give but four examples. These algorithms can be substituted for 7294 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 7295 or all existing algorithms can be transformed via the 7296 -D_GLIBCXX_PARALLEL macro. 7297 * Debug mode versions of classes in <unordered_set> and 7298 <unordered_map>. 7299 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 7300 now <backward/hash_set> and <backward/hash_map>. This code: 7301 #include <ext/hash_set> 7302 __gnu_cxx::hash_set<int> s; 7303 7304 Can be transformed (in order of preference) to: 7305 #include <tr1/unordered_set> 7306 std::tr1::unordered_set<int> s; 7307 7308 or 7309 #include <backward/hash_set> 7310 __gnu_cxx::hash_set<int> s; 7311 7312 Similar transformations apply to __gnu_cxx::hash_map, 7313 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 7314 __gnu_cxx::hash_multiset. 7315 7316 Fortran 7317 7318 * Due to the fact that the GMP and MPFR libraries are required for 7319 all languages, Fortran is no longer special in this regard and is 7320 available by default. 7321 * The [12]-fexternal-blas option has been added, which generates 7322 calls to BLAS routines for intrinsic matrix operations such as 7323 matmul rather than using the built-in algorithms. 7324 * Support to give a backtrace (compiler flag -fbacktrace or 7325 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 7326 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 7327 run-time error occured. 7328 * GNU Fortran now defines __GFORTRAN__ when it runs the C 7329 preprocessor (CPP). 7330 * The [13]-finit-local-zero, -finit-real, -finit-integer, 7331 -finit-character, and -finit-logical options have been added, which 7332 can be used to initialize local variables. 7333 * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added, 7334 which calculate the Gamma function and its logarithm. Use EXTERNAL 7335 gamma if you want to use your own gamma function. 7336 * GNU Fortran now regards the backslash character as literal (as 7337 required by the Fortran 2003 standard); using [16]-fbackslash GNU 7338 Fortran interprets backslashes as C-style escape characters. 7339 * The [17]interpretation of binary, octal and hexadecimal (BOZ) 7340 literal constants has been changed. Before they were always 7341 interpreted as integer; now they are bit-wise transferred as 7342 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 7343 2003 standard, and for real and complex variables in DATA 7344 statements or when directly assigned to real and complex variables. 7345 Everywhere else and especially in expressions they are still 7346 regarded as integer constants. 7347 * Fortran 2003 support has been extended: 7348 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 7349 + Pointer intent 7350 + Intrinsic module ISO_ENV_FORTRAN 7351 + Interoperability with C (ISO C Bindings) 7352 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 7353 attribute) 7354 + Fortran 2003 BOZ 7355 7356 Java (GCJ) 7357 7358 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 7359 This enables the use of all 1.5 language features, and fixes most 7360 existing front end bugs. 7361 * libgcj now supports all 1.5 language features which require runtime 7362 support: foreach, enum, annotations, generics, and auto-boxing. 7363 * We've made many changes to the tools shipped with gcj. 7364 + The old jv-scan tool has been removed. This tool never really 7365 worked properly. There is no replacement. 7366 + gcjh has been rewritten. Some of its more obscure options no 7367 longer work, but are still recognized in an attempt at 7368 compatibility. gjavah is a new program with similar 7369 functionality but different command-line options. 7370 + grmic and grmiregistry have been rewritten. grmid has been 7371 added. 7372 + gjar replaces the old fastjar. 7373 + gjarsigner (used for signing jars), gkeytool (used for key 7374 management), gorbd (for CORBA), gserialver (computes 7375 serialization UIDs), and gtnameserv (also for CORBA) are now 7376 installed. 7377 * The ability to dump the contents of the java run time heap to a 7378 file for off-line analysis has been added. The heap dumps may be 7379 analyzed with the new gc-analyze tool. They may be generated on 7380 out-of-memory conditions or on demand and are controlled by the new 7381 run time class gnu.gcj.util.GCInfo. 7382 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 7383 provide correct, updated, timezone information. This means that 7384 packagers no longer have to update libgcj when a time zone change 7385 is published. 7386 7387 New Targets and Target Specific Improvements 7388 7389 IA-32/x86-64 7390 7391 * Tuning for Intel Core 2 processors is available via -mtune=core2 7392 and -march=core2. 7393 * Tuning for AMD Geode processors is available via -mtune=geode and 7394 -march=geode. 7395 * Code generation of block move (memcpy) and block set (memset) was 7396 rewritten. GCC can now pick the best algorithm (loop, unrolled 7397 loop, instruction with rep prefix or a library call) based on the 7398 size of the block being copied and the CPU being optimized for. A 7399 new option -minline-stringops-dynamically has been added. With this 7400 option string operations of unknown size are expanded such that 7401 small blocks are copied by in-line code, while for large blocks a 7402 library call is used. This results in faster code than 7403 -minline-all-stringops when the library implementation is capable 7404 of using cache hierarchy hints. The heuristic choosing the 7405 particular algorithm can be overwritten via -mstringop-strategy. 7406 Newly also memset of values different from 0 is inlined. 7407 * GCC no longer places the cld instruction before string operations. 7408 Both i386 and x86-64 ABI documents mandate the direction flag to be 7409 clear at the entry of a function. It is now invalid to set the flag 7410 in asm statement without reseting it afterward. 7411 * Support for SSSE3 built-in functions and code generation are 7412 available via -mssse3. 7413 * Support for SSE4.1 built-in functions and code generation are 7414 available via -msse4.1. 7415 * Support for SSE4.2 built-in functions and code generation are 7416 available via -msse4.2. 7417 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 7418 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 7419 allow explicit control of x87 floating point precision. 7420 * Support for __float128 (TFmode) IEEE quad type and corresponding 7421 TCmode IEEE complex quad type is available via the soft-fp library 7422 on x86_64 targets. This includes basic arithmetic operations 7423 (addition, subtraction, negation, multiplication and division) on 7424 __float128 real and TCmode complex values, the full set of IEEE 7425 comparisons between __float128 values, conversions to and from 7426 float, double and long double floating point types, as well as 7427 conversions to and from signed or unsigned integer, signed or 7428 unsigned long integer and signed or unsigned quad (TImode) integer 7429 types. Additionally, all operations generate the full set of IEEE 7430 exceptions and support the full set of IEEE rounding modes. 7431 * GCC can now utilize the ACML library for vectorizing calls to a set 7432 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 7433 link to an ACML ABI compatible library. 7434 7435 ARM 7436 7437 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 7438 has been added. 7439 7440 CRIS 7441 7442 New features 7443 7444 * Compiler and Library support for the CRIS v32 architecture, as 7445 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 7446 added. 7447 7448 Configuration changes 7449 7450 * The cris-*-elf target now includes support for CRIS v32, including 7451 libraries, through the -march=v32 option. 7452 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 7453 * A new crisv32-*-linux* target defaults to generate code for CRIS 7454 v32. 7455 * The cris-*-aout target has been obsoleted. 7456 7457 Improved support for built-in functions 7458 7459 * GCC can now use the lz and swapwbr instructions to implement the 7460 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 7461 * __builtin_bswap32 is now implemented using the swapwb instruction, 7462 when available. 7463 7464 m68k and ColdFire 7465 7466 New features 7467 7468 * Support for several new ColdFire processors has been added. You can 7469 generate code for them using the new -mcpu option. 7470 * All targets now support ColdFire processors. 7471 * m68k-uclinux targets have improved support for C++ constructors and 7472 destructors, and for shared libraries. 7473 * It is now possible to set breakpoints on the first or last line of 7474 a function, even if there are no statements on that line. 7475 7476 Optimizations 7477 7478 * Support for sibling calls has been added. 7479 * More use is now made of the ColdFire mov3q instruction. 7480 * __builtin_clz is now implemented using the ff1 ColdFire 7481 instruction, when available. 7482 * GCC now honors the -m68010 option. 68010 code now uses clr rather 7483 than move to zero volatile memory. 7484 * 68020 targets and above can now use symbol(index.size*scale) 7485 addresses for indexed array accesses. Earlier compilers would 7486 always load the symbol into a base register first. 7487 7488 Configuration changes 7489 7490 * All m68k and ColdFire targets now allow the default processor to be 7491 set at configure time using --with-cpu. 7492 * A --with-arch configuration option has been added. This option 7493 allows you to restrict a target to ColdFire or non-ColdFire 7494 processors. 7495 7496 Preprocessor macros 7497 7498 * An __mcfv*__ macro is now defined for all ColdFire targets. 7499 (Earlier versions of GCC only defined __mcfv4e__.) 7500 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 7501 * All targets now define __mc68010 and __mc68010__ when generating 7502 68010 code. 7503 7504 Command-line changes 7505 7506 * New command-line options -march, -mcpu, -mtune and -mhard-float 7507 have been added. These options apply to both m68k and ColdFire 7508 targets. 7509 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 7510 versions of -mshort, etc. 7511 * -fforce-addr has been removed. It is now ignored by the compiler. 7512 7513 Other improvements 7514 7515 * ColdFire targets now try to maintain a 4-byte-aligned stack where 7516 possible. 7517 * m68k-uclinux targets now try to avoid situations that lead to the 7518 load-time error: BINFMT_FLAT: reloc outside program. 7519 7520 MIPS 7521 7522 Changes to existing configurations 7523 7524 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 7525 and n64. Every GNU/Linux configuration now builds these libraries 7526 by default. 7527 * GNU/Linux configurations now generate -mno-shared code unless 7528 overridden by -fpic, -fPIC, -fpie or -fPIE. 7529 * mipsisa32*-linux-gnu configurations now generate hard-float code by 7530 default, just like other mipsisa32* and mips*-linux-gnu 7531 configurations. You can build a soft-float version of any 7532 mips*-linux-gnu configuration by passing --with-float=soft to 7533 configure. 7534 * mips-wrs-vxworks now supports run-time processes (RTPs). 7535 7536 Changes to existing command-line options 7537 7538 * The -march and -mtune options no longer accept 24k as a processor 7539 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 7540 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 7541 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 7542 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 7543 24kx, 24kex and 34kx. 7544 7545 New configurations 7546 7547 GCC now supports the following configurations: 7548 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 7549 default. Earlier releases also recognized this configuration, but 7550 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 7551 you can customize any mips*-linux-gnu* configuration to a 7552 particular ISA or processor by passing an appropriate --with-arch 7553 option to configure. 7554 * mipsisa*-sde-elf*, which provides compatibility with MIPS 7555 Technologies' SDE toolchains. The configuration uses the SDE 7556 libraries by default, but you can use it like other newlib-based 7557 ELF configurations by passing --with-newlib to configure. It is the 7558 only configuration besides mips64vr*-elf* to build MIPS16 as well 7559 as non-MIPS16 libraries. 7560 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 7561 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 7562 and 64-bit forms of the EABI. 7563 7564 New processors and application-specific extensions 7565 7566 * Support for the SmartMIPS ASE is available through the new 7567 -msmartmips option. 7568 * Support for revision 2 of the DSP ASE is available through the new 7569 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 7570 indicates the revision of the ASE in use. 7571 * Support for the 4KS and 74K families of processors is available 7572 through the -march and -mtune options. 7573 7574 Improved support for built-in functions 7575 7576 * GCC can now use load-linked, store-conditional and sync 7577 instructions to implement atomic built-in functions such as 7578 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 7579 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 7580 * GCC can now use the clz and dclz instructions to implement the 7581 __builtin_ctz and __builtin_ffs families of functions. 7582 * There is a new __builtin___clear_cache function for flushing the 7583 instruction cache. GCC expands this function inline on MIPS32 7584 revision 2 targets, otherwise it calls the function specified by 7585 -mcache-flush-func. 7586 7587 MIPS16 improvements 7588 7589 * GCC can now compile objects that contain a mixture of MIPS16 and 7590 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 7591 for specifying which mode a function should use. 7592 * A new option called -minterlink-mips16 makes non-MIPS16 code 7593 link-compatible with MIPS16 code. 7594 * After many bug fixes, the long-standing MIPS16 -mhard-float support 7595 should now work fairly reliably. 7596 * GCC can now use the MIPS16e save and restore instructions. 7597 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 7598 with -G0 -fsection-anchors is often smaller than code compiled with 7599 -G8. However, please note that you must usually compile all objects 7600 in your application with the same -G option; see the documentation 7601 of -G for details. 7602 * A new option called-mcode-readable specifies which instructions are 7603 allowed to load from the code segment. -mcode-readable=yes is the 7604 default and says that any instruction may load from the code 7605 segment. The other alternatives are -mcode-readable=pcrel, which 7606 says that only PC-relative MIPS16 instructions may load from the 7607 code segment, and -mcode-readable=no, which says that no 7608 instruction may do so. Please see the documentation for more 7609 details, including example uses. 7610 7611 Small-data improvements 7612 7613 There are three new options for controlling small data: 7614 * -mno-extern-sdata, which disables small-data accesses for 7615 externally-defined variables. Code compiled with -Gn 7616 -mno-extern-sdata will be link-compatible with any -G setting 7617 between -G0 and -Gn inclusive. 7618 * -mno-local-sdata, which disables the use of small-data sections for 7619 data that is not externally visible. This option can be a useful 7620 way of reducing small-data usage in less performance-critical parts 7621 of an application. 7622 * -mno-gpopt, which disables the use of the $gp register while still 7623 honoring the -G limit when placing externally-visible data. This 7624 option implies -mno-extern-sdata and -mno-local-sdata and it can be 7625 useful in situations where $gp does not necessarily hold the 7626 expected value. 7627 7628 Miscellaneous improvements 7629 7630 * There is a new option called -mbranch-cost for tweaking the 7631 perceived cost of branches. 7632 * If GCC is configured to use a version of GAS that supports the 7633 .gnu_attribute directive, it will use that directive to record 7634 certain properties of the output code. .gnu_attribute is new to GAS 7635 2.18. 7636 * There are two new function attributes, near and far, for overriding 7637 the command-line setting of -mlong-calls on a function-by-function 7638 basis. 7639 * -mfp64, which previously required a 64-bit target, now works with 7640 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 7641 mipsisa*-sde-elf* configurations provide suitable library support. 7642 * GCC now recognizes the -mdmx and -mmt options and passes them down 7643 to the assembler. It does nothing else with the options at present. 7644 7645 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 7646 (BEA) 7647 7648 * Support has been added for this new architecture. 7649 7650 RS6000 (POWER/PowerPC) 7651 7652 * Support for the PowerPC 750CL paired-single instructions has been 7653 added with a new powerpc-*-linux*paired* target configuration. It 7654 is enabled by an associated -mpaired option and can be accessed 7655 using new built-in functions. 7656 * Support for auto-detecting architecture and system configuration to 7657 auto-select processor optimization tuning. 7658 * Support for VMX on AIX 5.3 has been added. 7659 * Support for AIX Version 6.1 has been added. 7660 7661 S/390, zSeries and System z9 7662 7663 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 7664 added. When using the -march=z9-ec option, the compiler will 7665 generate code making use of instructions provided by the decimal 7666 floating point facility and the floating point conversion facility 7667 (pfpo). Besides the instructions used to implement decimal floating 7668 point operations these facilities also contain instructions to move 7669 between general purpose and floating point registers and to modify 7670 and copy the sign-bit of floating point values. 7671 * When the -march=z9-ec option is used the new 7672 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 7673 decimal floating point hardware instructions will be used or not. 7674 If none of them is given the hardware support is enabled by 7675 default. 7676 * The -mstack-guard option can now be omitted when using stack 7677 checking via -mstack-size in order to let GCC choose a sensible 7678 stack guard value according to the frame size of each function. 7679 * Various changes to improve performance of generated code have been 7680 implemented, including: 7681 + The condition code set by an add logical with carry 7682 instruction is now available for overflow checks like: a + b + 7683 carry < b. 7684 + The test data class instruction is now used to implement 7685 sign-bit and infinity checks of binary and decimal floating 7686 point numbers. 7687 7688 SPARC 7689 7690 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 7691 added. 7692 7693 Xtensa 7694 7695 * Stack unwinding for exception handling now uses by default a 7696 specialized version of DWARF unwinding. This is not 7697 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 7698 Xtensa with previous versions of GCC. 7699 * For Xtensa processors that include the Conditional Store option, 7700 the built-in functions for atomic memory access are now implemented 7701 using S32C1I instructions. 7702 * If the Xtensa NSA option is available, GCC will use it to implement 7703 the __builtin_ctz and __builtin_clz functions. 7704 7705 Documentation improvements 7706 7707 * Existing libstdc++ documentation has been edited and restructured 7708 into a single DocBook XML manual. The results can be viewed online 7709 [18]here. 7710 7711 Other significant improvements 7712 7713 * The compiler's --help command-line option has been extended so that 7714 it now takes an optional set of arguments. These arguments restrict 7715 the information displayed to specific classes of command-line 7716 options, and possibly only a subset of those options. It is also 7717 now possible to replace the descriptive text associated with each 7718 displayed option with an indication of its current value, or for 7719 binary options, whether it has been enabled or disabled. 7720 Here are some examples. The following will display all the options 7721 controlling warning messages: 7722 --help=warnings 7723 7724 Whereas this will display all the undocumented, target specific 7725 options: 7726 --help=target,undocumented 7727 7728 This sequence of commands will display the binary optimizations 7729 that are enabled by -O3: 7730 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 7731 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 7732 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 7733 7734 * The configure options --with-pkgversion and --with-bugurl have been 7735 added. These allow distributors of GCC to include a 7736 distributor-specific string in manuals and --version output and to 7737 specify the URL for reporting bugs in their versions of GCC. 7738 7739 GCC 4.3.1 7740 7741 This is the [19]list of problem reports (PRs) from GCC's bug tracking 7742 system that are known to be fixed in the 4.3.1 release. This list might 7743 not be complete (that is, it is possible that some PRs that have been 7744 fixed are not listed here). 7745 7746 Target Specific Changes 7747 7748 IA-32/x86-64 7749 7750 ABI changes 7751 7752 * Starting with GCC 4.3.1, decimal floating point variables are 7753 aligned to their natural boundaries when they are passed on the 7754 stack for i386. 7755 7756 Command-line changes 7757 7758 * Starting with GCC 4.3.1, the -mcld option has been added to 7759 automatically generate a cld instruction in the prologue of 7760 functions that use string instructions. This option is used for 7761 backward compatibility on some operating systems and can be enabled 7762 by default for 32-bit x86 targets by configuring GCC with the 7763 --enable-cld configure option. 7764 7765 GCC 4.3.2 7766 7767 This is the [20]list of problem reports (PRs) from GCC's bug tracking 7768 system that are known to be fixed in the 4.3.2 release. This list might 7769 not be complete (that is, it is possible that some PRs that have been 7770 fixed are not listed here). 7771 7772 GCC 4.3.3 7773 7774 This is the [21]list of problem reports (PRs) from GCC's bug tracking 7775 system that are known to be fixed in the 4.3.3 release. This list might 7776 not be complete (that is, it is possible that some PRs that have been 7777 fixed are not listed here). 7778 7779 GCC 4.3.4 7780 7781 This is the [22]list of problem reports (PRs) from GCC's bug tracking 7782 system that are known to be fixed in the 4.3.4 release. This list might 7783 not be complete (that is, it is possible that some PRs that have been 7784 fixed are not listed here). 7785 7786 GCC 4.3.5 7787 7788 This is the [23]list of problem reports (PRs) from GCC's bug tracking 7789 system that are known to be fixed in the 4.3.5 release. This list might 7790 not be complete (that is, it is possible that some PRs that have been 7791 fixed are not listed here). 7792 7793 GCC 4.3.6 7794 7795 This is the [24]list of problem reports (PRs) from GCC's bug tracking 7796 system that are known to be fixed in the 4.3.6 release. This list might 7797 not be complete (that is, it is possible that some PRs that have been 7798 fixed are not listed here). 7799 7800 7801 For questions related to the use of GCC, please consult these web 7802 pages and the [25]GCC manuals. If that fails, the 7803 [26]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7804 web pages and the development of GCC are welcome on our developer 7805 list at [27]gcc (a] gcc.gnu.org. All of [28]our lists have public 7806 archives. 7807 7808 Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and 7809 distribution of this entire article is permitted in any medium, 7810 provided this notice is preserved. 7811 7812 These pages are [30]maintained by the GCC team. Last modified 7813 2018-09-30[31]. 7814 7815 References 7816 7817 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 7818 2. https://gmplib.org/ 7819 3. https://www.mpfr.org/ 7820 4. https://gcc.gnu.org/install/prerequisites.html 7821 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 7822 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 7823 7. http://gcc.gnu.org/gcc-4.3/porting_to.html 7824 8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 7825 9. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 7826 10. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 7827 11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 7828 12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 7829 13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 7830 14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 7831 15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 7832 16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 7833 17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 7834 18. https://gcc.gnu.org/onlinedocs/libstdc++/ 7835 19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 7836 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 7837 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 7838 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 7839 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 7840 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 7841 25. https://gcc.gnu.org/onlinedocs/ 7842 26. mailto:gcc-help (a] gcc.gnu.org 7843 27. mailto:gcc (a] gcc.gnu.org 7844 28. https://gcc.gnu.org/lists.html 7845 29. https://www.fsf.org/ 7846 30. https://gcc.gnu.org/about.html 7847 31. http://validator.w3.org/check/referer 7848 ====================================================================== 7849 http://gcc.gnu.org/gcc-4.2/index.html 7850 GCC 4.2 Release Series 7851 7852 May 19, 2008 7853 7854 The [1]GNU project and the GCC developers are pleased to announce the 7855 release of GCC 4.2.4. 7856 7857 This release is a bug-fix release, containing fixes for regressions in 7858 GCC 4.2.3 relative to previous releases of GCC. 7859 7860 This release series is no longer maintained. 7861 7862 Release History 7863 7864 GCC 4.2.4 7865 May 19, 2008 ([2]changes) 7866 7867 GCC 4.2.3 7868 February 1, 2008 ([3]changes) 7869 7870 GCC 4.2.2 7871 October 7, 2007 ([4]changes) 7872 7873 GCC 4.2.1 7874 July 18, 2007 ([5]changes) 7875 7876 GCC 4.2.0 7877 May 13, 2007 ([6]changes) 7878 7879 References and Acknowledgements 7880 7881 GCC used to stand for the GNU C Compiler, but since the compiler 7882 supports several other languages aside from C, it now stands for the 7883 GNU Compiler Collection. 7884 7885 A list of [7]successful builds is updated as new information becomes 7886 available. 7887 7888 The GCC developers would like to thank the numerous people that have 7889 contributed new features, improvements, bug fixes, and other changes as 7890 well as test results to GCC. This [8]amazing group of volunteers is 7891 what makes GCC successful. 7892 7893 For additional information about GCC please refer to the [9]GCC project 7894 web site or contact the [10]GCC development mailing list. 7895 7896 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 7897 7898 7899 For questions related to the use of GCC, please consult these web 7900 pages and the [13]GCC manuals. If that fails, the 7901 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7902 web pages and the development of GCC are welcome on our developer 7903 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 7904 archives. 7905 7906 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 7907 distribution of this entire article is permitted in any medium, 7908 provided this notice is preserved. 7909 7910 These pages are [18]maintained by the GCC team. Last modified 7911 2018-09-30[19]. 7912 7913 References 7914 7915 1. http://www.gnu.org/ 7916 2. http://gcc.gnu.org/gcc-4.2/changes.html 7917 3. http://gcc.gnu.org/gcc-4.2/changes.html 7918 4. http://gcc.gnu.org/gcc-4.2/changes.html 7919 5. http://gcc.gnu.org/gcc-4.2/changes.html 7920 6. http://gcc.gnu.org/gcc-4.2/changes.html 7921 7. http://gcc.gnu.org/gcc-4.2/buildstat.html 7922 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7923 9. http://gcc.gnu.org/index.html 7924 10. mailto:gcc (a] gcc.gnu.org 7925 11. http://gcc.gnu.org/mirrors.html 7926 12. http://gcc.gnu.org/svn.html 7927 13. https://gcc.gnu.org/onlinedocs/ 7928 14. mailto:gcc-help (a] gcc.gnu.org 7929 15. mailto:gcc (a] gcc.gnu.org 7930 16. https://gcc.gnu.org/lists.html 7931 17. https://www.fsf.org/ 7932 18. https://gcc.gnu.org/about.html 7933 19. http://validator.w3.org/check/referer 7934 ====================================================================== 7935 http://gcc.gnu.org/gcc-4.2/changes.html 7936 GCC 4.2 Release Series 7937 Changes, New Features, and Fixes 7938 7939 Caveats 7940 7941 * GCC no longer accepts the -fshared-data option. This option has had 7942 no effect in any GCC 4 release; the targets to which the option 7943 used to apply had been removed before GCC 4.0. 7944 7945 General Optimizer Improvements 7946 7947 * New command-line options specify the possible relationships among 7948 parameters and between parameters and global data. For example, 7949 -fargument-noalias-anything specifies that arguments do not alias 7950 any other storage. 7951 Each language will automatically use whatever option is required by 7952 the language standard. You should not need to use these options 7953 yourself. 7954 7955 New Languages and Language specific improvements 7956 7957 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 7958 * New command-line options -fstrict-overflow and -Wstrict-overflow 7959 have been added. -fstrict-overflow tells the compiler that it may 7960 assume that the program follows the strict signed overflow 7961 semantics permitted for the language: for C and C++ this means that 7962 the compiler may assume that signed overflow does not occur. For 7963 example, a loop like 7964 for (i = 1; i > 0; i *= 2) 7965 7966 is presumably intended to continue looping until i overflows. With 7967 -fstrict-overflow, the compiler may assume that signed overflow 7968 will not occur, and transform this into an infinite loop. 7969 -fstrict-overflow is turned on by default at -O2, and may be 7970 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 7971 be used to warn about cases where the compiler assumes that signed 7972 overflow will not occur. It takes five different levels: 7973 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 7974 -Wstrict-overflow=1 is enabled by -Wall. 7975 * The new command-line option -fno-toplevel-reorder directs GCC to 7976 emit top-level functions, variables, and asm statements in the same 7977 order that they appear in the input file. This is intended to 7978 support existing code which relies on a particular ordering (for 7979 example, code which uses top-level asm statements to switch 7980 sections). For new code, it is generally better to use function and 7981 variable attributes. The -fno-toplevel-reorder option may be used 7982 for most cases which currently use -fno-unit-at-a-time. The 7983 -fno-unit-at-a-time option will be removed in some future version 7984 of GCC. If you know of a case which requires -fno-unit-at-a-time 7985 which is not fixed by -fno-toplevel-reorder, please open a bug 7986 report. 7987 7988 C family 7989 7990 * The pragma redefine_extname will now macro expand its tokens for 7991 compatibility with SunPRO. 7992 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 7993 GCC to handle inline functions as specified in the C99 standard. In 7994 preparation for this, GCC 4.2 will warn about any use of non-static 7995 inline functions in gnu99 or c99 mode. This new warning may be 7996 disabled with the new gnu_inline function attribute or the new 7997 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 7998 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 7999 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 8000 in the current compilation. 8001 * A new command-line option -Waddress has been added to warn about 8002 suspicious uses of memory addresses as, for example, using the 8003 address of a function in a conditional expression, and comparisons 8004 against the memory address of a string literal. This warning is 8005 enabled by -Wall. 8006 8007 C++ 8008 8009 * C++ visibility handling has been overhauled. 8010 Restricted visiblity is propagated from classes to members, from 8011 functions to local statics, and from templates and template 8012 arguments to instantiations, unless the latter has explicitly 8013 declared visibility. 8014 The visibility attribute for a class must come between the 8015 class-key and the name, not after the closing brace. 8016 Attributes are now allowed for enums and elaborated-type-specifiers 8017 that only declare a type. 8018 Members of the anonymous namespace are now local to a particular 8019 translation unit, along with any other declarations which use them, 8020 though they are still treated as having external linkage for 8021 language semantics. 8022 * The (undocumented) extension which permitted templates with default 8023 arguments to be bound to template template parameters with fewer 8024 parameters has been removed. For example: 8025 template <template <typename> class C> 8026 void f(C<double>) {} 8027 8028 template <typename T, typename U = int> 8029 struct S {}; 8030 8031 template void f(S<double>); 8032 8033 is no longer accepted by G++. The reason this code is not accepted 8034 is that S is a template with two parameters; therefore, it cannot 8035 be bound to C which has only one parameter. 8036 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 8037 releases, have been removed. 8038 * The command-line option -fconst-strings, deprecated in previous GCC 8039 releases, has been removed. 8040 * The configure variable enable-__cxa_atexit is now enabled by 8041 default for more targets. Enabling this variable is necessary in 8042 order for static destructors to be executed in the correct order, 8043 but it depends upon the presence of a non-standard C library in the 8044 target library in order to work. The variable is now enabled for 8045 more targets which are known to have suitable C libraries. 8046 * -Wextra will produce warnings for if statements with a semicolon as 8047 the only body, to catch code like: 8048 if (a); 8049 return 1; 8050 return 0; 8051 8052 To suppress the warning in valid cases, use { } instead. 8053 * The C++ front end now also produces strict aliasing warnings when 8054 -fstrict-aliasing -Wstrict-aliasing is in effect. 8055 8056 Runtime Library (libstdc++) 8057 8058 * Added support for TR1 <random>, <complex>, and C compatibility 8059 headers. In addition, a lock-free version of shared_ptr was 8060 contributed as part of Phillip Jordan's Google Summer of Code 8061 project on lock-free containers. 8062 * In association with the Summer of Code work on lock-free 8063 containers, the interface for atomic builtins was adjusted, 8064 creating simpler alternatives for non-threaded code paths. Also, 8065 usage was consolidated and all elements were moved from namespace 8066 std to namespace__gnu_cxx. Affected interfaces are the functions 8067 __exchange_and_add, __atomic_add, and the objects __mutex, 8068 __recursive_mutex, and __scoped_lock. 8069 * Support for versioning weak symbol names via namespace association 8070 was added. However, as this changes the names of exported symbols, 8071 this is turned off by default in the current ABI. Intrepid users 8072 can enable this feature by using 8073 --enable-symvers=gnu-versioned-namespace during configuration. 8074 * Revised, simplified, and expanded policy-based associative 8075 containers, including data types for tree and trie forms 8076 (basic_tree, tree, trie), lists (list_update), and both 8077 collision-chaining and probing hash-based containers 8078 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 8079 the [3]documentation. 8080 * The implementation of the debug mode was modified, whereby the 8081 debug namespaces were nested inside of namespace std and namespace 8082 __gnu_cxx in order to resolve some long standing corner cases 8083 involving name lookup. Debug functionality from the policy-based 8084 data structures was consolidated and enabled with the single macro, 8085 _GLIBCXX_DEBUG. See PR 26142 for more information. 8086 * Added extensions for type traits: __conditional_type, 8087 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 8088 * Added a typelist implementation for compile-time meta-programming. 8089 Elements for typelist construction and operation can be found 8090 within namespace __gnu_cxx::typelist. 8091 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 8092 exception-safety. 8093 * Enabled library-wide visibility control, allowing -fvisibility to 8094 be used. 8095 * Consolidated all nested namespaces and the conversion of 8096 __gnu_internal implementation-private details to anonymous 8097 namespaces whenever possible. 8098 * Implemented LWG resolutions DR 431 and DR 538. 8099 8100 Fortran 8101 8102 * Support for allocatable components has been added (TR 15581 and 8103 Fortran 2003). 8104 * Support for the Fortran 2003 streaming IO extension has been added. 8105 * The GNU Fortran compiler now uses 4-byte record markers by default 8106 for unformatted files to be compatible with g77 and most other 8107 compilers. The implementation allows for records greater than 2 GB 8108 and is compatible with several other compilers. Older versions of 8109 gfortran used 8-byte record markers by default (on most systems). 8110 In order to change the length of the record markers, e.g. to read 8111 unformatted files created by older gfortran versions, the 8112 [4]-frecord-marker=8 option can be used. 8113 8114 Java (GCJ) 8115 8116 * A new command-line option -static-libgcj has been added for targets 8117 that use a linker compatible with GNU Binutils. As its name 8118 implies, this causes libgcj to be linked statically. In some cases 8119 this causes the resulting executable to start faster and use less 8120 memory than if the shared version of libgcj were used. However 8121 caution should be used as it can also cause essential parts of the 8122 library to be omitted. Some of these issues are discussed in: 8123 [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj 8124 * fastjar is no longer bundled with GCC. To build libgcj, you will 8125 need either InfoZIP (both zip and unzip) or an external jar 8126 program. In the former case, the GCC build will install a jar shell 8127 script that is based on InfoZIP and provides the same functionality 8128 as fastjar. 8129 8130 New Targets and Target Specific Improvements 8131 8132 IA-32/x86-64 8133 8134 * -mtune=generic can now be used to generate code running well on 8135 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 8136 Pentium-M, Intel Pentium 4 and Intel Core 2. 8137 * -mtune=native and -march=native will produce code optimized for the 8138 host architecture as detected using the cpuid instruction. 8139 * Added a new command-line option -fstackrealign and and 8140 __attribute__ ((force_align_arg_pointer)) to realign the stack at 8141 runtime. This allows functions compiled with a vector-aligned stack 8142 to be invoked from legacy objects that keep only word-alignment. 8143 8144 SPARC 8145 8146 * The default CPU setting has been changed from V7 to V9 in 32-bit 8147 mode on Solaris 7 and above. This is already the case in 64-bit 8148 mode. It can be overridden by specifying --with-cpu at configure 8149 time. 8150 * Back-end support of built-in functions for atomic memory access has 8151 been implemented. 8152 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 8153 added. 8154 8155 M32C 8156 8157 * Various bug fixes have made some functions (notably, functions 8158 returning structures) incompatible with previous releases. 8159 Recompiling all libraries is recommended. Note that code quality 8160 has considerably improved since 4.1, making a recompile even more 8161 beneficial. 8162 8163 MIPS 8164 8165 * Added support for the Broadcom SB-1A core. 8166 8167 IA-64 8168 8169 * Added support for IA-64 data and control speculation. By default 8170 speculation is enabled only during second scheduler pass. A number 8171 of machine flags was introduced to control the usage of speculation 8172 for both scheduler passes. 8173 8174 HPPA 8175 8176 * Added Java language support (libffi and libjava) for 32-bit HP-UX 8177 11 target. 8178 8179 Obsolete Systems 8180 8181 Documentation improvements 8182 8183 PDF Documentation 8184 8185 * A make pdf target has been added to the top-level makefile, 8186 enabling automated production of PDF documentation files. 8187 (Front-ends external to GCC should modify their Make-lang.in file 8188 to add a lang.pdf: target.) 8189 8190 Other significant improvements 8191 8192 Build system improvements 8193 8194 * All the components of the compiler are now bootstrapped by default. 8195 This improves the resilience to bugs in the system compiler or 8196 binary compatibility problems, as well as providing better testing 8197 of GCC 4.2 itself. In addition, if you build the compiler from a 8198 combined tree, the assembler, linker, etc. will also be 8199 bootstrapped (i.e. built with themselves). 8200 You can disable this behavior, and go back to the pre-GCC 4.2 set 8201 up, by configuring GCC with --disable-bootstrap. 8202 * The rules that configure follows to find target tools resemble more 8203 closely the locations that the built compiler will search. In 8204 addition, you can use the new configure option --with-target-tools 8205 to specify where to find the target tools used during the build, 8206 without affecting what the built compiler will use. 8207 This can be especially useful when building packages of GCC. For 8208 example, you may want to build GCC with GNU as or ld, even if the 8209 resulting compiler to work with the native assembler and linker. To 8210 do so, you can use --with-target-tools to point to the native 8211 tools. 8212 8213 Incompatible changes to the build system 8214 8215 * Front-ends external to GCC should modify their Make-lang.in file to 8216 replace double-colon rules (e.g. dvi::) with normal rules (like 8217 lang.dvi:). Front-end makefile hooks do not use double-colon rules 8218 anymore. 8219 * Up to GCC 4.1, a popular way to specify the target tools used 8220 during the build was to create directories named gas, binutils, 8221 etc. in the build tree, and create links to the tools from there. 8222 This does not work any more when the compiler is bootstrapped. The 8223 new configure option --with-target-tools provides a better way to 8224 achieve the same effect, and works for all native and cross 8225 settings. 8226 8227 8228 For questions related to the use of GCC, please consult these web 8229 pages and the [6]GCC manuals. If that fails, the 8230 [7]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8231 web pages and the development of GCC are welcome on our developer 8232 list at [8]gcc (a] gcc.gnu.org. All of [9]our lists have public archives. 8233 8234 Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and 8235 distribution of this entire article is permitted in any medium, 8236 provided this notice is preserved. 8237 8238 These pages are [11]maintained by the GCC team. Last modified 8239 2018-09-30[12]. 8240 8241 References 8242 8243 1. http://gcc.gnu.org/projects/gomp/ 8244 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 8245 3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 8246 4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 8247 5. https://gcc.gnu.org/wiki/Statically_linking_libgcj 8248 6. https://gcc.gnu.org/onlinedocs/ 8249 7. mailto:gcc-help (a] gcc.gnu.org 8250 8. mailto:gcc (a] gcc.gnu.org 8251 9. https://gcc.gnu.org/lists.html 8252 10. https://www.fsf.org/ 8253 11. https://gcc.gnu.org/about.html 8254 12. http://validator.w3.org/check/referer 8255 ====================================================================== 8256 http://gcc.gnu.org/gcc-4.1/index.html 8257 GCC 4.1 Release Series 8258 8259 February 13, 2007 8260 8261 The [1]GNU project and the GCC developers are pleased to announce the 8262 release of GCC 4.1.2. 8263 8264 This release is a bug-fix release, containing fixes for regressions in 8265 GCC 4.1.1 relative to previous releases of GCC. 8266 8267 This release series is no longer maintained. 8268 8269 Release History 8270 8271 GCC 4.1.2 8272 February 13, 2007 ([2]changes) 8273 8274 GCC 4.1.1 8275 May 24, 2006 ([3]changes) 8276 8277 GCC 4.1.0 8278 February 28, 2006 ([4]changes) 8279 8280 References and Acknowledgements 8281 8282 GCC used to stand for the GNU C Compiler, but since the compiler 8283 supports several other languages aside from C, it now stands for the 8284 GNU Compiler Collection. 8285 8286 A list of [5]successful builds is updated as new information becomes 8287 available. 8288 8289 The GCC developers would like to thank the numerous people that have 8290 contributed new features, improvements, bug fixes, and other changes as 8291 well as test results to GCC. This [6]amazing group of volunteers is 8292 what makes GCC successful. 8293 8294 For additional information about GCC please refer to the [7]GCC project 8295 web site or contact the [8]GCC development mailing list. 8296 8297 To obtain GCC please use [9]our mirror sites or [10]our SVN server. 8298 8299 8300 For questions related to the use of GCC, please consult these web 8301 pages and the [11]GCC manuals. If that fails, the 8302 [12]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8303 web pages and the development of GCC are welcome on our developer 8304 list at [13]gcc (a] gcc.gnu.org. All of [14]our lists have public 8305 archives. 8306 8307 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 8308 distribution of this entire article is permitted in any medium, 8309 provided this notice is preserved. 8310 8311 These pages are [16]maintained by the GCC team. Last modified 8312 2018-09-30[17]. 8313 8314 References 8315 8316 1. http://www.gnu.org/ 8317 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 8318 3. http://gcc.gnu.org/gcc-4.1/changes.html 8319 4. http://gcc.gnu.org/gcc-4.1/changes.html 8320 5. http://gcc.gnu.org/gcc-4.1/buildstat.html 8321 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8322 7. http://gcc.gnu.org/index.html 8323 8. mailto:gcc (a] gcc.gnu.org 8324 9. http://gcc.gnu.org/mirrors.html 8325 10. http://gcc.gnu.org/svn.html 8326 11. https://gcc.gnu.org/onlinedocs/ 8327 12. mailto:gcc-help (a] gcc.gnu.org 8328 13. mailto:gcc (a] gcc.gnu.org 8329 14. https://gcc.gnu.org/lists.html 8330 15. https://www.fsf.org/ 8331 16. https://gcc.gnu.org/about.html 8332 17. http://validator.w3.org/check/referer 8333 ====================================================================== 8334 http://gcc.gnu.org/gcc-4.1/changes.html 8335 GCC 4.1 Release Series 8336 Changes, New Features, and Fixes 8337 8338 The latest release in the 4.1 release series is [1]GCC 4.1.2. 8339 8340 Caveats 8341 8342 General Optimizer Improvements 8343 8344 * GCC now has infrastructure for inter-procedural optimizations and 8345 the following inter-procedural optimizations are implemented: 8346 + Profile guided inlining. When doing profile feedback guided 8347 optimization, GCC can now use the profile to make better 8348 informed decisions on whether inlining of a function is 8349 profitable or not. This means that GCC will no longer inline 8350 functions at call sites that are not executed very often, and 8351 that functions at hot call sites are more likely to be 8352 inlined. 8353 A new parameter min-inline-recursive-probability is also now 8354 available to throttle recursive inlining of functions with 8355 small average recursive depths. 8356 + Discovery of pure and const functions, a form of side-effects 8357 analysis. While older GCC releases could also discover such 8358 special functions, the new IPA-based pass runs earlier so that 8359 the results are available to more optimizers. The pass is also 8360 simply more powerful than the old one. 8361 + Analysis of references to static variables and type escape 8362 analysis, also forms of side-effects analysis. The results of 8363 these passes allow the compiler to be less conservative about 8364 call-clobbered variables and references. This results in more 8365 redundant loads being eliminated and in making static 8366 variables candidates for register promotion. 8367 + Improvement of RTL-based alias analysis. The results of type 8368 escape analysis are fed to the RTL type-based alias analyzer, 8369 allowing it to disambiguate more memory references. 8370 + Interprocedural constant propagation and function versioning. 8371 This pass looks for functions that are always called with the 8372 same constant value for one or more of the function arguments, 8373 and propagates those constants into those functions. 8374 + GCC will now eliminate static variables whose usage was 8375 optimized out. 8376 + -fwhole-program --combine can now be used to make all 8377 functions in program static allowing whole program 8378 optimization. As an exception, the main function and all 8379 functions marked with the new externally_visible attribute are 8380 kept global so that programs can link with runtime libraries. 8381 * GCC can now do a form of partial dead code elimination (PDCE) that 8382 allows code motion of expressions to the paths where the result of 8383 the expression is actually needed. This is not always a win, so the 8384 pass has been limited to only consider profitable cases. Here is an 8385 example: 8386 int foo (int *, int *); 8387 int 8388 bar (int d) 8389 { 8390 int a, b, c; 8391 b = d + 1; 8392 c = d + 2; 8393 a = b + c; 8394 if (d) 8395 { 8396 foo (&b, &c); 8397 a = b + c; 8398 } 8399 printf ("%d\n", a); 8400 } 8401 8402 The a = b + c can be sunk to right before the printf. Normal code 8403 sinking will not do this, it will sink the first one above into the 8404 else-branch of the conditional jump, which still gives you two 8405 copies of the code. 8406 * GCC now has a value range propagation pass. This allows the 8407 compiler to eliminate bounds checks and branches. The results of 8408 the pass can also be used to accurately compute branch 8409 probabilities. 8410 * The pass to convert PHI nodes to straight-line code (a form of 8411 if-conversion for GIMPLE) has been improved significantly. The two 8412 most significant improvements are an improved algorithm to 8413 determine the order in which the PHI nodes are considered, and an 8414 improvement that allow the pass to consider if-conversions of basic 8415 blocks with more than two predecessors. 8416 * Alias analysis improvements. GCC can now differentiate between 8417 different fields of structures in Tree-SSA's virtual operands form. 8418 This lets stores/loads from non-overlapping structure fields not 8419 conflict. A new algorithm to compute points-to sets was contributed 8420 that can allows GCC to see now that p->a and p->b, where p is a 8421 pointer to a structure, can never point to the same field. 8422 * Various enhancements to auto-vectorization: 8423 + Incrementally preserve SSA form when vectorizing. 8424 + Incrementally preserve loop-closed form when vectorizing. 8425 + Improvements to peeling for alignment: generate better code 8426 when the misalignment of an access is known at compile time, 8427 or when different accesses are known to have the same 8428 misalignment, even if the misalignment amount itself is 8429 unknown. 8430 + Consider dependence distance in the vectorizer. 8431 + Externalize generic parts of data reference analysis to make 8432 this analysis available to other passes. 8433 + Vectorization of conditional code. 8434 + Reduction support. 8435 * GCC can now partition functions in sections of hot and cold code. 8436 This can significantly improve performance due to better 8437 instruction cache locality. This feature works best together with 8438 profile feedback driven optimization. 8439 * A new pass to avoid saving of unneeded arguments to the stack in 8440 vararg functions if the compiler can prove that they will not be 8441 needed. 8442 * Transition of basic block profiling to tree level implementation 8443 has been completed. The new implementation should be considerably 8444 more reliable (hopefully avoiding profile mismatch errors when 8445 using -fprofile-use or -fbranch-probabilities) and can be used to 8446 drive higher level optimizations, such as inlining. 8447 The -ftree-based-profiling command-line option was removed and 8448 -fprofile-use now implies disabling old RTL level loop optimizer 8449 (-fno-loop-optimize). Speculative prefetching optimization 8450 (originally enabled by -fspeculative-prefetching) was removed. 8451 8452 New Languages and Language specific improvements 8453 8454 C and Objective-C 8455 8456 * The old Bison-based C and Objective-C parser has been replaced by a 8457 new, faster hand-written recursive-descent parser. 8458 8459 Ada 8460 8461 * The build infrastructure for the Ada runtime library and tools has 8462 been changed to be better integrated with the rest of the build 8463 infrastructure of GCC. This should make doing cross builds of Ada a 8464 bit easier. 8465 8466 C++ 8467 8468 * ARM-style name-injection of friend declarations is no longer the 8469 default. For example: 8470 struct S { 8471 friend void f(); 8472 }; 8473 8474 void g() { f(); } 8475 will not be accepted; instead a declaration of f will need to be 8476 present outside of the scope of S. The new -ffriend-injection 8477 option will enable the old behavior. 8478 * The (undocumented) extension which permitted templates with default 8479 arguments to be bound to template template parameters with fewer 8480 parameters has been deprecated, and will be removed in the next 8481 major release of G++. For example: 8482 template <template <typename> class C> 8483 void f(C<double>) {} 8484 8485 template <typename T, typename U = int> 8486 struct S {}; 8487 8488 template void f(S<double>); 8489 8490 makes use of the deprecated extension. The reason this code is not 8491 valid ISO C++ is that S is a template with two parameters; 8492 therefore, it cannot be bound to C which has only one parameter. 8493 8494 Runtime Library (libstdc++) 8495 8496 * Optimization work: 8497 + A new implementation of std::search_n is provided, better 8498 performing in case of random access iterators. 8499 + Added further efficient specializations of istream functions, 8500 i.e., character array and string extractors. 8501 + Other smaller improvements throughout. 8502 * Policy-based associative containers, designed for high-performance, 8503 flexibility and semantic safety are delivered in ext/pb_assoc. 8504 * A versatile string class, __gnu_cxx::__versa_string, providing 8505 facilities conforming to the standard requirements for 8506 basic_string, is delivered in <ext/vstring.h>. In particular: 8507 + Two base classes are provided: the default one avoids 8508 reference counting and is optimized for short strings; the 8509 alternate one, still uses it while improving in a few low 8510 level areas (e.g., alignment). See vstring_fwd.h for some 8511 useful typedefs. 8512 + Various algorithms have been rewritten (e.g., replace), the 8513 code streamlined and simple optimizations added. 8514 + Option 3 of DR 431 is implemented for both available bases, 8515 thus improving the support for stateful allocators. 8516 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 8517 libstdc++/23953) and LWG resolutions put into effect for the first 8518 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 8519 Issue 6.19). The implementation status of TR1 is now tracked in the 8520 docs in tr1.html. 8521 8522 Objective-C++ 8523 8524 * A new language front end for Objective-C++ has been added. This 8525 language allows users to mix the object oriented features of 8526 Objective-C with those of C++. 8527 8528 Java (GCJ) 8529 8530 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 8531 features (plus some 0.20 bug-fixes) 8532 + Networking 8533 o The java.net.HttpURLConnection implementation no longer 8534 buffers the entire response body in memory. This means 8535 that response bodies larger than available memory can now 8536 be handled. 8537 + (N)IO 8538 o NIO FileChannel.map implementation, fast bulk put 8539 implementation for DirectByteBuffer (speeds up this 8540 method 10x). 8541 o FileChannel.lock() and FileChannel.force() implemented. 8542 + XML 8543 o gnu.xml fix for nodes created outside a namespace 8544 context. 8545 o Add support for output indenting and 8546 cdata-section-elements output instruction in 8547 xml.transform. 8548 o xml.xpath corrections for cases where elements/attributes 8549 might have been created in non-namespace-aware mode. 8550 Corrections to handling of XSL variables and minor 8551 conformance updates. 8552 + AWT 8553 o GNU JAWT implementation, the AWT Native Interface, which 8554 allows direct access to native screen resources from 8555 within a Canvas's paint method. GNU Classpath Examples 8556 comes with a Demo, see libjava/classpath/examples/README. 8557 o awt.datatransfer updated to 1.5 with support for 8558 FlavorEvents. The gtk+ awt peers now allow copy/paste of 8559 text, images, URIs/files and serialized objects with 8560 other applications and tracking clipboard change events 8561 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 8562 objects are supported). A GNU Classpath Examples 8563 datatransfer Demo was added to show the new 8564 functionality. 8565 o Split gtk+ awt peers event handling in two threads and 8566 improve gdk lock handling (solves several awt lock ups). 8567 o Speed up awt Image loading. 8568 o Better gtk+ scrollbar peer implementation when using gtk+ 8569 >= 2.6. 8570 o Handle image loading errors correctly for gdkpixbuf and 8571 MediaTracker. 8572 o Better handle GDK lock. Properly prefix gtkpeer native 8573 functions (cp_gtk). 8574 o GdkGraphics2D has been updated to use Cairo 0.5.x or 8575 higher. 8576 o BufferedImage and GtkImage rewrites. All image drawing 8577 operations should now work correctly (flipping requires 8578 gtk+ >= 2.6) 8579 o Future Graphics2D, image and text work is documented at: 8580 [2]http://developer.classpath.org/mediation/ClasspathGrap 8581 hicsImagesText 8582 o When gtk+ 2.6 or higher is installed the default log 8583 handler will produce stack traces whenever a WARNING, 8584 CRITICAL or ERROR message is produced. 8585 + Free Swing 8586 o The RepaintManager has been reworked for more efficient 8587 painting, especially for large GUIs. 8588 o The layout manager OverlayLayout has been implemented, 8589 the BoxLayout has been rewritten to make use of the 8590 SizeRequirements utility class and caching for more 8591 efficient layout. 8592 o Improved accessibility support. 8593 o Significant progress has been made in the implementation 8594 of the javax.swing.plaf.metal package, with most UI 8595 delegates in a working state now. Please test this with 8596 your own applications and provide feedback that will help 8597 us to improve this package. 8598 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 8599 extended to highlight various features in our Free Swing 8600 implementation. And it includes a look and feel switcher 8601 for Metal (default), Ocean and GNU themes. 8602 o The javax.swing.plaf.multi package is now implemented. 8603 o Editing and several key actions for JTree and JTable were 8604 implemented. 8605 o Lots of icons and look and feel improvements for Free 8606 Swing basic and metal themes were added. Try running the 8607 GNU Classpath Swing Demo in examples 8608 (gnu.classpath.examples.swing.Demo) with: 8609 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 8610 l or 8611 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 8612 l 8613 o Start of styled text capabilites for java.swing.text. 8614 o DefaultMutableTreeNode pre-order, post-order, depth-first 8615 and breadth-first traversal enumerations implemented. 8616 o JInternalFrame colors and titlebar draw properly. 8617 o JTree is working up to par (icons, selection and keyboard 8618 traversal). 8619 o JMenus were made more compatible in visual and 8620 programmatic behavior. 8621 o JTable changeSelection and multiple selections 8622 implemented. 8623 o JButton and JToggleButton change states work properly 8624 now. 8625 o JFileChooser fixes. 8626 o revalidate() and repaint() fixes which make Free Swing 8627 much more responsive. 8628 o MetalIconFactory implemented. 8629 o Free Swing Top-Level Compatibility. JFrame, JDialog, 8630 JApplet, JInternalFrame, and JWindow are now 1.5 8631 compatible in the sense that you can call add() and 8632 setLayout() directly on them, which will have the same 8633 effect as calling getContentPane().add() and 8634 getContentPane().setLayout(). 8635 o The JTree interface has been completed. JTrees now 8636 recognizes mouse clicks and selections work. 8637 o BoxLayout works properly now. 8638 o Fixed GrayFilter to actually work. 8639 o Metal SplitPane implemented. 8640 o Lots of Free Swing text and editor stuff work now. 8641 + Free RMI and Corba 8642 o Andrew Watson, Vice President and Technical Director of 8643 the Object Management Group, has officially assigned us 8644 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 8645 mark remote classpath-specific system exceptions. 8646 Obtaining the VMCID means that GNU Classpath now is a 8647 recogniseable type of node in a highly interoperable 8648 CORBA world. 8649 o GNU Classpath now includes the first working draft to 8650 support the RMI over IIOP protocol. The current 8651 implementation is capable of remote invocations, 8652 transferring various Serializables and Externalizables 8653 via RMI-IIOP protocol. It can flatten graphs and, at 8654 least for the simple cases, is interoperable with 1.5 8655 JDKs. 8656 o org.omg.PortableInterceptor and related functionality in 8657 other packages is now implemented: 8658 # The sever and client interceptors work as required 8659 since 1.4. 8660 # The IOR interceptor works as needed for 1.5. 8661 o The org.omg.DynamicAny package is completed and passes 8662 the prepared tests. 8663 o The Portable Object Adapter should now support the output 8664 of the recent IDL to java compilers. These compilers now 8665 generate servants and not CORBA objects as before, making 8666 the output depend on the existing POA implementation. 8667 Completing POA means that such code can already be tried 8668 to run on Classpath. Our POA is tested for the following 8669 usager scenarios: 8670 # POA converts servant to the CORBA object. 8671 # Servant provides to the CORBA object. 8672 # POA activates new CORBA object with the given Object 8673 Id (byte array) that is later accessible for the 8674 servant. 8675 # During the first call, the ServantActivator provides 8676 servant for this and all subsequent calls on the 8677 current object. 8678 # During each call, the ServantLocator provides 8679 servant for this call only. 8680 # ServantLocator or ServantActivator forwards call to 8681 another server. 8682 # POA has a single servant, responsible for all 8683 objects. 8684 # POA has a default servant, but some objects are 8685 explicitly connected to they specific servants. 8686 The POA is verified using tests from the former 8687 cost.omg.org. 8688 o The CORBA implementation is now a working prototype that 8689 should support features up to 1.3 inclusive. We invite 8690 groups writing CORBA dependent applications to try 8691 Classpath implementation, reporting any possible bugs. 8692 The CORBA prototype is interoperable with Sun's 8693 implementation v 1.4, transferring object references, 8694 primitive types, narrow and wide strings, arrays, 8695 structures, trees, abstract interfaces and value types 8696 (feature of CORBA 2.3) between these two platforms. 8697 Remote exceptions are transferred and handled correctly. 8698 The stringified object references (IORs) from various 8699 sources are parsed as required. The transient (for 8700 current session) and permanent (till jre restart) 8701 redirections work. Both Little and Big Endian encoded 8702 messages are accepted. The implementation is verified 8703 using tests from the former cost.omg.org. The current 8704 release includes working examples (see the examples 8705 directory), demonstrating the client-server 8706 communication, using either CORBA Request or IDL-based 8707 stub (usually generated by a IDL to java compiler). These 8708 examples also show how to use the Classpath CORBA naming 8709 service. The IDL to java compiler is not yet written, but 8710 as our library must be compatible, it naturally accepts 8711 the output of other idlj implementations. 8712 + Misc 8713 o Updated TimeZone data against Olson tzdata2005l. 8714 o Make zip and jar packages UTF-8 clean. 8715 o "native" code builds and compiles (warning free) on 8716 Darwin and Solaris. 8717 o java.util.logging.FileHandler now rotates files. 8718 o Start of a generic JDWP framework in gnu/classpath/jdwp. 8719 This is unfinished, but feedback (at classpath (a] gnu.org) 8720 from runtime hackers is greatly appreciated. Although 8721 most of the work is currently being done around gcj/gij 8722 we want this framework to be as VM neutral as possible. 8723 Early design is described in: 8724 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 8725 o QT4 AWT peers, enable by giving configure 8726 --enable-qt-peer. Included, but not ready for production 8727 yet. They are explicitly disabled and not supported. But 8728 if you want to help with the development of these new 8729 features we are interested in feedback. You will have to 8730 explicitly enable them to try them out (and they will 8731 most likely contain bugs). 8732 o Documentation fixes all over the place. See 8733 [4]http://developer.classpath.org/doc/ 8734 8735 New Targets and Target Specific Improvements 8736 8737 IA-32/x86-64 8738 8739 * The x86-64 medium model (that allows building applications whose 8740 data segment exceeds 4GB) was redesigned to match latest ABI draft. 8741 New implementation split large datastructures into separate segment 8742 improving performance of accesses to small datastructures and also 8743 allows linking of small model libraries into medium model programs 8744 as long as the libraries are not accessing the large datastructures 8745 directly. Medium model is also supported in position independent 8746 code now. 8747 The ABI change results in partial incompatibility among medium 8748 model objects. Linking medium model libraries (or objects) compiled 8749 with new compiler into medium model program compiled with older 8750 will likely result in exceeding ranges of relocations. 8751 Binutils 2.16.91 or newer are required for compiling medium model 8752 now. 8753 8754 RS6000 (POWER/PowerPC) 8755 8756 * The AltiVec vector primitives in <altivec.h> are now implemented in 8757 a way that puts a smaller burden on the preprocessor, instead 8758 processing the "overloading" in the front ends. This should benefit 8759 compilation speed on AltiVec vector code. 8760 * AltiVec initializers now are generated more efficiently. 8761 * The popcountb instruction available on POWER5 now is generated. 8762 * The floating point round to integer instructions available on 8763 POWER5+ now is generated. 8764 * Floating point divides can be synthesized using the floating point 8765 reciprocal estimate instructions. 8766 * Double precision floating point constants are initialized as single 8767 precision values if they can be represented exactly. 8768 8769 S/390, zSeries and System z9 8770 8771 * Support for the IBM System z9 109 processor has been added. When 8772 using the -march=z9-109 option, the compiler will generate code 8773 making use of instructions provided by the extended immediate 8774 facility. 8775 * Support for 128-bit IEEE floating point has been added. When using 8776 the -mlong-double-128 option, the compiler will map the long double 8777 data type to 128-bit IEEE floating point. Using this option 8778 constitutes an ABI change, and requires glibc support. 8779 * Various changes to improve performance of generated code have been 8780 implemented, including: 8781 + In functions that do not require a literal pool, register %r13 8782 (which is traditionally reserved as literal pool pointer), can 8783 now be freely used for other purposes by the compiler. 8784 + More precise tracking of register use allows the compiler to 8785 generate more efficient function prolog and epilog code in 8786 certain cases. 8787 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 8788 instructions are now used to implement C string functions. 8789 + The MOVE CHARACTER instruction with single byte overlap is now 8790 used to implement the memset function with non-zero fill byte. 8791 + The LOAD ZERO instructions are now used where appropriate. 8792 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 8793 and INSERT IMMEDIATE instructions are now used more frequently 8794 to optimize bitfield operations. 8795 + The BRANCH ON COUNT instruction is now used more frequently. 8796 In particular, the fact that a loop contains a subroutine call 8797 no longer prevents the compiler from using this instruction. 8798 + The compiler is now aware that all shift and rotate 8799 instructions implicitly truncate the shift count to six bits. 8800 * Back-end support for the following generic features has been 8801 implemented: 8802 + The full set of [5]built-in functions for atomic memory 8803 access. 8804 + The -fstack-protector feature. 8805 + The optimization pass avoiding unnecessary stores of incoming 8806 argument registers in functions with variable argument list. 8807 8808 SPARC 8809 8810 * The default code model in 64-bit mode has been changed from 8811 Medium/Anywhere to Medium/Middle on Solaris. 8812 * TLS support is disabled by default on Solaris prior to release 10. 8813 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 8814 and later) by specifying --enable-tls at configure time. 8815 8816 MorphoSys 8817 8818 * Support has been added for this new architecture. 8819 8820 Obsolete Systems 8821 8822 Documentation improvements 8823 8824 Other significant improvements 8825 8826 * GCC can now emit code for protecting applications from 8827 stack-smashing attacks. The protection is realized by buffer 8828 overflow detection and reordering of stack variables to avoid 8829 pointer corruption. 8830 * Some built-in functions have been fortified to protect them against 8831 various buffer overflow (and format string) vulnerabilities. 8832 Compared to the mudflap bounds checking feature, the safe builtins 8833 have far smaller overhead. This means that programs built using 8834 safe builtins should not experience any measurable slowdown. 8835 8836 GCC 4.1.2 8837 8838 This is the [6]list of problem reports (PRs) from GCC's bug tracking 8839 system that are known to be fixed in the 4.1.2 release. This list might 8840 not be complete (that is, it is possible that some PRs that have been 8841 fixed are not listed here). 8842 8843 When generating code for a shared library, GCC now recognizes that 8844 global functions may be replaced when the program runs. Therefore, it 8845 is now more conservative in deducing information from the bodies of 8846 functions. For example, in this example: 8847 void f() {} 8848 void g() { 8849 try { f(); } 8850 catch (...) { 8851 cout << "Exception"; 8852 } 8853 } 8854 8855 G++ would previously have optimized away the catch clause, since it 8856 would have concluded that f cannot throw exceptions. Because users may 8857 replace f with another function in the main body of the program, this 8858 optimization is unsafe, and is no longer performed. If you wish G++ to 8859 continue to optimize as before, you must add a throw() clause to the 8860 declaration of f to make clear that it does not throw exceptions. 8861 8862 8863 For questions related to the use of GCC, please consult these web 8864 pages and the [7]GCC manuals. If that fails, the 8865 [8]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8866 web pages and the development of GCC are welcome on our developer 8867 list at [9]gcc (a] gcc.gnu.org. All of [10]our lists have public 8868 archives. 8869 8870 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and 8871 distribution of this entire article is permitted in any medium, 8872 provided this notice is preserved. 8873 8874 These pages are [12]maintained by the GCC team. Last modified 8875 2018-09-30[13]. 8876 8877 References 8878 8879 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 8880 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText 8881 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 8882 4. http://developer.classpath.org/doc/ 8883 5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 8884 6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 8885 7. https://gcc.gnu.org/onlinedocs/ 8886 8. mailto:gcc-help (a] gcc.gnu.org 8887 9. mailto:gcc (a] gcc.gnu.org 8888 10. https://gcc.gnu.org/lists.html 8889 11. https://www.fsf.org/ 8890 12. https://gcc.gnu.org/about.html 8891 13. http://validator.w3.org/check/referer 8892 ====================================================================== 8893 http://gcc.gnu.org/gcc-4.0/index.html 8894 GCC 4.0 Release Series 8895 8896 January 31, 2007 8897 8898 The [1]GNU project and the GCC developers are pleased to announce the 8899 release of GCC 4.0.4. 8900 8901 This release is a bug-fix release, containing fixes for regressions in 8902 GCC 4.0.3 relative to previous releases of GCC. 8903 8904 This release series is no longer maintained. 8905 8906 Release History 8907 8908 GCC 4.0.4 8909 January 31, 2007 ([2]changes) 8910 8911 GCC 4.0.3 8912 March 10, 2006 ([3]changes) 8913 8914 GCC 4.0.2 8915 September 28, 2005 ([4]changes) 8916 8917 GCC 4.0.1 8918 July 7, 2005 ([5]changes) 8919 8920 GCC 4.0.0 8921 April 20, 2005 ([6]changes) 8922 8923 References and Acknowledgements 8924 8925 GCC used to stand for the GNU C Compiler, but since the compiler 8926 supports several other languages aside from C, it now stands for the 8927 GNU Compiler Collection. 8928 8929 A list of [7]successful builds is updated as new information becomes 8930 available. 8931 8932 The GCC developers would like to thank the numerous people that have 8933 contributed new features, improvements, bug fixes, and other changes as 8934 well as test results to GCC. This [8]amazing group of volunteers is 8935 what makes GCC successful. 8936 8937 For additional information about GCC please refer to the [9]GCC project 8938 web site or contact the [10]GCC development mailing list. 8939 8940 To obtain GCC please use [11]our mirror sites, or [12]our SVN server. 8941 8942 8943 For questions related to the use of GCC, please consult these web 8944 pages and the [13]GCC manuals. If that fails, the 8945 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8946 web pages and the development of GCC are welcome on our developer 8947 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 8948 archives. 8949 8950 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 8951 distribution of this entire article is permitted in any medium, 8952 provided this notice is preserved. 8953 8954 These pages are [18]maintained by the GCC team. Last modified 8955 2018-09-30[19]. 8956 8957 References 8958 8959 1. http://www.gnu.org/ 8960 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 8961 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 8962 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 8963 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 8964 6. http://gcc.gnu.org/gcc-4.0/changes.html 8965 7. http://gcc.gnu.org/gcc-4.0/buildstat.html 8966 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8967 9. http://gcc.gnu.org/index.html 8968 10. mailto:gcc (a] gcc.gnu.org 8969 11. http://gcc.gnu.org/mirrors.html 8970 12. http://gcc.gnu.org/svn.html 8971 13. https://gcc.gnu.org/onlinedocs/ 8972 14. mailto:gcc-help (a] gcc.gnu.org 8973 15. mailto:gcc (a] gcc.gnu.org 8974 16. https://gcc.gnu.org/lists.html 8975 17. https://www.fsf.org/ 8976 18. https://gcc.gnu.org/about.html 8977 19. http://validator.w3.org/check/referer 8978 ====================================================================== 8979 http://gcc.gnu.org/gcc-4.0/changes.html 8980 GCC 4.0 Release Series 8981 Changes, New Features, and Fixes 8982 8983 The latest release in the 4.0 release series is [1]GCC 4.0.4. 8984 8985 Caveats 8986 8987 * GCC now generates location lists by default when compiling with 8988 debug info and optimization. 8989 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 8990 or later is needed to debug binaries containing location 8991 lists. 8992 + When you are trying to view a value of a variable in a part of 8993 a function where it has no location (for example when the 8994 variable is no longer used and thus its location was used for 8995 something else) GDB will say that it is not available. 8996 You can disable generating location lists by -fno-var-tracking. 8997 * GCC no longer accepts the -fwritable-strings option. Use named 8998 character arrays when you need a writable string. 8999 * The options -freduce-all-givs and -fmove-all-movables have been 9000 discontinued. They were used to circumvent a shortcoming in the 9001 heuristics of the old loop optimization code with respect to common 9002 Fortran constructs. The new (tree) loop optimizer works differently 9003 and doesn't need those work-arounds. 9004 * The graph-coloring register allocator, formerly enabled by the 9005 option -fnew-ra, has been discontinued. 9006 * -I- has been deprecated. -iquote is meant to replace the need for 9007 this option. 9008 * The MIPS -membedded-pic and -mrnames options have been removed. 9009 * All MIPS targets now require the GNU assembler. In particular, IRIX 9010 configurations can no longer use the MIPSpro assemblers, although 9011 they do still support the MIPSpro linkers. 9012 * The SPARC option -mflat has been removed. 9013 * English-language diagnostic messages will now use Unicode quotation 9014 marks in UTF-8 locales. (Non-English messages already used the 9015 quotes appropriate for the language in previous releases.) If your 9016 terminal does not support UTF-8 but you are using a UTF-8 locale 9017 (such locales are the default on many GNU/Linux systems) then you 9018 should set LC_CTYPE=C in the environment to disable that locale. 9019 Programs that parse diagnostics and expect plain ASCII 9020 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 9021 explanation of Unicode quotation marks for more information. 9022 * The specs file is no longer installed on most platforms. Most users 9023 will be totally unaffected. However, if you are accustomed to 9024 editing the specs file yourself, you will now have to use the 9025 -dumpspecs option to generate the specs file, and then edit the 9026 resulting file. 9027 9028 General Optimizer Improvements 9029 9030 * The [3]tree ssa branch has been merged. This merge has brought in a 9031 completely new optimization framework based on a higher level 9032 intermediate representation than the existing RTL representation. 9033 Numerous new code transformations based on the new framework are 9034 available in GCC 4.0, including: 9035 + Scalar replacement of aggregates 9036 + Constant propagation 9037 + Value range propagation 9038 + Partial redundancy elimination 9039 + Load and store motion 9040 + Strength reduction 9041 + Dead store elimination 9042 + Dead and unreachable code elimination 9043 + [4]Autovectorization 9044 + Loop interchange 9045 + Tail recursion by accumulation 9046 Many of these passes outperform their counterparts from previous 9047 GCC releases. 9048 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 9049 scheduling optimization intended for loops that perform heavy 9050 computations. 9051 9052 New Languages and Language specific improvements 9053 9054 C family 9055 9056 * The sentinel attribute has been added to GCC. This function 9057 attribute allows GCC to warn when variadic functions such as execl 9058 are not NULL terminated. See the GCC manual for a complete 9059 description of its behavior. 9060 * Given __attribute__((alias("target"))) it is now an error if target 9061 is not a symbol, defined in the same translation unit. This also 9062 applies to aliases created by #pragma weak alias=target. This is 9063 because it's meaningless to define an alias to an undefined symbol. 9064 On Solaris, the native assembler would have caught this error, but 9065 GNU as does not. 9066 9067 C and Objective-C 9068 9069 * The -Wstrict-aliasing=2 option has been added. This warning catches 9070 all unsafe cases, but it may also give a warning for some cases 9071 that are safe. 9072 * The cast-as-lvalue, conditional-expression-as-lvalue and 9073 compound-expression-as-lvalue extensions, which were deprecated in 9074 3.3.4 and 3.4, have been removed. 9075 * The -fwritable-strings option, which was deprecated in 3.4, has 9076 been removed. 9077 * #pragma pack() semantics have been brought closer to those used by 9078 other compilers. This also applies to C++. 9079 * Taking the address of a variable with register storage is invalid 9080 in C. GCC now issues an error instead of a warning. 9081 * Arrays of incomplete element type are invalid in C. GCC now issues 9082 an error for such arrays. Declarations such as extern struct s x[]; 9083 (where struct s has not been defined) can be moved after the 9084 definition of struct s. Function parameters declared as arrays of 9085 incomplete type can instead be declared as pointers. 9086 9087 C++ 9088 9089 * When compiling without optimizations (-O0), the C++ front end is 9090 much faster than in any previous versions of GCC. Independent 9091 testers have measured speed-ups up to 25% in real-world production 9092 code, compared to the 3.4 family (which was already the fastest 9093 version to date). Upgrading from older versions might show even 9094 bigger improvements. 9095 * ELF visibility attributes can now be applied to a class type, so 9096 that it affects every member function of a class at once, without 9097 having to specify each individually: 9098 class __attribute__ ((visibility("hidden"))) Foo 9099 { 9100 int foo1(); 9101 void foo2(); 9102 }; 9103 The syntax is deliberately similar to the __declspec() system used 9104 by Microsoft Windows based compilers, allowing cross-platform 9105 projects to easily reuse their existing macro system for denoting 9106 exports and imports. By explicitly marking internal classes never 9107 used outside a binary as hidden, one can completely avoid PLT 9108 indirection overheads during their usage by the compiler. You can 9109 find out more about the advantages of this at 9110 [6]https://www.akkadia.org/drepper/dsohowto.pdf 9111 * The -fvisibility-inlines-hidden option has been added which marks 9112 all inlineable functions as having hidden ELF visibility, thus 9113 removing their symbol and typeinfo from the exported symbol table 9114 of the output ELF binary. Using this option can reduce the exported 9115 symbol count of template-heavy code by up to 40% with no code 9116 change at all, thus notably improving link and load times for the 9117 binary as well as a reduction in size of up to 10%. Also, check the 9118 new [7]-fvisibility option. 9119 * The compiler now uses the library interface specified by the [8]C++ 9120 ABI for thread-safe initialization of function-scope static 9121 variables. Most users should leave this alone, but embedded 9122 programmers may want to disable this by specifying 9123 -fno-threadsafe-statics for a small savings in code size. 9124 * Taking the address of an explicit register variable is no longer 9125 supported. Note that C++ allows taking the address of variables 9126 with register storage so this will continue to compile with a 9127 warning. For example, assuming that r0 is a machine register: 9128 register int foo asm ("r0"); 9129 register int bar; 9130 &foo; // error, no longer accepted 9131 &bar; // OK, with a warning 9132 * G++ has an undocumented extension to virtual function covariancy 9133 rules that allowed the overrider to return a type that was 9134 implicitly convertable to the overridden function's return type. 9135 For instance a function returning void * could be overridden by a 9136 function returning T *. This is now deprecated and will be removed 9137 in a future release. 9138 * The G++ minimum and maximum operators (<? and >?) and their 9139 compound forms (<?=) and >?=) have been deprecated and will be 9140 removed in a future version. Code using these operators should be 9141 modified to use std::min and std::max instead. 9142 * Declaration of nested classes of class templates as friends are 9143 supported: 9144 template <typename T> struct A { 9145 class B {}; 9146 }; 9147 class C { 9148 template <typename T> friend class A<T>::B; 9149 }; 9150 This complements the feature member functions of class templates as 9151 friends introduced in GCC 3.4.0. 9152 * When declaring a friend class using an unqualified name, classes 9153 outside the innermost non-class scope are not searched: 9154 class A; 9155 namespace N { 9156 class B { 9157 friend class A; // Refer to N::A which has not been declared yet 9158 // because name outside namespace N are not searched 9159 friend class ::A; // Refer to ::A 9160 }; 9161 } 9162 Hiding the friend name until declaration is still not implemented. 9163 * Friends of classes defined outside their namespace are correctly 9164 handled: 9165 namespace N { 9166 class A; 9167 } 9168 class N::A { 9169 friend class B; // Refer to N::B in GCC 4.0.0 9170 // but ::B in earlier versions of GCC 9171 }; 9172 9173 Runtime Library (libstdc++) 9174 9175 * Optimization work: 9176 + Added efficient specializations of istream functions for char 9177 and wchar_t. 9178 + Further performance tuning of strings, in particular wrt 9179 single-char append and getline. 9180 + iter_swap - and therefore most of the mutating algorithms - 9181 now makes an unqualified call to swap when the value_type of 9182 the two iterators is the same. 9183 * A large subset of the features in Technical Report 1 (TR1 for 9184 short) is experimentally delivered (i.e., no guarantees about the 9185 implementation are provided. In particular it is not promised that 9186 the library will remain link-compatible when code using TR1 is 9187 used): 9188 + General utilities such as reference_wrapper and shared_ptr. 9189 + Function objects, i.e., result_of, mem_fn, bind, function. 9190 + Support for metaprogramming. 9191 + New containers such as tuple, array, unordered_set, 9192 unordered_map, unordered_multiset, unordered_multimap. 9193 * As usual, many bugs have been fixed and LWG resolutions implemented 9194 for the first time (e.g., DR 409). 9195 9196 Java 9197 9198 * In order to prevent naming conflicts with other implementations of 9199 these tools, some GCJ binaries have been renamed: 9200 + rmic is now grmic, 9201 + rmiregistry is now grmiregistry, and 9202 + jar is now fastjar. 9203 In particular, these names were problematic for the jpackage.org 9204 packaging conventions which install symlinks in /usr/bin that point 9205 to the preferred versions of these tools. 9206 * The -findirect-dispatch argument to the compiler now works and 9207 generates code following a new "binary compatibility" ABI. Code 9208 compiled this way follows the binary compatibility rules of the 9209 Java Language Specification. 9210 * libgcj now has support for using GCJ as a JIT, using the 9211 gnu.gcj.jit family of system properties. 9212 * libgcj can now find a shared library corresponding to the bytecode 9213 representation of a class. See the documentation for the new 9214 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 9215 property. 9216 * There have been many improvements to the class library. Here are 9217 some highlights: 9218 + Much more of AWT and Swing exist. 9219 + Many new packages and classes were added, including 9220 java.util.regex, java.net.URI, javax.crypto, 9221 javax.crypto.interfaces, javax.crypto.spec, javax.net, 9222 javax.net.ssl, javax.security.auth, 9223 javax.security.auth.callback, javax.security.auth.login, 9224 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 9225 javax.imageio, javax.imageio.event, javax.imageio.spi, 9226 javax.print, javax.print.attribute, 9227 javax.print.attribute.standard, javax.print.event, and 9228 javax.xml 9229 + Updated SAX and DOM, and imported GNU JAXP 9230 9231 Fortran 9232 9233 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 9234 front end. The new front end supports Fortran 90 and Fortran 95. It 9235 may not yet be as stable as the old Fortran front end. 9236 9237 Ada 9238 9239 * Ada (with tasking and Zero Cost Exceptions) is now available on 9240 many more targets, including but not limited to: alpha-linux, 9241 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 9242 s390x-linux, sparc-linux. 9243 * Some of the new Ada 2005 features are now implemented like 9244 Wide_Wide_Character and Ada.Containers. 9245 * Many bugs have been fixed, tools and documentation improved. 9246 * To compile Ada from the sources, install an older working Ada 9247 compiler and then use --enable-languages=ada at configuration time, 9248 since the Ada front end is not currently activated by default. See 9249 the [10]Installing GCC for details. 9250 9251 New Targets and Target Specific Improvements 9252 9253 H8/300 9254 9255 * The frame layout has changed. In the new layout, the prologue of a 9256 function first saves registers and then allocate space for locals, 9257 resulting in an 1% improvement on code size. 9258 9259 IA-32/x86-64 (AMD64) 9260 9261 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 9262 log1p, log2, logb and tan mathematical builtins (and their float 9263 and long double variants) are now implemented as inline x87 9264 intrinsics when using -ffast-math. 9265 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 9266 (and their float and long double variants) are now implemented as 9267 inline x87 intrinsics when using -ffast-math. 9268 * The x87's fsincos instruction is now used automatically with 9269 -ffast-math when calculating both the sin and cos of the same 9270 argument. 9271 * Instruction selection for multiplication and division by constants 9272 has been improved. 9273 9274 IA-64 9275 9276 * Floating point division, integer division and sqrt are now inlined, 9277 resulting in significant performance improvements on some codes. 9278 9279 MIPS 9280 9281 * Division by zero checks now use conditional traps if the target 9282 processor supports them. This decreases code size by one word per 9283 division operation. The old behavior (branch and break) can be 9284 obtained either at configure time by passing --with-divide=breaks 9285 to configure or at runtime by passing -mdivide-breaks to GCC. 9286 * Support for MIPS64 paired-single instructions has been added. It is 9287 enabled by -mpaired-single and can be accessed using both the 9288 target-independent vector extensions and new MIPS-specific built-in 9289 functions. 9290 * Support for the MIPS-3D ASE has been added. It is enabled by 9291 -mips3d and provides new MIPS-3D-specific built-in functions. 9292 * The -mexplicit-relocs option now supports static n64 code (as is 9293 used, for example, in 64-bit linux kernels). -mexplicit-relocs 9294 should now be feature-complete and is enabled by default when GCC 9295 is configured to use a compatible assembler. 9296 * Support for the NEC VR4130 series has been added. This support 9297 includes the use of VR-specific instructions and a new VR4130 9298 scheduler. Full VR4130 support can be selected with -march=vr4130 9299 while code for any ISA can be tuned for the VR4130 using 9300 -mtune=vr4130. There is also a new -mvr4130-align option that 9301 produces better schedules at the cost of increased code size. 9302 * Support for the Broadcom SB-1 has been extended. There is now an 9303 SB-1 scheduler as well as support for the SB-1-specific 9304 paired-single instructions. Full SB-1 support can be selected with 9305 -march=sb1 while code for any ISA can be optimized for the SB-1 9306 using -mtune=sb1. 9307 * The compiler can now work around errata in R4000, R4400, VR4120 and 9308 VR4130 processors. These workarounds are enabled by -mfix-r4000, 9309 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 9310 and VR4130 workarounds need binutils 2.16 or above. 9311 * IRIX shared libraries are now installed into the standard library 9312 directories: o32 libraries go into lib/, n32 libraries go into 9313 lib32/ and n64 libraries go into lib64/. 9314 * The compiler supports a new -msym32 option. It can be used to 9315 optimize n64 code in which all symbols are known to have 32-bit 9316 values. 9317 9318 S/390 and zSeries 9319 9320 * New command-line options help to generate code intended to run in 9321 an environment where stack space is restricted, e.g. Linux kernel 9322 code: 9323 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 9324 warnings for single functions that require large or dynamic 9325 stack frames. 9326 + -mstack-size and -mstack-guard generate code that checks for 9327 stack overflow at run time. 9328 + -mpacked-stack generates code that reduces the stack frame 9329 size of many functions by reusing unneeded parts of the stack 9330 bias area. 9331 * The -msoft-float option now ensures that generated code never 9332 accesses floating point registers. 9333 * The s390x-ibm-tpf target now fully supports C++, including 9334 exceptions and threads. 9335 * Various changes to improve performance of the generated code have 9336 been implemented, including: 9337 + GCC now uses sibling calls where possible. 9338 + Condition code handling has been optimized, allowing GCC to 9339 omit redundant comparisons in certain cases. 9340 + The cost function guiding many optimizations has been refined 9341 to more accurately represent the z900 and z990 processors. 9342 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 9343 instructions are now used to avoid conditional branches in 9344 certain cases. 9345 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 9346 optimize address arithmetic required to access large stack 9347 frames. 9348 + GCC now makes more efficient use of memory-to-memory type 9349 instructions (MVC, CLC, ...). 9350 + More precise tracking of special register use allows better 9351 instruction scheduling, in particular of the function prologue 9352 and epilogue sequences. 9353 + The Java front end now generates inline code to implement 9354 integer division, instead of calling library routines. 9355 9356 SPARC 9357 9358 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 9359 -mf934 have been removed. They have been replaced with -mcpu=xxx. 9360 * The internal model used to estimate the relative cost of each 9361 instruction has been updated. It is expected to give better results 9362 on recent UltraSPARC processors. 9363 * Code generation for function prologues and epilogues has been 9364 improved, resulting in better scheduling and allowing multiple exit 9365 points in functions. 9366 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 9367 It is enabled by -mvis and provides new built-in functions for VIS 9368 instructions on UltraSPARC processors. 9369 * The option -mapp-regs has been turned on by default on Solaris too. 9370 9371 NetWare 9372 9373 * Novell NetWare (on ix86, no other hardware platform was ever really 9374 supported by this OS) has been re-enabled and the ABI supported by 9375 GCC has been brought into sync with that of MetroWerks CodeWarrior 9376 (the ABI previously supported was that of some Unix systems, which 9377 NetWare never tried to support). 9378 9379 Obsolete Systems 9380 9381 Support for a number of older systems has been declared obsolete in GCC 9382 4.0. Unless there is activity to revive them, the next release of GCC 9383 will have their sources permanently removed. 9384 9385 All GCC ports for the following processor architectures have been 9386 declared obsolete: 9387 * Intel i860 9388 * Ubicom IP2022 9389 * National Semiconductor NS32K (ns32k) 9390 * Texas Instruments TMS320C[34]x 9391 9392 Also, those for some individual systems have been obsoleted: 9393 * SPARC family 9394 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 9395 sparc86x-*-elf) 9396 + OpenBSD 32-bit (sparc-*-openbsd*) 9397 9398 Documentation improvements 9399 9400 Other significant improvements 9401 9402 * Location lists are now generated by default when compiling with 9403 debug info and optimization. Location lists provide more accurate 9404 debug info about locations of variables and they allow debugging 9405 code compiled with -fomit-frame-pointer. 9406 * The -fvisibility option has been added which allows the default ELF 9407 visibility of all symbols to be set per compilation and the new 9408 #pragma GCC visibility preprocessor command allows the setting of 9409 default ELF visibility for a region of code. Using 9410 -fvisibility=hidden especially in combination with the new 9411 -fvisibility-inlines-hidden can yield substantial improvements in 9412 output binary quality including avoiding PLT indirection overheads, 9413 reduction of the exported symbol count by up to 60% (with resultant 9414 improvements to link and load times), better scope for the 9415 optimizer to improve code and up to a 20% reduction in binary size. 9416 Using these options correctly yields a binary with a similar symbol 9417 count to a Windows DLL. 9418 Perhaps more importantly, this new feature finally allows (with 9419 careful planning) complete avoidance of symbol clashes when 9420 manually loading shared objects with RTLD_GLOBAL, thus finally 9421 solving problems many projects such as python were forced to use 9422 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 9423 can find more information about using these options at 9424 [11]https://gcc.gnu.org/wiki/Visibility. 9425 __________________________________________________________________ 9426 9427 GCC 4.0.1 9428 9429 This is the [12]list of problem reports (PRs) from GCC's bug tracking 9430 system that are known to be fixed in the 4.0.1 release. This list might 9431 not be complete (that is, it is possible that some PRs that have been 9432 fixed are not listed here). 9433 9434 GCC 4.0.2 9435 9436 This is the [13]list of problem reports (PRs) from GCC's bug tracking 9437 system that are known to be fixed in the 4.0.2 release. This list might 9438 not be complete (that is, it is possible that some PRs that have been 9439 fixed are not listed here). 9440 9441 Unfortunately, due to a release engineering failure, this release has a 9442 regression on Solaris that will affect some C++ programs. We suggest 9443 that Solaris users apply a [14]patch that corrects the problem. Users 9444 who do not wish to apply the patch should explicitly link C++ programs 9445 with the -pthreads option, even if they do not use threads. This 9446 problem has been corrected in the current 4.0 branch sources and will 9447 not be present in GCC 4.0.3. 9448 9449 GCC 4.0.3 9450 9451 Starting with this release, the function getcontext is recognized by 9452 the compiler as having the same semantics as the setjmp function. In 9453 particular, the compiler will ensure that all registers are dead before 9454 calling such a function and will emit a warning about the variables 9455 that may be clobbered after the second return from the function. 9456 9457 GCC 4.0.4 9458 9459 This is the [15]list of problem reports (PRs) from GCC's bug tracking 9460 system that are known to be fixed in the 4.0.4 release. This list might 9461 not be complete (that is, it is possible that some PRs that have been 9462 fixed are not listed here). 9463 9464 The 4.0.4 release is provided for those that require a high degree of 9465 binary compatibility with previous 4.0.x releases. For most users, the 9466 GCC team recommends that version 4.1.1 or later be used instead." 9467 9468 9469 For questions related to the use of GCC, please consult these web 9470 pages and the [16]GCC manuals. If that fails, the 9471 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9472 web pages and the development of GCC are welcome on our developer 9473 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 9474 archives. 9475 9476 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 9477 distribution of this entire article is permitted in any medium, 9478 provided this notice is preserved. 9479 9480 These pages are [21]maintained by the GCC team. Last modified 9481 2018-09-30[22]. 9482 9483 References 9484 9485 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 9486 2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 9487 3. http://gcc.gnu.org/projects/tree-ssa/ 9488 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html 9489 5. http://gcc.gnu.org/news/sms.html 9490 6. https://www.akkadia.org/drepper/dsohowto.pdf 9491 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility 9492 8. https://itanium-cxx-abi.github.io/cxx-abi/ 9493 9. http://gcc.gnu.org/fortran/ 9494 10. https://gcc.gnu.org/install/ 9495 11. https://gcc.gnu.org/wiki/Visibility 9496 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 9497 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 9498 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 9499 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 9500 16. https://gcc.gnu.org/onlinedocs/ 9501 17. mailto:gcc-help (a] gcc.gnu.org 9502 18. mailto:gcc (a] gcc.gnu.org 9503 19. https://gcc.gnu.org/lists.html 9504 20. https://www.fsf.org/ 9505 21. https://gcc.gnu.org/about.html 9506 22. http://validator.w3.org/check/referer 9507 ====================================================================== 9508 http://gcc.gnu.org/gcc-3.4/index.html 9509 GCC 3.4 Release Series 9510 9511 May 26, 2006 9512 9513 The [1]GNU project and the GCC developers are pleased to announce the 9514 release of GCC 3.4.6. 9515 9516 This release is a bug-fix release, containing fixes for regressions in 9517 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 9518 3.4.x series. 9519 9520 The GCC 3.4 release series includes numerous [2]new features, 9521 improvements, bug fixes, and other changes, thanks to an [3]amazing 9522 group of volunteers. 9523 9524 This release series is no longer maintained. 9525 9526 Release History 9527 9528 GCC 3.4.6 9529 March 6, 2006 ([4]changes) 9530 9531 GCC 3.4.5 9532 November 30, 2005 ([5]changes) 9533 9534 GCC 3.4.4 9535 May 18, 2005 ([6]changes) 9536 9537 GCC 3.4.3 9538 November 4, 2004 ([7]changes) 9539 9540 GCC 3.4.2 9541 September 6, 2004 ([8]changes) 9542 9543 GCC 3.4.1 9544 July 1, 2004 ([9]changes) 9545 9546 GCC 3.4.0 9547 April 18, 2004 ([10]changes) 9548 9549 References and Acknowledgements 9550 9551 GCC used to stand for the GNU C Compiler, but since the compiler 9552 supports several other languages aside from C, it now stands for the 9553 GNU Compiler Collection. 9554 9555 A list of [11]successful builds is updated as new information becomes 9556 available. 9557 9558 The GCC developers would like to thank the numerous people that have 9559 contributed new features, improvements, bug fixes, and other changes as 9560 well as test results to GCC. This [12]amazing group of volunteers is 9561 what makes GCC successful. 9562 9563 For additional information about GCC please refer to the [13]GCC 9564 project web site or contact the [14]GCC development mailing list. 9565 9566 To obtain GCC please use [15]our mirror sites, or [16]our SVN server. 9567 9568 9569 For questions related to the use of GCC, please consult these web 9570 pages and the [17]GCC manuals. If that fails, the 9571 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9572 web pages and the development of GCC are welcome on our developer 9573 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 9574 archives. 9575 9576 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 9577 distribution of this entire article is permitted in any medium, 9578 provided this notice is preserved. 9579 9580 These pages are [22]maintained by the GCC team. Last modified 9581 2018-09-30[23]. 9582 9583 References 9584 9585 1. http://www.gnu.org/ 9586 2. http://gcc.gnu.org/gcc-3.4/changes.html 9587 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9588 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 9589 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 9590 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 9591 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 9592 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 9593 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 9594 10. http://gcc.gnu.org/gcc-3.4/changes.html 9595 11. http://gcc.gnu.org/gcc-3.4/buildstat.html 9596 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9597 13. http://gcc.gnu.org/index.html 9598 14. mailto:gcc (a] gcc.gnu.org 9599 15. http://gcc.gnu.org/mirrors.html 9600 16. http://gcc.gnu.org/svn.html 9601 17. https://gcc.gnu.org/onlinedocs/ 9602 18. mailto:gcc-help (a] gcc.gnu.org 9603 19. mailto:gcc (a] gcc.gnu.org 9604 20. https://gcc.gnu.org/lists.html 9605 21. https://www.fsf.org/ 9606 22. https://gcc.gnu.org/about.html 9607 23. http://validator.w3.org/check/referer 9608 ====================================================================== 9609 http://gcc.gnu.org/gcc-3.4/changes.html 9610 GCC 3.4 Release Series 9611 Changes, New Features, and Fixes 9612 9613 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 9614 is now closed. 9615 9616 GCC 3.4 has [2]many improvements in the C++ front end. Before reporting 9617 a bug, please make sure it's really GCC, and not your code, that is 9618 broken. 9619 9620 Caveats 9621 9622 * GNU Make is now required to build GCC. 9623 * With -nostdinc the preprocessor used to ignore both standard 9624 include paths and include paths contained in environment variables. 9625 It was neither documented nor intended that environment variable 9626 paths be ignored, so this has been corrected. 9627 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 9628 -fvolatile-static. It is unlikely that they worked correctly in any 9629 3.x release. 9630 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 9631 * Support for all the systems [3]obsoleted in GCC 3.3 has been 9632 removed from GCC 3.4. See below for a [4]list of systems which are 9633 obsoleted in this release. 9634 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 9635 compilers will not work. 9636 * The implementation of the [5]MIPS ABIs has changed. As a result, 9637 the code generated for certain MIPS targets will not be binary 9638 compatible with earlier releases. 9639 * In previous releases, the MIPS port had a fake "hilo" register with 9640 the user-visible name accum. This register has been removed. 9641 * The implementation of the [6]SPARC ABIs has changed. As a result, 9642 the code generated will not be binary compatible with earlier 9643 releases in certain cases. 9644 * The configure option --enable-threads=pthreads has been removed; 9645 use --enable-threads=posix instead, which should have the same 9646 effect. 9647 * Code size estimates used by inlining heuristics for C, Objective-C, 9648 C++ and Java have been redesigned significantly. As a result the 9649 parameters of -finline-insns, --param max-inline-insns-single and 9650 --param max-inline-insns-auto need to be reconsidered. 9651 * --param max-inline-slope and --param min-inline-insns have been 9652 removed; they are not needed for the new bottom-up inlining 9653 heuristics. 9654 * The new unit-at-a-time compilation scheme has several compatibility 9655 issues: 9656 + The order in which functions, variables, and top-level asm 9657 statements are emitted may have changed. Code relying on some 9658 particular ordering needs to be updated. The majority of such 9659 top-level asm statements can be replaced by section 9660 attributes. 9661 + Unreferenced static variables and functions are removed. This 9662 may result in undefined references when an asm statement 9663 refers to the variable/function directly. In that case either 9664 the variable/function shall be listed in asm statement operand 9665 or in the case of top-level asm statements the attribute used 9666 shall be used to force function/variable to be always output 9667 and considered as a possibly used by unknown code. 9668 For variables the attribute is accepted only by GCC 3.4 and 9669 newer, while for earlier versions it is sufficient to use 9670 unused to silence warnings about the variables not being 9671 referenced. To keep code portable across different GCC 9672 versions, you can use appropriate preprocessor conditionals. 9673 + Static functions now can use non-standard passing conventions 9674 that may break asm statements calling functions directly. 9675 Again the attribute used shall be used to prevent this 9676 behavior. 9677 As a temporary workaround, -fno-unit-at-a-time can be used, but 9678 this scheme may not be supported by future releases of GCC. 9679 * GCC 3.4 automatically places zero-initialized variables in the .bss 9680 section on some operating systems. Versions of GNU Emacs up to (and 9681 including) 21.3 will not work correctly when using this 9682 optimization; you can use -fno-zero-initialized-in-bss to disable 9683 it. 9684 * If GCC 3.4 is configured with --enable-threads=posix (the default 9685 on most targets that support pthreads) then _REENTRANT will be 9686 defined unconditionally by some libstdc++ headers. C++ code which 9687 relies on that macro to detect whether multi-threaded code is being 9688 compiled might change in meaning, possibly resulting in linker 9689 errors for single-threaded programs. Affected users of [7]Boost 9690 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 9691 See Bugzilla for [8]more information. 9692 9693 General Optimizer Improvements 9694 9695 * Usability of the profile feedback and coverage testing has been 9696 improved. 9697 + Performance of profiled programs has been improved by faster 9698 profile merging code. 9699 + Better use of the profile feedback for optimization (loop 9700 unrolling and loop peeling). 9701 + File locking support allowing fork() calls and parallel runs 9702 of profiled programs. 9703 + Coverage file format has been redesigned. 9704 + gcov coverage tool has been improved. 9705 + make profiledbootstrap available to build a faster compiler. 9706 Experiments made on i386 hardware showed an 11% speedup on -O0 9707 and a 7.5% speedup on -O2 compilation of a [9]large C++ 9708 testcase. 9709 + New value profiling pass enabled via -fprofile-values 9710 + New value profile transformations pass enabled via -fvpt aims 9711 to optimize some code sequences by exploiting knowledge about 9712 value ranges or other properties of the operands. At the 9713 moment a conversion of expensive divisions into cheaper 9714 operations has been implemented. 9715 + New -fprofile-generate and -fprofile-use command-line options 9716 to simplify the use of profile feedback. 9717 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 9718 Java which is enabled via -funit-at-a-time (and implied by -O2). In 9719 this scheme a whole file is parsed first and optimized later. The 9720 following basic inter-procedural optimizations are implemented: 9721 + Removal of unreachable functions and variables 9722 + Discovery of local functions (functions with static linkage 9723 whose address is never taken) 9724 + On i386, these local functions use register parameter passing 9725 conventions. 9726 + Reordering of functions in topological order of the call graph 9727 to enable better propagation of optimizing hints (such as the 9728 stack alignments needed by functions) in the back end. 9729 + Call graph based out-of-order inlining heuristics which allows 9730 to limit overall compilation unit growth (--param 9731 inline-unit-growth). 9732 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 9733 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 9734 CPU). 9735 * More realistic code size estimates used by inlining for C, 9736 Objective-C, C++ and Java. The growth of large functions can now be 9737 limited via --param large-function-insns and --param 9738 large-function-growth. 9739 * A new cfg-level loop optimizer pass replaces the old loop unrolling 9740 pass and adds two other loop transformations -- loop peeling and 9741 loop unswitching -- and also uses the profile feedback to limit 9742 code growth. (The three optimizations are enabled by 9743 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 9744 respectively). 9745 The old loop unroller still can be enabled by -fold-unroll-loops 9746 and may produce better code in some cases, especially when the 9747 webizer optimization pass is not run. 9748 * A new web construction pass enabled via -fweb (and implied by -O3) 9749 improves the quality of register allocation, CSE, first scheduling 9750 pass and some other optimization passes by avoiding re-use of 9751 pseudo registers with non-overlapping live ranges. The pass almost 9752 always improves code quality but does make debugging difficult and 9753 thus is not enabled by default by -O2 9754 The pass is especially effective as cleanup after code duplication 9755 passes, such as the loop unroller or the tracer. 9756 * Experimental implementations of superblock or trace scheduling in 9757 the second scheduling pass can be enabled via 9758 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 9759 9760 New Languages and Language specific improvements 9761 9762 Ada 9763 9764 * The Ada front end has been updated to include numerous bug fixes 9765 and enhancements. These include: 9766 + Improved project file support 9767 + Additional set of warnings about potential wrong code 9768 + Improved error messages 9769 + Improved code generation 9770 + Improved cross reference information 9771 + Improved inlining 9772 + Better run-time check elimination 9773 + Better error recovery 9774 + More efficient implementation of unbounded strings 9775 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 9776 ... 9777 + New GNAT.xxxx packages (e.g. GNAT.Strings, 9778 GNAT.Exception_Action) 9779 + New pragmas 9780 + New -gnatS switch replacing gnatpsta 9781 + Implementation of new Ada features (in particular limited 9782 with, limited aggregates) 9783 9784 C/Objective-C/C++ 9785 9786 * Precompiled headers are now supported. Precompiled headers can 9787 dramatically speed up compilation of some projects. There are some 9788 known defects in the current precompiled header implementation that 9789 will result in compiler crashes in relatively rare situations. 9790 Therefore, precompiled headers should be considered a "technology 9791 preview" in this release. Read the manual for details about how to 9792 use precompiled headers. 9793 * File handling in the preprocessor has been rewritten. GCC no longer 9794 gets confused by symlinks and hardlinks, and now has a correct 9795 implementation of #import and #pragma once. These two directives 9796 have therefore been un-deprecated. 9797 * The undocumented extension that allowed C programs to have a label 9798 at the end of a compound statement, which has been deprecated since 9799 GCC 3.0, has been removed. 9800 * The cast-as-lvalue extension has been removed for C++ and 9801 deprecated for C and Objective-C. In particular, code like this: 9802 int i; 9803 (char) i = 5; 9804 9805 or this: 9806 char *p; 9807 ((int *) p)++; 9808 9809 is no longer accepted for C++ and will not be accepted for C and 9810 Objective-C in a future version. 9811 * The conditional-expression-as-lvalue extension has been deprecated 9812 for C and Objective-C. In particular, code like this: 9813 int a, b, c; 9814 (a ? b : c) = 2; 9815 9816 will not be accepted for C and Objective-C in a future version. 9817 * The compound-expression-as-lvalue extension has been deprecated for 9818 C and Objective-C. In particular, code like this: 9819 int a, b; 9820 (a, b) = 2; 9821 9822 will not be accepted for C and Objective-C in a future version. A 9823 possible non-intrusive workaround is the following: 9824 (*(a, &b)) = 2; 9825 9826 * Several [10]built-in functions such as __builtin_popcount for 9827 counting bits, finding the highest and lowest bit in a word, and 9828 parity have been added. 9829 * The -fwritable-strings option has been deprecated and will be 9830 removed. 9831 * Many C math library functions are now recognized as built-ins and 9832 optimized. 9833 * The C, C++, and Objective-C compilers can now handle source files 9834 written in any character encoding supported by the host C library. 9835 The default input character set is taken from the current locale, 9836 and may be overridden with the -finput-charset command line option. 9837 In the future we will add support for inline encoding markers. 9838 9839 C++ 9840 9841 * G++ is now much closer to full conformance to the ISO/ANSI C++ 9842 standard. This means, among other things, that a lot of invalid 9843 constructs which used to be accepted in previous versions will now 9844 be rejected. It is very likely that existing C++ code will need to 9845 be fixed. This document lists some of the most common issues. 9846 * A hand-written recursive-descent C++ parser has replaced the 9847 YACC-derived C++ parser from previous GCC releases. The new parser 9848 contains much improved infrastructure needed for better parsing of 9849 C++ source codes, handling of extensions, and clean separation 9850 (where possible) between proper semantics analysis and parsing. The 9851 new parser fixes many bugs that were found in the old parser. 9852 * You must now use the typename and template keywords to disambiguate 9853 dependent names, as required by the C++ standard. 9854 struct K { 9855 typedef int mytype_t; 9856 }; 9857 9858 template <class T1> struct A { 9859 template <class T2> struct B { 9860 void callme(void); 9861 }; 9862 9863 template <int N> void bar(void) 9864 { 9865 // Use 'typename' to tell the parser that T1::mytype_t names 9866 // a type. This is needed because the name is dependent (in 9867 // this case, on template parameter T1). 9868 typename T1::mytype_t x; 9869 x = 0; 9870 } 9871 }; 9872 9873 template <class T> void template_func(void) 9874 { 9875 // Use 'template' to prefix member templates within 9876 // dependent types (a has type A<T>, which depends on 9877 // the template parameter T). 9878 A<T> a; 9879 a.template bar<0>(); 9880 9881 // Use 'template' to tell the parser that B is a nested 9882 // template class (dependent on template parameter T), and 9883 // 'typename' because the whole A<T>::B<int> is 9884 // the name of a type (again, dependent). 9885 typename A<T>::template B<int> b; 9886 b.callme(); 9887 } 9888 9889 void non_template_func(void) 9890 { 9891 // Outside of any template class or function, no names can be 9892 // dependent, so the use of the keyword 'typename' and 'template' 9893 // is not needed (and actually forbidden). 9894 A<K> a; 9895 a.bar<0>(); 9896 A<K>::B<float> b; 9897 b.callme(); 9898 } 9899 * In a template definition, unqualified names will no longer find 9900 members of a dependent base (as specified by [temp.dep]/3 in the 9901 C++ standard). For example, 9902 template <typename T> struct B { 9903 int m; 9904 int n; 9905 int f (); 9906 int g (); 9907 }; 9908 int n; 9909 int g (); 9910 template <typename T> struct C : B<T> { 9911 void h () 9912 { 9913 m = 0; // error 9914 f (); // error 9915 n = 0; // ::n is modified 9916 g (); // ::g is called 9917 } 9918 }; 9919 You must make the names dependent, e.g. by prefixing them with 9920 this->. Here is the corrected definition of C<T>::h, 9921 template <typename T> void C<T>::h () 9922 { 9923 this->m = 0; 9924 this->f (); 9925 this->n = 0 9926 this->g (); 9927 } 9928 As an alternative solution (unfortunately not backwards compatible 9929 with GCC 3.3), you may use using declarations instead of this->: 9930 template <typename T> struct C : B<T> { 9931 using B<T>::m; 9932 using B<T>::f; 9933 using B<T>::n; 9934 using B<T>::g; 9935 void h () 9936 { 9937 m = 0; 9938 f (); 9939 n = 0; 9940 g (); 9941 } 9942 }; 9943 * In templates, all non-dependent names are now looked up and bound 9944 at definition time (while parsing the code), instead of later when 9945 the template is instantiated. For instance: 9946 void foo(int); 9947 9948 template <int> struct A { 9949 static void bar(void){ 9950 foo('a'); 9951 } 9952 }; 9953 9954 void foo(char); 9955 9956 int main() 9957 { 9958 A<0>::bar(); // Calls foo(int), used to call foo(char). 9959 } 9960 9961 * In an explicit instantiation of a class template, you must use 9962 class or struct before the template-id: 9963 template <int N> 9964 class A {}; 9965 9966 template A<0>; // error, not accepted anymore 9967 template class A<0>; // OK 9968 * The "named return value" and "implicit typename" extensions have 9969 been removed. 9970 * Default arguments in function types have been deprecated and will 9971 be removed. 9972 * ARM-style name-injection of friend declarations has been deprecated 9973 and will be removed. For example: struct S { friend void f(); }; 9974 void g() { f(); } will not be accepted by future versions of G++; 9975 instead a declaration of "f" will need to be present outside of the 9976 scope of "S". 9977 * Covariant returns are implemented for all but varadic functions 9978 that require an adjustment. 9979 * When -pedantic is used, G++ now issues errors about spurious 9980 semicolons. For example, 9981 namespace N {}; // Invalid semicolon. 9982 void f() {}; // Invalid semicolon. 9983 * G++ no longer accepts attributes for a declarator after the 9984 initializer associated with that declarator. For example, 9985 X x(1) __attribute__((...)); 9986 is no longer accepted. Instead, use: 9987 X x __attribute__((...)) (1); 9988 * Inside the scope of a template class, the name of the class itself 9989 can be treated as either a class or a template. So GCC used to 9990 accept the class name as argument of type template, and template 9991 template parameter. However this is not C++ standard compliant. Now 9992 the name is not treated as a valid template template argument 9993 unless you qualify the name by its scope. For example, the code 9994 below no longer compiles. 9995 template <template <class> class TT> class X {}; 9996 template <class T> class Y { 9997 X<Y> x; // Invalid, Y is always a type template parameter. 9998 }; 9999 The valid code for the above example is 10000 X< ::Y> x; // Valid. 10001 (Notice the space between < and : to prevent GCC to interpret this 10002 as a digraph for [.) 10003 * Friend declarations that refer to template specializations are 10004 rejected if the template has not already been declared. For 10005 example, 10006 template <typename T> 10007 class C { 10008 friend void f<> (C&); 10009 }; 10010 is rejected. You must first declare f as a template, 10011 template <typename T> 10012 void f(T); 10013 * In case of friend declarations, every name used in the friend 10014 declaration must be accessible at the point of that declaration. 10015 Previous versions of G++ used to be less strict about this and 10016 allowed friend declarations for private class members, for example. 10017 See the ISO C++ Standard Committee's [11]defect report #209 for 10018 details. 10019 * Declaration of member functions of class templates as friends are 10020 supported. For example, 10021 template <typename T> struct A { 10022 void f(); 10023 }; 10024 class C { 10025 template <typename T> friend void A<T>::f(); 10026 }; 10027 * You must use template <> to introduce template specializations, as 10028 required by the standard. For example, 10029 template <typename T> 10030 struct S; 10031 10032 struct S<int> { }; 10033 is rejected. You must write, 10034 template <> struct S<int> {}; 10035 * G++ used to accept code like this, 10036 struct S { 10037 int h(); 10038 void f(int i = g()); 10039 int g(int i = h()); 10040 }; 10041 This behavior is not mandated by the standard. Now G++ issues an 10042 error about this code. To avoid the error, you must move the 10043 declaration of g before the declaration of f. The default arguments 10044 for g must be visible at the point where it is called. 10045 * The C++ ABI Section 3.3.3 specifications for the array construction 10046 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 10047 NULL when the allocator argument returns NULL. These changes are 10048 incorporated into the libstdc++ runtime library. 10049 * Using a name introduced by a typedef in a friend declaration or in 10050 an explicit instantiation is now rejected, as specified by the ISO 10051 C++ standard. 10052 class A; 10053 typedef A B; 10054 class C { 10055 friend class B; // error, no typedef name here 10056 friend B; // error, friend always needs class/struct/enum 10057 friend class A; // OK 10058 }; 10059 10060 template <int> class Q {}; 10061 typedef Q<0> R; 10062 template class R; // error, no typedef name here 10063 template class Q<0>; // OK 10064 * When allocating an array with a new expression, GCC used to allow 10065 parentheses around the type name. This is actually ill-formed and 10066 it is now rejected: 10067 int* a = new (int)[10]; // error, not accepted anymore 10068 int* a = new int[10]; // OK 10069 * When binding an rvalue of class type to a reference, the copy 10070 constructor of the class must be accessible. For instance, consider 10071 the following code: 10072 class A 10073 { 10074 public: 10075 A(); 10076 10077 private: 10078 A(const A&); // private copy ctor 10079 }; 10080 10081 A makeA(void); 10082 void foo(const A&); 10083 10084 void bar(void) 10085 { 10086 foo(A()); // error, copy ctor is not accessible 10087 foo(makeA()); // error, copy ctor is not accessible 10088 10089 A a1; 10090 foo(a1); // OK, a1 is a lvalue 10091 } 10092 This might be surprising at first sight, especially since most 10093 popular compilers do not correctly implement this rule ([12]further 10094 details). 10095 * When forming a pointer to member or a pointer to member function, 10096 access checks for class visibility (public, protected, private) are 10097 now performed using the qualifying scope of the name itself. This 10098 is better explained with an example: 10099 class A 10100 { 10101 public: 10102 void pub_func(); 10103 protected: 10104 void prot_func(); 10105 private: 10106 void priv_func(); 10107 }; 10108 10109 class B : public A 10110 { 10111 public: 10112 void foo() 10113 { 10114 &A::pub_func; // OK, pub_func is accessible through A 10115 &A::prot_func; // error, cannot access prot_func through A 10116 &A::priv_func; // error, cannot access priv_func through A 10117 10118 &B::pub_func; // OK, pub_func is accessible through B 10119 &B::prot_func; // OK, can access prot_func through B (within B) 10120 &B::priv_func; // error, cannot access priv_func through B 10121 } 10122 }; 10123 10124 Runtime Library (libstdc++) 10125 10126 * Optimization work: 10127 + Streamlined streambuf, filebuf, separate synched with C 10128 Standard I/O streambuf. 10129 + All formatted I/O now uses cached locale information. 10130 + STL optimizations (memory/speed for list, red-black trees as 10131 used by sets and maps). 10132 + More use of GCC builtins. 10133 + String optimizations (avoid contention on 10134 increment/decrement-and-test of the reference count in the 10135 empty-string object, constructor from input_iterators 10136 speedup). 10137 * Static linkage size reductions. 10138 * Large File Support (files larger than 2 GB on 32-bit systems). 10139 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 10140 * Generic character traits. 10141 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 10142 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 10143 * The allocator class is now standard-conformant, and two additional 10144 extension allocators have been added, mt_alloc and 10145 bitmap_allocator. 10146 * PCH support: -include bits/stdc++.h (2x compile speedup). 10147 * Rewrote __cxa_demangle with support for C++ style allocators. 10148 * New debug modes for STL containers and iterators. 10149 * Testsuite rewrite: five times as many tests, plus increasingly 10150 sophisticated tests, including I/O, MT, multi-locale, wide and 10151 narrow characters. 10152 * Use current versions of GNU "autotools" for build/configuration. 10153 10154 Objective-C 10155 10156 * The Objective-C front end has been updated to include the numerous 10157 bug fixes and enhancements previously available only in Apple's 10158 version of GCC. These include: 10159 + Structured exception (@try... @catch... @finally, @throw) and 10160 synchronization (@synchronized) support. These are accessible 10161 via the -fobjc-exceptions switch; as of this writing, they may 10162 only be used in conjunction with -fnext-runtime on Mac OS X 10163 10.3 and later. See [13]Options Controlling Objective-C 10164 Dialect for more information. 10165 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 10166 may now be encoded as 'B'. In addition, the back-end/codegen 10167 dependencies have been removed. 10168 + An overhaul of message dispatch construction, ensuring that 10169 the various receiver types (and casts thereof) are handled 10170 properly, and that correct diagnostics are issued. 10171 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 10172 (-freplace-objc-classes) debugging modes, currently available 10173 on Mac OS X 10.3 and later. See [14]Options Controlling 10174 Objective-C Dialect for more information. 10175 + Access to optimized runtime entry points (-fno-nil-receivers ) 10176 on the assumption that message receivers are never nil. This 10177 is currently available on Mac OS X 10.3 and later. See 10178 [15]Options Controlling Objective-C Dialect for more 10179 information. 10180 10181 Java 10182 10183 * Compiling a .jar file will now cause non-.class entries to be 10184 automatically compiled as resources. 10185 * libgcj has been ported to Darwin. 10186 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 10187 to gcj. 10188 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 10189 code from shared libraries. 10190 * libgcj has been much more completely merged with [16]GNU Classpath. 10191 * Class loading is now much more correct; in particular the caller's 10192 class loader is now used when that is required. 10193 * [17]Eclipse 2.x will run out of the box using gij. 10194 * Parts of java.nio have been implemented. Direct and indirect 10195 buffers work, as do fundamental file and socket operations. 10196 * java.awt has been improved, though it is still not ready for 10197 general use. 10198 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 10199 method. 10200 * The MinGW port has matured. Enhancements include socket timeout 10201 support, thread interruption, improved Runtime.exec() handling and 10202 support for accented characters in filenames. 10203 10204 Fortran 10205 10206 * Fortran improvements are listed in the [18]Fortran documentation. 10207 10208 New Targets and Target Specific Improvements 10209 10210 Alpha 10211 10212 * Several [19]built-in functions have been added such as 10213 __builtin_alpha_zap to allow utilizing the more obscure 10214 instructions of the CPU. 10215 * Parameter passing of complex arguments has changed to match the 10216 ABI. This change is incompatible with previous GCC versions, but 10217 does fix compatibility with the Tru64 compiler and several corner 10218 cases where GCC was incompatible with itself. 10219 10220 ARM 10221 10222 * Nicolas Pitre has contributed his hand-coded floating-point support 10223 code for ARM. It is both significantly smaller and faster than the 10224 existing C-based implementation, even when building applications 10225 for Thumb. The arm-elf configuration has been converted to use the 10226 new code. 10227 * Support for the Intel's iWMMXt architecture, a second generation 10228 XScale processor, has been added. Enabled at run time with the 10229 -mcpu=iwmmxt command line switch. 10230 * A new ARM target has been added: arm-wince-pe. This is similar to 10231 the arm-pe target, but it defaults to using the APCS32 ABI. 10232 * The existing ARM pipeline description has been converted to the use 10233 the [20]DFA processor pipeline model. There is not much change in 10234 code performance, but the description is now [21]easier to 10235 understand. 10236 * Support for the Cirrus EP9312 Maverick floating point co-processor 10237 added. Enabled at run time with the -mcpu=ep9312 command line 10238 switch. Note however that the multilibs to support this chip are 10239 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 10240 enable their production you will have to uncomment the entries in 10241 that file. 10242 10243 H8/300 10244 10245 * Support for long long has been added. 10246 * Support for saveall attribute has been added. 10247 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 10248 for H8/300H and H8S, which is much faster than the previous 10249 implementation. 10250 * A lot of small performance improvements. 10251 10252 IA-32/AMD64 (x86-64) 10253 10254 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 10255 -march=k8 and -mcpu=k8. 10256 * Scalar SSE code generation carefully avoids reformatting penalties, 10257 hidden dependencies and minimizes the number of uops generated on 10258 both Intel and AMD CPUs. 10259 * Vector MMX and SSE operands are now passed in registers to improve 10260 performance and match the argument passing convention used by the 10261 Intel C++ Compiler. As a result it is not possible to call 10262 functions accepting vector arguments compiled by older GCC version. 10263 * Conditional jump elimination is now more aggressive on modern CPUs. 10264 * The Athlon ports has been converted to use the DFA processor 10265 pipeline description. 10266 * Optimization of indirect tail calls is now possible in a similar 10267 fashion as direct sibcall optimization. 10268 * Further small performance improvements. 10269 * -m128bit-long-double is now less buggy. 10270 * __float128 support in 64-bit compilation. 10271 * Support for data structures exceeding 2GB in 64-bit mode. 10272 * -mcpu has been renamed to -mtune. 10273 10274 IA-64 10275 10276 * Tuning code for the Itanium 2 processor has been added. The 10277 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 10278 enabled by default now. To generate code tuned for Itanium 1 the 10279 option -mtune=itanium1 should be used. 10280 * [22]DFA processor pipeline descriptions for the IA-64 processors 10281 have been added. This resulted in about 3% improvement on the 10282 SPECInt2000 benchmark for Itanium 2. 10283 * Instruction bundling for the IA-64 processors has been rewritten 10284 using the DFA pipeline hazard recognizer. It resulted in about 60% 10285 compiler speedup on the SPECInt2000 C programs. 10286 10287 M32R 10288 10289 * Support for the M32R/2 processor has been added by Renesas. 10290 * Support for an M32R GNU/Linux target and PIC code generation has 10291 been added by Renesas. 10292 10293 M68000 10294 10295 * Bernardo Innocenti (Develer S.r.l.) has contributed the 10296 m68k-uclinux target, based on former work done by Paul Dale 10297 (SnapGear Inc.). Code generation for the ColdFire processors family 10298 has been enhanced and extended to support the MCF 53xx and MCF 54xx 10299 cores, integrating former work done by Peter Barada (Motorola). 10300 10301 MIPS 10302 10303 Processor-specific changes 10304 10305 * Support for the RM7000 and RM9000 processors has been added. It can 10306 be selected using the -march compiler option and should work with 10307 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 10308 * Support for revision 2 of the MIPS32 ISA has been added. It can be 10309 selected with the command-line option -march=mips32r2. 10310 * There is a new option, -mfix-sb1, to work around certain SB-1 10311 errata. 10312 10313 Configuration 10314 10315 * It is possible to customize GCC using the following configure-time 10316 options: 10317 + --with-arch, which specifies the default value of the -march 10318 option. 10319 + --with-tune, which specifies the default value of the -mtune 10320 option. 10321 + --with-abi, which specifies the default ABI. 10322 + --with-float=soft, which tells GCC to use software floating 10323 point by default. 10324 + --with-float=hard, which tells GCC to use hardware floating 10325 point by default. 10326 * A 64-bit GNU/Linux port has been added. The associated 10327 configurations are mips64-linux-gnu and mips64el-linux-gnu. 10328 * The 32-bit GNU/Linux port now supports Java. 10329 * The IRIX 6 configuration now supports the o32 ABI and will build 10330 o32 multilibs by default. This support is compatible with both 10331 binutils and the SGI tools, but note that several features, 10332 including debugging information and DWARF2 exception handling, are 10333 only available when using the GNU assembler. Use of the GNU 10334 assembler and linker (version 2.15 or above) is strongly 10335 recommended. 10336 * The IRIX 6 configuration now supports 128-bit long doubles. 10337 * There are two new RTEMS-specific configurations, mips-rtems and 10338 mipsel-rtems. 10339 * There are two new *-elf configurations, mipsisa32r2-elf and 10340 mipsisa32r2el-elf. 10341 10342 General 10343 10344 * Several [23]ABI bugs have been fixed. Unfortunately, these changes 10345 will break binary compatibility with earlier releases. 10346 * GCC can now use explicit relocation operators when generating 10347 -mabicalls code. This behavior is controlled by -mexplicit-relocs 10348 and can have several performance benefits. For example: 10349 + It allows for more optimization of GOT accesses, including 10350 better scheduling and redundancy elimination. 10351 + It allows sibling calls to be implemented as jumps. 10352 + n32 and n64 leaf functions can use a call-clobbered global 10353 pointer instead of $28. 10354 + The code to set up $gp can be removed from functions that 10355 don't need it. 10356 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 10357 option is equivalent to the assembler's -xgot option and should be 10358 used instead of -Wa,-xgot. 10359 * Frame pointer elimination is now supported when generating 64-bit 10360 MIPS16 code. 10361 * Inline block moves have been optimized to take more account of 10362 alignment information. 10363 * Many internal changes have been made to the MIPS port, mostly aimed 10364 at reducing the reliance on assembler macros. 10365 10366 PowerPC 10367 10368 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 10369 [24]ABI incompatibilities regarding the way parameters are passed 10370 during functions calls. These changes may result in incompatibility 10371 between code compiled with GCC 3.3 and GCC 3.4. 10372 10373 PowerPC Darwin 10374 10375 * Support for shared/dylib gcc libraries has been added. It is 10376 enabled by default on powerpc-apple-darwin7.0.0 and up. 10377 * Libgcj is enabled by default. On systems older than 10378 powerpc-apple-darwin7.0.0 you need to install dlcompat. 10379 * 128-bit IBM extended precision format support added for long 10380 double. 10381 10382 PowerPC64 GNU/Linux 10383 10384 * By default, PowerPC64 GNU/Linux now uses natural alignment of 10385 structure elements. The old four byte alignment for double, with 10386 special rules for a struct starting with a double, can be chosen 10387 with -malign-power. This change may result in incompatibility 10388 between code compiled with GCC 3.3 and GCC 3.4. 10389 * -mabi=altivec is now the default rather than -mabi=no-altivec. 10390 * 128-bit IBM extended precision format support added for long 10391 double. 10392 10393 S/390 and zSeries 10394 10395 * New command-line options allow to specify the intended execution 10396 environment for generated code: 10397 + -mesa/-mzarch allows to specify whether to generate code 10398 running in ESA/390 mode or in z/Architecture mode (this is 10399 applicable to 31-bit code only). 10400 + -march allows to specify a minimum processor architecture 10401 level (g5, g6, z900, or z990). 10402 + -mtune allows to specify which processor to tune for. 10403 * It is possible to customize GCC using the following configure-time 10404 options: 10405 + --with-mode, which specifies whether to default to assuming 10406 ESA/390 or z/Architecture mode. 10407 + --with-arch, which specifies the default value of the -march 10408 option. 10409 + --with-tune, which specifies the default value of the -mtune 10410 option. 10411 * Support for the z990 processor has been added, and can be selected 10412 using -march=z990 or -mtune=z990. This includes instruction 10413 scheduling tuned for the superscalar instruction pipeline of the 10414 z990 processor as well as support for all new instructions provided 10415 by the long-displacement facility. 10416 * Support to generate 31-bit code optimized for zSeries processors 10417 (running in ESA/390 or in z/Architecture mode) has been added. This 10418 can be selected using -march=z900 and -mzarch respectively. 10419 * Instruction scheduling for the z900 and z990 processors now uses 10420 the DFA pipeline hazard recognizer. 10421 * GCC no longer generates code to maintain a stack backchain, 10422 previously used to generate stack backtraces for debugging 10423 purposes. As replacement that does not incur runtime overhead, 10424 DWARF-2 call frame information is provided by GCC; this is 10425 supported by GDB 6.1. The old behavior can be restored using the 10426 -mbackchain option. 10427 * The stack frame size of functions may now exceed 2 GB in 64-bit 10428 code. 10429 * A port for the 64-bit IBM TPF operating system has been added; the 10430 configuration is s390x-ibm-tpf. This configuration is supported as 10431 cross-compilation target only. 10432 * Various changes to improve the generated code have been 10433 implemented, including: 10434 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 10435 instructions to significantly speed up many floating-point 10436 applications. 10437 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 10438 WITH BORROW instructions to speed up long long arithmetic. 10439 + GCC now uses the SEARCH STRING instruction to implement 10440 strlen(). 10441 + In many cases, function call overhead for 31-bit code has been 10442 reduced by placing the literal pool after the function code 10443 instead of after the function prolog. 10444 + Register 14 is no longer reserved in 64-bit code. 10445 + Handling of global register variables has been improved. 10446 10447 SPARC 10448 10449 * The option -mflat is deprecated. 10450 * Support for large (> 2GB) frames has been added to the 64-bit port. 10451 * Several [25]ABI bugs have been fixed. Unfortunately, these changes 10452 will break binary compatibility with earlier releases. 10453 * The default debugging format has been switched from STABS to 10454 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 10455 the default debugging format for 64-bit code on Solaris. 10456 10457 SuperH 10458 10459 * Support for the SH2E processor has been added. Enabled at run time 10460 with the -m2e command line switch, or at configure time by 10461 specifying sh2e as the machine part of the target triple. 10462 10463 V850 10464 10465 * Support for the Mitsubishi V850E1 processor has been added. This is 10466 a variant of the V850E processor with some additional debugging 10467 instructions. 10468 10469 Xtensa 10470 10471 * Several ABI bugs have been fixed. Unfortunately, these changes 10472 break binary compatibility with earlier releases. 10473 + For big-endian processors, the padding of aggregate return 10474 values larger than a word has changed. If the size of an 10475 aggregate return value is not a multiple of 32 bits, previous 10476 versions of GCC inserted padding in the most-significant bytes 10477 of the first return value register. Aggregates larger than a 10478 word are now padded in the least-significant bytes of the last 10479 return value register used. Aggregates smaller than a word are 10480 still padded in the most-significant bytes. The return value 10481 padding has not changed for little-endian processors. 10482 + Function arguments with 16-byte alignment are now properly 10483 aligned. 10484 + The implementation of the va_list type has changed. A va_list 10485 value created by va_start from a previous release cannot be 10486 used with va_arg from this release, or vice versa. 10487 * More processor configuration options for Xtensa processors are 10488 supported: 10489 + the ABS instruction is now optional; 10490 + the ADDX* and SUBX* instructions are now optional; 10491 + an experimental CONST16 instruction can be used to synthesize 10492 constants instead of loading them from constant pools. 10493 These and other Xtensa processor configuration options can no 10494 longer be enabled or disabled by command-line options; the 10495 processor configuration must be specified by the xtensa-config.h 10496 header file when building GCC. Additionally, the 10497 -mno-serialize-volatile option is no longer supported. 10498 10499 Obsolete Systems 10500 10501 Support for a number of older systems has been declared obsolete in GCC 10502 3.4. Unless there is activity to revive them, the next release of GCC 10503 will have their sources permanently removed. 10504 10505 All configurations of the following processor architectures have been 10506 declared obsolete: 10507 * Mitsubishi D30V, d30v-* 10508 * AT&T DSP1600 and DSP1610, dsp16xx-* 10509 * Intel 80960, i960 10510 10511 Also, some individual systems have been obsoleted: 10512 * ARM Family 10513 + Support for generating code for operation in APCS/26 mode 10514 (-mapcs-26). 10515 * IBM ESA/390 10516 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 10517 maintained and supported.) 10518 * Intel 386 family 10519 + MOSS, i?86-moss-msdos and i?86-*-moss* 10520 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 10521 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 10522 i?86-*-freebsd2* 10523 + GNU/Linux with a.out object format, i?86-linux*aout* 10524 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 10525 + Interix versions before Interix 3, i?86-*-interix 10526 + Mach microkernel, i?86-mach* 10527 + SCO UnixWare with UDK, i?86-*-udk* 10528 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 10529 + VSTa microkernel, i386-*-vsta 10530 * Motorola M68000 family 10531 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 10532 + NetBSD with a.out object format (before NetBSD 1.4), 10533 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 10534 + Generic System V r.4, m68k-*-sysv4* 10535 * VAX 10536 + Generic VAX, vax-*-* (This is generic VAX only; we have not 10537 obsoleted any VAX triples for specific operating systems.) 10538 10539 Documentation improvements 10540 10541 Other significant improvements 10542 10543 * The build system has undergone several significant cleanups. 10544 Subdirectories will only be configured if they are being built, and 10545 all subdirectory configures are run from the make command. The top 10546 level has been autoconfiscated. 10547 * Building GCC no longer writes to its source directory. This should 10548 help those wishing to share a read-only source directory over NFS 10549 or build from a CD. The exceptions to this feature are if you 10550 configure with either --enable-maintainer-mode or 10551 --enable-generated-files-in-srcdir. 10552 * The -W warning option has been renamed to -Wextra, which is more 10553 easily understood. The older spelling will be retained for 10554 backwards compatibility. 10555 * Substantial improvements in compile time have been made, 10556 particularly for non-optimizing compilations. 10557 __________________________________________________________________ 10558 10559 GCC 3.4.0 10560 10561 Bug Fixes 10562 10563 A vast number of bugs have been fixed in 3.4.0, too many to publish a 10564 complete list here. [26]Follow this link to query the Bugzilla database 10565 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 10566 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 10567 regressions. 10568 __________________________________________________________________ 10569 10570 GCC 3.4.1 10571 10572 Bug Fixes 10573 10574 This section lists the problem reports (PRs) from GCC's bug tracking 10575 system that are known to be fixed in the 3.4.1 release. This list might 10576 not be complete (that is, it is possible that some PRs that have been 10577 fixed are not listed here). 10578 10579 Bootstrap failures 10580 10581 * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 10582 emitted - PIC related 10583 * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 10584 * [29]14760 A bug in configure.in prevents using both 10585 --program-suffix and --program-prefix 10586 * [30]14671 [hppa64] bootstrap fails: ICE in 10587 save_call_clobbered_regs, in caller_save.c 10588 * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on 10589 Alpha 10590 * [32]15178 Solaris 9/x86 fails linking after stage 3 10591 10592 Multi-platform internal compiler errors (ICEs) 10593 10594 * [33]12753 (preprocessor) Memory corruption in preprocessor on bad 10595 input 10596 * [34]13985 ICE in gcc.c-torture/compile/930621-1.c 10597 * [35]14810 (c++) tree check failures with invalid code involving 10598 templates 10599 * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 10600 cp/parser.c 10601 * [37]15044 (c++) ICE on syntax error, template header 10602 * [38]15057 (c++) Compiling of conditional value throw constructs 10603 cause a segmentation violation 10604 * [39]15064 (c++) typeid of template parameter gives ICE 10605 * [40]15142 (c++) ICE when passing a string where a char* is expected 10606 in a throw statement 10607 * [41]15159 ICE in rtl_verify_flow_info_1 10608 * [42]15165 (c++) ICE in instantiate_template 10609 * [43]15193 Unary minus using pointer to V4SF vector causes 10610 -fforce-mem to exhaust all memory 10611 * [44]15209 (c++) Runs out of memory with packed structs 10612 * [45]15227 (c++) Trouble with invalid function definition 10613 * [46]15285 (c++) instantiate_type ICE when forming pointer to 10614 template function 10615 * [47]15299 (c++) ICE in resolve_overloaded_unification 10616 * [48]15329 (c++) ICE on constructor of member template 10617 * [49]15550 ICE in extract_insn, in recog.c 10618 * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c 10619 * [51]15640 (c++) ICE on invalid code in arg_assoc, in 10620 cp/name-lookup.c 10621 * [52]15666 [unit-at-a-time] Gcc abort on valid code 10622 * [53]15696 (c++) ICE with bad pointer-to-member code 10623 * [54]15701 (c++) ICE with friends and template template parameter 10624 * [55]15761 ICE in do_SUBST, in combine.c 10625 * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 10626 10627 Ada 10628 10629 * [57]14538 All RTEMS targets broken for gnat 10630 10631 C front end 10632 10633 * [58]12391 missing warning about assigning to an incomplete type 10634 * [59]14649 atan(1.0) should not be a constant expression 10635 * [60]15004 [unit-at-a-time] no warning for unused paramater in 10636 static function 10637 * [61]15749 --pedantic-errors behaves differently from --pedantic 10638 with C-compiler on GNU/Linux 10639 10640 C++ compiler and library 10641 10642 * [62]10646 non-const reference is incorrectly matched in a "const T" 10643 partial specialization 10644 * [63]12077 wcin.rdbuf()->in_avail() return value too high 10645 * [64]13598 enc_filebuf doesn't work 10646 * [65]14211 const_cast returns lvalue but should be rvalue 10647 * [66]14220 num_put::do_put() undesired float/double behavior 10648 * [67]14245 problem with user-defined allocators in std::basic_string 10649 * [68]14340 libstdc++ Debug mode: failure to convert iterator to 10650 const_iterator 10651 * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 10652 FILE* 10653 * [70]14668 no warning anymore for reevaluation of declaration 10654 * [71]14775 LFS (large file support) tests missing 10655 * [72]14821 Duplicate namespace alias declaration should not conflict 10656 * [73]14930 Friend declaration ignored 10657 * [74]14932 cannot use offsetof to get offsets of array elements in 10658 g++ 3.4.0 10659 * [75]14950 [non unit-at-a-time] always_inline does not mix with 10660 templates and -O0 10661 * [76]14962 g++ ignores #pragma redefine_extname 10662 * [77]14975 Segfault on low-level write error during imbue 10663 * [78]15002 Linewise stream input is unusably slow (std::string slow) 10664 * [79]15025 compiler accepts redeclaration of template as 10665 non-template 10666 * [80]15046 [arm] Math functions misdetected by cross configuration 10667 * [81]15069 a bit test on a variable of enum type is miscompiled 10668 * [82]15074 g++ -lsupc++ still links against libstdc++ 10669 * [83]15083 spurious "statement has no effect" warning 10670 * [84]15096 parse error with templates and pointer to const member 10671 * [85]15287 combination of operator[] and operator .* fails in 10672 templates 10673 * [86]15317 __attribute__ unused in first parameter of constructor 10674 gives error 10675 * [87]15337 sizeof on incomplete type diagnostic 10676 * [88]15361 bitset<>::_Find_next fails 10677 * [89]15412 _GLIBCXX_ symbols symbols defined and used in different 10678 namespaces 10679 * [90]15427 valid code results in incomplete type error 10680 * [91]15471 Incorrect member pointer offsets in anonymous 10681 structs/unions 10682 * [92]15503 nested template problem 10683 * [93]15507 compiler hangs while laying out union 10684 * [94]15542 operator & and template definitions 10685 * [95]15565 SLES9: leading + sign for unsigned int with showpos 10686 * [96]15625 friend defined inside a template fails to find static 10687 function 10688 * [97]15629 Function templates, overloads, and friend name injection 10689 * [98]15742 'noreturn' attribute ignored in method of template 10690 functions. 10691 * [99]15775 Allocator::pointer consistently ignored 10692 * [100]15821 Duplicate namespace alias within namespace rejected 10693 * [101]15862 'enum yn' fails (confict with undeclared builtin) 10694 * [102]15875 rejects pointer to member in template 10695 * [103]15877 valid code using templates and anonymous enums is 10696 rejected 10697 * [104]15947 Puzzling error message for wrong destructor declaration 10698 in template class 10699 * [105]16020 cannot copy __gnu_debug::bitset 10700 * [106]16154 input iterator concept too restrictive 10701 * [107]16174 deducing top-level consts 10702 10703 Java 10704 10705 * [108]14315 Java compiler is not parallel make safe 10706 10707 Fortran 10708 10709 * [109]15151 [g77] incorrect logical i/o in 64-bit mode 10710 10711 Objective-C 10712 10713 * [110]7993 private variables cannot be shadowed in subclasses 10714 10715 Optimization bugs 10716 10717 * [111]15228 useless copies of floating point operands 10718 * [112]15345 [non-unit-at-a-time] unreferenced nested inline 10719 functions not optimized away 10720 * [113]15945 Incorrect floating point optimization 10721 * [114]15526 ftrapv aborts on 0 * (-1) 10722 * [115]14690 Miscompiled POOMA tests 10723 * [116]15112 GCC generates code to write to unchanging memory 10724 10725 Preprocessor 10726 10727 * [117]15067 Minor glitch in the source of cpp 10728 10729 Main driver program bugs 10730 10731 * [118]1963 collect2 interprets -oldstyle_liblookup as -o 10732 ldstyle_liblookup 10733 10734 x86-specific (Intel/AMD) 10735 10736 * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 10737 section} 10738 10739 HPPA-specific 10740 10741 * [120]14782 GCC produces an unaligned data access at -O2 10742 * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 10743 * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c 10744 10745 IA64-specific 10746 10747 * [123]14610 __float80 constants incorrectly emitted 10748 * [124]14813 init_array sections are initialized in the wrong order 10749 * [125]14857 GCC segfault on duplicated asm statement 10750 * [126]15598 Gcc 3.4 ICE on valid code 10751 * [127]15653 Gcc 3.4 ICE on valid code 10752 10753 MIPS-specific 10754 10755 * [128]15189 wrong filling of delay slot with -march=mips1 -G0 10756 -mno-split-addresses -mno-explicit-relocs 10757 * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 10758 2.14.91 10759 * [130]16144 Bogus reference to __divdf3 when -O1 10760 * [131]16176 Miscompilation of unaligned data in MIPS backend 10761 10762 PowerPC-specific 10763 10764 * [132]11591 ICE in gcc.dg/altivec-5.c 10765 * [133]12028 powerpc-eabispe produces bad sCOND operation 10766 * [134]14478 rs6000 geu/ltu patterns generate incorrect code 10767 * [135]14567 long double and va_arg complex args 10768 * [136]14715 Altivec stack layout may overlap gpr save with stack 10769 temps 10770 * [137]14902 (libstdc++) Stream checking functions fail when -pthread 10771 option is used. 10772 * [138]14924 Compiler ICE on valid code 10773 * [139]14960 -maltivec affects vector return with -mabi=no-altivec 10774 * [140]15106 vector varargs failure passing from altivec to 10775 non-altivec code for -m32 10776 * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 10777 half-word operation 10778 * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 10779 and stvx 10780 * [143]15662 Segmentation fault when an exception is thrown - even if 10781 try and catch are specified 10782 10783 s390-specific 10784 10785 * [144]15054 Bad code due to overlapping stack temporaries 10786 10787 SPARC-specific 10788 10789 * [145]15783 ICE with union assignment in 64-bit mode 10790 * [146]15626 GCC 3.4 emits "ld: warning: relocation error: 10791 R_SPARC_UA32" 10792 10793 x86-64-specific 10794 10795 * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 10796 * [148]14723 Backported -march=nocona from mainline 10797 * [149]15290 __float128 failed to pass to function properly 10798 10799 Cygwin/Mingw32-specific 10800 10801 * [150]15250 Option -mms-bitfields support on GCC 3.4 is not 10802 conformant to MS layout 10803 * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 10804 worker on windows32 targets 10805 10806 Bugs specific to embedded processors 10807 10808 * [152]8309 [m68k] -m5200 produces erroneous SImode set of short 10809 varaible on stack 10810 * [153]13250 [SH] Gcc code for rotation clobbers the register, but 10811 gcc continues to use the register as if it was not clobbered 10812 * [154]13803 [coldfire] movqi operand constraints too restrictivefor 10813 TARGET_COLDFIRE 10814 * [155]14093 [SH] ICE for code when using -mhitachi option in SH 10815 * [156]14457 [m6811hc] ICE with simple c++ source 10816 * [157]14542 [m6811hc] ICE on simple source 10817 * [158]15100 [SH] cc1plus got hang-up on 10818 libstdc++-v3/testsuite/abi_check.cc 10819 * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on 10820 cris-* 10821 * [160]15396 [SH] ICE with -O2 -fPIC 10822 * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 10823 ColdFire 10824 10825 Testsuite problems (compiler not affected) 10826 10827 * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely 10828 * [163]15488 (libstdc++) possibly insufficient file permissions for 10829 executing test suite 10830 * [164]15489 (libstdc++) testsuite_files determined incorrectly 10831 10832 Documentation bugs 10833 10834 * [165]13928 (libstdc++) no whatis info in some man pages generated 10835 by doxygen 10836 * [166]14150 Ada documentation out of date 10837 * [167]14949 (c++) Need to document method visibility changes 10838 * [168]15123 libstdc++-doc: Allocators.3 manpage is empty 10839 __________________________________________________________________ 10840 10841 GCC 3.4.2 10842 10843 Bug Fixes 10844 10845 This section lists the problem reports (PRs) from GCC's bug tracking 10846 system that are known to be fixed in the 3.4.2 release. This list might 10847 not be complete (that is, it is possible that some PRs that have been 10848 fixed are not listed here). 10849 10850 Bootstrap failures and issues 10851 10852 * [169]16469 [mips-sgi-irix5.3] bootstrap fails in 10853 libstdc++-v3/testsuite 10854 * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by 10855 profiledbootstrap does not work with the built compiler 10856 * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 10857 10858 Multi-platform internal compiler errors (ICEs) 10859 10860 * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 10861 cp_parser_class_specifier, in cp/parser.c 10862 * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 10863 * [174]15461 (c++) ICE due to NRV and inlining 10864 * [175]15890 (c++) ICE in c_expand_expr, in c-common.c 10865 * [176]16180 ICE: segmentation fault in RTL optimization 10866 * [177]16224 (c++) ICE in write_unscoped_name (template/namespace) 10867 * [178]16408 ICE: in delete_insn, in cfgrtl.c 10868 * [179]16529 (c++) ICE for: namespace-alias shall not be declared as 10869 the name of any other entity 10870 * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw 10871 * [181]16706 (c++) ICE in finish_member_declaration, in 10872 cp/semantics.c 10873 * [182]16810 (c++) Legal C++ program with cast gives ICE in 10874 build_ptrmemfunc 10875 * [183]16851 (c++) ICE when throwing a comma expression 10876 * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 10877 * [185]16904 (c++) ICE in finish_class_member_access_expr, in 10878 cp/typeck.c 10879 * [186]16905 (c++) ICE (segfault) with exceptions 10880 * [187]16964 (c++) ICE in cp_parser_class_specifier due to 10881 redefinition 10882 * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x' 10883 (identifier_node) in dependent_template_p, in cp/pt.c 10884 10885 Preprocessor bugs 10886 10887 * [189]16366 Preprocessor option -remap causes memory corruption 10888 10889 Optimization 10890 10891 * [190]15345 unreferenced nested inline functions not optimized away 10892 * [191]16590 Incorrect execution when compiling with -O2 10893 * [192]16693 Bitwise AND is lost when used within a cast to an enum 10894 of the same precision 10895 * [193]17078 Jump into if(0) substatement fails 10896 10897 Problems in generated debug information 10898 10899 * [194]13956 incorrect stabs for nested local variables 10900 10901 C front end bugs 10902 10903 * [195]16684 GCC should not warn about redundant redeclarations of 10904 built-ins 10905 10906 C++ compiler and library 10907 10908 * [196]12658 Thread safety problems in locale::global() and 10909 locale::locale() 10910 * [197]13092 g++ accepts invalid pointer-to-member conversion 10911 * [198]15320 Excessive memory consumption 10912 * [199]16246 Incorrect template argument deduction 10913 * [200]16273 Memory exhausted when using nested classes and virtual 10914 functions 10915 * [201]16401 ostringstream in gcc 3.4.x very slow for big data 10916 * [202]16411 undefined reference to 10917 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 10918 >::file() 10919 * [203]16489 G++ incorrectly rejects use of a null constant integral 10920 expression as a null constant pointer 10921 * [204]16618 offsetof fails with constant member 10922 * [205]16637 syntax error reported for valid input code 10923 * [206]16717 __attribute__((constructor)) broken in C++ 10924 * [207]16813 compiler error in DEBUG version of range insertion 10925 std::map::insert 10926 * [208]16853 pointer-to-member initialization from incompatible one 10927 accepted 10928 * [209]16889 ambiguity is not detected 10929 * [210]16959 Segmentation fault in ios_base::sync_with_stdio 10930 10931 Java compiler and library 10932 10933 * [211]7587 direct threaded interpreter not thread-safe 10934 * [212]16473 ServerSocket accept() leaks file descriptors 10935 * [213]16478 Hash synchronization deadlock with finalizers 10936 10937 Alpha-specific 10938 10939 * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 10940 * [215]16974 could not split insn (ice in final_scan_insn, in 10941 final.c) 10942 10943 x86-specific 10944 10945 * [216]16298 ICE in output_operand 10946 * [217]17113 ICE with SSE2 intrinsics 10947 10948 x86-64 specific 10949 10950 * [218]14697 libstdc++ couldn't find 32bit libgcc_s 10951 10952 MIPS-specific 10953 10954 * [219]15869 [mips64] No NOP after LW (with -mips1 -O0) 10955 * [220]16325 [mips64] value profiling clobbers gp on mips 10956 * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 10957 char[]s 10958 * [222]16380 [mips64] Use of uninitialised register after dbra 10959 conversion 10960 * [223]16407 [mips64] Unaligned access to local variables 10961 * [224]16643 [mips64] verify_local_live_at_start ICE after 10962 crossjumping & cfgcleanup 10963 10964 ARM-specific 10965 10966 * [225]15927 THUMB -O2: strength-reduced iteration variable ends up 10967 off by 1 10968 * [226]15948 THUMB: ICE with non-commutative cbranch 10969 * [227]17019 THUMB: bad switch statement in md code for 10970 addsi3_cbranch_scratch 10971 10972 IA64-specific 10973 10974 * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 10975 (-mtune=merced) 10976 * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 10977 (-mtune=itanium) 10978 * [230]16278 Gcc failed to build Linux kernel with -mtune=merced 10979 * [231]16414 ICE on valid code: typo in comparison of asm_noperands 10980 result 10981 * [232]16445 ICE on valid code: don't count ignored insns 10982 * [233]16490 ICE (segfault) while compiling with -fprofile-use 10983 * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 10984 10985 PowerPC-specific 10986 10987 * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 10988 * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 10989 issue) 10990 10991 SPARC-specific 10992 10993 * [237]16199 ICE while compiling apache 2.0.49 10994 * [238]16416 -m64 doesn't imply -mcpu=v9 anymore 10995 * [239]16430 ICE when returning non-C aggregates larger than 16 bytes 10996 10997 Bugs specific to embedded processors 10998 10999 * [240]16379 [m32r] can't output large model function call of memcpy 11000 * [241]17093 [m32r] ICE with -msdata=use -O0 11001 * [242]17119 [m32r] ICE at switch case 0x8000 11002 11003 DJGPP-specific 11004 11005 * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 11006 11007 Alpha Tru64-specific 11008 11009 * [244]16210 libstdc++ gratuitously omits "long long" I/O 11010 11011 Testsuite, documentation issues (compiler is not affected): 11012 11013 * [245]15488 (libstdc++) possibly insufficient file permissions for 11014 executing test suite 11015 * [246]16250 ada/doctools runs makeinfo even in release tarball 11016 __________________________________________________________________ 11017 11018 GCC 3.4.3 11019 11020 This is the [247]list of problem reports (PRs) from GCC's bug tracking 11021 system that are known to be fixed in the 3.4.3 release. This list might 11022 not be complete (that is, it is possible that some PRs that have been 11023 fixed are not listed here). 11024 11025 Bootstrap failures 11026 11027 * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 11028 * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 11029 when undeclared 11030 11031 Internal compiler errors (ICEs) affecting multiple platforms 11032 11033 * [250]13948 (java) GCJ segmentation fault while compiling GL4Java 11034 .class files 11035 * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 11036 * [252]16301 (c++) ICE when "strong" attribute is attached to a using 11037 directive 11038 * [253]16566 ICE with flexible arrays 11039 * [254]17023 ICE with nested functions in parameter declaration 11040 * [255]17027 ICE with noreturn function in loop at -O2 11041 * [256]17524 ICE in grokdeclarator, in cp/decl.c 11042 * [257]17826 (c++) ICE in cp_tree_equal 11043 11044 C and optimization bugs 11045 11046 * [258]15526 -ftrapv aborts on 0 * (-1) 11047 * [259]16999 #ident stopped working 11048 * [260]17503 quadratic behaviour in invalid_mode_change_p 11049 * [261]17581 Long long arithmetic fails inside a switch/case 11050 statement when compiled with -O2 11051 * [262]18129 -fwritable-strings doesn't work 11052 11053 C++ compiler and library bugs 11054 11055 * [263]10975 incorrect initial ostringstream::tellp() 11056 * [264]11722 Unbuffered filebuf::sgetn is slow 11057 * [265]14534 Unrecognizing static function as a template parameter 11058 when its return value is also templated 11059 * [266]15172 Copy constructor optimization in aggregate 11060 initialization 11061 * [267]15786 Bad error message for frequently occuring error. 11062 * [268]16162 Rejects valid member-template-definition 11063 * [269]16612 empty basic_strings can't live in shared memory 11064 * [270]16715 std::basic_iostream is instantiated when used, even 11065 though instantiations are already contained in libstdc++ 11066 * [271]16848 code in /ext/demangle.h appears broken 11067 * [272]17132 GCC fails to eliminate function template specialization 11068 when argument deduction fails 11069 * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 11070 in ropeimpl.h 11071 * [274]17327 use of `enumeral_type' in template type unification 11072 * [275]17393 "unused variable '._0'" warning with -Wall 11073 * [276]17501 Confusion with member templates 11074 * [277]17537 g++ not passing -lstdc++ to linker when all command line 11075 arguments are libraries 11076 * [278]17585 usage of unqualified name of static member from within 11077 class not allowed 11078 * [279]17821 Poor diagnostic for using "." instead of "->" 11079 * [280]17829 wrong error: call of overloaded function is ambiguous 11080 * [281]17851 Misleading diagnostic for invalid function declarations 11081 with undeclared types 11082 * [282]17976 Destructor is called twice 11083 * [283]18020 rejects valid definition of enum value in template 11084 * [284]18093 bogus conflict in namespace aliasing 11085 * [285]18140 C++ parser bug when using >> in templates 11086 11087 Fortran 11088 11089 * [286]17541 data statements with double precision constants fail 11090 11091 x86-specific 11092 11093 * [287]17853 -O2 ICE for MMX testcase 11094 11095 SPARC-specific 11096 11097 * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c 11098 11099 Darwin-specific 11100 11101 * [289]17167 FATAL:Symbol L_foo$stub already defined. 11102 11103 AIX-specific 11104 11105 * [290]17277 could not catch an exception when specified -maix64 11106 11107 Solaris-specific 11108 11109 * [291]17505 <cmath> calls acosf(), ceilf(), and other functions 11110 missing from system libraries 11111 11112 HP/UX specific: 11113 11114 * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 11115 11116 ARM-specific 11117 11118 * [293]17384 ICE with mode attribute on structures 11119 11120 MIPS-specific 11121 11122 * [294]17770 No NOP after LWL with -mips1 11123 11124 Other embedded target specific 11125 11126 * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 11127 * [296]14064 [avr-elf] -fdata-sections triggers ICE 11128 * [297]14678 [m68hc11-elf] gcc ICE 11129 * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 11130 * [299]15790 [i686-coff] Alignment error building gcc with i686-coff 11131 target 11132 * [300]15886 [SH] Miscompilation with -O2 -fPIC 11133 * [301]16884 [avr-elf] [fweb related] bug while initializing 11134 variables 11135 11136 Bugs relating to debugger support 11137 11138 * [302]13841 missing debug info for _Complex function arguments 11139 * [303]15860 [big-endian targets] No DW_AT_location debug info is 11140 emitted for formal arguments to a function that uses "register" 11141 qualifiers 11142 11143 Testsuite issues (compiler not affected) 11144 11145 * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 11146 * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 11147 * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 11148 testsuite 11149 11150 Documentation 11151 11152 * [307]15498 typo in gcc manual: non-existing locale example en_UK, 11153 should be en_GB 11154 * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 11155 document broken shell 11156 * [309]16406 USE_LD_AS_NEEDED undocumented 11157 __________________________________________________________________ 11158 11159 GCC 3.4.4 11160 11161 This is the [310]list of problem reports (PRs) from GCC's bug tracking 11162 system that are known to be fixed in the 3.4.4 release. This list might 11163 not be complete (that is, it is possible that some PRs that have been 11164 fixed are not listed here). 11165 __________________________________________________________________ 11166 11167 GCC 3.4.5 11168 11169 This is the [311]list of problem reports (PRs) from GCC's bug tracking 11170 system that are known to be fixed in the 3.4.5 release. This list might 11171 not be complete (that is, it is possible that some PRs that have been 11172 fixed are not listed here). 11173 11174 Bootstrap issues 11175 11176 * [312]24688 sco_math fixincl breaks math.h 11177 11178 C compiler bugs 11179 11180 * [313]17188 struct Foo { } redefinition 11181 * [314]20187 wrong code for ((unsigned char)(unsigned long 11182 long)((a?a:1)&(a*b)))?0:1) 11183 * [315]21873 infinite warning loop on bad array initializer 11184 * [316]21899 enum definition accepts values to be overriden 11185 * [317]22061 ICE in find_function_data, in function.c 11186 * [318]22308 Failure to diagnose violation of constraint 6.516p2 11187 * [319]22458 ICE on missing brace 11188 * [320]22589 ICE casting to long long 11189 * [321]24101 Segfault with preprocessed source 11190 11191 C++ compiler and library bugs 11192 11193 * [322]10611 operations on vector mode not recognized in C++ 11194 * [323]13377 unexpected behavior of namespace usage directive 11195 * [324]16002 Strange error message with new parser 11196 * [325]17413 local classes as template argument 11197 * [326]17609 spurious error message after using keyword 11198 * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 11199 * [328]18124 ICE with invalid template template parameter 11200 * [329]18155 typedef in template declaration not rejected 11201 * [330]18177 ICE with const_cast for undeclared variable 11202 * [331]18368 C++ error message regression 11203 * [332]16378 ICE when returning a copy of a packed member 11204 * [333]18466 int ::i; accepted 11205 * [334]18512 ICE on invalid usage of template base class 11206 * [335]18454 ICE when returning undefined type 11207 * [336]18738 typename not allowed with non-dependent qualified name 11208 * [337]18803 rejects access to operator() in template 11209 * [338]19004 ICE in uses_template_parms, in cp/pt.c 11210 * [339]19208 Spurious error about variably modified type 11211 * [340]18253 bad error message / ICE for invalid template parameter 11212 * [341]19608 ICE after friend function definition in local class 11213 * [342]19884 ICE on explicit instantiation of a non-template 11214 constructor 11215 * [343]20153 ICE when C++ template function contains anonymous union 11216 * [344]20563 Infinite loop in diagnostic (and ice after error 11217 message) 11218 * [345]20789 ICE with incomplete type in template 11219 * [346]21336 Internal compiler error when using custom new operators 11220 * [347]21768 ICE in error message due to violation of coding 11221 conventions 11222 * [348]21853 constness of pointer to data member ignored 11223 * [349]21903 Default argument of template function causes a 11224 compile-time error 11225 * [350]21983 multiple diagnostics 11226 * [351]21987 New testsuite failure 11227 g++.dg/warn/conversion-function-1.C 11228 * [352]22153 ICE on invalid template specialization 11229 * [353]22172 Internal compiler error, seg fault. 11230 * [354]21286 filebuf::xsgetn vs pipes 11231 * [355]22233 ICE with wrong number of template parameters 11232 * [356]22508 ICE after invalid operator new 11233 * [357]22545 ICE with pointer to class member & user defined 11234 conversion operator 11235 * [358]23528 Wrong default allocator in ext/hash_map 11236 * [359]23550 char_traits requirements/1.cc test bad math 11237 * [360]23586 Bad diagnostic for invalid namespace-name 11238 * [361]23624 ICE in invert_truthvalue, in fold-const.c 11239 * [362]23639 Bad error message: not a member of '<declaration error>' 11240 * [363]23797 ICE on typename outside template 11241 * [364]23965 Bogus error message: no matching function for call to 11242 'foo(<type error>)' 11243 * [365]24052 &#`label_decl' not supported by dump_expr#<expression 11244 error> 11245 * [366]24580 virtual base class cause exception not to be caught 11246 11247 Problems in generated debug information 11248 11249 * [367]24267 Bad DWARF for altivec vectors 11250 11251 Optimizations issues 11252 11253 * [368]17810 ICE in verify_local_live_at_start 11254 * [369]17860 Wrong generated code for loop with varying bound 11255 * [370]21709 ICE on compile-time complex NaN 11256 * [371]21964 broken tail call at -O2 or more 11257 * [372]22167 Strange optimization bug when using -Os 11258 * [373]22619 Compilation failure for real_const_1.f and 11259 real_const_2.f90 11260 * [374]23241 Invalid code generated for comparison of uchar to 255 11261 * [375]23478 Miscompilation due to reloading of a var that is also 11262 used in EH pad 11263 * [376]24470 segmentation fault in cc1plus when compiling with -O 11264 * [377]24950 ICE in operand_subword_force 11265 11266 Precompiled headers problems 11267 11268 * [378]14400 Cannot compile qt-x11-free-3.3.0 11269 * [379]14940 PCH largefile test fails on various platforms 11270 11271 Preprocessor bugs 11272 11273 * [380]20239 ICE on empty preprocessed input 11274 * [381]15220 "gcc -E -MM -MG" reports missing system headers in 11275 source directory 11276 11277 Testsuite issues 11278 11279 * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 11280 i686-pc-linux-gnu 11281 11282 Alpha specific 11283 11284 * [383]21888 bootstrap failure with linker relaxation enabled 11285 11286 ARM specific 11287 11288 * [384]15342 [arm-linux]: ICE in verify_local_live_at_start 11289 * [385]23985 Memory aliasing information incorrect in inlined memcpy 11290 11291 ColdFile specific 11292 11293 * [386]16719 Illegal move of byte into address register causes 11294 compiler to ICE 11295 11296 HPPA specific 11297 11298 * [387]21723 ICE while building libgfortran 11299 * [388]21841 -mhp-ld/-mgnu-ld documentation 11300 11301 IA-64 specific 11302 11303 * [389]23644 IA-64 hardware models and configuration options 11304 documentation error 11305 * [390]24718 Shared libgcc not used for linking by default 11306 11307 M68000 specific 11308 11309 * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c 11310 11311 MIPS specific 11312 11313 * [392]20621 ICE in change_address_1, in emit-rtl.c 11314 11315 PowerPC and PowerPC64 specific 11316 11317 * [393]18583 error on valid code: const 11318 __attribute__((altivec(vector__))) doesn't work in arrays 11319 * [394]20191 ICE in reload_cse_simplify_operands 11320 * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 11321 * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 11322 * [397]23404 gij trashes args of functions with more than 8 fp args 11323 * [398]23539 C & C++ compiler generating misaligned references 11324 regardless of compiler flags 11325 * [399]24102 floatdisf2_internal2 broken 11326 * [400]24465 -mminimal-toc miscompilation of __thread vars 11327 11328 Solaris specific 11329 11330 * [401]19933 Problem with define of HUGE_VAL in math_c99 11331 * [402]21889 Native Solaris assembler cannot grok DTP-relative debug 11332 symbols 11333 11334 SPARC specific 11335 11336 * [403]19300 PCH failures on sparc-linux 11337 * [404]20301 Assembler labels have a leading "-" 11338 * [405]20673 C PCH testsuite assembly comparison failure 11339 11340 x86 and x86_64 specific 11341 11342 * [406]18582 ICE with arrays of type V2DF 11343 * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 11344 -fsched2-use-traces 11345 * [408]21716 ICE in reg-stack.c's swap_rtx_condition 11346 * [409]24315 amd64 fails -fpeephole2 11347 __________________________________________________________________ 11348 11349 GCC 3.4.6 11350 11351 This is the [410]list of problem reports (PRs) from GCC's bug tracking 11352 system that are known to be fixed in the 3.4.6 release. This list might 11353 not be complete (that is, it is possible that some PRs that have been 11354 fixed are not listed here). 11355 11356 11357 For questions related to the use of GCC, please consult these web 11358 pages and the [411]GCC manuals. If that fails, the 11359 [412]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11360 web pages and the development of GCC are welcome on our developer 11361 list at [413]gcc (a] gcc.gnu.org. All of [414]our lists have public 11362 archives. 11363 11364 Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and 11365 distribution of this entire article is permitted in any medium, 11366 provided this notice is preserved. 11367 11368 These pages are [416]maintained by the GCC team. Last modified 11369 2018-09-30[417]. 11370 11371 References 11372 11373 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 11374 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 11375 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 11376 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 11377 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html 11378 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 11379 7. https://www.boost.org/ 11380 8. https://gcc.gnu.org/PR11953 11381 9. https://gcc.gnu.org/PR8361 11382 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other Builtins 11383 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 11384 12. http://gcc.gnu.org/bugs/#cxx_rvalbind 11385 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 11386 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 11387 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 11388 16. http://www.gnu.org/software/classpath/ 11389 17. http://www.eclipse.org/ 11390 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 11391 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 11392 20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 11393 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 11394 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 11395 23. http://gcc.gnu.org/gcc-3.4/mips-abi.html 11396 24. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html 11397 25. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 11398 26. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=\[3\.4.*[Rr]egression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED 11399 27. https://gcc.gnu.org/PR10129 11400 28. https://gcc.gnu.org/PR14576 11401 29. https://gcc.gnu.org/PR14760 11402 30. https://gcc.gnu.org/PR14671 11403 31. https://gcc.gnu.org/PR15093 11404 32. https://gcc.gnu.org/PR15178 11405 33. https://gcc.gnu.org/PR12753 11406 34. https://gcc.gnu.org/PR13985 11407 35. https://gcc.gnu.org/PR14810 11408 36. https://gcc.gnu.org/PR14883 11409 37. https://gcc.gnu.org/PR15044 11410 38. https://gcc.gnu.org/PR15057 11411 39. https://gcc.gnu.org/PR15064 11412 40. https://gcc.gnu.org/PR15142 11413 41. https://gcc.gnu.org/PR15159 11414 42. https://gcc.gnu.org/PR15165 11415 43. https://gcc.gnu.org/PR15193 11416 44. https://gcc.gnu.org/PR15209 11417 45. https://gcc.gnu.org/PR15227 11418 46. https://gcc.gnu.org/PR15285 11419 47. https://gcc.gnu.org/PR15299 11420 48. https://gcc.gnu.org/PR15329 11421 49. https://gcc.gnu.org/PR15550 11422 50. https://gcc.gnu.org/PR15554 11423 51. https://gcc.gnu.org/PR15640 11424 52. https://gcc.gnu.org/PR15666 11425 53. https://gcc.gnu.org/PR15696 11426 54. https://gcc.gnu.org/PR15701 11427 55. https://gcc.gnu.org/PR15761 11428 56. https://gcc.gnu.org/PR15829 11429 57. https://gcc.gnu.org/PR14538 11430 58. https://gcc.gnu.org/PR12391 11431 59. https://gcc.gnu.org/PR14649 11432 60. https://gcc.gnu.org/PR15004 11433 61. https://gcc.gnu.org/PR15749 11434 62. https://gcc.gnu.org/PR10646 11435 63. https://gcc.gnu.org/PR12077 11436 64. https://gcc.gnu.org/PR13598 11437 65. https://gcc.gnu.org/PR14211 11438 66. https://gcc.gnu.org/PR14220 11439 67. https://gcc.gnu.org/PR14245 11440 68. https://gcc.gnu.org/PR14340 11441 69. https://gcc.gnu.org/PR14600 11442 70. https://gcc.gnu.org/PR14668 11443 71. https://gcc.gnu.org/PR14775 11444 72. https://gcc.gnu.org/PR14821 11445 73. https://gcc.gnu.org/PR14930 11446 74. https://gcc.gnu.org/PR14932 11447 75. https://gcc.gnu.org/PR14950 11448 76. https://gcc.gnu.org/PR14962 11449 77. https://gcc.gnu.org/PR14975 11450 78. https://gcc.gnu.org/PR15002 11451 79. https://gcc.gnu.org/PR15025 11452 80. https://gcc.gnu.org/PR15046 11453 81. https://gcc.gnu.org/PR15069 11454 82. https://gcc.gnu.org/PR15074 11455 83. https://gcc.gnu.org/PR15083 11456 84. https://gcc.gnu.org/PR15096 11457 85. https://gcc.gnu.org/PR15287 11458 86. https://gcc.gnu.org/PR15317 11459 87. https://gcc.gnu.org/PR15337 11460 88. https://gcc.gnu.org/PR15361 11461 89. https://gcc.gnu.org/PR15412 11462 90. https://gcc.gnu.org/PR15427 11463 91. https://gcc.gnu.org/PR15471 11464 92. https://gcc.gnu.org/PR15503 11465 93. https://gcc.gnu.org/PR15507 11466 94. https://gcc.gnu.org/PR15542 11467 95. https://gcc.gnu.org/PR15565 11468 96. https://gcc.gnu.org/PR15625 11469 97. https://gcc.gnu.org/PR15629 11470 98. https://gcc.gnu.org/PR15742 11471 99. https://gcc.gnu.org/PR15775 11472 100. https://gcc.gnu.org/PR15821 11473 101. https://gcc.gnu.org/PR15862 11474 102. https://gcc.gnu.org/PR15875 11475 103. https://gcc.gnu.org/PR15877 11476 104. https://gcc.gnu.org/PR15947 11477 105. https://gcc.gnu.org/PR16020 11478 106. https://gcc.gnu.org/PR16154 11479 107. https://gcc.gnu.org/PR16174 11480 108. https://gcc.gnu.org/PR14315 11481 109. https://gcc.gnu.org/PR15151 11482 110. https://gcc.gnu.org/PR7993 11483 111. https://gcc.gnu.org/PR15228 11484 112. https://gcc.gnu.org/PR15345 11485 113. https://gcc.gnu.org/PR15945 11486 114. https://gcc.gnu.org/PR15526 11487 115. https://gcc.gnu.org/PR14690 11488 116. https://gcc.gnu.org/PR15112 11489 117. https://gcc.gnu.org/PR15067 11490 118. https://gcc.gnu.org/PR1963 11491 119. https://gcc.gnu.org/PR15717 11492 120. https://gcc.gnu.org/PR14782 11493 121. https://gcc.gnu.org/PR14828 11494 122. https://gcc.gnu.org/PR15202 11495 123. https://gcc.gnu.org/PR14610 11496 124. https://gcc.gnu.org/PR14813 11497 125. https://gcc.gnu.org/PR14857 11498 126. https://gcc.gnu.org/PR15598 11499 127. https://gcc.gnu.org/PR15653 11500 128. https://gcc.gnu.org/PR15189 11501 129. https://gcc.gnu.org/PR15331 11502 130. https://gcc.gnu.org/PR16144 11503 131. https://gcc.gnu.org/PR16176 11504 132. https://gcc.gnu.org/PR11591 11505 133. https://gcc.gnu.org/PR12028 11506 134. https://gcc.gnu.org/PR14478 11507 135. https://gcc.gnu.org/PR14567 11508 136. https://gcc.gnu.org/PR14715 11509 137. https://gcc.gnu.org/PR14902 11510 138. https://gcc.gnu.org/PR14924 11511 139. https://gcc.gnu.org/PR14960 11512 140. https://gcc.gnu.org/PR15106 11513 141. https://gcc.gnu.org/PR16026 11514 142. https://gcc.gnu.org/PR15191 11515 143. https://gcc.gnu.org/PR15662 11516 144. https://gcc.gnu.org/PR15054 11517 145. https://gcc.gnu.org/PR15783 11518 146. https://gcc.gnu.org/PR15626 11519 147. https://gcc.gnu.org/PR14326 11520 148. https://gcc.gnu.org/PR14723 11521 149. https://gcc.gnu.org/PR15290 11522 150. https://gcc.gnu.org/PR15250 11523 151. https://gcc.gnu.org/PR15551 11524 152. https://gcc.gnu.org/PR8309 11525 153. https://gcc.gnu.org/PR13250 11526 154. https://gcc.gnu.org/PR13803 11527 155. https://gcc.gnu.org/PR14093 11528 156. https://gcc.gnu.org/PR14457 11529 157. https://gcc.gnu.org/PR14542 11530 158. https://gcc.gnu.org/PR15100 11531 159. https://gcc.gnu.org/PR15296 11532 160. https://gcc.gnu.org/PR15396 11533 161. https://gcc.gnu.org/PR15782 11534 162. https://gcc.gnu.org/PR11610 11535 163. https://gcc.gnu.org/PR15488 11536 164. https://gcc.gnu.org/PR15489 11537 165. https://gcc.gnu.org/PR13928 11538 166. https://gcc.gnu.org/PR14150 11539 167. https://gcc.gnu.org/PR14949 11540 168. https://gcc.gnu.org/PR15123 11541 169. https://gcc.gnu.org/PR16469 11542 170. https://gcc.gnu.org/PR16344 11543 171. https://gcc.gnu.org/PR16842 11544 172. https://gcc.gnu.org/PR12608 11545 173. https://gcc.gnu.org/PR14492 11546 174. https://gcc.gnu.org/PR15461 11547 175. https://gcc.gnu.org/PR15890 11548 176. https://gcc.gnu.org/PR16180 11549 177. https://gcc.gnu.org/PR16224 11550 178. https://gcc.gnu.org/PR16408 11551 179. https://gcc.gnu.org/PR16529 11552 180. https://gcc.gnu.org/PR16698 11553 181. https://gcc.gnu.org/PR16706 11554 182. https://gcc.gnu.org/PR16810 11555 183. https://gcc.gnu.org/PR16851 11556 184. https://gcc.gnu.org/PR16870 11557 185. https://gcc.gnu.org/PR16904 11558 186. https://gcc.gnu.org/PR16905 11559 187. https://gcc.gnu.org/PR16964 11560 188. https://gcc.gnu.org/PR17068 11561 189. https://gcc.gnu.org/PR16366 11562 190. https://gcc.gnu.org/PR15345 11563 191. https://gcc.gnu.org/PR16590 11564 192. https://gcc.gnu.org/PR16693 11565 193. https://gcc.gnu.org/PR17078 11566 194. https://gcc.gnu.org/PR13956 11567 195. https://gcc.gnu.org/PR16684 11568 196. https://gcc.gnu.org/PR12658 11569 197. https://gcc.gnu.org/PR13092 11570 198. https://gcc.gnu.org/PR15320 11571 199. https://gcc.gnu.org/PR16246 11572 200. https://gcc.gnu.org/PR16273 11573 201. https://gcc.gnu.org/PR16401 11574 202. https://gcc.gnu.org/PR16411 11575 203. https://gcc.gnu.org/PR16489 11576 204. https://gcc.gnu.org/PR16618 11577 205. https://gcc.gnu.org/PR16637 11578 206. https://gcc.gnu.org/PR16717 11579 207. https://gcc.gnu.org/PR16813 11580 208. https://gcc.gnu.org/PR16853 11581 209. https://gcc.gnu.org/PR16889 11582 210. https://gcc.gnu.org/PR16959 11583 211. https://gcc.gnu.org/PR7587 11584 212. https://gcc.gnu.org/PR16473 11585 213. https://gcc.gnu.org/PR16478 11586 214. https://gcc.gnu.org/PR10695 11587 215. https://gcc.gnu.org/PR16974 11588 216. https://gcc.gnu.org/PR16298 11589 217. https://gcc.gnu.org/PR17113 11590 218. https://gcc.gnu.org/PR14697 11591 219. https://gcc.gnu.org/PR15869 11592 220. https://gcc.gnu.org/PR16325 11593 221. https://gcc.gnu.org/PR16357 11594 222. https://gcc.gnu.org/PR16380 11595 223. https://gcc.gnu.org/PR16407 11596 224. https://gcc.gnu.org/PR16643 11597 225. https://gcc.gnu.org/PR15927 11598 226. https://gcc.gnu.org/PR15948 11599 227. https://gcc.gnu.org/PR17019 11600 228. https://gcc.gnu.org/PR16130 11601 229. https://gcc.gnu.org/PR16142 11602 230. https://gcc.gnu.org/PR16278 11603 231. https://gcc.gnu.org/PR16414 11604 232. https://gcc.gnu.org/PR16445 11605 233. https://gcc.gnu.org/PR16490 11606 234. https://gcc.gnu.org/PR16683 11607 235. https://gcc.gnu.org/PR16195 11608 236. https://gcc.gnu.org/PR16239 11609 237. https://gcc.gnu.org/PR16199 11610 238. https://gcc.gnu.org/PR16416 11611 239. https://gcc.gnu.org/PR16430 11612 240. https://gcc.gnu.org/PR16379 11613 241. https://gcc.gnu.org/PR17093 11614 242. https://gcc.gnu.org/PR17119 11615 243. https://gcc.gnu.org/PR15928 11616 244. https://gcc.gnu.org/PR16210 11617 245. https://gcc.gnu.org/PR15488 11618 246. https://gcc.gnu.org/PR16250 11619 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 11620 248. https://gcc.gnu.org/PR17369 11621 249. https://gcc.gnu.org/PR17850 11622 250. https://gcc.gnu.org/PR13948 11623 251. https://gcc.gnu.org/PR14492 11624 252. https://gcc.gnu.org/PR16301 11625 253. https://gcc.gnu.org/PR16566 11626 254. https://gcc.gnu.org/PR17023 11627 255. https://gcc.gnu.org/PR17027 11628 256. https://gcc.gnu.org/PR17524 11629 257. https://gcc.gnu.org/PR17826 11630 258. https://gcc.gnu.org/PR15526 11631 259. https://gcc.gnu.org/PR16999 11632 260. https://gcc.gnu.org/PR17503 11633 261. https://gcc.gnu.org/PR17581 11634 262. https://gcc.gnu.org/PR18129 11635 263. https://gcc.gnu.org/PR10975 11636 264. https://gcc.gnu.org/PR11722 11637 265. https://gcc.gnu.org/PR14534 11638 266. https://gcc.gnu.org/PR15172 11639 267. https://gcc.gnu.org/PR15786 11640 268. https://gcc.gnu.org/PR16162 11641 269. https://gcc.gnu.org/PR16612 11642 270. https://gcc.gnu.org/PR16715 11643 271. https://gcc.gnu.org/PR16848 11644 272. https://gcc.gnu.org/PR17132 11645 273. https://gcc.gnu.org/PR17259 11646 274. https://gcc.gnu.org/PR17327 11647 275. https://gcc.gnu.org/PR17393 11648 276. https://gcc.gnu.org/PR17501 11649 277. https://gcc.gnu.org/PR17537 11650 278. https://gcc.gnu.org/PR17585 11651 279. https://gcc.gnu.org/PR17821 11652 280. https://gcc.gnu.org/PR17829 11653 281. https://gcc.gnu.org/PR17851 11654 282. https://gcc.gnu.org/PR17976 11655 283. https://gcc.gnu.org/PR18020 11656 284. https://gcc.gnu.org/PR18093 11657 285. https://gcc.gnu.org/PR18140 11658 286. https://gcc.gnu.org/PR17541 11659 287. https://gcc.gnu.org/PR17853 11660 288. https://gcc.gnu.org/PR17245 11661 289. https://gcc.gnu.org/PR17167 11662 290. https://gcc.gnu.org/PR17277 11663 291. https://gcc.gnu.org/PR17505 11664 292. https://gcc.gnu.org/PR17684 11665 293. https://gcc.gnu.org/PR17384 11666 294. https://gcc.gnu.org/PR17770 11667 295. https://gcc.gnu.org/PR11476 11668 296. https://gcc.gnu.org/PR14064 11669 297. https://gcc.gnu.org/PR14678 11670 298. https://gcc.gnu.org/PR15583 11671 299. https://gcc.gnu.org/PR15790 11672 300. https://gcc.gnu.org/PR15886 11673 301. https://gcc.gnu.org/PR16884 11674 302. https://gcc.gnu.org/PR13841 11675 303. https://gcc.gnu.org/PR15860 11676 304. https://gcc.gnu.org/PR17465 11677 305. https://gcc.gnu.org/PR17469 11678 306. https://gcc.gnu.org/PR18138 11679 307. https://gcc.gnu.org/PR15498 11680 308. https://gcc.gnu.org/PR15747 11681 309. https://gcc.gnu.org/PR16406 11682 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 11683 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 11684 312. https://gcc.gnu.org/PR24688 11685 313. https://gcc.gnu.org/PR17188 11686 314. https://gcc.gnu.org/PR20187 11687 315. https://gcc.gnu.org/PR21873 11688 316. https://gcc.gnu.org/PR21899 11689 317. https://gcc.gnu.org/PR22061 11690 318. https://gcc.gnu.org/PR22208 11691 319. https://gcc.gnu.org/PR22458 11692 320. https://gcc.gnu.org/PR22589 11693 321. https://gcc.gnu.org/PR24101 11694 322. https://gcc.gnu.org/PR10611 11695 323. https://gcc.gnu.org/PR13377 11696 324. https://gcc.gnu.org/PR16002 11697 325. https://gcc.gnu.org/PR17413 11698 326. https://gcc.gnu.org/PR17609 11699 327. https://gcc.gnu.org/PR17618 11700 328. https://gcc.gnu.org/PR18124 11701 329. https://gcc.gnu.org/PR18155 11702 330. https://gcc.gnu.org/PR18177 11703 331. https://gcc.gnu.org/PR18368 11704 332. https://gcc.gnu.org/PR18378 11705 333. https://gcc.gnu.org/PR18466 11706 334. https://gcc.gnu.org/PR18512 11707 335. https://gcc.gnu.org/PR18545 11708 336. https://gcc.gnu.org/PR18738 11709 337. https://gcc.gnu.org/PR18803 11710 338. https://gcc.gnu.org/PR19004 11711 339. https://gcc.gnu.org/PR19208 11712 340. https://gcc.gnu.org/PR19253 11713 341. https://gcc.gnu.org/PR19608 11714 342. https://gcc.gnu.org/PR19884 11715 343. https://gcc.gnu.org/PR20153 11716 344. https://gcc.gnu.org/PR20563 11717 345. https://gcc.gnu.org/PR20789 11718 346. https://gcc.gnu.org/PR21336 11719 347. https://gcc.gnu.org/PR21768 11720 348. https://gcc.gnu.org/PR21853 11721 349. https://gcc.gnu.org/PR21903 11722 350. https://gcc.gnu.org/PR21983 11723 351. https://gcc.gnu.org/PR21987 11724 352. https://gcc.gnu.org/PR22153 11725 353. https://gcc.gnu.org/PR22172 11726 354. https://gcc.gnu.org/PR21286 11727 355. https://gcc.gnu.org/PR22233 11728 356. https://gcc.gnu.org/PR22508 11729 357. https://gcc.gnu.org/PR22545 11730 358. https://gcc.gnu.org/PR23528 11731 359. https://gcc.gnu.org/PR23550 11732 360. https://gcc.gnu.org/PR23586 11733 361. https://gcc.gnu.org/PR23624 11734 362. https://gcc.gnu.org/PR23639 11735 363. https://gcc.gnu.org/PR23797 11736 364. https://gcc.gnu.org/PR23965 11737 365. https://gcc.gnu.org/PR24052 11738 366. https://gcc.gnu.org/PR24580 11739 367. https://gcc.gnu.org/PR24267 11740 368. https://gcc.gnu.org/PR17810 11741 369. https://gcc.gnu.org/PR17860 11742 370. https://gcc.gnu.org/PR21709 11743 371. https://gcc.gnu.org/PR21964 11744 372. https://gcc.gnu.org/PR22167 11745 373. https://gcc.gnu.org/PR22619 11746 374. https://gcc.gnu.org/PR23241 11747 375. https://gcc.gnu.org/PR23478 11748 376. https://gcc.gnu.org/PR24470 11749 377. https://gcc.gnu.org/PR24950 11750 378. https://gcc.gnu.org/PR14400 11751 379. https://gcc.gnu.org/PR14940 11752 380. https://gcc.gnu.org/PR20239 11753 381. https://gcc.gnu.org/PR15220 11754 382. https://gcc.gnu.org/PR19275 11755 383. https://gcc.gnu.org/PR21888 11756 384. https://gcc.gnu.org/PR15342 11757 385. https://gcc.gnu.org/PR23985 11758 386. https://gcc.gnu.org/PR16719 11759 387. https://gcc.gnu.org/PR21723 11760 388. https://gcc.gnu.org/PR21841 11761 389. https://gcc.gnu.org/PR23644 11762 390. https://gcc.gnu.org/PR24718 11763 391. https://gcc.gnu.org/PR18421 11764 392. https://gcc.gnu.org/PR20621 11765 393. https://gcc.gnu.org/PR18583 11766 394. https://gcc.gnu.org/PR20191 11767 395. https://gcc.gnu.org/PR22083 11768 396. https://gcc.gnu.org/PR23070 11769 397. https://gcc.gnu.org/PR23404 11770 398. https://gcc.gnu.org/PR23539 11771 399. https://gcc.gnu.org/PR24102 11772 400. https://gcc.gnu.org/PR24465 11773 401. https://gcc.gnu.org/PR19933 11774 402. https://gcc.gnu.org/PR21889 11775 403. https://gcc.gnu.org/PR19300 11776 404. https://gcc.gnu.org/PR20301 11777 405. https://gcc.gnu.org/PR20673 11778 406. https://gcc.gnu.org/PR18582 11779 407. https://gcc.gnu.org/PR19340 11780 408. https://gcc.gnu.org/PR21716 11781 409. https://gcc.gnu.org/PR24315 11782 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 11783 411. https://gcc.gnu.org/onlinedocs/ 11784 412. mailto:gcc-help (a] gcc.gnu.org 11785 413. mailto:gcc (a] gcc.gnu.org 11786 414. https://gcc.gnu.org/lists.html 11787 415. https://www.fsf.org/ 11788 416. https://gcc.gnu.org/about.html 11789 417. http://validator.w3.org/check/referer 11790 ====================================================================== 11791 http://gcc.gnu.org/gcc-3.3/index.html 11792 GCC 3.3 Release Series 11793 11794 May 03, 2005 11795 11796 The [1]GNU project and the GCC developers are pleased to announce the 11797 release of GCC 3.3.6. 11798 11799 This release is a bug-fix release, containing fixes for regressions in 11800 GCC 3.3.5 relative to previous releases of GCC. 11801 11802 This release is the last of the series 3.3.x. 11803 11804 The GCC 3.3 release series includes numerous [2]new features, 11805 improvements, bug fixes, and other changes, thanks to an [3]amazing 11806 group of volunteers. 11807 11808 This release series is no longer maintained. 11809 11810 Release History 11811 11812 GCC 3.3.6 11813 May 3, 2005 ([4]changes) 11814 11815 GCC 3.3.5 11816 September 30, 2004 ([5]changes) 11817 11818 GCC 3.3.4 11819 May 31, 2004 ([6]changes) 11820 11821 GCC 3.3.3 11822 February 14, 2004 ([7]changes) 11823 11824 GCC 3.3.2 11825 October 16, 2003 ([8]changes) 11826 11827 GCC 3.3.1 11828 August 8, 2003 ([9]changes) 11829 11830 GCC 3.3 11831 May 14, 2003 ([10]changes) 11832 11833 References and Acknowledgements 11834 11835 GCC used to stand for the GNU C Compiler, but since the compiler 11836 supports several other languages aside from C, it now stands for the 11837 GNU Compiler Collection. 11838 11839 A list of [11]successful builds is updated as new information becomes 11840 available. 11841 11842 The GCC developers would like to thank the numerous people that have 11843 contributed new features, improvements, bug fixes, and other changes as 11844 well as test results to GCC. This [12]amazing group of volunteers is 11845 what makes GCC successful. 11846 11847 For additional information about GCC please refer to the [13]GCC 11848 project web site or contact the [14]GCC development mailing list. 11849 11850 To obtain GCC please use [15]our mirror sites, or our CVS server. 11851 11852 11853 For questions related to the use of GCC, please consult these web 11854 pages and the [16]GCC manuals. If that fails, the 11855 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11856 web pages and the development of GCC are welcome on our developer 11857 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 11858 archives. 11859 11860 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 11861 distribution of this entire article is permitted in any medium, 11862 provided this notice is preserved. 11863 11864 These pages are [21]maintained by the GCC team. Last modified 11865 2018-09-30[22]. 11866 11867 References 11868 11869 1. http://www.gnu.org/ 11870 2. http://gcc.gnu.org/gcc-3.3/changes.html 11871 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11872 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 11873 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 11874 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 11875 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 11876 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 11877 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 11878 10. http://gcc.gnu.org/gcc-3.3/changes.html 11879 11. http://gcc.gnu.org/gcc-3.3/buildstat.html 11880 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11881 13. http://gcc.gnu.org/index.html 11882 14. mailto:gcc (a] gcc.gnu.org 11883 15. http://gcc.gnu.org/mirrors.html 11884 16. https://gcc.gnu.org/onlinedocs/ 11885 17. mailto:gcc-help (a] gcc.gnu.org 11886 18. mailto:gcc (a] gcc.gnu.org 11887 19. https://gcc.gnu.org/lists.html 11888 20. https://www.fsf.org/ 11889 21. https://gcc.gnu.org/about.html 11890 22. http://validator.w3.org/check/referer 11891 ====================================================================== 11892 http://gcc.gnu.org/gcc-3.3/changes.html 11893 GCC 3.3 Release Series 11894 Changes, New Features, and Fixes 11895 11896 The latest release in the 3.3 release series is [1]GCC 3.3.6. 11897 11898 Caveats 11899 11900 * The preprocessor no longer accepts multi-line string literals. They 11901 were deprecated in 3.0, 3.1, and 3.2. 11902 * The preprocessor no longer supports the -A- switch when appearing 11903 alone. -A- followed by an assertion is still supported. 11904 * Support for all the systems [2]obsoleted in GCC 3.1 has been 11905 removed from GCC 3.3. See below for a [3]list of systems which are 11906 obsoleted in this release. 11907 * Checking for null format arguments has been decoupled from the rest 11908 of the format checking mechanism. Programs which use the format 11909 attribute may regain this functionality by using the new [4]nonnull 11910 function attribute. Note that all functions for which GCC has a 11911 built-in format attribute, an appropriate built-in nonnull 11912 attribute is also applied. 11913 * The DWARF (version 1) debugging format has been deprecated and will 11914 be removed in a future version of GCC. Version 2 of the DWARF 11915 debugging format will continue to be supported for the foreseeable 11916 future. 11917 * The C and Objective-C compilers no longer accept the "Naming Types" 11918 extension (typedef foo = bar); it was already unavailable in C++. 11919 Code which uses it will need to be changed to use the "typeof" 11920 extension instead: typedef typeof(bar) foo. (We have removed this 11921 extension without a period of deprecation because it has caused the 11922 compiler to crash since version 3.0 and no one noticed until very 11923 recently. Thus we conclude it is not in widespread use.) 11924 * The -traditional C compiler option has been removed. It was 11925 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 11926 available.) The <varargs.h> header, used for writing variadic 11927 functions in traditional C, still exists but will produce an error 11928 message if used. 11929 * GCC 3.3.1 automatically places zero-initialized variables in the 11930 .bss section on some operating systems. Versions of GNU Emacs up to 11931 (and including) 21.3 will not work correctly when using this 11932 optimization; you can use -fno-zero-initialized-in-bss to disable 11933 it. 11934 11935 General Optimizer Improvements 11936 11937 * A new scheme for accurately describing processor pipelines, the 11938 [5]DFA scheduler, has been added. 11939 * Pavel Nejedly, Charles University Prague, has contributed new file 11940 format used by the edge coverage profiler (-fprofile-arcs). 11941 The new format is robust and diagnoses common mistakes where 11942 profiles from different versions (or compilations) of the program 11943 are combined resulting in nonsensical profiles and slow code to 11944 produced with profile feedback. Additionally this format allows 11945 extra data to be gathered. Currently, overall statistics are 11946 produced helping optimizers to identify hot spots of a program 11947 globally replacing the old intra-procedural scheme and resulting in 11948 better code. Note that the gcov tool from older GCC versions will 11949 not be able to parse the profiles generated by GCC 3.3 and vice 11950 versa. 11951 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 11952 pass enabled using -ftracer. This pass simplifies the control flow 11953 of functions allowing other optimizations to do better job. 11954 He also contributed the function reordering pass 11955 (-freorder-functions) to optimize function placement using profile 11956 feedback. 11957 11958 New Languages and Language specific improvements 11959 11960 C/ObjC/C++ 11961 11962 * The preprocessor now accepts directives within macro arguments. It 11963 processes them just as if they had not been within macro arguments. 11964 * The separate ISO and traditional preprocessors have been completely 11965 removed. The front end handles either type of preprocessed output 11966 if necessary. 11967 * In C99 mode preprocessor arithmetic is done in the precision of the 11968 target's intmax_t, as required by that standard. 11969 * The preprocessor can now copy comments inside macros to the output 11970 file when the macro is expanded. This feature, enabled using the 11971 -CC option, is intended for use by applications which place 11972 metadata or directives inside comments, such as lint. 11973 * The method of constructing the list of directories to be searched 11974 for header files has been revised. If a directory named by a -I 11975 option is a standard system include directory, the option is 11976 ignored to ensure that the default search order for system 11977 directories and the special treatment of system header files are 11978 not defeated. 11979 * A few more [6]ISO C99 features now work correctly. 11980 * A new function attribute, nonnull, has been added which allows 11981 pointer arguments to functions to be specified as requiring a 11982 non-null value. The compiler currently uses this information to 11983 issue a warning when it detects a null value passed in such an 11984 argument slot. 11985 * A new type attribute, may_alias, has been added. Accesses to 11986 objects with types with this attribute are not subjected to 11987 type-based alias analysis, but are instead assumed to be able to 11988 alias any other type of objects, just like the char type. 11989 11990 C++ 11991 11992 * Type based alias analysis has been implemented for C++ aggregate 11993 types. 11994 11995 Objective-C 11996 11997 * Generate an error if Objective-C objects are passed by value in 11998 function and method calls. 11999 * When -Wselector is used, check the whole list of selectors at the 12000 end of compilation, and emit a warning if a @selector() is not 12001 known. 12002 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 12003 * No longer need to include objc/objc-class.h to compile self calls 12004 in class methods (NeXT runtime only). 12005 * New -Wundeclared-selector option. 12006 * Removed selector bloating which was causing object files to be 10% 12007 bigger on average (GNU runtime only). 12008 * Using at run time @protocol() objects has been fixed in certain 12009 situations (GNU runtime only). 12010 * Type checking has been fixed and improved in many situations 12011 involving protocols. 12012 12013 Java 12014 12015 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 12016 1.4) API. 12017 * The JDK 1.4 assert facility has been implemented. 12018 * The bytecode interpreter is now direct threaded and thus faster. 12019 12020 Fortran 12021 12022 * Fortran improvements are listed in [7]the Fortran documentation. 12023 12024 Ada 12025 12026 * Ada tasking now works with glibc 2.3.x threading libraries. 12027 12028 New Targets and Target Specific Improvements 12029 12030 * The following changes have been made to the HP-PA port: 12031 + The port now defaults to scheduling for the PA8000 series of 12032 processors. 12033 + Scheduling support for the PA7300 processor has been added. 12034 + The 32-bit port now supports weak symbols under HP-UX 11. 12035 + The handling of initializers and finalizers has been improved 12036 under HP-UX 11. The 64-bit port no longer uses collect2. 12037 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 12038 + ABI fixes to correct the passing of small structures by value. 12039 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 12040 use the DFA processor pipeline description. 12041 * The following NetBSD configurations for the SuperH processor family 12042 have been added: 12043 + SH3, big-endian, sh-*-netbsdelf* 12044 + SH3, little-endian, shle-*-netbsdelf* 12045 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 12046 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 12047 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 12048 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 12049 * The following changes have been made to the IA-32/x86-64 port: 12050 + SSE2 and 3dNOW! intrinsics are now supported. 12051 + Support for thread local storage has been added to the IA-32 12052 and x86-64 ports. 12053 + The x86-64 port has been significantly improved. 12054 * The following changes have been made to the MIPS port: 12055 + All configurations now accept the -mabi switch. Note that you 12056 will need appropriate multilibs for this option to work 12057 properly. 12058 + ELF configurations will always pass an ABI flag to the 12059 assembler, except when the MIPS EABI is selected. 12060 + -mabi=64 no longer selects MIPS IV code. 12061 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 12062 been removed from this release. 12063 + -march now changes the core ISA level. In previous releases, 12064 it would change the use of processor-specific extensions, but 12065 would leave the core ISA unchanged. For example, mips64-elf 12066 -march=r8000 will now generate MIPS IV code. 12067 + Under most configurations, -mipsN now acts as a synonym for 12068 -march. 12069 + There are some new preprocessor macros to describe the -march 12070 and -mtune settings. See the documentation of those options 12071 for details. 12072 + Support for the NEC VR-Series processors has been added. This 12073 includes the 54xx, 5500, and 41xx series. 12074 + Support for the Sandcraft sr71k processor has been added. 12075 * The following changes have been made to the S/390 port: 12076 + Support to build the Java runtime libraries has been added. 12077 Java is now enabled by default on s390-*-linux* and 12078 s390x-*-linux* targets. 12079 + Multilib support for the s390x-*-linux* target has been added; 12080 this allows to build 31-bit binaries using the -m31 option. 12081 + Support for thread local storage has been added. 12082 + Inline assembler code may now use the 'Q' constraint to 12083 specify memory operands without index register. 12084 + Various platform-specific performance improvements have been 12085 implemented; in particular, the compiler now uses the BRANCH 12086 ON COUNT family of instructions and makes more frequent use of 12087 the TEST UNDER MASK family of instructions. 12088 * The following changes have been made to the PowerPC port: 12089 + Support for IBM Power4 processor added. 12090 + Support for Motorola e500 SPE added. 12091 + Support for AIX 5.2 added. 12092 + Function and Data sections now supported on AIX. 12093 + Sibcall optimizations added. 12094 * The support for H8 Tiny is added to the H8/300 port with -mn. 12095 12096 Obsolete Systems 12097 12098 Support for a number of older systems has been declared obsolete in GCC 12099 3.3. Unless there is activity to revive them, the next release of GCC 12100 will have their sources permanently removed. 12101 12102 All configurations of the following processor architectures have been 12103 declared obsolete: 12104 * Matsushita MN10200, mn10200-*-* 12105 * Motorola 88000, m88k-*-* 12106 * IBM ROMP, romp-*-* 12107 12108 Also, some individual systems have been obsoleted: 12109 * Alpha 12110 + Interix, alpha*-*-interix* 12111 + Linux libc1, alpha*-*-linux*libc1* 12112 + Linux ECOFF, alpha*-*-linux*ecoff* 12113 * ARM 12114 + Generic a.out, arm*-*-aout* 12115 + Conix, arm*-*-conix* 12116 + "Old ABI," arm*-*-oabi 12117 + StrongARM/COFF, strongarm-*-coff* 12118 * HPPA (PA-RISC) 12119 + Generic OSF, hppa1.0-*-osf* 12120 + Generic BSD, hppa1.0-*-bsd* 12121 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 12122 + HiUX, hppa*-*-hiux* 12123 + Mach Lites, hppa*-*-lites* 12124 * Intel 386 family 12125 + Windows NT 3.x, i?86-*-win32 12126 * MC68000 family 12127 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 12128 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 12129 m68k-sun-mach* 12130 + AT&T systems, m68000-att-sysv* 12131 + Atari systems, m68k-atari-sysv* 12132 + Motorola systems, m68k-motorola-sysv* 12133 + NCR systems, m68k-ncr-sysv* 12134 + Plexus systems, m68k-plexus-sysv* 12135 + Commodore systems, m68k-cbm-sysv* 12136 + Citicorp TTI, m68k-tti-* 12137 + Unos, m68k-crds-unos* 12138 + Concurrent RTU, m68k-ccur-rtu* 12139 + Linux a.out, m68k-*-linux*aout* 12140 + Linux libc1, m68k-*-linux*libc1* 12141 + pSOS, m68k-*-psos* 12142 * MIPS 12143 + Generic ECOFF, mips*-*-ecoff* 12144 + SINIX, mips-sni-sysv4 12145 + Orion RTEMS, mips64orion-*-rtems* 12146 * National Semiconductor 32000 12147 + OpenBSD, ns32k-*-openbsd* 12148 * POWER (aka RS/6000) and PowerPC 12149 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 12150 + Bull BOSX, rs6000-bull-bosx 12151 + Generic Mach, rs6000-*-mach* 12152 + Generic SysV, powerpc*-*-sysv* 12153 + Linux libc1, powerpc*-*-linux*libc1* 12154 * Sun SPARC 12155 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 12156 sparclite-*-aout*, and sparc86x-*-aout* 12157 + NetBSD a.out, sparc-*-netbsd*aout* 12158 + Generic BSD, sparc-*-bsd* 12159 + ChorusOS, sparc-*-chorusos* 12160 + Linux a.out, sparc-*-linux*aout* 12161 + Linux libc1, sparc-*-linux*libc1* 12162 + LynxOS, sparc-*-lynxos* 12163 + Solaris on HAL hardware, sparc-hal-solaris2* 12164 + SunOS versions 3 and 4, sparc-*-sunos[34]* 12165 * NEC V850 12166 + RTEMS, v850-*-rtems* 12167 * VAX 12168 + VMS, vax-*-vms* 12169 12170 Documentation improvements 12171 12172 Other significant improvements 12173 12174 * Almost all front-end dependencies in the compiler have been 12175 separated out into a set of language hooks. This should make adding 12176 a new front end clearer and easier. 12177 * One effect of removing the separate preprocessor is a small 12178 increase in the robustness of the compiler in general, and the 12179 maintainability of target descriptions. Previously target-specific 12180 built-in macros and others, such as __FAST_MATH__, had to be 12181 handled with so-called specs that were hard to maintain. Often they 12182 would fail to behave properly when conflicting options were 12183 supplied on the command line, and define macros in the user's 12184 namespace even when strict ISO compliance was requested. 12185 Integrating the preprocessor has cleanly solved these issues. 12186 * The Makefile suite now supports redirection of make install by 12187 means of the variable DESTDIR. 12188 __________________________________________________________________ 12189 12190 GCC 3.3 12191 12192 Detailed release notes for the GCC 3.3 release follow. 12193 12194 Bug Fixes 12195 12196 bootstrap failures 12197 12198 * [8]10140 cross compiler build failures: missing __mempcpy (DUP: 12199 [9]10198,[10]10338) 12200 12201 Internal compiler errors (multi-platform) 12202 12203 * [11]3581 large string causes segmentation fault in cc1 12204 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 12205 * [13]5533 (c++) ICE when processing std::accumulate(begin, end, 12206 init, invalid_op) 12207 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 12208 * [15]6412 (c++) ICE in retrieve_specialization 12209 * [16]6620 (c++) partial template specialization causes an ICE 12210 (segmentation fault) 12211 * [17]6663 (c++) ICE with attribute aligned 12212 * [18]7068 ICE with incomplete types 12213 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation 12214 * [20]7647 (c++) ICE when data member has the name of the enclosing 12215 class 12216 * [21]7675 ICE in fixup_var_refs_1 12217 * [22]7718 'complex' template instantiation causes ICE 12218 * [23]8116 (c++) ICE in member template function 12219 * [24]8358 (ada) Ada compiler accesses freed memory, crashes 12220 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 12221 fault 12222 * [26]8564 (c++) ICE in find_function_data, in function.c 12223 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 12224 * [28]8766 (c++) ICE after failed initialization of static template 12225 variable 12226 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c 12227 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 12228 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class 12229 definition 12230 * [32]9216 (c++) ICE on missing template parameter 12231 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c 12232 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 12233 loop 12234 * [35]9429 (c++) ICE in template instantiation with a pointered new 12235 operator 12236 * [36]9516 Internal error when using a big array 12237 * [37]9600 (c++) ICE with typedefs in template class 12238 * [38]9629 (c++) virtual inheritance segfault 12239 * [39]9672 (c++) ICE: Error reporting routines re-entered 12240 * [40]9749 (c++) ICE in write_expression on invalid function 12241 prototype 12242 * [41]9794 (fortran) ICE: floating point exception during constant 12243 folding 12244 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE 12245 * [43]9916 (c++) ICE with noreturn function in ?: statement 12246 * [44]9936 ICE with local function and variable-length 2d array 12247 * [45]10262 (c++) cc1plus crashes with large generated code 12248 * [46]10278 (c++) ICE in parser for invalid code 12249 * [47]10446 (c++) ICE on definition of nonexistent member function of 12250 nested class in a class template 12251 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable 12252 declaration 12253 * [49]10506 (c++) ICE in build_new at cp/init.c with 12254 -fkeep-inline-functions and multiple inheritance 12255 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the 12256 precision of the declared type 12257 12258 Optimization bugs 12259 12260 * [51]2001 Inordinately long compile times in reload CSE regs 12261 * [52]2391 Exponential compilation time explosion in combine 12262 * [53]2960 Duplicate loop conditions even with -Os 12263 * [54]4046 redundant conditional branch 12264 * [55]6405 Loop-unrolling related performance regressions 12265 * [56]6798 very long compile time with large case-statement 12266 * [57]6871 const objects shouldn't be moved to .bss 12267 * [58]6909 problem w/ -Os on modified loop-2c.c test case 12268 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of 12269 non-void function'' warning 12270 * [60]7642 optimization problem with signbit() 12271 * [61]8634 incorrect code for inlining of memcpy under -O2 12272 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as 12273 regular function call 12274 12275 C front end 12276 12277 * [63]2161 long if-else cascade overflows parser stack 12278 * [64]4319 short accepted on typedef'd char 12279 * [65]8602 incorrect line numbers in warning messages when using 12280 inline functions 12281 * [66]9177 -fdump-translation-unit: C front end deletes function_decl 12282 AST nodes and breaks debugging dumps 12283 * [67]9853 miscompilation of non-constant structure initializer 12284 12285 c++ compiler and library 12286 12287 * [68]45 legal template specialization code is rejected (DUP: 12288 [69]3784) 12289 * [70]764 lookup failure: friend operator and dereferencing a pointer 12290 and templates (DUP: [71]5116) 12291 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP: 12292 2863) 12293 * [73]3663 G++ doesn't check access control during template 12294 instantiation 12295 * [74]3797 gcc fails to emit explicit specialization of a template 12296 member 12297 * [75]3948 Two destructors are called when no copy destructor is 12298 defined (ABI change) 12299 * [76]4137 Conversion operator within template is not accepted 12300 * [77]4361 bogus ambiguity taking the address of a member template 12301 * [78]4802 g++ accepts illegal template code (access to private 12302 member; DUP: [79]5837) 12303 * [80]4803 inline function is used but never defined, and g++ does 12304 not object 12305 * [81]5094 Partial specialization cannot be friend? 12306 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 12307 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 12308 time 12309 * [84]7015 certain __asm__ constructs rejected 12310 * [85]7086 compile time regression (quadratic behavior in 12311 fixup_var_refs) 12312 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and 12313 std::abort 12314 * [87]7247 copy constructor missing when inlining enabled (invalid 12315 optimization?) 12316 * [88]7441 string array initialization compilation time regression 12317 from seconds to minutes 12318 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong 12319 * [90]7804 bad printing of floating point constant in warning message 12320 * [91]8099 Friend classes and template specializations 12321 * [92]8117 member function pointers and multiple inheritance 12322 * [93]8205 using declaration and multiple inheritance 12323 * [94]8645 unnecessary non-zero checks in stl_tree.h 12324 * [95]8724 explicit destructor call for incomplete class allowed 12325 * [96]8805 compile time regression with many member variables 12326 * [97]8691 -O3 and -fno-implicit-templates are incompatible 12327 * [98]8700 unhelpful error message for binding temp to reference 12328 * [99]8724 explicit destructor call for incomplete class allowed 12329 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems 12330 * [101]9016 Failure to consistently constant fold "constant" C++ 12331 objects 12332 * [102]9053 g++ confused about ambiguity of overloaded function 12333 templates 12334 * [103]9152 undefined virtual thunks 12335 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out 12336 * [105]9297 data corruption due to codegen bug (when copying.) 12337 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken 12338 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 12339 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in 12340 local classes 12341 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters 12342 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439) 12343 * [112]9474 GCC freezes in compiling a weird code mixing <iostream> 12344 and <iostream.h> 12345 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1) 12346 [114][DR 231] 12347 * [115]9555 ostream inserters fail to set badbit on exception 12348 * [116]9561 ostream inserters rethrow exception of wrong type 12349 * [117]9563 ostream::sentry returns true after a failed preparation 12350 * [118]9582 one-definition rule violation in std::allocator 12351 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors 12352 * [120]9683 bug in initialization chains for static const variables 12353 from template classes 12354 * [121]9791 -Woverloaded-virtual reports hiding of destructor 12355 * [122]9817 collate::compare doesn't handle nul characters 12356 * [123]9825 filebuf::sputbackc breaks sbumpc 12357 * [124]9826 operator>>(basic_istream, basic_string) fails to compile 12358 with custom traits 12359 * [125]9924 Multiple using statements for builtin functions not 12360 allowed 12361 * [126]9946 destructor is not called for temporary object 12362 * [127]9964 filebuf::close() sometimes fails to close file 12363 * [128]9988 filebuf::overflow writes EOF to file 12364 * [129]10033 optimization breaks polymorphic references w/ typeid 12365 operator 12366 * [130]10097 filebuf::underflow drops characters 12367 * [131]10132 filebuf destructor can throw exceptions 12368 * [132]10180 gcc fails to warn about non-inlined function 12369 * [133]10199 method parametrized by template does not work everywhere 12370 * [134]10300 use of array-new (nothrow) in segfaults on NULL return 12371 * [135]10427 Stack corruption with variable-length automatic arrays 12372 and virtual destructors 12373 * [136]10503 Compilation never stops in fixed_type_or_null 12374 12375 Objective-C 12376 12377 * [137]5956 selectors aren't matched properly when added to the 12378 selector table 12379 12380 Fortran compiler and library 12381 12382 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't 12383 detect 12384 * [139]3924 g77 generates code that is rejected by GAS if COFF debug 12385 info requested 12386 * [140]5634 doc: explain that configure --prefix=~/... does not work 12387 * [141]6367 multiple repeat counts confuse namelist read into array 12388 * [142]6491 Logical operations error on logicals when using 12389 -fugly-logint 12390 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C" 12391 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 12392 on irix6.5 12393 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 12394 assume a direct access file 12395 * [146]7278 g77 "bug"; the executable misbehaves (with -O2 12396 -fno-automatic) 12397 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows 12398 * [148]7388 Incorrect output with 0-based array of characters 12399 * [149]8587 Double complex zero ** double precision number -> NaN 12400 instead of zero 12401 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 12402 unknown register name line-length-none 12403 * [151]10197 Direct access files not unformatted by default 12404 12405 Java compiler and library 12406 12407 * [152]6005 gcj fails to build rhug on alpha 12408 * [153]6389 System.getProperty("") should always throw an 12409 IllegalArgumentException 12410 * [154]6576 java.util.ResourceBundle.getResource ignores locale 12411 * [155]6652 new java.io.File("").getCanonicalFile() throws exception 12412 * [156]7060 getMethod() doesn't search super interface 12413 * [157]7073 bytecode interpreter gives wrong answer for interface 12414 getSuperclass() 12415 * [158]7180 possible bug in 12416 javax.naming.spi.NamingManager.getPlusPath() 12417 * [159]7416 java.security startup refs "GNU libgcj.security" 12418 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent 12419 env (DUP: [161]7578) 12420 * [162]7611 Internal error while compiling libjava with -O 12421 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry 12422 * [164]7766 ZipInputStream.available returns 0 immediately after 12423 construction 12424 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 12425 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 12426 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens' 12427 * [168]8234 ZipInputStream chokes when InputStream.read() returns 12428 small chunks 12429 * [169]8415 reflection bug: exception info for Method 12430 * [170]8481 java.Random.nextInt(int) may return negative 12431 * [171]8593 Error reading GZIPped files with BufferedReader 12432 * [172]8759 java.beans.Introspector has no flushCaches() or 12433 flushFromCaches() methods 12434 * [173]8997 spin() calls Thread.sleep 12435 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 12436 instead of the root content of C: 12437 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns 12438 wrong return codes 12439 * [176]9271 Severe bias in java.security.SecureRandom 12440 12441 Ada compiler and library 12442 12443 * [177]6767 make gnatlib-shared fails on -laddr2line 12444 * [178]9911 gnatmake fails to link when GCC configured with 12445 --with-sjlj-exceptions=yes 12446 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled 12447 * [180]10546 Ada tasking not working on Red Hat 9 12448 12449 preprocessor 12450 12451 * [181]7029 preprocessor should ignore #warning with -M 12452 12453 ARM-specific 12454 12455 * [182]2903 [arm] Optimization bug with long long arithmetic 12456 * [183]7873 arm-linux-gcc fails when assigning address to a bit field 12457 12458 FreeBSD-specific 12459 12460 * [184]7680 float functions undefined in math.h/cmath with #define 12461 _XOPEN_SOURCE 12462 12463 HP-UX or HP-PA-specific 12464 12465 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 12466 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 12467 fputc_unlocked 12468 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 12469 12470 m68hc11-specific 12471 12472 * [188]6744 Bad assembler code generated: reference to pseudo 12473 register z 12474 * [189]7361 Internal compiler error in reload_cse_simplify_operands, 12475 in reload1.c 12476 12477 MIPS-specific 12478 12479 * [190]9496 [mips-linux] bug in optimizer? 12480 12481 PowerPC-specific 12482 12483 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 12484 space 12485 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux 12486 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg 12487 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c 12488 12489 SPARC-specific 12490 12491 * [195]10267 (documentation) Wrong build instructions for 12492 *-*-solaris2* 12493 12494 x86-specific (Intel/AMD) 12495 12496 * [196]7916 ICE in instantiate_virtual_register_1 12497 * [197]7926 (c++) i486 instructions in header files make c++ programs 12498 crash on i386 12499 * [198]8555 ICE in gen_split_1231 12500 * [199]8994 ICE with -O -march=pentium4 12501 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs 12502 * [201]9806 ICE in inline assembly with -fPIC flag 12503 * [202]10077 gcc -msse2 generates movd to move dwords between xmm 12504 regs 12505 * [203]10233 64-bit comparison only comparing bottom 32-bits 12506 * [204]10286 type-punning doesn't work with __m64 and -O 12507 * [205]10308 [x86] ICE with -O -fgcse or -O2 12508 __________________________________________________________________ 12509 12510 GCC 3.3.1 12511 12512 Bug Fixes 12513 12514 This section lists the problem reports (PRs) from GCC's bug tracking 12515 system that are known to be fixed in the 3.3.1 release. This list might 12516 not be complete (that is, it is possible that some PRs that have been 12517 fixed are not listed here). 12518 12519 Bootstrap failures 12520 12521 * [206]11272 [Solaris] make bootstrap fails while building libstdc++ 12522 12523 Internal compiler errors (multi-platform) 12524 12525 * [207]5754 ICE on invalid nested template class 12526 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 12527 and --enable-checking 12528 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c 12529 * [210]7053 (c++) ICE when declaring a function already defined as a 12530 friend method of a template class 12531 * [211]8164 (c++) ICE when using different const expressions as 12532 template parameter 12533 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c 12534 * [213]9559 (c++) ICE with invalid initialization of a static const 12535 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 12536 when redeclaring a static member variable 12537 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in 12538 dwarfout.c with -g -O -finline-functions 12539 * [216]10432 (c++) ICE in poplevel, in cp/decl.c 12540 * [217]10475 ICE in subreg_highpart_offset for code with long long 12541 * [218]10635 (c++) ICE when dereferencing an incomplete type casted 12542 from a void pointer 12543 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 12544 instantiating static member variables 12545 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets 12546 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 12547 * [222]10796 (c++) ICE when defining an enum with two values: -1 and 12548 MAX_INT_64BIT 12549 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 12550 sched.c 12551 * [224]10939 (c++) ICE with template code 12552 * [225]10956 (c++) ICE when specializing a template member function 12553 of a template class, in tsubst, in cp/pt.c 12554 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*() 12555 defined) 12556 * [227]11059 (c++) ICE with empty union 12557 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 12558 -O2 -fnon-call-exceptions 12559 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type 12560 * [230]11149 (c++) ICE on error when instantiation with call function 12561 of a base type 12562 * [231]11228 (c++) ICE on new-expression using array operator new and 12563 default-initialization 12564 * [232]11282 (c++) Infinite memory usage after syntax error 12565 * [233]11301 (fortran) ICE with -fno-globals 12566 * [234]11308 (c++) ICE when using an enum type name as if it were a 12567 class or namespace 12568 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from 12569 an empty struct 12570 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR 12571 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 12572 template member functions 12573 12574 Optimization bugs 12575 12576 * [238]11198 -O2 -frename-registers generates wrong code (aliasing 12577 problem) 12578 * [239]11304 Wrong code production with -fomit-frame-pointer 12579 * [240]11381 volatile memory access optimized away 12580 * [241]11536 [strength-reduce] -O2 optimization produces wrong code 12581 * [242]11557 constant folding bug generates wrong code 12582 12583 C front end 12584 12585 * [243]5897 No warning for statement after return 12586 * [244]11279 DWARF-2 output mishandles large enums 12587 12588 Preprocessor bugs 12589 12590 * [245]11022 no warning for non-compatible macro redefinition 12591 12592 C++ compiler and library 12593 12594 * [246]2330 static_cast<>() to a private base is allowed 12595 * [247]5388 Incorrect message "operands to ?: have different types" 12596 * [248]5390 Libiberty fails to demangle multi-digit template 12597 parameters 12598 * [249]7877 Incorrect parameter passing to specializations of member 12599 function templates 12600 * [250]9393 Anonymous namespaces and compiling the same file twice 12601 * [251]10032 -pedantic converts some errors to warnings 12602 * [252]10468 const typeof(x) is non-const, but only in templates 12603 * [253]10527 confused error message with "new int()" parameter 12604 initializer 12605 * [254]10679 parameter MIN_INLINE_INSNS is not honored 12606 * [255]10682 gcc chokes on a typedef for an enum inside a class 12607 template 12608 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 12609 0. 12610 * [257]10845 template member function (with nested template as 12611 parameter) cannot be called anymore if another unrelated template 12612 member function is defined 12613 * [258]10849 Cannot define an out-of-class specialization of a 12614 private nested template class 12615 * [259]10888 Suppress -Winline warnings for system headers 12616 * [260]10929 -Winline warns about functions for which no definition 12617 is visible 12618 * [261]10931 valid conversion static_cast<const unsigned 12619 int&>(lvalue-of-type-int) is rejected 12620 * [262]10940 Bad code with explicit specialization 12621 * [263]10968 If member function implicitly instantiated, explicit 12622 instantiation of class fails to instantiate it 12623 * [264]10990 Cannot convert with dynamic_cast<> to a private base 12624 class from within a member function 12625 * [265]11039 Bad interaction between implicit typename deprecation 12626 and friendship 12627 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say 12628 "__unused__" instead 12629 * [267]11095 C++ iostream manipulator causes segfault when called 12630 with negative argument 12631 * [268]11098 g++ doesn't emit complete debugging information for 12632 local variables in destructors 12633 * [269]11137 GNU/Linux shared library constructors not called unless 12634 there's one global object 12635 * [270]11154 spurious ambiguity report for template class 12636 specialization 12637 * [271]11329 Compiler cannot find user defined implicit typecast 12638 * [272]11332 Spurious error with casts in ?: expression 12639 * [273]11431 static_cast behavior with subclasses when default 12640 constructor available 12641 * [274]11528 money_get facet does not accept "$.00" as valid 12642 * [275]11546 Type lookup problems in out-of-line definition of a 12643 class doubly nested from a template class 12644 * [276]11567 C++ code containing templated member function with same 12645 name as pure virtual member function results in linking failure 12646 * [277]11645 Failure to deal with using and private inheritance 12647 12648 Java compiler and library 12649 12650 * [278]5179 Qualified static field access doesn't initialize its 12651 class 12652 * [279]8204 gcj -O2 to native reorders certain instructions 12653 improperly 12654 * [280]10838 java.io.ObjectInputStream syntax error 12655 * [281]10886 The RMI registry that comes with GCJ does not work 12656 correctly 12657 * [282]11349 JNDI URL context factories not located correctly 12658 12659 x86-specific (Intel/AMD) 12660 12661 * [283]4823 ICE on inline assembly code 12662 * [284]8878 miscompilation with -O and SSE 12663 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3 12664 -masm=intel 12665 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 12666 in reload1.c 12667 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 12668 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source 12669 * [289]11044 [x86] out of range loop instructions for FP code on K6 12670 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE 12671 built-ins 12672 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 12673 is used 12674 12675 SPARC- or Solaris- specific 12676 12677 * [292]9362 solaris 'as' dies when fed .s and "-gstabs" 12678 * [293]10142 [SPARC64] gcc produces wrong code when passing 12679 structures by value 12680 * [294]10663 New configure check aborts with Sun tools. 12681 * [295]10835 combinatorial explosion in scheduler on HyperSPARC 12682 * [296]10876 ICE in calculate_giv_inc when building KDE 12683 * [297]10955 wrong code at -O3 for structure argument in context of 12684 structure return 12685 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25 12686 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 12687 Linux kernel 12688 12689 ia64 specific 12690 12691 * [300]10907 gcc violates the ia64 ABI (GP must be preserved) 12692 * [301]11320 scheduler bug (in machine depended reorganization pass) 12693 * [302]11599 bug with conditional and __builtin_prefetch 12694 12695 PowerPC specific 12696 12697 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 12698 during loop) 12699 * [304]10871 error in rs6000_stack_info save_size computation 12700 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 12701 cures it 12702 12703 m68k-specific 12704 12705 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx 12706 * [307]10557 [m68k] ICE in subreg_offset_representable_p 12707 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p 12708 12709 ARM-specific 12710 12711 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for 12712 functions with __attribute__ ((interrupt ("IRQ"))) 12713 * [310]10842 [arm] Clobbered link register is copied to pc under 12714 certain circumstances 12715 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes 12716 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 12717 (3.4) 12718 12719 MIPS-specific 12720 12721 * [313]11084 ICE in propagate_one_insn, in flow.c 12722 12723 SH-specific 12724 12725 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf 12726 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 12727 * [316]11096 i686-linux to sh-linux cross compiler fails to compile 12728 C++ files 12729 12730 GNU/Linux (or Hurd?) specific 12731 12732 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 12733 12734 UnixWare specific 12735 12736 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 12737 7.1.1 12738 12739 Cygwin (or mingw) specific 12740 12741 * [319]5287 ICE with dllimport attribute 12742 * [320]10148 [MingW/CygWin] Compiler dumps core 12743 12744 DJGPP specific 12745 12746 * [321]8787 GCC fails to emit .intel_syntax when invoked with 12747 -masm=intel on DJGPP 12748 12749 Darwin (and MacOS X) specific 12750 12751 * [322]10900 trampolines crash 12752 12753 Documentation 12754 12755 * [323]1607 (c++) Format attributes on methods undocumented 12756 * [324]4252 Invalid option `-fdump-translation-unit' 12757 * [325]4490 Clarify restrictions on -m96bit-long-double, 12758 -m128bit-long-double 12759 * [326]10355 document an issue with regparm attribute on some systems 12760 (e.g. Solaris) 12761 * [327]10726 (fortran) Documentation for function "IDate Intrinsic 12762 (Unix)" is wrong 12763 * [328]10805 document bug in old version of Sun assembler 12764 * [329]10815 warn against GNU binutils on AIX 12765 * [330]10877 document need for newer binutils on i?86-*-linux-gnu 12766 * [331]11280 Manual incorrect with respect to -freorder-blocks 12767 * [332]11466 Document -mlittle-endian and its restrictions for the 12768 sparc64 port 12769 12770 Testsuite bugs (compiler itself is not affected) 12771 12772 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 12773 report failure 12774 * [334]10810 gcc-3.3 fails make check: buffer overrun in 12775 test_demangle.c 12776 __________________________________________________________________ 12777 12778 GCC 3.3.2 12779 12780 Bug Fixes 12781 12782 This section lists the problem reports (PRs) from GCC's bug tracker 12783 that are known to be fixed in the 3.3.2 release. This list might not be 12784 complete (that is, it is possible that some PRs that have been fixed 12785 are not listed here). 12786 12787 Bootstrap failures and problems 12788 12789 * [335]8336 [SCO5] bootstrap config still tries to use COFF options 12790 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 12791 --enable-threads=posix 12792 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 12793 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 12794 7.1.1) 12795 * [339]11687 xstormy16-elf build fails in libf2c 12796 * [340]12263 [SGI IRIX] bootstrap fails during compile of 12797 libf2c/libI77/backspace.c 12798 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9 12799 fix-header processing) 12800 12801 Internal compiler errors (multi-platform) 12802 12803 * [342]7277 Casting integers to vector types causes ICE 12804 * [343]7939 (c++) ICE on invalid function template specialization 12805 * [344]11063 (c++) ICE on parsing initialization list of const array 12806 member 12807 * [345]11207 ICE with negative index in array element designator 12808 * [346]11522 (fortran) g77 dwarf-2 ICE in 12809 add_abstract_origin_attribute 12810 * [347]11595 (c++) ICE on duplicate label definition 12811 * [348]11646 (c++) ICE in commit_one_edge_insertion with 12812 -fnon-call-exceptions -fgcse -O 12813 * [349]11665 ICE in struct initializer when taking address 12814 * [350]11852 (c++) ICE with bad struct initializer. 12815 * [351]11878 (c++) ICE in cp_expr_size 12816 * [352]11883 ICE with any -O on mercury-generated C code 12817 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 12818 cp/typeck2.c when applying typeid operator to template template 12819 parameter 12820 * [354]12146 ICE in lookup_template_function, in cp/pt.c 12821 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions 12822 -fno-gcse -O2 12823 * [356]12369 (c++) ICE with templates and friends 12824 * [357]12446 ICE in emit_move_insn on complicated array reference 12825 * [358]12510 ICE in final_scan_insn 12826 * [359]12544 ICE with large parameters used in nested functions 12827 12828 C and optimization bugs 12829 12830 * [360]9862 spurious warnings with -W -finline-functions 12831 * [361]10962 lookup_field is a linear search on a linked list (can be 12832 slow if large struct) 12833 * [362]11370 -Wunreachable-code gives false complaints 12834 * [363]11637 invalid assembly with -fnon-call-exceptions 12835 * [364]11885 Problem with bitfields in packed structs 12836 * [365]12082 Inappropriate unreachable code warnings 12837 * [366]12180 Inline optimization fails for variadic function 12838 * [367]12340 loop unroller + gcse produces wrong code 12839 12840 C++ compiler and library 12841 12842 * [368]3907 nested template parameter collides with member name 12843 * [369]5293 confusing message when binding a temporary to a reference 12844 * [370]5296 [DR115] Pointers to functions and to template functions 12845 behave differently in deduction 12846 * [371]7939 ICE on function template specialization 12847 * [372]8656 Unable to assign function with __attribute__ and pointer 12848 return type to an appropriate variable 12849 * [373]10147 Confusing error message for invalid template function 12850 argument 12851 * [374]11400 std::search_n() makes assumptions about Size parameter 12852 * [375]11409 issues with using declarations, overloading, and 12853 built-in functions 12854 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 12855 multiple bits in mask 12856 * [377]11786 operator() call on variable in other namespace not 12857 recognized 12858 * [378]11867 static_cast ignores ambiguity 12859 * [379]11928 bug with conversion operators that are typedefs 12860 * [380]12114 Uninitialized memory accessed in dtor 12861 * [381]12163 static_cast + explicit constructor regression 12862 * [382]12181 Wrong code with comma operator and c++ 12863 * [383]12236 regparm and fastcall messes up parameters 12864 * [384]12266 incorrect instantiation of unneeded template during 12865 overload resolution 12866 * [385]12296 istream::peek() doesn't set eofbit 12867 * [386]12298 [sjlj exceptions] Stack unwind destroys 12868 not-yet-constructed object 12869 * [387]12369 ICE with templates and friends 12870 * [388]12337 apparently infinite loop in g++ 12871 * [389]12344 stdcall attribute ignored if function returns a pointer 12872 * [390]12451 missing(late) class forward declaration in cxxabi.h 12873 * [391]12486 g++ accepts invalid use of a qualified name 12874 12875 x86 specific (Intel/AMD) 12876 12877 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX 12878 builtins 12879 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 12880 -O2 12881 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture 12882 * [395]12116 [k6] Invalid assembly output values with X-MAME code 12883 * [396]12070 ICE converting between double and long double with 12884 -msoft-float 12885 12886 ia64-specific 12887 12888 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 12889 * [398]11535 __builtin_return_address may not work on ia64 12890 * [399]11693 [ia64] ICE in gen_nop_type 12891 * [400]12224 [ia64] Thread-local storage doesn't work 12892 12893 PowerPC-specific 12894 12895 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 12896 kernel 12897 * [402]11319 loop miscompiled on ppc32 12898 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code 12899 12900 SPARC-specific 12901 12902 * [404]11662 wrong code for expr. with cast to long long and 12903 exclusive or 12904 * [405]11965 invalid assembler code for a shift < 32 operation 12905 * [406]12301 (c++) stack corruption when a returned expression throws 12906 an exception 12907 12908 Alpha-specific 12909 12910 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of 12911 kernel 2.4.22-pre8 12912 12913 HPUX-specific 12914 12915 * [408]11313 problem with #pragma weak and static inline functions 12916 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore? 12917 12918 Solaris specific 12919 12920 * [410]12166 Profiled programs crash if PROFDIR is set 12921 12922 Solaris-x86 specific 12923 12924 * [411]12101 i386 Solaris no longer works with GNU as? 12925 12926 Miscellaneous embedded target-specific bugs 12927 12928 * [412]10988 [m32r-elf] wrong blockmove code with -O3 12929 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 12930 -O2 12931 * [414]11902 [sh4] spec file improperly inserts rpath even when none 12932 needed 12933 * [415]11903 [sh4] -pthread fails to link due to error in spec file 12934 on sh4 12935 __________________________________________________________________ 12936 12937 GCC 3.3.3 12938 12939 Minor features 12940 12941 In addition to the bug fixes documented below, this release contains 12942 few minor features such as: 12943 * Support for --with-sysroot 12944 * Support for automatic detection of executable stacks 12945 * Support for SSE3 instructions 12946 * Support for thread local storage debugging under GDB on S390 12947 12948 Bug Fixes 12949 12950 This section lists the problem reports (PRs) from GCC's bug tracker 12951 that are known to be fixed in the 3.3.3 release. This list might not be 12952 complete (that is, it is possible that some PRs that have been fixed 12953 are not listed here). 12954 12955 Bootstrap failures and issues 12956 12957 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 12958 * [417]12399 boehm-gc fails (when building a cross compiler): libtool 12959 unable to infer tagged configuration 12960 * [418]13068 mklibgcc.in doesn't handle multi-level multilib 12961 subdirectories properly 12962 12963 Internal compiler errors (multi-platform) 12964 12965 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to 12966 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 12967 * [420]10555 (c++) ICE on undefined template argument 12968 * [421]10706 (c++) ICE in mangle_class_name_for_template 12969 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops 12970 active 12971 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c 12972 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0 12973 * [425]12632 (fortran) -fbounds-check ICE 12974 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc 12975 3.3.2 12976 * [427]12726 (c++) ICE (segfault) on trivial code 12977 * [428]12890 (c++) ICE on compilation of class with throwing method 12978 * [429]12900 (c++) ICE in rtl_verify_flow_info_1 12979 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 12980 correct code with -O2 -fno-force-mem 12981 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive 12982 template 12983 * [432]13318 ICE: floating point exception in the loop optimizer 12984 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 12985 except.c 12986 * [434]13574 (c++) invalid array default initializer in class lets 12987 gcc consume all memory and die 12988 * [435]13475 ICE on SIMD variables with partial value initialization 12989 * [436]13797 (c++) ICE on invalid template parameter 12990 * [437]13824 (java) gcj SEGV with simple .java program 12991 12992 C and optimization bugs 12993 12994 * [438]8776 loop invariants are not removed (most likely) 12995 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 12996 strncmp by memcmp 12997 * [440]11350 undefined labels with -Os -fPIC 12998 * [441]12826 Optimizer removes reference through volatile pointer 12999 * [442]12500 stabs debug info: void no longer a predefined / builtin 13000 type 13001 * [443]12941 builtin-bitops-1.c miscompilation (latent bug) 13002 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix 13003 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled 13004 * [446]13507 spurious printf format warning 13005 * [447]13382 Type information for const pointer disappears during 13006 optimization. 13007 * [448]13394 noreturn attribute ignored on recursive invokation 13008 * [449]13400 Compiled code crashes storing to read-only location 13009 * [450]13521 Endless loop in calculate_global_regs_live 13010 13011 C++ compiler and library 13012 13013 Some of the bug fixes in this list were made to implement decisions 13014 that the ISO C++ standards committee has made concerning several defect 13015 reports (DRs). Links in the list below point to detailed discussion of 13016 the relevant defect report. 13017 * [451]2094 unimplemented: use of `ptrmem_cst' in template type 13018 unification 13019 * [452]2294 using declaration confusion 13020 * [453]5050 template instantiation depth exceeds limit: recursion 13021 problem? 13022 * [454]9371 Bad exception handling in 13023 i/ostream::operator>>/<<(streambuf*) 13024 * [455]9546 bad exception handling in ostream members 13025 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the 13026 face of unknown locales 13027 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work 13028 * [459]10095 istream::operator>>(int&) sets ios::badbit when 13029 ios::failbit is set. 13030 * [460]11554 Warning about reordering of initializers doesn't mention 13031 location of constructor 13032 * [461]12297 istream::sentry::sentry() handles eof() incorrectly. 13033 * [462]12352 Exception safety problems in src/localename.cc 13034 * [463]12438 Memory leak in locale::combine() 13035 * [464]12540 Memory leak in locale::locale(const char*) 13036 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented 13037 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented 13038 * [470]12696 memory eating infinite loop in diagnostics (error 13039 recovery problem) 13040 * [471]12815 Code compiled with optimization behaves unexpectedly 13041 * [472]12862 Conflicts between typedefs/enums and namespace member 13042 declarations 13043 * [473]12926 Wrong value after assignment in initialize list using 13044 bit-fields 13045 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented 13046 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented 13047 * [478]13007 basic_streambuf::pubimbue, imbue wrong 13048 * [479]13009 Implicitly-defined assignment operator writes to wrong 13049 memory 13050 * [480]13057 regparm attribute not applied to destructor 13051 * [481]13070 -Wformat option ignored in g++ 13052 * [482]13081 forward template declarations in <complex> let inlining 13053 fail 13054 * [483]13239 Assertion does not seem to work correctly anymore 13055 * [484]13262 "xxx is private within this context" when initializing a 13056 self-contained template class 13057 * [485]13290 simple typo in concept checking for std::generate_n 13058 * [486]13323 Template code does not compile in presence of typedef 13059 * [487]13369 __verify_grouping (and __add_grouping?) not correct 13060 * [488]13371 infinite loop with packed struct and inlining 13061 * [489]13445 Template argument replacement "dereferences" a typedef 13062 * [490]13461 Fails to access protected-ctor from public constant 13063 * [491]13462 Non-standard-conforming type set::pointer 13064 * [492]13478 gcc uses wrong constructor to initialize a const 13065 reference 13066 * [493]13544 "conflicting types" for enums in different scopes 13067 * [494]13650 string::compare should not (always) use 13068 traits_type::length() 13069 * [495]13683 bogus warning about passing non-PODs through ellipsis 13070 * [496]13688 Derived class is denied access to protected base class 13071 member class 13072 * [497]13774 Member variable cleared in virtual multiple inheritance 13073 class 13074 * [498]13884 Protect sstream.tcc from extern template use 13075 13076 Java compiler and library 13077 13078 * [499]10746 [win32] garbage collection crash in GCJ 13079 13080 Objective-C compiler and library 13081 13082 * [500]11433 Crash due to dereferencing null pointer when querying 13083 protocol 13084 13085 Fortran compiler and library 13086 13087 * [501]12633 logical expression gives incorrect result with 13088 -fugly-logint option 13089 * [502]13037 [gcse-lm] g77 generates incorrect code 13090 * [503]13213 Hex constant problem when compiling with -fugly-logint 13091 and -ftypeless-boz 13092 13093 x86-specific (Intel/AMD) 13094 13095 * [504]4490 ICE with -m128bit-long-double 13096 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have 13097 `reg' in make_field_assignment, in combine.c 13098 * [506]12441 ICE: can't find a register to spill 13099 * [507]12943 array static-init failure under -fpic, -fPIC 13100 * [508]13608 Incorrect code with -O3 -ffast-math 13101 13102 PowerPC-specific 13103 13104 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of 13105 __attribute__((aligned(16))) 13106 * [510]11793 ICE in extract_insn, in recog.c (const_vector's) 13107 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in 13108 altivec.md) 13109 * [512]12537 g++ generates writeable text sections 13110 13111 SPARC-specific 13112 13113 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0 13114 -m64 13115 * [514]12865 mprotect call to make trampoline executable may fail 13116 * [515]13354 ICE in sparc_emit_set_const32 13117 13118 ARM-specific 13119 13120 * [516]10467 [arm] ICE in pre_insert_copy_insn, 13121 13122 ia64-specific 13123 13124 * [517]11226 ICE passing struct arg with two floats 13125 * [518]11227 ICE for _Complex float, _Complex long double args 13126 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64 13127 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn 13128 * Various fixes for libunwind 13129 13130 Alpha-specific 13131 13132 * [521]12654 Incorrect comparison code generated for Alpha 13133 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 13134 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 13135 13136 HPPA-specific 13137 13138 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 13139 * [525]12158 [hppa] compilation does not terminate at -O1 13140 13141 S390-specific 13142 13143 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only 13144 (1<<24)-1 possible for CLCL-Instruction 13145 13146 SH-specific 13147 13148 * [527]9365 segfault in gen_far_branch (config/sh/sh.c) 13149 * [528]10392 optimizer generates faulty array indexing 13150 * [529]11322 SH profiler outputs multiple definitions of symbol 13151 * [530]13069 gcc/config/sh/rtems.h broken 13152 * [531]13302 Putting a va_list in a struct causes seg fault 13153 * [532]13585 Incorrect optimization of call to sfunc 13154 * Fix inappropriately exported libgcc functions from the shared 13155 library 13156 13157 Other embedded target specific 13158 13159 * [533]8916 [mcore] unsigned char assign gets hosed. 13160 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c 13161 * [535]13122 [h8300] local variable gets corrupted by function call 13162 when -fomit-frame-pointer is given 13163 * [536]13256 [cris] strict_low_part mistreated in delay slots 13164 * [537]13373 [mcore] optimization with -frerun-cse-after-loop 13165 -fexpensive-optimizations produces wrong code on mcore 13166 13167 GNU HURD-specific 13168 13169 * [538]12561 gcc/config/t-gnu needs updating to work with 13170 --with-sysroot 13171 13172 Tru64 Unix specific 13173 13174 * [539]6243 testsuite fails almost all tests due to no libintl in 13175 LD_LIBRARY_PATH during test. 13176 * [540]11397 weak aliases broken on Tru64 UNIX 13177 13178 AIX-specific 13179 13180 * [541]12505 build failure due to defines of uchar in cpphash.h and 13181 sys/types.h 13182 * [542]13150 WEAK symbols not exported by collect2 13183 13184 IRIX-specific 13185 13186 * [543]12666 fixincludes problem on IRIX 6.5.19m 13187 13188 Solaris-specific 13189 13190 * [544]12969 Including sys/byteorder.h breaks configure checks 13191 13192 Testsuite problems (compiler is not affected) 13193 13194 * [545]10819 testsuite creates CR+LF on compiler version lines in 13195 test summary files 13196 * [546]11612 abi_check not finding correct libgcc_s.so.1 13197 13198 Miscellaneous 13199 13200 * [547]13211 using -###, incorrect warnings about unused linker file 13201 are produced 13202 __________________________________________________________________ 13203 13204 GCC 3.3.4 13205 13206 This is the [548]list of problem reports (PRs) from GCC's bug tracking 13207 system that are known to be fixed in the 3.3.4 release. This list might 13208 not be complete (that is, it is possible that some PRs that have been 13209 fixed are not listed here). 13210 __________________________________________________________________ 13211 13212 GCC 3.3.5 13213 13214 This is the [549]list of problem reports (PRs) from GCC's bug tracking 13215 system that are known to be fixed in the 3.3.5 release. This list might 13216 not be complete (that is, it is possible that some PRs that have been 13217 fixed are not listed here). 13218 __________________________________________________________________ 13219 13220 GCC 3.3.6 13221 13222 This is the [550]list of problem reports (PRs) from GCC's bug tracking 13223 system that are known to be fixed in the 3.3.6 release. This list might 13224 not be complete (that is, it is possible that some PRs that have been 13225 fixed are not listed here). 13226 13227 13228 For questions related to the use of GCC, please consult these web 13229 pages and the [551]GCC manuals. If that fails, the 13230 [552]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13231 web pages and the development of GCC are welcome on our developer 13232 list at [553]gcc (a] gcc.gnu.org. All of [554]our lists have public 13233 archives. 13234 13235 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and 13236 distribution of this entire article is permitted in any medium, 13237 provided this notice is preserved. 13238 13239 These pages are [556]maintained by the GCC team. Last modified 13240 2018-09-30[557]. 13241 13242 References 13243 13244 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 13245 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 13246 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 13247 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 13248 5. http://gcc.gnu.org/news/dfa.html 13249 6. http://gcc.gnu.org/c99status.html 13250 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 13251 8. https://gcc.gnu.org/PR10140 13252 9. https://gcc.gnu.org/PR10198 13253 10. https://gcc.gnu.org/PR10338 13254 11. https://gcc.gnu.org/PR3581 13255 12. https://gcc.gnu.org/PR4382 13256 13. https://gcc.gnu.org/PR5533 13257 14. https://gcc.gnu.org/PR6387 13258 15. https://gcc.gnu.org/PR6412 13259 16. https://gcc.gnu.org/PR6620 13260 17. https://gcc.gnu.org/PR6663 13261 18. https://gcc.gnu.org/PR7068 13262 19. https://gcc.gnu.org/PR7083 13263 20. https://gcc.gnu.org/PR7647 13264 21. https://gcc.gnu.org/PR7675 13265 22. https://gcc.gnu.org/PR7718 13266 23. https://gcc.gnu.org/PR8116 13267 24. https://gcc.gnu.org/PR8358 13268 25. https://gcc.gnu.org/PR8511 13269 26. https://gcc.gnu.org/PR8564 13270 27. https://gcc.gnu.org/PR8660 13271 28. https://gcc.gnu.org/PR8766 13272 29. https://gcc.gnu.org/PR8803 13273 30. https://gcc.gnu.org/PR8846 13274 31. https://gcc.gnu.org/PR8906 13275 32. https://gcc.gnu.org/PR9216 13276 33. https://gcc.gnu.org/PR9261 13277 34. https://gcc.gnu.org/PR9263 13278 35. https://gcc.gnu.org/PR9429 13279 36. https://gcc.gnu.org/PR9516 13280 37. https://gcc.gnu.org/PR9600 13281 38. https://gcc.gnu.org/PR9629 13282 39. https://gcc.gnu.org/PR9672 13283 40. https://gcc.gnu.org/PR9749 13284 41. https://gcc.gnu.org/PR9794 13285 42. https://gcc.gnu.org/PR9829 13286 43. https://gcc.gnu.org/PR9916 13287 44. https://gcc.gnu.org/PR9936 13288 45. https://gcc.gnu.org/PR10262 13289 46. https://gcc.gnu.org/PR10278 13290 47. https://gcc.gnu.org/PR10446 13291 48. https://gcc.gnu.org/PR10451 13292 49. https://gcc.gnu.org/PR10506 13293 50. https://gcc.gnu.org/PR10549 13294 51. https://gcc.gnu.org/PR2001 13295 52. https://gcc.gnu.org/PR2391 13296 53. https://gcc.gnu.org/PR2960 13297 54. https://gcc.gnu.org/PR4046 13298 55. https://gcc.gnu.org/PR6405 13299 56. https://gcc.gnu.org/PR6798 13300 57. https://gcc.gnu.org/PR6871 13301 58. https://gcc.gnu.org/PR6909 13302 59. https://gcc.gnu.org/PR7189 13303 60. https://gcc.gnu.org/PR7642 13304 61. https://gcc.gnu.org/PR8634 13305 62. https://gcc.gnu.org/PR8750 13306 63. https://gcc.gnu.org/PR2161 13307 64. https://gcc.gnu.org/PR4319 13308 65. https://gcc.gnu.org/PR8602 13309 66. https://gcc.gnu.org/PR9177 13310 67. https://gcc.gnu.org/PR9853 13311 68. https://gcc.gnu.org/PR45 13312 69. https://gcc.gnu.org/PR3784 13313 70. https://gcc.gnu.org/PR764 13314 71. https://gcc.gnu.org/PR5116 13315 72. https://gcc.gnu.org/PR2862 13316 73. https://gcc.gnu.org/PR3663 13317 74. https://gcc.gnu.org/PR3797 13318 75. https://gcc.gnu.org/PR3948 13319 76. https://gcc.gnu.org/PR4137 13320 77. https://gcc.gnu.org/PR4361 13321 78. https://gcc.gnu.org/PR4802 13322 79. https://gcc.gnu.org/PR5837 13323 80. https://gcc.gnu.org/PR4803 13324 81. https://gcc.gnu.org/PR5094 13325 82. https://gcc.gnu.org/PR5730 13326 83. https://gcc.gnu.org/PR6713 13327 84. https://gcc.gnu.org/PR7015 13328 85. https://gcc.gnu.org/PR7086 13329 86. https://gcc.gnu.org/PR7099 13330 87. https://gcc.gnu.org/PR7247 13331 88. https://gcc.gnu.org/PR7441 13332 89. https://gcc.gnu.org/PR7768 13333 90. https://gcc.gnu.org/PR7804 13334 91. https://gcc.gnu.org/PR8099 13335 92. https://gcc.gnu.org/PR8117 13336 93. https://gcc.gnu.org/PR8205 13337 94. https://gcc.gnu.org/PR8645 13338 95. https://gcc.gnu.org/PR8724 13339 96. https://gcc.gnu.org/PR8805 13340 97. https://gcc.gnu.org/PR8691 13341 98. https://gcc.gnu.org/PR8700 13342 99. https://gcc.gnu.org/PR8724 13343 100. https://gcc.gnu.org/PR8949 13344 101. https://gcc.gnu.org/PR9016 13345 102. https://gcc.gnu.org/PR9053 13346 103. https://gcc.gnu.org/PR9152 13347 104. https://gcc.gnu.org/PR9182 13348 105. https://gcc.gnu.org/PR9297 13349 106. https://gcc.gnu.org/PR9318 13350 107. https://gcc.gnu.org/PR9320 13351 108. https://gcc.gnu.org/PR9400 13352 109. https://gcc.gnu.org/PR9424 13353 110. https://gcc.gnu.org/PR9425 13354 111. https://gcc.gnu.org/PR9439 13355 112. https://gcc.gnu.org/PR9474 13356 113. https://gcc.gnu.org/PR9548 13357 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 13358 115. https://gcc.gnu.org/PR9555 13359 116. https://gcc.gnu.org/PR9561 13360 117. https://gcc.gnu.org/PR9563 13361 118. https://gcc.gnu.org/PR9582 13362 119. https://gcc.gnu.org/PR9622 13363 120. https://gcc.gnu.org/PR9683 13364 121. https://gcc.gnu.org/PR9791 13365 122. https://gcc.gnu.org/PR9817 13366 123. https://gcc.gnu.org/PR9825 13367 124. https://gcc.gnu.org/PR9826 13368 125. https://gcc.gnu.org/PR9924 13369 126. https://gcc.gnu.org/PR9946 13370 127. https://gcc.gnu.org/PR9964 13371 128. https://gcc.gnu.org/PR9988 13372 129. https://gcc.gnu.org/PR10033 13373 130. https://gcc.gnu.org/PR10097 13374 131. https://gcc.gnu.org/PR10132 13375 132. https://gcc.gnu.org/PR10180 13376 133. https://gcc.gnu.org/PR10199 13377 134. https://gcc.gnu.org/PR10300 13378 135. https://gcc.gnu.org/PR10427 13379 136. https://gcc.gnu.org/PR10503 13380 137. https://gcc.gnu.org/PR5956 13381 138. https://gcc.gnu.org/PR1832 13382 139. https://gcc.gnu.org/PR3924 13383 140. https://gcc.gnu.org/PR5634 13384 141. https://gcc.gnu.org/PR6367 13385 142. https://gcc.gnu.org/PR6491 13386 143. https://gcc.gnu.org/PR6742 13387 144. https://gcc.gnu.org/PR7113 13388 145. https://gcc.gnu.org/PR7236 13389 146. https://gcc.gnu.org/PR7278 13390 147. https://gcc.gnu.org/PR7384 13391 148. https://gcc.gnu.org/PR7388 13392 149. https://gcc.gnu.org/PR8587 13393 150. https://gcc.gnu.org/PR9038 13394 151. https://gcc.gnu.org/PR10197 13395 152. https://gcc.gnu.org/PR6005 13396 153. https://gcc.gnu.org/PR6389 13397 154. https://gcc.gnu.org/PR6576 13398 155. https://gcc.gnu.org/PR6652 13399 156. https://gcc.gnu.org/PR7060 13400 157. https://gcc.gnu.org/PR7073 13401 158. https://gcc.gnu.org/PR7180 13402 159. https://gcc.gnu.org/PR7416 13403 160. https://gcc.gnu.org/PR7570 13404 161. https://gcc.gnu.org/PR7578 13405 162. https://gcc.gnu.org/PR7611 13406 163. https://gcc.gnu.org/PR7709 13407 164. https://gcc.gnu.org/PR7766 13408 165. https://gcc.gnu.org/PR7785 13409 166. https://gcc.gnu.org/PR7786 13410 167. https://gcc.gnu.org/PR8142 13411 168. https://gcc.gnu.org/PR8234 13412 169. https://gcc.gnu.org/PR8415 13413 170. https://gcc.gnu.org/PR8481 13414 171. https://gcc.gnu.org/PR8593 13415 172. https://gcc.gnu.org/PR8759 13416 173. https://gcc.gnu.org/PR8997 13417 174. https://gcc.gnu.org/PR9253 13418 175. https://gcc.gnu.org/PR9254 13419 176. https://gcc.gnu.org/PR9271 13420 177. https://gcc.gnu.org/PR6767 13421 178. https://gcc.gnu.org/PR9911 13422 179. https://gcc.gnu.org/PR10020 13423 180. https://gcc.gnu.org/PR10546 13424 181. https://gcc.gnu.org/PR7029 13425 182. https://gcc.gnu.org/PR2903 13426 183. https://gcc.gnu.org/PR7873 13427 184. https://gcc.gnu.org/PR7680 13428 185. https://gcc.gnu.org/PR8705 13429 186. https://gcc.gnu.org/PR9986 13430 187. https://gcc.gnu.org/PR10056 13431 188. https://gcc.gnu.org/PR6744 13432 189. https://gcc.gnu.org/PR7361 13433 190. https://gcc.gnu.org/PR9496 13434 191. https://gcc.gnu.org/PR7067 13435 192. https://gcc.gnu.org/PR8480 13436 193. https://gcc.gnu.org/PR8784 13437 194. https://gcc.gnu.org/PR10315 13438 195. https://gcc.gnu.org/PR10267 13439 196. https://gcc.gnu.org/PR7916 13440 197. https://gcc.gnu.org/PR7926 13441 198. https://gcc.gnu.org/PR8555 13442 199. https://gcc.gnu.org/PR8994 13443 200. https://gcc.gnu.org/PR9426 13444 201. https://gcc.gnu.org/PR9806 13445 202. https://gcc.gnu.org/PR10077 13446 203. https://gcc.gnu.org/PR10233 13447 204. https://gcc.gnu.org/PR10286 13448 205. https://gcc.gnu.org/PR10308 13449 206. https://gcc.gnu.org/PR11272 13450 207. https://gcc.gnu.org/PR5754 13451 208. https://gcc.gnu.org/PR6597 13452 209. https://gcc.gnu.org/PR6949 13453 210. https://gcc.gnu.org/PR7053 13454 211. https://gcc.gnu.org/PR8164 13455 212. https://gcc.gnu.org/PR8384 13456 213. https://gcc.gnu.org/PR9559 13457 214. https://gcc.gnu.org/PR9649 13458 215. https://gcc.gnu.org/PR9864 13459 216. https://gcc.gnu.org/PR10432 13460 217. https://gcc.gnu.org/PR10475 13461 218. https://gcc.gnu.org/PR10635 13462 219. https://gcc.gnu.org/PR10661 13463 220. https://gcc.gnu.org/PR10700 13464 221. https://gcc.gnu.org/PR10712 13465 222. https://gcc.gnu.org/PR10796 13466 223. https://gcc.gnu.org/PR10890 13467 224. https://gcc.gnu.org/PR10939 13468 225. https://gcc.gnu.org/PR10956 13469 226. https://gcc.gnu.org/PR11041 13470 227. https://gcc.gnu.org/PR11059 13471 228. https://gcc.gnu.org/PR11083 13472 229. https://gcc.gnu.org/PR11105 13473 230. https://gcc.gnu.org/PR11149 13474 231. https://gcc.gnu.org/PR11228 13475 232. https://gcc.gnu.org/PR11282 13476 233. https://gcc.gnu.org/PR11301 13477 234. https://gcc.gnu.org/PR11308 13478 235. https://gcc.gnu.org/PR11473 13479 236. https://gcc.gnu.org/PR11503 13480 237. https://gcc.gnu.org/PR11513 13481 238. https://gcc.gnu.org/PR11198 13482 239. https://gcc.gnu.org/PR11304 13483 240. https://gcc.gnu.org/PR11381 13484 241. https://gcc.gnu.org/PR11536 13485 242. https://gcc.gnu.org/PR11557 13486 243. https://gcc.gnu.org/PR5897 13487 244. https://gcc.gnu.org/PR11279 13488 245. https://gcc.gnu.org/PR11022 13489 246. https://gcc.gnu.org/PR2330 13490 247. https://gcc.gnu.org/PR5388 13491 248. https://gcc.gnu.org/PR5390 13492 249. https://gcc.gnu.org/PR7877 13493 250. https://gcc.gnu.org/PR9393 13494 251. https://gcc.gnu.org/PR10032 13495 252. https://gcc.gnu.org/PR10468 13496 253. https://gcc.gnu.org/PR10527 13497 254. https://gcc.gnu.org/PR10679 13498 255. https://gcc.gnu.org/PR10682 13499 256. https://gcc.gnu.org/PR10689 13500 257. https://gcc.gnu.org/PR10845 13501 258. https://gcc.gnu.org/PR10849 13502 259. https://gcc.gnu.org/PR10888 13503 260. https://gcc.gnu.org/PR10929 13504 261. https://gcc.gnu.org/PR10931 13505 262. https://gcc.gnu.org/PR10940 13506 263. https://gcc.gnu.org/PR10968 13507 264. https://gcc.gnu.org/PR10990 13508 265. https://gcc.gnu.org/PR11039 13509 266. https://gcc.gnu.org/PR11062 13510 267. https://gcc.gnu.org/PR11095 13511 268. https://gcc.gnu.org/PR11098 13512 269. https://gcc.gnu.org/PR11137 13513 270. https://gcc.gnu.org/PR11154 13514 271. https://gcc.gnu.org/PR11329 13515 272. https://gcc.gnu.org/PR11332 13516 273. https://gcc.gnu.org/PR11431 13517 274. https://gcc.gnu.org/PR11528 13518 275. https://gcc.gnu.org/PR11546 13519 276. https://gcc.gnu.org/PR11567 13520 277. https://gcc.gnu.org/PR11645 13521 278. https://gcc.gnu.org/PR5179 13522 279. https://gcc.gnu.org/PR8204 13523 280. https://gcc.gnu.org/PR10838 13524 281. https://gcc.gnu.org/PR10886 13525 282. https://gcc.gnu.org/PR11349 13526 283. https://gcc.gnu.org/PR4823 13527 284. https://gcc.gnu.org/PR8878 13528 285. https://gcc.gnu.org/PR9815 13529 286. https://gcc.gnu.org/PR10402 13530 287. https://gcc.gnu.org/PR10504 13531 288. https://gcc.gnu.org/PR10673 13532 289. https://gcc.gnu.org/PR11044 13533 290. https://gcc.gnu.org/PR11089 13534 291. https://gcc.gnu.org/PR11420 13535 292. https://gcc.gnu.org/PR9362 13536 293. https://gcc.gnu.org/PR10142 13537 294. https://gcc.gnu.org/PR10663 13538 295. https://gcc.gnu.org/PR10835 13539 296. https://gcc.gnu.org/PR10876 13540 297. https://gcc.gnu.org/PR10955 13541 298. https://gcc.gnu.org/PR11018 13542 299. https://gcc.gnu.org/PR11556 13543 300. https://gcc.gnu.org/PR10907 13544 301. https://gcc.gnu.org/PR11320 13545 302. https://gcc.gnu.org/PR11599 13546 303. https://gcc.gnu.org/PR9745 13547 304. https://gcc.gnu.org/PR10871 13548 305. https://gcc.gnu.org/PR11440 13549 306. https://gcc.gnu.org/PR7594 13550 307. https://gcc.gnu.org/PR10557 13551 308. https://gcc.gnu.org/PR11054 13552 309. https://gcc.gnu.org/PR10834 13553 310. https://gcc.gnu.org/PR10842 13554 311. https://gcc.gnu.org/PR11052 13555 312. https://gcc.gnu.org/PR11183 13556 313. https://gcc.gnu.org/PR11084 13557 314. https://gcc.gnu.org/PR10331 13558 315. https://gcc.gnu.org/PR10413 13559 316. https://gcc.gnu.org/PR11096 13560 317. https://gcc.gnu.org/PR2873 13561 318. https://gcc.gnu.org/PR3163 13562 319. https://gcc.gnu.org/PR5287 13563 320. https://gcc.gnu.org/PR10148 13564 321. https://gcc.gnu.org/PR8787 13565 322. https://gcc.gnu.org/PR10900 13566 323. https://gcc.gnu.org/PR1607 13567 324. https://gcc.gnu.org/PR4252 13568 325. https://gcc.gnu.org/PR4490 13569 326. https://gcc.gnu.org/PR10355 13570 327. https://gcc.gnu.org/PR10726 13571 328. https://gcc.gnu.org/PR10805 13572 329. https://gcc.gnu.org/PR10815 13573 330. https://gcc.gnu.org/PR10877 13574 331. https://gcc.gnu.org/PR11280 13575 332. https://gcc.gnu.org/PR11466 13576 333. https://gcc.gnu.org/PR10737 13577 334. https://gcc.gnu.org/PR10810 13578 335. https://gcc.gnu.org/PR8336 13579 336. https://gcc.gnu.org/PR9330 13580 337. https://gcc.gnu.org/PR9631 13581 338. https://gcc.gnu.org/PR9877 13582 339. https://gcc.gnu.org/PR11687 13583 340. https://gcc.gnu.org/PR12263 13584 341. https://gcc.gnu.org/PR12490 13585 342. https://gcc.gnu.org/PR7277 13586 343. https://gcc.gnu.org/PR7939 13587 344. https://gcc.gnu.org/PR11063 13588 345. https://gcc.gnu.org/PR11207 13589 346. https://gcc.gnu.org/PR11522 13590 347. https://gcc.gnu.org/PR11595 13591 348. https://gcc.gnu.org/PR11646 13592 349. https://gcc.gnu.org/PR11665 13593 350. https://gcc.gnu.org/PR11852 13594 351. https://gcc.gnu.org/PR11878 13595 352. https://gcc.gnu.org/PR11883 13596 353. https://gcc.gnu.org/PR11991 13597 354. https://gcc.gnu.org/PR12146 13598 355. https://gcc.gnu.org/PR12215 13599 356. https://gcc.gnu.org/PR12369 13600 357. https://gcc.gnu.org/PR12446 13601 358. https://gcc.gnu.org/PR12510 13602 359. https://gcc.gnu.org/PR12544 13603 360. https://gcc.gnu.org/PR9862 13604 361. https://gcc.gnu.org/PR10962 13605 362. https://gcc.gnu.org/PR11370 13606 363. https://gcc.gnu.org/PR11637 13607 364. https://gcc.gnu.org/PR11885 13608 365. https://gcc.gnu.org/PR12082 13609 366. https://gcc.gnu.org/PR12180 13610 367. https://gcc.gnu.org/PR12340 13611 368. https://gcc.gnu.org/PR3907 13612 369. https://gcc.gnu.org/PR5293 13613 370. https://gcc.gnu.org/PR5296 13614 371. https://gcc.gnu.org/PR7939 13615 372. https://gcc.gnu.org/PR8656 13616 373. https://gcc.gnu.org/PR10147 13617 374. https://gcc.gnu.org/PR11400 13618 375. https://gcc.gnu.org/PR11409 13619 376. https://gcc.gnu.org/PR11740 13620 377. https://gcc.gnu.org/PR11786 13621 378. https://gcc.gnu.org/PR11867 13622 379. https://gcc.gnu.org/PR11928 13623 380. https://gcc.gnu.org/PR12114 13624 381. https://gcc.gnu.org/PR12163 13625 382. https://gcc.gnu.org/PR12181 13626 383. https://gcc.gnu.org/PR12236 13627 384. https://gcc.gnu.org/PR12266 13628 385. https://gcc.gnu.org/PR12296 13629 386. https://gcc.gnu.org/PR12298 13630 387. https://gcc.gnu.org/PR12369 13631 388. https://gcc.gnu.org/PR12337 13632 389. https://gcc.gnu.org/PR12344 13633 390. https://gcc.gnu.org/PR12451 13634 391. https://gcc.gnu.org/PR12486 13635 392. https://gcc.gnu.org/PR8869 13636 393. https://gcc.gnu.org/PR9786 13637 394. https://gcc.gnu.org/PR11689 13638 395. https://gcc.gnu.org/PR12116 13639 396. https://gcc.gnu.org/PR12070 13640 397. https://gcc.gnu.org/PR11184 13641 398. https://gcc.gnu.org/PR11535 13642 399. https://gcc.gnu.org/PR11693 13643 400. https://gcc.gnu.org/PR12224 13644 401. https://gcc.gnu.org/PR11087 13645 402. https://gcc.gnu.org/PR11319 13646 403. https://gcc.gnu.org/PR11949 13647 404. https://gcc.gnu.org/PR11662 13648 405. https://gcc.gnu.org/PR11965 13649 406. https://gcc.gnu.org/PR12301 13650 407. https://gcc.gnu.org/PR11717 13651 408. https://gcc.gnu.org/PR11313 13652 409. https://gcc.gnu.org/PR11712 13653 410. https://gcc.gnu.org/PR12166 13654 411. https://gcc.gnu.org/PR12101 13655 412. https://gcc.gnu.org/PR10988 13656 413. https://gcc.gnu.org/PR11805 13657 414. https://gcc.gnu.org/PR11902 13658 415. https://gcc.gnu.org/PR11903 13659 416. https://gcc.gnu.org/PR11890 13660 417. https://gcc.gnu.org/PR12399 13661 418. https://gcc.gnu.org/PR13068 13662 419. https://gcc.gnu.org/PR10060 13663 420. https://gcc.gnu.org/PR10555 13664 421. https://gcc.gnu.org/PR10706 13665 422. https://gcc.gnu.org/PR11496 13666 423. https://gcc.gnu.org/PR11741 13667 424. https://gcc.gnu.org/PR12440 13668 425. https://gcc.gnu.org/PR12632 13669 426. https://gcc.gnu.org/PR12712 13670 427. https://gcc.gnu.org/PR12726 13671 428. https://gcc.gnu.org/PR12890 13672 429. https://gcc.gnu.org/PR12900 13673 430. https://gcc.gnu.org/PR13060 13674 431. https://gcc.gnu.org/PR13289 13675 432. https://gcc.gnu.org/PR13318 13676 433. https://gcc.gnu.org/PR13392 13677 434. https://gcc.gnu.org/PR13574 13678 435. https://gcc.gnu.org/PR13475 13679 436. https://gcc.gnu.org/PR13797 13680 437. https://gcc.gnu.org/PR13824 13681 438. https://gcc.gnu.org/PR8776 13682 439. https://gcc.gnu.org/PR10339 13683 440. https://gcc.gnu.org/PR11350 13684 441. https://gcc.gnu.org/PR12826 13685 442. https://gcc.gnu.org/PR12500 13686 443. https://gcc.gnu.org/PR12941 13687 444. https://gcc.gnu.org/PR12953 13688 445. https://gcc.gnu.org/PR13041 13689 446. https://gcc.gnu.org/PR13507 13690 447. https://gcc.gnu.org/PR13382 13691 448. https://gcc.gnu.org/PR13394 13692 449. https://gcc.gnu.org/PR13400 13693 450. https://gcc.gnu.org/PR13521 13694 451. https://gcc.gnu.org/PR2094 13695 452. https://gcc.gnu.org/PR2294 13696 453. https://gcc.gnu.org/PR5050 13697 454. https://gcc.gnu.org/PR9371 13698 455. https://gcc.gnu.org/PR9546 13699 456. https://gcc.gnu.org/PR10081 13700 457. https://gcc.gnu.org/PR10093 13701 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 13702 459. https://gcc.gnu.org/PR10095 13703 460. https://gcc.gnu.org/PR11554 13704 461. https://gcc.gnu.org/PR12297 13705 462. https://gcc.gnu.org/PR12352 13706 463. https://gcc.gnu.org/PR12438 13707 464. https://gcc.gnu.org/PR12540 13708 465. https://gcc.gnu.org/PR12594 13709 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 13710 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 13711 468. https://gcc.gnu.org/PR12657 13712 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 13713 470. https://gcc.gnu.org/PR12696 13714 471. https://gcc.gnu.org/PR12815 13715 472. https://gcc.gnu.org/PR12862 13716 473. https://gcc.gnu.org/PR12926 13717 474. https://gcc.gnu.org/PR12967 13718 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 13719 476. https://gcc.gnu.org/PR12971 13720 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 13721 478. https://gcc.gnu.org/PR13007 13722 479. https://gcc.gnu.org/PR13009 13723 480. https://gcc.gnu.org/PR13057 13724 481. https://gcc.gnu.org/PR13070 13725 482. https://gcc.gnu.org/PR13081 13726 483. https://gcc.gnu.org/PR13239 13727 484. https://gcc.gnu.org/PR13262 13728 485. https://gcc.gnu.org/PR13290 13729 486. https://gcc.gnu.org/PR13323 13730 487. https://gcc.gnu.org/PR13369 13731 488. https://gcc.gnu.org/PR13371 13732 489. https://gcc.gnu.org/PR13445 13733 490. https://gcc.gnu.org/PR13461 13734 491. https://gcc.gnu.org/PR13462 13735 492. https://gcc.gnu.org/PR13478 13736 493. https://gcc.gnu.org/PR13544 13737 494. https://gcc.gnu.org/PR13650 13738 495. https://gcc.gnu.org/PR13683 13739 496. https://gcc.gnu.org/PR13688 13740 497. https://gcc.gnu.org/PR13774 13741 498. https://gcc.gnu.org/PR13884 13742 499. https://gcc.gnu.org/PR10746 13743 500. https://gcc.gnu.org/PR11433 13744 501. https://gcc.gnu.org/PR12633 13745 502. https://gcc.gnu.org/PR13037 13746 503. https://gcc.gnu.org/PR13213 13747 504. https://gcc.gnu.org/PR4490 13748 505. https://gcc.gnu.org/PR12292 13749 506. https://gcc.gnu.org/PR12441 13750 507. https://gcc.gnu.org/PR12943 13751 508. https://gcc.gnu.org/PR13608 13752 509. https://gcc.gnu.org/PR11598 13753 510. https://gcc.gnu.org/PR11793 13754 511. https://gcc.gnu.org/PR12467 13755 512. https://gcc.gnu.org/PR12537 13756 513. https://gcc.gnu.org/PR12496 13757 514. https://gcc.gnu.org/PR12865 13758 515. https://gcc.gnu.org/PR13354 13759 516. https://gcc.gnu.org/PR10467 13760 517. https://gcc.gnu.org/PR11226 13761 518. https://gcc.gnu.org/PR11227 13762 519. https://gcc.gnu.org/PR12644 13763 520. https://gcc.gnu.org/PR13149 13764 521. https://gcc.gnu.org/PR12654 13765 522. https://gcc.gnu.org/PR12965 13766 523. https://gcc.gnu.org/PR13031 13767 524. https://gcc.gnu.org/PR11634 13768 525. https://gcc.gnu.org/PR12158 13769 526. https://gcc.gnu.org/PR11992 13770 527. https://gcc.gnu.org/PR9365 13771 528. https://gcc.gnu.org/PR10392 13772 529. https://gcc.gnu.org/PR11322 13773 530. https://gcc.gnu.org/PR13069 13774 531. https://gcc.gnu.org/PR13302 13775 532. https://gcc.gnu.org/PR13585 13776 533. https://gcc.gnu.org/PR8916 13777 534. https://gcc.gnu.org/PR11576 13778 535. https://gcc.gnu.org/PR13122 13779 536. https://gcc.gnu.org/PR13256 13780 537. https://gcc.gnu.org/PR13373 13781 538. https://gcc.gnu.org/PR12561 13782 539. https://gcc.gnu.org/PR6243 13783 540. https://gcc.gnu.org/PR11397 13784 541. https://gcc.gnu.org/PR12505 13785 542. https://gcc.gnu.org/PR13150 13786 543. https://gcc.gnu.org/PR12666 13787 544. https://gcc.gnu.org/PR12969 13788 545. https://gcc.gnu.org/PR10819 13789 546. https://gcc.gnu.org/PR11612 13790 547. https://gcc.gnu.org/PR13211 13791 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 13792 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 13793 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 13794 551. https://gcc.gnu.org/onlinedocs/ 13795 552. mailto:gcc-help (a] gcc.gnu.org 13796 553. mailto:gcc (a] gcc.gnu.org 13797 554. https://gcc.gnu.org/lists.html 13798 555. https://www.fsf.org/ 13799 556. https://gcc.gnu.org/about.html 13800 557. http://validator.w3.org/check/referer 13801 ====================================================================== 13802 http://gcc.gnu.org/gcc-3.2/index.html 13803 GCC 3.2 Release Series 13804 13805 April 25, 2003 13806 13807 The [1]GNU project and the GCC developers are pleased to announce the 13808 release of GCC 3.2.3. 13809 13810 The purpose of the GCC 3.2 release series is to provide a stable 13811 platform for OS distributors to use building their next releases. A 13812 primary objective was to stabilize the C++ ABI; we believe that the 13813 interface to the compiler and the C++ standard library are now 13814 relatively stable. 13815 13816 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 13817 interoperate with code compiled by GCC 3.1.1 or earlier. 13818 13819 Please refer to our [2]detailed list of news, caveats, and bug-fixes 13820 for further information. 13821 13822 This release series is no longer maintained. 13823 13824 Release History 13825 13826 GCC 3.2.3 13827 April 25, 2003 ([3]changes) 13828 13829 GCC 3.2.2 13830 February 5, 2003 ([4]changes) 13831 13832 GCC 3.2.1 13833 November 19, 2002 ([5]changes) 13834 13835 GCC 3.2 13836 August 14, 2002 ([6]changes) 13837 13838 References and Acknowledgements 13839 13840 GCC used to stand for the GNU C Compiler, but since the compiler 13841 supports several other languages aside from C, it now stands for the 13842 GNU Compiler Collection. 13843 13844 A list of [7]successful builds is updated as new information becomes 13845 available. 13846 13847 The GCC developers would like to thank the numerous people that have 13848 contributed new features, improvements, bug fixes, and other changes as 13849 well as test results to GCC. This [8]amazing group of volunteers is 13850 what makes GCC successful. 13851 13852 For additional information about GCC please refer to the [9]GCC project 13853 web site or contact the [10]GCC development mailing list. 13854 13855 To obtain GCC please use [11]our mirror sites, or our CVS server. 13856 13857 13858 For questions related to the use of GCC, please consult these web 13859 pages and the [12]GCC manuals. If that fails, the 13860 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13861 web pages and the development of GCC are welcome on our developer 13862 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 13863 archives. 13864 13865 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 13866 distribution of this entire article is permitted in any medium, 13867 provided this notice is preserved. 13868 13869 These pages are [17]maintained by the GCC team. Last modified 13870 2018-09-30[18]. 13871 13872 References 13873 13874 1. http://www.gnu.org/ 13875 2. http://gcc.gnu.org/gcc-3.2/changes.html 13876 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 13877 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 13878 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 13879 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2 13880 7. http://gcc.gnu.org/gcc-3.2/buildstat.html 13881 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13882 9. http://gcc.gnu.org/index.html 13883 10. mailto:gcc (a] gcc.gnu.org 13884 11. http://gcc.gnu.org/mirrors.html 13885 12. https://gcc.gnu.org/onlinedocs/ 13886 13. mailto:gcc-help (a] gcc.gnu.org 13887 14. mailto:gcc (a] gcc.gnu.org 13888 15. https://gcc.gnu.org/lists.html 13889 16. https://www.fsf.org/ 13890 17. https://gcc.gnu.org/about.html 13891 18. http://validator.w3.org/check/referer 13892 ====================================================================== 13893 http://gcc.gnu.org/gcc-3.2/changes.html 13894 GCC 3.2 Release Series 13895 Changes, New Features, and Fixes 13896 13897 The latest release in the 3.2 release series is [1]GCC 3.2.3. 13898 13899 Caveats and New Features 13900 13901 Caveats 13902 13903 * The C++ compiler does not correctly zero-initialize 13904 pointers-to-data members. You must explicitly initialize them. For 13905 example: int S::*m(0); will work, but depending on 13906 default-initialization to zero will not work. This bug cannot be 13907 fixed in GCC 3.2 without inducing unacceptable risks. It will be 13908 fixed in GCC 3.3. 13909 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 13910 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 13911 a number of C++ ABI fixes which make its C++ compiler generate 13912 binary code which is incompatible with the C++ compilers found in 13913 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 13914 13915 Frontend Enhancements 13916 13917 C/C++/Objective-C 13918 13919 * The method of constructing the list of directories to be searched 13920 for header files has been revised. If a directory named by a -I 13921 option is a standard system include directory, the option is 13922 ignored to ensure that the default search order for system 13923 directories and the special treatment of system header files are 13924 not defeated. 13925 * The C and Objective-C compilers no longer accept the "Naming Types" 13926 extension (typedef foo = bar); it was already unavailable in C++. 13927 Code which uses it will need to be changed to use the "typeof" 13928 extension instead: typedef typeof(bar) foo. (We have removed this 13929 extension without a period of deprecation because it has caused the 13930 compiler to crash since version 3.0 and no one noticed until very 13931 recently. Thus we conclude it is not in widespread use.) 13932 13933 C++ 13934 13935 * GCC 3.2 fixed serveral differences between the C++ ABI implemented 13936 in GCC and the multi-vendor standard, but more have been found 13937 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 13938 code which is affected by these bugs. We will fix these bugs in 13939 some future release, once we are confident that all have been 13940 found; until then, it is our intention to make changes to the ABI 13941 only if they are necessary for correct compilation of C++, as 13942 opposed to conformance to the ABI documents. 13943 * For details on how to build an ABI compliant compiler for GNU/Linux 13944 systems, check the [3]common C++ ABI page. 13945 13946 New Targets and Target Specific Improvements 13947 13948 IA-32 13949 13950 * Fixed a number of bugs in SSE and MMX intrinsics. 13951 * Fixed common compiler crashes with SSE instruction set enabled 13952 (implied by -march=pentium3, pentium4, athlon-xp) 13953 * __m128 and __m128i is not 128bit aligned when used in structures. 13954 13955 x86-64 13956 13957 * A bug whereby the compiler could generate bad code for bzero has 13958 been fixed. 13959 * ABI fixes (implying ABI incompatibilities with previous version in 13960 some corner cases) 13961 * Fixed prefetch code generation 13962 __________________________________________________________________ 13963 13964 GCC 3.2.3 13965 13966 3.2.3 is a bug fix release only; there are no new features that were 13967 not present in GCC 3.2.2. 13968 13969 Bug Fixes 13970 13971 This section lists the problem reports (PRs) from GCC's bug tracking 13972 system that are known to be fixed in the 3.2.3 release. This list might 13973 not be complete (that is, it is possible that some PRs that have been 13974 fixed are not listed here), and some of the titles have been changed to 13975 make them more clear. 13976 13977 Internal Compiler Errors (multi-platform) 13978 13979 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in 13980 cc1plus 13981 * [5]6440: (c++) template specializations cause ICE 13982 * [6]7050: (c++) ICE on: (i ? get_string() : throw) 13983 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 13984 * [8]7982: (c++) ICE due to infinite recursion (using STL set) 13985 * [9]8068: exceedingly high (infinite) memory usage 13986 * [10]8178: ICE with __builtin_ffs 13987 * [11]8396: ICE in copy_to_mode_reg, in explow.c 13988 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 13989 * [13]9768: ICE when optimizing inline code at -O2 13990 * [14]9798: (c++) Infinite recursion (segfault) in 13991 cp/decl.c:push_using_directive with recursive using directives 13992 * [15]9799: mismatching structure initializer with nested flexible 13993 array member: ICE 13994 * [16]9928: ICE on duplicate enum declaration 13995 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 13996 sparc, alpha) 13997 * [18]10352: ICE in find_reloads_toplev 13998 * [19]10336: ICE with -Wunreachable-code 13999 14000 C/optimizer bugs: 14001 14002 * [20]8224: Incorrect joining of signed and unsigned division 14003 * [21]8613: -O2 produces wrong code with builtin strlen and 14004 postincrements 14005 * [22]8828: gcc reports some code is unreachable when it is not 14006 * [23]9226: GCSE breaking argument passing 14007 * [24]9853: miscompilation of non-constant structure initializer 14008 * [25]9797: C99-style struct initializers are miscompiled 14009 * [26]9967: Some standard C function calls should not be replaced 14010 when optimizing for size 14011 * [27]10116: ce2: invalid merge of join_bb in the context of switch 14012 statements 14013 * [28]10171: wrong code for inlined function 14014 * [29]10175: -Wunreachable-code doesn't work for single lines 14015 14016 C++ compiler and library: 14017 14018 * [30]8316: Confusing diagnostic for code that misuses conversion 14019 operators 14020 * [31]9169: filebuf output fails if codecvt<>::out returns noconv 14021 * [32]9420: incomplete type incorrectly reported 14022 * [33]9459: typeof in return type specification of template not 14023 supported 14024 * [34]9507: filebuf::open handles ios_base::ate incorrectly 14025 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc 14026 * [36]9602: Total confusion about template/friend/virtual/abstract 14027 * [37]9993: destructor not called for local object created within and 14028 returned from infinite loop 14029 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 14030 system 14031 14032 Java compiler and library: 14033 14034 * [39]9652: libgcj build fails on irix6.5.1[78] 14035 * [40]10144: gas on solaris complains about bad .stabs lines for 14036 java, native as unaffected 14037 14038 x86-specific (Intel/AMD): 14039 14040 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86 14041 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions 14042 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 14043 failed 14044 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 14045 failed 14046 14047 SPARC-specific: 14048 14049 * [45]7784: [Sparc] ICE in extract_insn, in recog.c 14050 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in 14051 unroll.c 14052 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 14053 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 14054 execute/loop-2d.c 14055 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 14056 * [50]9414: Scheduling bug on Ultrasparc 14057 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64 14058 14059 m68k-specific: 14060 14061 * [52]7248: broken "inclusive or" code 14062 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 14063 14064 PowerPC-specific: 14065 14066 * [54]9732: Wrong code with -O2 -fPIC 14067 * [55]10073: ICE: powerpc cannot split insn 14068 14069 Alpha-specific: 14070 14071 * [56]7702: optimization problem on a DEC alpha under OSF1 14072 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 14073 14074 HP-specific: 14075 14076 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 14077 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 14078 (missing symbol) 14079 * [60]10271: Floating point args don't get reloaded across function 14080 calls with -O2 14081 14082 MIPS specific: 14083 14084 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 14085 compile/920501-4.c 14086 14087 CRIS specific: 14088 14089 * [62]10377: gcc-3.2.2 creates bad assembler code for cris 14090 14091 Miscellaneous and minor bugs: 14092 14093 * [63]6955: collect2 says "core dumped" when there is no core 14094 __________________________________________________________________ 14095 14096 GCC 3.2.2 14097 14098 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 14099 install by means of the DESTDIR variable. Parts of the GCC tree have 14100 featured that support long before, but now it is available even from 14101 the top level. 14102 14103 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 14104 features that were not present in GCC 3.2.1. 14105 14106 Bug Fixes 14107 14108 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 14109 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 14110 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 14111 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 14112 change, and thus restores ABI-compatibility with previous releases 14113 (except GCC 3.2.1) on these platforms. 14114 14115 This section lists the problem reports (PRs) from GCC's bug tracking 14116 system that are known to be fixed in the 3.2.2 release. This list might 14117 not be complete (that is, it is possible that some PRs that have been 14118 fixed are not listed here) and some of the titles have been changed to 14119 make them more clear. 14120 14121 Internal Compiler Errors (multi-platform) 14122 14123 * [64]5919: (c++) ICE when passing variable array to template 14124 function 14125 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 14126 * [66]7507: ICE with -O2 when address of called function is a 14127 complicated expression 14128 * [67]7622: ICE with nested inline functions if function's address is 14129 taken 14130 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 14131 [69]9258) 14132 * [70]8031: (c++) ICE in code comparing typeids and casting from 14133 virtual base 14134 * [71]8275: ICE in simplify_subreg 14135 * [72]8332: (c++) builtin strlen/template interaction causes ICE 14136 * [73]8372: (c++) ICE on explicit call of destructor 14137 * [74]8439: (c, not c++) empty struct causes ICE 14138 * [75]8442: (c++) ICE with nested template classes 14139 * [76]8518: ICE when compiling mplayer ("extern inline" issue) 14140 * [77]8615: (c++) ICE with out-of-range character constant template 14141 argument 14142 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 14143 * [79]8799: (c++) ICE: error reporting routines re-entered 14144 * [80]9328: (c++) ICE with typeof(X) for overloaded X 14145 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes 14146 14147 C++ (compiler and library) bugs 14148 14149 * [82]47: scoping in nested classes is broken 14150 * [83]6745: problems with iostream rdbuf() member function 14151 * [84]8214: conversion from const char* const to char* sometimes 14152 accepted illegally 14153 * [85]8493: builtin strlen and overload resolution (same bug as 14154 [86]8332) 14155 * [87]8503: strange behaviour of function types 14156 * [88]8727: compiler confused by inheritance from an anonymous struct 14157 * [89]7445: poor performance of std::locale::classic() in 14158 multi-threaded applications 14159 * [90]8230: mishandling of overflow in vector<T>::resize 14160 * [91]8399: sync_with_stdio(false) breaks unformatted input 14161 * [92]8662: illegal access of private member of unnamed class is 14162 accepted 14163 * [93]8707: "make distclean" fails in libstdc++-v3 directory 14164 * [94]8708: __USE_MALLOC doesn't work 14165 * [95]8790: Use of non-thread-safe strtok in src/localename.cc 14166 * [96]8887: Bug in date formats with --enable-clocale=generic 14167 * [97]9076: Call Frame Instructions are not handled correctly during 14168 unwind operation 14169 * [98]9151: std::setprecision limited to 16 digits when outputting a 14170 double to a stream 14171 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers 14172 * [100]9269: libstdc++ headers: explicit specialization of function 14173 must precede its first use 14174 * [101]9322: return value of basic_streambuf<>::getloc affected by 14175 locale::global 14176 * [102]9433: segfault in runtime support for dynamic_cast 14177 14178 C and optimizer bugs 14179 14180 * [103]8032: GCC incorrectly initializes static structs that have 14181 flexible arrays 14182 * [104]8639: simple arithmetic expression broken 14183 * [105]8794: optimization improperly eliminates certain expressions 14184 * [106]8832: traditional "asm volatile" code is illegally optimized 14185 * [107]8988: loop optimizer bug: with -O2, code is generated that 14186 segfaults (found on i386, bug present for all platforms) 14187 * [108]9492: structure copy clobbers subsequent stores to structure 14188 14189 Objective-C bugs 14190 14191 * [109]9267: Objective-C parser won't build with newer bison versions 14192 (e.g. 1.875) 14193 14194 Ada bugs 14195 14196 * [110]8344: Ada build problem due to conflict between gcc/final.o, 14197 gcc/ada/final.o 14198 14199 Preprocessor bugs 14200 14201 * [111]8524: _Pragma within macros is improperly expanded 14202 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 14203 -fshort-wchar 14204 14205 ARM-specific 14206 14207 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95 14208 14209 x86-specific (Intel/AMD) 14210 14211 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 14212 * [115]8599: loop unroll bug with -march=k6-3 14213 * [116]9506: ABI breakage in structure return (affects BSD and 14214 Cygwin, but not GNU/Linux) 14215 14216 FreeBSD 5.0 specific 14217 14218 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 14219 14220 RTEMS-specific 14221 14222 * [118]9292: hppa1.1-rtems configurery problems 14223 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 14224 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue 14225 * [121]9296: gthr-rtems regression 14226 * [122]9316: powerpc-rtems: extending multilibs 14227 14228 HP-PA specific 14229 14230 * [123]9493: ICE with -O2 when building a simple function 14231 14232 Documentation 14233 14234 * [124]7341: hyperlink to gcov in GCC documentation doesn't work 14235 * [125]8947: Please add a warning about "-malign-double" in docs 14236 * [126]7448, [127]8882: typo cleanups 14237 __________________________________________________________________ 14238 14239 GCC 3.2.1 14240 14241 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 14242 generates code that is known not to be binary-compatible with the 14243 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 14244 in the distribution, for details. 14245 14246 This release also removes an old GCC extension, "naming types", and the 14247 documentation now directs users to use a different GCC extension, 14248 __typeof__, instead. The feature had evidently been broken for a while. 14249 14250 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 14251 the new warning there are no new features that were not present in GCC 14252 3.2. 14253 14254 In addition, the previous fix for [128]PR 7445 (poor performance of 14255 std::locale::classic() in multi-threaded applications) was reverted 14256 ("unfixed"), because the "fix" was not thread-safe. 14257 14258 Bug Fixes 14259 14260 This section lists the problem reports (PRs) from GCC's bug tracking 14261 system that are known to be fixed in the 3.2.1 release. This list might 14262 not be complete (that is, it is possible that some PRs that have been 14263 fixed are not listed here). As you can see, the number of bug fixes is 14264 quite large, so it is strongly recommended that users of earlier GCC 14265 3.x releases upgrade to GCC 3.2.1. 14266 14267 Internal Compiler Errors (multi-platform) 14268 14269 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 14270 * [130]5661: (c++) ICE instantiating template on array of unknown 14271 size (bad code) 14272 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 14273 64-bit platforms 14274 * [132]6994: (c++) ICE in find_function_data 14275 * [133]7150: preprocessor: GCC -dM -E gives an ICE 14276 * [134]7160: ICE when optimizing branches without a return value 14277 * [135]7228: (c++) ICE when using member template and template 14278 function 14279 * [136]7266: (c++) ICE with -pedantic on missing typename 14280 * [137]7353: ICE from use of "Naming Types" extension, see above 14281 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c 14282 * [139]7478: (c++) ICE on static_cast inside template 14283 * [140]7526: preprocessor core dump when _Pragma implies #pragma 14284 dependency 14285 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803 14286 is a duplicate) 14287 * [143]7754: (c++) ICE on union with template parameter 14288 * [144]7788: (c++) redeclaring a definition as an incomplete class 14289 causes ICE 14290 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c 14291 * [146]8055: preprocessor dies with SIG11 when building FreeBSD 14292 kernel 14293 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 14294 variables 14295 * [148]8134: (c++) ICE in force_store_init_value on legal code 14296 * [149]8149: (c++) ICE on incomplete type 14297 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 14298 initialization 14299 14300 C++ (compiler and library) bugs 14301 14302 * [151]5607: No pointer adjustment in covariant return types 14303 * [152]6579: Infinite loop with statement expressions in member 14304 initialization 14305 * [153]6803: Default copy constructor bug in GCC 3.1 14306 * [154]7176: g++ confused by friend and static member with same name 14307 * [155]7188: Segfault with template class and recursive (incorrect) 14308 initializer list 14309 * [156]7306: Regression: GCC 3.x fails to compile code with virtual 14310 inheritance if a method has a variable number of arguments 14311 * [157]7461: ctype<char>::classic_table() returns offset array on 14312 Cygwin 14313 * [158]7524: f(const float arg[3]) fails 14314 * [159]7584: Erroneous ambiguous base error on using declaration 14315 * [160]7676: Member template overloading problem 14316 * [161]7679: infinite loop when a right parenthesis is missing 14317 * [162]7811: default locale not taken from environment 14318 * [163]7961: compare( char *) implemented incorrectly in 14319 basic_string<> 14320 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if 14321 streambuf::underflow() leaves gptr() NULL (dups: [165]8127, 14322 [166]6745) 14323 * [167]8096: deque::at() throws std::range_error instead of 14324 std::out_of_range 14325 * [168]8127: cout << cin.rdbuf() infinite loop 14326 * [169]8218: Excessively large memory consumed for classes with large 14327 array members 14328 * [170]8287: GCC 3.2: Destructor called for non-constructed local 14329 object 14330 * [171]8347: empty vector range used in string construction causes 14331 core dump 14332 * [172]8348: fail() flag is set in istringstream when eof() flag is 14333 set 14334 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file) 14335 14336 C and optimizer bugs 14337 14338 * [174]6627: -fno-align-functions doesn't seem to disable function 14339 alignment 14340 * [175]6631: life_analysis misoptimizes code to initialize fields of 14341 a structure 14342 * [176]7102: unsigned char division results in floating exception 14343 * [177]7120: Run once loop should *always* be unrolled 14344 (pessimization) 14345 * [178]7209: Bug involving array referencing and ?: operator 14346 * [179]7515: invalid inlining of global function with -O3 14347 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test 14348 * [181]8467: bug in sibling call optimization 14349 14350 Preprocessor bugs 14351 14352 * [182]4890: incorrect line markers from the traditional preprocessor 14353 * [183]7357: -M option omits system headers files (making it the same 14354 as -MM) 14355 * [184]7358: Changes to Sun's make Dependencies 14356 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 14357 C headers 14358 * [186]7862: Interrupting GCC -MD removes .d file but not .o 14359 * [187]8190: Failed compilation deletes -MD dependency file 14360 * [188]8524: _Pragma within macro is improperly expanded 14361 14362 x86 specific (Intel/AMD) 14363 14364 * [189]5351: (i686-only) function pass-by-value structure copy 14365 corrupts stack ([190]7591 is a duplicate) 14366 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with 14367 -march=pentium3/pentium2/athlon (these are all the same underlying 14368 bug, in MMX register use) 14369 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same 14370 as above?) 14371 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 14372 * [199]6981: wrong code in 64-bit manipulation on x86 14373 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 14374 macro 14375 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 14376 intrinsics are broken 14377 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 14378 -march=pentium4 14379 * [203]7693: Typo in i386 mmintrin.h header 14380 * [204]7723: ICE - Pentium3 sse - GCC 3.2 14381 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 14382 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 14383 14384 PowerPC specific 14385 14386 * [207]5967: GCC bug when profiling nested functions on powerpc 14387 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while 14388 loop on PowerPC 14389 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 14390 * [210]7130: miscompiled code for GCC-3.1 on 14391 powerpc-unknown-linux-gnu with -funroll-all-loops 14392 * [211]7133: PowerPC ICE: unrecognizable insn 14393 * [212]7380: ICE in extract_insn, at recog.c:2148 14394 * [213]8252: ICE on Altivec code with optimization turned on 14395 * [214]8451: Altivec ICE in GCC 3.2 14396 14397 HP/PA specific 14398 14399 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa 14400 14401 SPARC specific 14402 14403 * [216]6668: when using --disable-multilib, libgcc_s.so is installed 14404 in the wrong place on sparc-solaris 14405 * [217]7151: ICE when compiling for UltraSPARC 14406 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 14407 double and -O1 14408 * [219]7842: [REGRESSION] SPARC code gen bug 14409 14410 ARM specific 14411 14412 * [220]7856: [arm] invalid offset in constant pool reference 14413 * [221]7967: optimization produces wrong code (ARM) 14414 14415 Alpha specific 14416 14417 * [222]7374: __builtin_fabsl broken on alpha 14418 14419 IBM s390 specific 14420 14421 * [223]7370: ICE in fixup_var_refs_1 on s390x 14422 * [224]7409: loop optimization bug on s390x-linux-gnu 14423 * [225]8232: s390x: ICE when using bcmp with int length argument 14424 14425 SCO specific 14426 14427 * [226]7623: SCO OpenServer build fails with machmode.def: undefined 14428 symbol: BITS_PER_UNIT 14429 14430 m68k/Coldfire specific 14431 14432 * [227]8314: crtbegin, crtend need to be multilib'ed for this 14433 platform 14434 14435 Documentation 14436 14437 * [228]761: Document some undocumented options 14438 * [229]5610: Fix documentation about invoking SSE instructions 14439 (-mfpmath=sse) 14440 * [230]7484: List -Wmissing-declarations as C-only option 14441 * [231]7531: -mcmodel not documented for x86-64 14442 * [232]8120: Update documentation of bad use of ## 14443 __________________________________________________________________ 14444 14445 GCC 3.2 14446 14447 3.2 is a small bug fix release, but there is a change to the 14448 application binary interface (ABI), hence the change to the second part 14449 of the version number. 14450 14451 The main purpose of the 3.2 release is to correct a couple of problems 14452 in the C++ ABI, with the intention of providing a stable interface 14453 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 14454 14455 Bug Fixes 14456 14457 C++ 14458 14459 * [233]7320: g++ 3.2 relocation problem 14460 * [234]7470: vtable: virtual function pointers not in declaration 14461 order 14462 14463 libstdc++ 14464 14465 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t 14466 * [236]6503, [237]6642, [238]7186: Problems with comparing or 14467 subtracting various types of const and non-const iterators 14468 * [239]7216: ambiguity with basic_iostream::traits_type 14469 * [240]7220: problem with basic_istream::ignore(0,delimiter) 14470 * [241]7222: locale::operator==() doesn't work on std::locale("") 14471 * [242]7286: placement operator delete issue 14472 * [243]7442: cxxabi.h does not match the C++ ABI 14473 * [244]7445: poor performance of std::locale::classic() in 14474 multi-threaded applications 14475 14476 x86-64 specific 14477 14478 * [245]7291: off-by-one in generated inline bzero code for x86-64 14479 14480 14481 For questions related to the use of GCC, please consult these web 14482 pages and the [246]GCC manuals. If that fails, the 14483 [247]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14484 web pages and the development of GCC are welcome on our developer 14485 list at [248]gcc (a] gcc.gnu.org. All of [249]our lists have public 14486 archives. 14487 14488 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and 14489 distribution of this entire article is permitted in any medium, 14490 provided this notice is preserved. 14491 14492 These pages are [251]maintained by the GCC team. Last modified 14493 2018-09-30[252]. 14494 14495 References 14496 14497 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 14498 2. http://gcc.gnu.org/gcc-3.1/changes.html 14499 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html 14500 4. https://gcc.gnu.org/PR3782 14501 5. https://gcc.gnu.org/PR6440 14502 6. https://gcc.gnu.org/PR7050 14503 7. https://gcc.gnu.org/PR7741 14504 8. https://gcc.gnu.org/PR7982 14505 9. https://gcc.gnu.org/PR8068 14506 10. https://gcc.gnu.org/PR8178 14507 11. https://gcc.gnu.org/PR8396 14508 12. https://gcc.gnu.org/PR8674 14509 13. https://gcc.gnu.org/PR9768 14510 14. https://gcc.gnu.org/PR9798 14511 15. https://gcc.gnu.org/PR9799 14512 16. https://gcc.gnu.org/PR9928 14513 17. https://gcc.gnu.org/PR10114 14514 18. https://gcc.gnu.org/PR10352 14515 19. https://gcc.gnu.org/PR10336 14516 20. https://gcc.gnu.org/PR8224 14517 21. https://gcc.gnu.org/PR8613 14518 22. https://gcc.gnu.org/PR8828 14519 23. https://gcc.gnu.org/PR9226 14520 24. https://gcc.gnu.org/PR9853 14521 25. https://gcc.gnu.org/PR9797 14522 26. https://gcc.gnu.org/PR9967 14523 27. https://gcc.gnu.org/PR10116 14524 28. https://gcc.gnu.org/PR10171 14525 29. https://gcc.gnu.org/PR10175 14526 30. https://gcc.gnu.org/PR8316 14527 31. https://gcc.gnu.org/PR9169 14528 32. https://gcc.gnu.org/PR9420 14529 33. https://gcc.gnu.org/PR9459 14530 34. https://gcc.gnu.org/PR9507 14531 35. https://gcc.gnu.org/PR9538 14532 36. https://gcc.gnu.org/PR9602 14533 37. https://gcc.gnu.org/PR9993 14534 38. https://gcc.gnu.org/PR10167 14535 39. https://gcc.gnu.org/PR9652 14536 40. https://gcc.gnu.org/PR10144 14537 41. https://gcc.gnu.org/PR8746 14538 42. https://gcc.gnu.org/PR9888 14539 43. https://gcc.gnu.org/PR9638 14540 44. https://gcc.gnu.org/PR9954 14541 45. https://gcc.gnu.org/PR7784 14542 46. https://gcc.gnu.org/PR7796 14543 47. https://gcc.gnu.org/PR8281 14544 48. https://gcc.gnu.org/PR8366 14545 49. https://gcc.gnu.org/PR8726 14546 50. https://gcc.gnu.org/PR9414 14547 51. https://gcc.gnu.org/PR10067 14548 52. https://gcc.gnu.org/PR7248 14549 53. https://gcc.gnu.org/PR8343 14550 54. https://gcc.gnu.org/PR9732 14551 55. https://gcc.gnu.org/PR10073 14552 56. https://gcc.gnu.org/PR7702 14553 57. https://gcc.gnu.org/PR9671 14554 58. https://gcc.gnu.org/PR8694 14555 59. https://gcc.gnu.org/PR9953 14556 60. https://gcc.gnu.org/PR10271 14557 61. https://gcc.gnu.org/PR6362 14558 62. https://gcc.gnu.org/PR10377 14559 63. https://gcc.gnu.org/PR6955 14560 64. https://gcc.gnu.org/PR5919 14561 65. https://gcc.gnu.org/PR7129 14562 66. https://gcc.gnu.org/PR7507 14563 67. https://gcc.gnu.org/PR7622 14564 68. https://gcc.gnu.org/PR7681 14565 69. https://gcc.gnu.org/PR9528 14566 70. https://gcc.gnu.org/PR8031 14567 71. https://gcc.gnu.org/PR8275 14568 72. https://gcc.gnu.org/PR8332 14569 73. https://gcc.gnu.org/PR8372 14570 74. https://gcc.gnu.org/PR8439 14571 75. https://gcc.gnu.org/PR8442 14572 76. https://gcc.gnu.org/PR8518 14573 77. https://gcc.gnu.org/PR8615 14574 78. https://gcc.gnu.org/PR8663 14575 79. https://gcc.gnu.org/PR8799 14576 80. https://gcc.gnu.org/PR9328 14577 81. https://gcc.gnu.org/PR9465 14578 82. https://gcc.gnu.org/PR47 14579 83. https://gcc.gnu.org/PR6745 14580 84. https://gcc.gnu.org/PR8214 14581 85. https://gcc.gnu.org/PR8493 14582 86. https://gcc.gnu.org/PR8332 14583 87. https://gcc.gnu.org/PR8503 14584 88. https://gcc.gnu.org/PR8727 14585 89. https://gcc.gnu.org/PR7445 14586 90. https://gcc.gnu.org/PR8230 14587 91. https://gcc.gnu.org/PR8399 14588 92. https://gcc.gnu.org/PR8662 14589 93. https://gcc.gnu.org/PR8707 14590 94. https://gcc.gnu.org/PR8708 14591 95. https://gcc.gnu.org/PR8790 14592 96. https://gcc.gnu.org/PR8887 14593 97. https://gcc.gnu.org/PR9076 14594 98. https://gcc.gnu.org/PR9151 14595 99. https://gcc.gnu.org/PR9168 14596 100. https://gcc.gnu.org/PR9269 14597 101. https://gcc.gnu.org/PR9322 14598 102. https://gcc.gnu.org/PR9433 14599 103. https://gcc.gnu.org/PR8032 14600 104. https://gcc.gnu.org/PR8639 14601 105. https://gcc.gnu.org/PR8794 14602 106. https://gcc.gnu.org/PR8832 14603 107. https://gcc.gnu.org/PR8988 14604 108. https://gcc.gnu.org/PR9492 14605 109. https://gcc.gnu.org/PR9267 14606 110. https://gcc.gnu.org/PR8344 14607 111. https://gcc.gnu.org/PR8524 14608 112. https://gcc.gnu.org/PR8880 14609 113. https://gcc.gnu.org/PR9090 14610 114. https://gcc.gnu.org/PR8588 14611 115. https://gcc.gnu.org/PR8599 14612 116. https://gcc.gnu.org/PR9506 14613 117. https://gcc.gnu.org/PR9484 14614 118. https://gcc.gnu.org/PR9292 14615 119. https://gcc.gnu.org/PR9293 14616 120. https://gcc.gnu.org/PR9295 14617 121. https://gcc.gnu.org/PR9296 14618 122. https://gcc.gnu.org/PR9316 14619 123. https://gcc.gnu.org/PR9493 14620 124. https://gcc.gnu.org/PR7341 14621 125. https://gcc.gnu.org/PR8947 14622 126. https://gcc.gnu.org/PR7448 14623 127. https://gcc.gnu.org/PR8882 14624 128. https://gcc.gnu.org/PR7445 14625 129. https://gcc.gnu.org/PR2521 14626 130. https://gcc.gnu.org/PR5661 14627 131. https://gcc.gnu.org/PR6419 14628 132. https://gcc.gnu.org/PR6994 14629 133. https://gcc.gnu.org/PR7150 14630 134. https://gcc.gnu.org/PR7160 14631 135. https://gcc.gnu.org/PR7228 14632 136. https://gcc.gnu.org/PR7266 14633 137. https://gcc.gnu.org/PR7353 14634 138. https://gcc.gnu.org/PR7411 14635 139. https://gcc.gnu.org/PR7478 14636 140. https://gcc.gnu.org/PR7526 14637 141. https://gcc.gnu.org/PR7721 14638 142. https://gcc.gnu.org/PR7803 14639 143. https://gcc.gnu.org/PR7754 14640 144. https://gcc.gnu.org/PR7788 14641 145. https://gcc.gnu.org/PR8031 14642 146. https://gcc.gnu.org/PR8055 14643 147. https://gcc.gnu.org/PR8067 14644 148. https://gcc.gnu.org/PR8134 14645 149. https://gcc.gnu.org/PR8149 14646 150. https://gcc.gnu.org/PR8160 14647 151. https://gcc.gnu.org/PR5607 14648 152. https://gcc.gnu.org/PR6579 14649 153. https://gcc.gnu.org/PR6803 14650 154. https://gcc.gnu.org/PR7176 14651 155. https://gcc.gnu.org/PR7188 14652 156. https://gcc.gnu.org/PR7306 14653 157. https://gcc.gnu.org/PR7461 14654 158. https://gcc.gnu.org/PR7524 14655 159. https://gcc.gnu.org/PR7584 14656 160. https://gcc.gnu.org/PR7676 14657 161. https://gcc.gnu.org/PR7679 14658 162. https://gcc.gnu.org/PR7811 14659 163. https://gcc.gnu.org/PR7961 14660 164. https://gcc.gnu.org/PR8071 14661 165. https://gcc.gnu.org/PR8127 14662 166. https://gcc.gnu.org/PR6745 14663 167. https://gcc.gnu.org/PR8096 14664 168. https://gcc.gnu.org/PR8127 14665 169. https://gcc.gnu.org/PR8218 14666 170. https://gcc.gnu.org/PR8287 14667 171. https://gcc.gnu.org/PR8347 14668 172. https://gcc.gnu.org/PR8348 14669 173. https://gcc.gnu.org/PR8391 14670 174. https://gcc.gnu.org/PR6627 14671 175. https://gcc.gnu.org/PR6631 14672 176. https://gcc.gnu.org/PR7102 14673 177. https://gcc.gnu.org/PR7120 14674 178. https://gcc.gnu.org/PR7209 14675 179. https://gcc.gnu.org/PR7515 14676 180. https://gcc.gnu.org/PR7814 14677 181. https://gcc.gnu.org/PR8467 14678 182. https://gcc.gnu.org/PR4890 14679 183. https://gcc.gnu.org/PR7357 14680 184. https://gcc.gnu.org/PR7358 14681 185. https://gcc.gnu.org/PR7602 14682 186. https://gcc.gnu.org/PR7862 14683 187. https://gcc.gnu.org/PR8190 14684 188. https://gcc.gnu.org/PR8524 14685 189. https://gcc.gnu.org/PR5351 14686 190. https://gcc.gnu.org/PR7591 14687 191. https://gcc.gnu.org/PR6845 14688 192. https://gcc.gnu.org/PR7034 14689 193. https://gcc.gnu.org/PR7124 14690 194. https://gcc.gnu.org/PR7174 14691 195. https://gcc.gnu.org/PR7134 14692 196. https://gcc.gnu.org/PR7375 14693 197. https://gcc.gnu.org/PR7390 14694 198. https://gcc.gnu.org/PR6890 14695 199. https://gcc.gnu.org/PR6981 14696 200. https://gcc.gnu.org/PR7242 14697 201. https://gcc.gnu.org/PR7396 14698 202. https://gcc.gnu.org/PR7630 14699 203. https://gcc.gnu.org/PR7693 14700 204. https://gcc.gnu.org/PR7723 14701 205. https://gcc.gnu.org/PR7951 14702 206. https://gcc.gnu.org/PR8146 14703 207. https://gcc.gnu.org/PR5967 14704 208. https://gcc.gnu.org/PR6984 14705 209. https://gcc.gnu.org/PR7114 14706 210. https://gcc.gnu.org/PR7130 14707 211. https://gcc.gnu.org/PR7133 14708 212. https://gcc.gnu.org/PR7380 14709 213. https://gcc.gnu.org/PR8252 14710 214. https://gcc.gnu.org/PR8451 14711 215. https://gcc.gnu.org/PR7250 14712 216. https://gcc.gnu.org/PR6668 14713 217. https://gcc.gnu.org/PR7151 14714 218. https://gcc.gnu.org/PR7335 14715 219. https://gcc.gnu.org/PR7842 14716 220. https://gcc.gnu.org/PR7856 14717 221. https://gcc.gnu.org/PR7967 14718 222. https://gcc.gnu.org/PR7374 14719 223. https://gcc.gnu.org/PR7370 14720 224. https://gcc.gnu.org/PR7409 14721 225. https://gcc.gnu.org/PR8232 14722 226. https://gcc.gnu.org/PR7623 14723 227. https://gcc.gnu.org/PR8314 14724 228. https://gcc.gnu.org/PR761 14725 229. https://gcc.gnu.org/PR5610 14726 230. https://gcc.gnu.org/PR7484 14727 231. https://gcc.gnu.org/PR7531 14728 232. https://gcc.gnu.org/PR8120 14729 233. https://gcc.gnu.org/PR7320 14730 234. https://gcc.gnu.org/PR7470 14731 235. https://gcc.gnu.org/PR6410 14732 236. https://gcc.gnu.org/PR6503 14733 237. https://gcc.gnu.org/PR6642 14734 238. https://gcc.gnu.org/PR7186 14735 239. https://gcc.gnu.org/PR7216 14736 240. https://gcc.gnu.org/PR7220 14737 241. https://gcc.gnu.org/PR7222 14738 242. https://gcc.gnu.org/PR7286 14739 243. https://gcc.gnu.org/PR7442 14740 244. https://gcc.gnu.org/PR7445 14741 245. https://gcc.gnu.org/PR7291 14742 246. https://gcc.gnu.org/onlinedocs/ 14743 247. mailto:gcc-help (a] gcc.gnu.org 14744 248. mailto:gcc (a] gcc.gnu.org 14745 249. https://gcc.gnu.org/lists.html 14746 250. https://www.fsf.org/ 14747 251. https://gcc.gnu.org/about.html 14748 252. http://validator.w3.org/check/referer 14749 ====================================================================== 14750 http://gcc.gnu.org/gcc-3.1/index.html 14751 GCC 3.1 14752 14753 July 27, 2002 14754 14755 The [1]GNU project and the GCC developers are pleased to announce the 14756 release of GCC 3.1.1. 14757 14758 The links below still apply to GCC 3.1.1. 14759 14760 This release series is no longer maintained. 14761 14762 May 15, 2002 14763 14764 The [2]GNU project and the GCC developers are pleased to announce the 14765 release of GCC 3.1. 14766 14767 GCC used to stand for the GNU C Compiler, but since the compiler 14768 supports several other languages aside from C, it now stands for the 14769 GNU Compiler Collection. 14770 14771 A list of [3]successful builds is updated as new information becomes 14772 available. 14773 14774 The GCC developers would like to thank the numerous people that have 14775 contributed [4]new features, improvements, bug fixes, and other changes 14776 as well as test results to GCC. This [5]amazing group of volunteers is 14777 what makes GCC successful. 14778 14779 For additional information about GCC please refer to the [6]GCC project 14780 web site or contact the [7]GCC development mailing list. 14781 14782 To obtain GCC please use [8]our mirror sites, or our CVS server. 14783 __________________________________________________________________ 14784 14785 14786 For questions related to the use of GCC, please consult these web 14787 pages and the [9]GCC manuals. If that fails, the 14788 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14789 web pages and the development of GCC are welcome on our developer 14790 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 14791 archives. 14792 14793 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 14794 distribution of this entire article is permitted in any medium, 14795 provided this notice is preserved. 14796 14797 These pages are [14]maintained by the GCC team. Last modified 14798 2018-09-30[15]. 14799 14800 References 14801 14802 1. http://www.gnu.org/ 14803 2. http://www.gnu.org/ 14804 3. http://gcc.gnu.org/gcc-3.1/buildstat.html 14805 4. http://gcc.gnu.org/gcc-3.1/changes.html 14806 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14807 6. http://gcc.gnu.org/index.html 14808 7. mailto:gcc (a] gcc.gnu.org 14809 8. http://gcc.gnu.org/mirrors.html 14810 9. https://gcc.gnu.org/onlinedocs/ 14811 10. mailto:gcc-help (a] gcc.gnu.org 14812 11. mailto:gcc (a] gcc.gnu.org 14813 12. https://gcc.gnu.org/lists.html 14814 13. https://www.fsf.org/ 14815 14. https://gcc.gnu.org/about.html 14816 15. http://validator.w3.org/check/referer 14817 ====================================================================== 14818 http://gcc.gnu.org/gcc-3.1/changes.html 14819 GCC 3.1 Release Series 14820 Changes, New Features, and Fixes 14821 14822 Additional changes in GCC 3.1.1 14823 14824 * A bug related to how structures and unions are returned has been 14825 fixed for powerpc-*-netbsd*. 14826 * An important bug in the implementation of -fprefetch-loop-arrays 14827 has been fixed. Previously the optimization prefetched random 14828 blocks of memory for most targets except for i386. 14829 * The Java compiler now compiles Java programs much faster and also 14830 works with parallel make. 14831 * Nested functions have been fixed for mips*-*-netbsd*. 14832 * Some missing floating point support routines have beed added for 14833 mips*-*-netbsd*. 14834 * This [1]message gives additional information about the bugs fixed 14835 in this release. 14836 14837 Caveats 14838 14839 * The -traditional C compiler option has been deprecated and will be 14840 removed in GCC 3.3. (It remains possible to preprocess non-C code 14841 with the traditional preprocessor.) 14842 * The default debugging format for most ELF platforms (including 14843 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 14844 from stabs to DWARF2. This requires GDB 5.1.1 or later. 14845 14846 General Optimizer Improvements 14847 14848 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 14849 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 14850 for profile driven optimizations. 14851 Options -fprofile-arcs and -fbranch-probabilities can now be used 14852 to improve speed of the generated code by profiling the actual 14853 program behaviour on typical runs. In the absence of profile info 14854 the compiler attempts to guess the profile statically. 14855 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 14856 monitor performance of the generated code. 14857 According to the SPECInt2000 results on an AMD Athlon CPU, the code 14858 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 14859 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 14860 is about 2.1% faster compared to 2.95.3. Tests were done using the 14861 -O2 -march=athlon command-line options. 14862 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 14863 infrastructure developed by CodeSourcery, LLC for the C++ front 14864 end, so that it is now used in the C front end too. Inlining 14865 functions as trees exposes them earlier to the compiler, giving it 14866 more opportunities for optimization. 14867 * Support for data prefetching instructions has been added to the GCC 14868 back end and several targets. A new __builtin_prefetch intrinsic is 14869 available to explicitly insert prefetch instructions and 14870 experimental support for loop array prefetching has been added (see 14871 -fprefetch-loop-array documentation). 14872 * Support for emitting debugging information for macros has been 14873 added for DWARF2. It is activated using -g3. 14874 14875 New Languages and Language specific improvements 14876 14877 C/C++ 14878 14879 * A few more [4]ISO C99 features. 14880 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 14881 * The preprocessor's symbol table has been merged with the symbol 14882 table of the C, C++ and Objective-C front ends. 14883 * The preprocessor consumes less memory than the preprocessor in GCC 14884 3.0, often significantly so. On normal input files, it typically 14885 consumes less memory than pre-3.0 cccp-based GCC, too. 14886 14887 C++ 14888 14889 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 14890 was a workaround to allow std compliant code to work with the 14891 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 14892 * The C++ ABI has been fixed so that void (A::*)() const is mangled 14893 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 14894 affects pointer to cv-qualified member function types. 14895 * The C++ ABI has been changed to correctly handle this code: 14896 struct A { 14897 void operator delete[] (void *, size_t); 14898 }; 14899 14900 struct B : public A { 14901 }; 14902 14903 new B[10]; 14904 14905 The amount of storage allocated for the array will be greater than 14906 it was in 3.0, in order to store the number of elements in the 14907 array, so that the correct size can be passed to operator delete[] 14908 when the array is deleted. Previously, the value passed to operator 14909 delete[] was unpredictable. 14910 This change will only affect code that declares a two-argument 14911 operator delete[] with a second parameter of type size_t in a base 14912 class, and does not override that definition in a derived class. 14913 * The C++ ABI has been changed so that: 14914 struct A { 14915 void operator delete[] (void *, size_t); 14916 void operator delete[] (void *); 14917 }; 14918 14919 does not cause unnecessary storage to be allocated when an array of 14920 A objects is allocated. 14921 This change will only affect code that declares both of these forms 14922 of operator delete[], and declared the two-argument form before the 14923 one-argument form. 14924 * The C++ ABI has been changed so that when a parameter is passed by 14925 value, any cleanup for that parameter is performed in the caller, 14926 as specified by the ia64 C++ ABI, rather than the called function 14927 as before. As a result, classes with a non-trivial destructor but a 14928 trivial copy constructor will be passed and returned by invisible 14929 reference, rather than by bitwise copy as before. 14930 * G++ now supports the "named return value optimization": for code 14931 like 14932 A f () { 14933 A a; 14934 ... 14935 return a; 14936 } 14937 14938 G++ will allocate a in the return value slot, so that the return 14939 becomes a no-op. For this to work, all return statements in the 14940 function must return the same variable. 14941 * Improvements to the C++ library are listed in [5]the libstdc++-v3 14942 FAQ. 14943 14944 Objective-C 14945 14946 * Annoying linker warnings (due to incorrect code being generated) 14947 have been fixed. 14948 * If a class method cannot be found, the compiler no longer issues a 14949 warning if a corresponding instance method exists in the root 14950 class. 14951 * Forward @protocol declarations have been fixed. 14952 * Loading of categories has been fixed in certain situations (GNU run 14953 time only). 14954 * The class lookup in the run-time library has been rewritten so that 14955 class method dispatch is more than twice as fast as it used to be 14956 (GNU run time only). 14957 14958 Java 14959 14960 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 14961 javax.transaction. 14962 * Property files and other system resources can be compiled into 14963 executables which use libgcj using the new gcj --resource feature. 14964 * libgcj has been ported to more platforms. In particular there is 14965 now a mostly-functional mingw32 (Windows) target port. 14966 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 14967 Java code can now be called from a C/C++ application. 14968 * gcj can now use builtin functions for certain known methods, for 14969 instance Math.cos. 14970 * gcj can now automatically remove redundant array-store checks in 14971 some common cases. 14972 * The --no-store-checks optimization option was added. This can be 14973 used to omit runtime store checks for code which is known not to 14974 throw ArrayStoreException 14975 * The following third party interface standards were added to libgcj: 14976 org.w3c.dom and org.xml.sax. 14977 * java.security has been merged with GNU Classpath. The new package 14978 is now JDK 1.2 compliant, and much more complete. 14979 * A bytecode verifier was added to the libgcj interpreter. 14980 * java.lang.Character was rewritten to comply with the Unicode 3.0 14981 standard, and improve performance. 14982 * Partial support for many more locales was added to libgcj. 14983 * Socket timeouts have been implemented. 14984 * libgcj has been merged into a single shared library. There are no 14985 longer separate shared libraries for the garbage collector and 14986 zlib. 14987 * Several performance improvements were made to gcj and libgcj: 14988 + Hash synchronization (thin locks) 14989 + A special allocation path for finalizer-free objects 14990 + Thread-local allocation 14991 + Parallel GC, and other GC tweaks 14992 14993 Fortran 14994 14995 Fortran improvements are listed in [6]the Fortran documentation. 14996 14997 Ada 14998 14999 [7]AdaCore, has contributed its GNAT Ada 95 front end and associated 15000 tools. The GNAT compiler fully implements the Ada language as defined 15001 by the ISO/IEC 8652 standard. 15002 15003 Please note that the integration of the Ada front end is still work in 15004 progress. 15005 15006 New Targets and Target Specific Improvements 15007 15008 * Hans-Peter Nilsson has contributed a port to MMIX, the CPU 15009 architecture used in new editions of Donald E. Knuth's The Art of 15010 Computer Programming. 15011 * Axis Communications has contributed its port to the CRIS CPU 15012 architecture, used in the ETRAX system-on-a-chip series. 15013 * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH 15014 SH5 64-bit RISC microprocessor architecture, extending the existing 15015 SH port. 15016 * UltraSPARC is fully supported in 64-bit mode. The option -m64 15017 enables it. 15018 * For compatibility with the Sun compiler #pragma redefine_extname 15019 has been implemented on Solaris. 15020 * The x86 back end has had some noticeable work done to it. 15021 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 15022 Jaeger have contributed a port to the AMD x86-64 architecture. 15023 For more information on x86-64 see http://www.x86-64.org. 15024 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 15025 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 15026 enable the respective instruction sets. Intel C++ compatible 15027 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 15028 will be added in next major release. 15029 + Following those improvements, targets for Pentium MMX, K6-2, 15030 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 15031 added. Refer to the documentation on -march= and -mcpu= 15032 options for details. 15033 + For those targets that support it, -mfpmath=sse will cause the 15034 compiler to generate SSE/SSE2 instructions for floating point 15035 math instead of x87 instructions. Usually, this will lead to 15036 quicker code especially on the Pentium 4. Note that only 15037 scalar floating point instructions are used and GCC does not 15038 exploit SIMD features yet. 15039 + Prefetch support has been added to the Pentium III, Pentium 4, 15040 K6-2, K6-3, and Athlon series. 15041 + Code generated for floating point to integer conversions has 15042 been improved leading to better performance of many 3D 15043 applications. 15044 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 15045 * C++ support for AIX has been improved. 15046 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 15047 PowerPC port supporting the AltiVec programming model (SIMD). The 15048 support, though presently useful, is experimental and is expected 15049 to stabilize for 3.2. The support is written to conform to 15050 Motorola's AltiVec specs. See -maltivec. 15051 15052 Obsolete Systems 15053 15054 Support for a number of older systems has been declared obsolete in GCC 15055 3.1. Unless there is activity to revive them, the next release of GCC 15056 will have their sources permanently removed. 15057 15058 All configurations of the following processor architectures have been 15059 declared obsolete: 15060 * MIL-STD-1750A, 1750a-*-* 15061 * AMD A29k, a29k-*-* 15062 * Convex, c*-convex-* 15063 * Clipper, clipper-*-* 15064 * Elxsi, elxsi-*-* 15065 * Intel i860, i860-*-* 15066 * Sun picoJava, pj-*-* and pjl-*-* 15067 * Western Electric 32000, we32k-*-* 15068 15069 Most configurations of the following processor architectures have been 15070 declared obsolete, but we are preserving a few systems which may have 15071 active developers. It is unlikely that the remaining systems will 15072 survive much longer unless we see definite signs of port activity. 15073 * Motorola 88000 except 15074 + Generic a.out, m88k-*-aout* 15075 + Generic SVR4, m88k-*-sysv4 15076 + OpenBSD, m88k-*-openbsd* 15077 * NS32k except 15078 + NetBSD, ns32k-*-netbsd* 15079 + OpenBSD, ns32k-*-openbsd*. 15080 * ROMP except 15081 + OpenBSD, romp-*-openbsd*. 15082 15083 Finally, only some configurations of these processor architectures are 15084 being obsoleted. 15085 * Alpha: 15086 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 15087 alpha*-*-osf[45], are still supported.) 15088 * ARM: 15089 + RISCiX, arm-*-riscix*. 15090 * i386: 15091 + 386BSD, i?86-*-bsd* 15092 + Chorus, i?86-*-chorusos* 15093 + DG/UX, i?86-*-dgux* 15094 + FreeBSD 1.x, i?86-*-freebsd1.* 15095 + IBM AIX, i?86-*-aix* 15096 + ISC UNIX, i?86-*-isc* 15097 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 15098 + NEXTstep, i?86-next-* 15099 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 15100 + RTEMS/coff, i?86-*-rtemscoff* 15101 + RTEMS/go32, i?86-go32-rtems* 15102 + Sequent/BSD, i?86-sequent-bsd* 15103 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 15104 i?86-sequent-sysv3* 15105 + SunOS, i?86-*-sunos* 15106 * Motorola 68000: 15107 + Altos, m68[k0]*-altos-* 15108 + Apollo, m68[k0]*-apollo-* 15109 + Apple A/UX, m68[k0]*-apple-* 15110 + Bull, m68[k0]*-bull-* 15111 + Convergent, m68[k0]*-convergent-* 15112 + Generic SVR3, m68[k0]*-*-sysv3* 15113 + ISI, m68[k0]*-isi-* 15114 + LynxOS, m68[k0]*-*-lynxos* 15115 + NEXT, m68[k0]*-next-* 15116 + RTEMS/coff, m68[k0]*-*-rtemscoff* 15117 + Sony, m68[k0]*-sony-* 15118 * MIPS: 15119 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 15120 + Generic BSD, mips-*-bsd* 15121 + Generic System V, mips-*-sysv* 15122 + IRIX before version 5, mips-sgi-irix[1234]* 15123 + RiscOS, mips-*-riscos* 15124 + Sony, mips-sony-* 15125 + Tandem, mips-tandem-* 15126 * SPARC: 15127 + RTEMS/a.out, sparc-*-rtemsaout*. 15128 15129 Documentation improvements 15130 15131 * The old manual ("Using and Porting the GNU Compiler Collection") 15132 has been replaced by a users manual ("Using the GNU Compiler 15133 Collection") and a separate internals reference manual ("GNU 15134 Compiler Collection Internals"). 15135 * More complete and much improved documentation about GCC's internal 15136 representation used by the C and C++ front ends. 15137 * Many cleanups and improvements in general. 15138 15139 15140 For questions related to the use of GCC, please consult these web 15141 pages and the [8]GCC manuals. If that fails, the 15142 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15143 web pages and the development of GCC are welcome on our developer 15144 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 15145 archives. 15146 15147 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 15148 distribution of this entire article is permitted in any medium, 15149 provided this notice is preserved. 15150 15151 These pages are [13]maintained by the GCC team. Last modified 15152 2018-09-30[14]. 15153 15154 References 15155 15156 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 15157 2. http://gcc.gnu.org/news/profiledriven.html 15158 3. http://gcc.gnu.org/benchmarks/ 15159 4. http://gcc.gnu.org/c99status.html 15160 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 15161 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 15162 7. https://www.adacore.com/ 15163 8. https://gcc.gnu.org/onlinedocs/ 15164 9. mailto:gcc-help (a] gcc.gnu.org 15165 10. mailto:gcc (a] gcc.gnu.org 15166 11. https://gcc.gnu.org/lists.html 15167 12. https://www.fsf.org/ 15168 13. https://gcc.gnu.org/about.html 15169 14. http://validator.w3.org/check/referer 15170 ====================================================================== 15171 http://gcc.gnu.org/gcc-3.0/index.html 15172 GCC 3.0.4 15173 15174 February 20, 2002 15175 15176 The [1]GNU project and the GCC developers are pleased to announce the 15177 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 15178 series. 15179 15180 This release series is no longer maintained. 15181 15182 GCC used to stand for the GNU C Compiler, but since the compiler 15183 supports several other languages aside from C, it now stands for the 15184 GNU Compiler Collection. 15185 15186 GCC 3.0.x has several new optimizations, new targets, new languages and 15187 many other new features, relative to GCC 2.95.x. See the [2]new 15188 features page for a more complete list. 15189 15190 A list of [3]successful builds is updated as new information becomes 15191 available. 15192 15193 The GCC developers would like to thank the numerous people that have 15194 contributed new features, test results, bug fixes, etc to GCC. This 15195 [4]amazing group of volunteers is what makes GCC successful. 15196 15197 And finally, we can't in good conscience fail to mention some 15198 [5]caveats to using GCC 3.0.x. 15199 15200 For additional information about GCC please refer to the [6]GCC project 15201 web site or contact the [7]GCC development mailing list. 15202 15203 To obtain GCC please use [8]our mirror sites, or our CVS server. 15204 __________________________________________________________________ 15205 15206 Previous 3.0.x Releases 15207 15208 December 20, 2001: GCC 3.0.3 has been released. 15209 October 25, 2001: GCC 3.0.2 has been released. 15210 August 20, 2001: GCC 3.0.1 has been released. 15211 June 18, 2001: GCC 3.0 has been released. 15212 15213 15214 For questions related to the use of GCC, please consult these web 15215 pages and the [9]GCC manuals. If that fails, the 15216 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15217 web pages and the development of GCC are welcome on our developer 15218 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 15219 archives. 15220 15221 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 15222 distribution of this entire article is permitted in any medium, 15223 provided this notice is preserved. 15224 15225 These pages are [14]maintained by the GCC team. Last modified 15226 2018-09-30[15]. 15227 15228 References 15229 15230 1. http://www.gnu.org/ 15231 2. http://gcc.gnu.org/gcc-3.0/features.html 15232 3. http://gcc.gnu.org/gcc-3.0/buildstat.html 15233 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15234 5. http://gcc.gnu.org/gcc-3.0/caveats.html 15235 6. http://gcc.gnu.org/index.html 15236 7. mailto:gcc (a] gcc.gnu.org 15237 8. http://gcc.gnu.org/mirrors.html 15238 9. https://gcc.gnu.org/onlinedocs/ 15239 10. mailto:gcc-help (a] gcc.gnu.org 15240 11. mailto:gcc (a] gcc.gnu.org 15241 12. https://gcc.gnu.org/lists.html 15242 13. https://www.fsf.org/ 15243 14. https://gcc.gnu.org/about.html 15244 15. http://validator.w3.org/check/referer 15245 ====================================================================== 15246 http://gcc.gnu.org/gcc-3.0/features.html 15247 GCC 3.0 New Features 15248 15249 Additional changes in GCC 3.0.4 15250 15251 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 15252 system, which use the ELF object file format, on x86 processors. 15253 * Correct debugging information is generated from functions that have 15254 lines from multiple files (e.g. yacc output). 15255 * A fix for whitespace handling in the -traditional preprocessor, 15256 which can affect Fortran. 15257 * Fixes to the exception handling runtime. 15258 * More fixes for bad code generation in C++. 15259 * A fix for shared library generation under AIX 4.3. 15260 * Documentation updates. 15261 * Port of GCC to Tensilica's Xtensa processor contributed. 15262 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 15263 15264 Additional changes in GCC 3.0.3 15265 15266 * A fix to correct an accidental change to the PowerPC ABI. 15267 * Fixes for bad code generation on a variety of architectures. 15268 * Improvements to the debugging information generated for C++ 15269 classes. 15270 * Fixes for bad code generation in C++. 15271 * A fix to avoid crashes in the C++ demangler. 15272 * A fix to the C++ standard library to avoid buffer overflows. 15273 * Miscellaneous improvements for a variety of architectures. 15274 15275 Additional changes in GCC 3.0.2 15276 15277 * Fixes for bad code generation during loop unrolling. 15278 * Fixes for bad code generation by the sibling call optimization. 15279 * Minor improvements to x86 code generation. 15280 * Implementation of function descriptors in C++ vtables for IA64. 15281 * Numerous minor bug-fixes. 15282 15283 Additional changes in GCC 3.0.1 15284 15285 * C++ fixes for incorrect code-generation. 15286 * Improved cross-compiling support for the C++ standard library. 15287 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 15288 in GCC 3.0. 15289 * Fixes for various exception-handling bugs. 15290 * A port to the S/390 architecture. 15291 15292 General Optimizer Improvements 15293 15294 * [2]Basic block reordering pass. 15295 * New if-conversion pass with support for conditional (predicated) 15296 execution. 15297 * New tail call and sibling call elimination optimizations. 15298 * New register renaming pass. 15299 * New (experimental) [3]static single assignment (SSA) representation 15300 support. 15301 * New dead-code elimination pass implemented using the SSA 15302 representation. 15303 * [4]Global null pointer test elimination. 15304 * [5]Global code hoisting/unification. 15305 * More builtins and optimizations for stdio.h, string.h and old BSD 15306 functions, as well as for ISO C99 functions. 15307 * New builtin __builtin_expect for giving hints to the branch 15308 predictor. 15309 15310 New Languages and Language specific improvements 15311 15312 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 15313 and supported, including the run-time library containing most 15314 common non-GUI Java classes, a bytecode interpreter, and the Boehm 15315 conservative garbage collector. Many bugs have been fixed. GCJ can 15316 compile Java source or Java bytecodes to either native code or Java 15317 class files, and supports native methods written in either the 15318 standard JNI or the more efficient and convenient CNI. 15319 * Here is a [6]partial list of C++ improvements, both new features 15320 and those no longer supported. 15321 * New C++ ABI. On the IA-64 platform GCC is capable of 15322 inter-operating with other IA-64 compilers. 15323 * The new ABI also significantly reduces the size of symbol and debug 15324 information. 15325 * New C++ support library and many C++ bug fixes, vastly improving 15326 our conformance to the ISO C++ standard. 15327 * New [7]inliner for C++. 15328 * Rewritten C preprocessor, integrated into the C, C++ and Objective 15329 C compilers, with very many improvements including ISO C99 support 15330 and [8]improvements to dependency generation. 15331 * Support for more [9]ISO C99 features. 15332 * Many improvements to support for checking calls to format functions 15333 such as printf and scanf, including support for ISO C99 format 15334 features, extensions from the Single Unix Specification and GNU 15335 libc 2.2, checking of strfmon formats and features to assist in 15336 auditing for format string security bugs. 15337 * New warnings for C code that may have undefined semantics because 15338 of violations of sequence point rules in the C standard (such as a 15339 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 15340 * Additional warning option -Wfloat-equal. 15341 * Improvements to -Wtraditional. 15342 * Fortran improvements are listed in [10]the Fortran documentation. 15343 15344 New Targets and Target Specific Improvements 15345 15346 * New x86 back end, generating much improved code. 15347 * Support for a generic i386-elf target contributed. 15348 * New option to emit x86 assembly code using Intel style syntax 15349 (-mintel-syntax). 15350 * HPUX 11 support contributed. 15351 * Improved PowerPC code generation, including scheduled prologue and 15352 epilogue. 15353 * Port of GCC to Intel's IA-64 processor contributed. 15354 * Port of GCC to Motorola's MCore 210 and 340 contributed. 15355 * New unified back-end for Arm, Thumb and StrongArm contributed. 15356 * Port of GCC to Intel's XScale processor contributed. 15357 * Port of GCC to Atmel's AVR microcontrollers contributed. 15358 * Port of GCC to Mitsubishi's D30V processor contributed. 15359 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 15360 processor family) contributed. 15361 * Port of GCC to Fujitsu's FR30 processor contributed. 15362 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 15363 * Port of GCC to Sun's picoJava processor core contributed. 15364 15365 Documentation improvements 15366 15367 * Substantially rewritten and improved C preprocessor manual. 15368 * Many improvements to other documentation. 15369 * Manpages for gcc, cpp and gcov are now generated automatically from 15370 the master Texinfo manual, eliminating the problem of manpages 15371 being out of date. (The generated manpages are only extracts from 15372 the full manual, which is provided in Texinfo form, from which 15373 info, HTML, other formats and a printed manual can be generated.) 15374 * Generated info files are included in the release tarballs alongside 15375 their Texinfo sources, avoiding problems on some platforms with 15376 building makeinfo as part of the GCC distribution. 15377 15378 Other significant improvements 15379 15380 * Garbage collection used internally by the compiler for most memory 15381 allocation instead of obstacks. 15382 * Lengauer and Tarjan algorithm used for computing dominators in the 15383 CFG. This algorithm can be significantly faster and more space 15384 efficient than our older algorithm. 15385 * gccbug script provided to assist in submitting bug reports to our 15386 bug tracking system. (Bug reports previously submitted directly to 15387 our mailing lists, for which you received no bug tracking number, 15388 should be submitted again using gccbug if you can reproduce the 15389 problem with GCC 3.0.) 15390 * The internal libgcc library is [11]built as a shared library on 15391 systems that support it. 15392 * Extensive testsuite included with GCC, with many new tests. In 15393 addition to tests for GCC bugs that have been fixed, many tests 15394 have been added for language features, compiler warnings and 15395 builtin functions. 15396 * Additional language-independent warning options -Wpacked, -Wpadded, 15397 -Wunreachable-code and -Wdisabled-optimization. 15398 * Target-independent options -falign-functions, -falign-loops and 15399 -falign-jumps. 15400 15401 Plus a great many bug fixes and almost all the [12]features found in 15402 GCC 2.95. 15403 15404 15405 For questions related to the use of GCC, please consult these web 15406 pages and the [13]GCC manuals. If that fails, the 15407 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15408 web pages and the development of GCC are welcome on our developer 15409 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 15410 archives. 15411 15412 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 15413 distribution of this entire article is permitted in any medium, 15414 provided this notice is preserved. 15415 15416 These pages are [18]maintained by the GCC team. Last modified 15417 2018-09-30[19]. 15418 15419 References 15420 15421 1. http://www.netbsd.org/ 15422 2. http://gcc.gnu.org/news/reorder.html 15423 3. http://gcc.gnu.org/news/ssa.html 15424 4. http://gcc.gnu.org/news/null.html 15425 5. http://gcc.gnu.org/news/unify.html 15426 6. http://gcc.gnu.org/gcc-3.0/c++features.html 15427 7. http://gcc.gnu.org/news/inlining.html 15428 8. http://gcc.gnu.org/news/dependencies.html 15429 9. http://gcc.gnu.org/c99status.html 15430 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 15431 11. http://gcc.gnu.org/gcc-3.0/libgcc.html 15432 12. http://gcc.gnu.org/gcc-2.95/features.html 15433 13. https://gcc.gnu.org/onlinedocs/ 15434 14. mailto:gcc-help (a] gcc.gnu.org 15435 15. mailto:gcc (a] gcc.gnu.org 15436 16. https://gcc.gnu.org/lists.html 15437 17. https://www.fsf.org/ 15438 18. https://gcc.gnu.org/about.html 15439 19. http://validator.w3.org/check/referer 15440 ====================================================================== 15441 http://gcc.gnu.org/gcc-3.0/caveats.html 15442 GCC 3.0 Caveats 15443 15444 * -fstrict-aliasing is now part of -O2 and higher optimization 15445 levels. This allows the compiler to assume the strictest aliasing 15446 rules applicable to the language being compiled. For C and C++, 15447 this activates optimizations based on the type of expressions. This 15448 optimization may thus break old, non-compliant code. 15449 * Enumerations are now properly promoted to int in function 15450 parameters and function returns. Normally this change is not 15451 visible, but when using -fshort-enums this is an ABI change. 15452 * The undocumented extension that allowed C programs to have a label 15453 at the end of a compound statement has been deprecated and may be 15454 removed in a future version. Programs that now generate a warning 15455 about this may be fixed by adding a null statement (a single 15456 semicolon) after the label. 15457 * The poorly documented extension that allowed string constants in C, 15458 C++ and Objective C to contain unescaped newlines has been 15459 deprecated and may be removed in a future version. Programs using 15460 this extension may be fixed in several ways: the bare newline may 15461 be replaced by \n, or preceded by \n\, or string concatenation may 15462 be used with the bare newline preceded by \n" and " placed at the 15463 start of the next line. 15464 * The Chill compiler is not included in GCC 3.0, because of the lack 15465 of a volunteer to convert it to use garbage collection. 15466 * Certain non-standard iostream methods from earlier versions of 15467 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 15468 ostream::form, and istream::gets. 15469 * The new C++ ABI is not yet fully supported by current (as of 15470 2001-07-01) releases and development versions of GDB, or any 15471 earlier versions. There is a problem setting breakpoints by line 15472 number, and other related issues that have been fixed in GCC 3.0 15473 but not yet handled in GDB: 15474 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 15475 15476 15477 For questions related to the use of GCC, please consult these web 15478 pages and the [2]GCC manuals. If that fails, the 15479 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15480 web pages and the development of GCC are welcome on our developer 15481 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 15482 15483 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 15484 distribution of this entire article is permitted in any medium, 15485 provided this notice is preserved. 15486 15487 These pages are [7]maintained by the GCC team. Last modified 15488 2018-09-30[8]. 15489 15490 References 15491 15492 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 15493 2. https://gcc.gnu.org/onlinedocs/ 15494 3. mailto:gcc-help (a] gcc.gnu.org 15495 4. mailto:gcc (a] gcc.gnu.org 15496 5. https://gcc.gnu.org/lists.html 15497 6. https://www.fsf.org/ 15498 7. https://gcc.gnu.org/about.html 15499 8. http://validator.w3.org/check/referer 15500 ====================================================================== 15501 http://gcc.gnu.org/gcc-2.95/index.html 15502 GCC 2.95 15503 15504 March 16, 2001: The GNU project and the GCC developers are pleased to 15505 announce the release of GCC version 2.95.3. 15506 15507 This release is no longer maintained. 15508 15509 Release History 15510 15511 GCC 2.95.3 15512 March 16, 2001 15513 15514 GCC 2.95.2 15515 October 27, 1999 15516 15517 GCC 2.95.1 15518 August 19, 1999 15519 15520 GCC 2.95 15521 July 31, 1999. This is the first release of GCC since the April 15522 1999 GCC/EGCS reunification and includes nearly a year's worth 15523 of new development and bugfixes. 15524 15525 References and Acknowledgements 15526 15527 GCC used to stand for the GNU C Compiler, but since the compiler 15528 supports several other languages aside from C, it now stands for the 15529 GNU Compiler Collection. 15530 15531 The whole suite has been extensively [1]regression tested and 15532 [2]package tested. It should be reliable and suitable for widespread 15533 use. 15534 15535 The compiler has several new optimizations, new targets, new languages 15536 and other new features. See the [3]new features page for a more 15537 complete list of new features found in the GCC 2.95 releases. 15538 15539 The sources include installation instructions in both HTML and 15540 plaintext forms in the install directory in the distribution. However, 15541 the most up to date installation instructions and [4]build/test status 15542 are on the web pages. We will update those pages as new information 15543 becomes available. 15544 15545 The GCC developers would like to thank the numerous people that have 15546 contributed new features, test results, bugfixes, etc to GCC. This 15547 [5]amazing group of volunteers is what makes GCC successful. 15548 15549 And finally, we can't in good conscience fail to mention some 15550 [6]caveats to using GCC 2.95. 15551 15552 Download GCC 2.95 from one of our many [7]mirror sites. 15553 15554 For additional information about GCC please see the [8]GCC project web 15555 server or contact the [9]GCC development mailing list. 15556 15557 15558 For questions related to the use of GCC, please consult these web 15559 pages and the [10]GCC manuals. If that fails, the 15560 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15561 web pages and the development of GCC are welcome on our developer 15562 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 15563 archives. 15564 15565 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 15566 distribution of this entire article is permitted in any medium, 15567 provided this notice is preserved. 15568 15569 These pages are [15]maintained by the GCC team. Last modified 15570 2018-09-30[16]. 15571 15572 References 15573 15574 1. http://gcc.gnu.org/gcc-2.95/regress.html 15575 2. http://gcc.gnu.org/gcc-2.95/othertest.html 15576 3. http://gcc.gnu.org/gcc-2.95/features.html 15577 4. http://gcc.gnu.org/gcc-2.95/buildstat.html 15578 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15579 6. http://gcc.gnu.org/gcc-2.95/caveats.html 15580 7. http://gcc.gnu.org/mirrors.html 15581 8. http://gcc.gnu.org/index.html 15582 9. mailto:gcc (a] gcc.gnu.org 15583 10. https://gcc.gnu.org/onlinedocs/ 15584 11. mailto:gcc-help (a] gcc.gnu.org 15585 12. mailto:gcc (a] gcc.gnu.org 15586 13. https://gcc.gnu.org/lists.html 15587 14. https://www.fsf.org/ 15588 15. https://gcc.gnu.org/about.html 15589 16. http://validator.w3.org/check/referer 15590 ====================================================================== 15591 http://gcc.gnu.org/gcc-2.95/features.html 15592 GCC 2.95 New Features 15593 15594 * General Optimizer Improvements: 15595 + [1]Localized register spilling to improve speed and code 15596 density especially on small register class machines. 15597 + [2]Global CSE using lazy code motion algorithms. 15598 + [3]Improved global constant/copy propagation. 15599 + [4]Improved control flow graph analysis and manipulation. 15600 + [5]Local dead store elimination. 15601 + [6]Memory Load hoisting/store sinking in loops. 15602 + [7]Type based alias analysis is enabled by default. Note this 15603 feature will expose bugs in the Linux kernel. Please refer to 15604 the FAQ (as shipped with GCC 2.95) for additional information 15605 on this issue. 15606 + Major revamp of GIV detection, combination and simplification 15607 to improve loop performance. 15608 + Major improvements to register allocation and reloading. 15609 * New Languages and Language specific improvements 15610 + [8]Many C++ improvements. 15611 + [9]Many Fortran improvements. 15612 + [10]Java front-end has been integrated. A [11]runtime library 15613 is available separately. 15614 + [12]ISO C99 support 15615 + [13]Chill front-end and runtime has been integrated. 15616 + Boehm garbage collector support in libobjc. 15617 + More support for various pragmas which appear in vendor 15618 include files 15619 * New Targets and Target Specific Improvements 15620 + [14]SPARC backend rewrite. 15621 + -mschedule=8000 will optimize code for PA8000 class 15622 processors; -mpa-risc-2-0 will generate code for PA2.0 15623 processors 15624 + Various micro-optimizations for the ia32 port. K6 15625 optimizations 15626 + Compiler will attempt to align doubles in the stack on the 15627 ia32 port 15628 + Alpha EV6 support 15629 + PowerPC 750 15630 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 15631 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 15632 + c3x, c4x 15633 + HyperSPARC 15634 + SparcLite86x 15635 + sh4 15636 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 15637 arm-linux) 15638 + vxWorks targets include support for vxWorks threads 15639 + StrongARM 110 and ARM9 support added. ARM Scheduling 15640 parameters rewritten. 15641 + Various changes to the MIPS port to avoid assembler macros, 15642 which in turn improves performance 15643 + Various performance improvements to the i960 port. 15644 + Major rewrite of ns32k port 15645 * Other significant improvements 15646 + [15]Ability to dump cfg information and display it using vcg. 15647 + The new faster scheme for fixing vendor header files is 15648 enabled by default. 15649 + Experimental internationalization support. 15650 + multibyte character support 15651 + Some compile-time speedups for pathological problems 15652 + Better support for complex types 15653 * Plus the usual mountain of bugfixes 15654 * Core compiler is based on the gcc2 development tree from Sept 30, 15655 1998, so we have all of the [16]features found in GCC 2.8. 15656 15657 Additional Changes in GCC 2.95.1 15658 15659 * Generic bugfixes and improvements 15660 + Various documentation fixes related to the GCC/EGCS merger. 15661 + Fix memory management bug which could lead to spurious aborts, 15662 core dumps or random parsing errors in the compiler. 15663 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 15664 support. 15665 + Fix infinite loop in the CSE optimizer. 15666 + Avoid undefined behavior in compiler FP emulation code 15667 + Fix install problem when prefix is overridden on the make 15668 install command. 15669 + Fix problem with unwanted installation of assert.h on some 15670 systems. 15671 + Fix problem with finding the wrong assembler in a single tree 15672 build. 15673 + Avoid increasing the known alignment of a register that is 15674 already known to be a pointer. 15675 * Platform specific bugfixes and improvements 15676 + Codegen bugfix for prologue/epilogue for cpu32 target. 15677 + Fix long long code generation bug for the Coldfire target. 15678 + Fix various aborts in the SH compiler. 15679 + Fix bugs in libgcc support library for the SH. 15680 + Fix alpha ev6 code generation bug. 15681 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 15682 AIX platforms. 15683 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 15684 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 15685 targets. 15686 + Fix weak symbol handling for rs6000/ppc svr4 targets. 15687 + Fix various problems with 64bit code generation for the 15688 rs6000/ppc port. 15689 + Fix codegen bug which caused tetex to be mis-compiled on the 15690 x86. 15691 + Fix compiler abort in new cfg code exposed by x86 port. 15692 + Fix out of range array reference in code convert flat 15693 registers to the x87 stacked FP register file. 15694 + Fix minor vxworks configuration bug. 15695 + Fix return type of bsearch for SunOS 4.x. 15696 * Language & Runtime specific fixes. 15697 + The G++ signature extension has been deprecated. It will be 15698 removed in the next major release of G++. Use of signatures 15699 will result in a warning from the compiler. 15700 + Several bugs relating to templates and namespaces were fixed. 15701 + A bug that caused crashes when combining templates with -g on 15702 DWARF1 platforms was fixed. 15703 + Pointers-to-members, virtual functions, and multiple 15704 inheritance should now work together correctly. 15705 + Some code-generation bugs relating to function try blocks were 15706 fixed. 15707 + G++ is a little bit more lenient with certain archaic 15708 constructs than in GCC 2.95. 15709 + Fix to prevent shared library version #s from bring truncated 15710 to 1 digit 15711 + Fix missing std:: in the libstdc++ library. 15712 + Fix stream locking problems in libio. 15713 + Fix problem in java compiler driver. 15714 15715 Additional Changes in GCC 2.95.2 15716 15717 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 15718 the optimizations performed by -fstrict-aliasing are valid according to 15719 the C and C++ standards, the optimization have caused some problems, 15720 particularly with old non-conforming code. 15721 15722 The GCC developers are experimenting with ways to warn users about code 15723 which violates the C/C++ standards, but those warnings are not ready 15724 for widespread use at this time. Rather than wait for those warnings 15725 the GCC developers have chosen to disable -fstrict-aliasing by default 15726 for the GCC 2.95.2 release. 15727 15728 We strongly encourage developers to find and fix code which violates 15729 the C/C++ standards as -fstrict-aliasing may be enabled by default in 15730 future releases. Use the option -fstrict-aliasing to re-enable these 15731 optimizations. 15732 * Generic bugfixes and improvements 15733 + Fix incorrectly optimized memory reference in global common 15734 subexpression elimination (GCSE) optimization pass. 15735 + Fix code generation bug in regmove.c in which it could 15736 incorrectly change a "const" value. 15737 + Fix bug in optimization of conditionals involving volatile 15738 memory references. 15739 + Avoid over-allocation of stack space for some procedures. 15740 + Fixed bug in the compiler which caused incorrect optimization 15741 of an obscure series of bit manipulations, shifts and 15742 arithmetic. 15743 + Fixed register allocator bug which caused teTeX to be 15744 mis-compiled on SPARC targets. 15745 + Avoid incorrect optimization of degenerate case statements for 15746 certain targets such as the ARM. 15747 + Fix out of range memory reference in the jump optimizer. 15748 + Avoid dereferencing null pointer in fix-header. 15749 + Fix test for GCC specific features so that it is possible to 15750 bootstrap with gcc-2.6.2 and older versions of GCC. 15751 + Fix typo in scheduler which could potentially cause out of 15752 range memory accesses. 15753 + Avoid incorrect loop reversal which caused incorrect code for 15754 certain loops on PowerPC targets. 15755 + Avoid incorrect optimization of switch statements on certain 15756 targets (for example the ARM). 15757 * Platform specific bugfixes and improvements 15758 + Work around bug in Sun V5.0 compilers which caused bootstrap 15759 comparison failures on SPARC targets. 15760 + Fix SPARC backend bug which caused aborts in final.c. 15761 + Fix sparc-hal-solaris2* configuration fragments. 15762 + Fix bug in sparc block profiling. 15763 + Fix obscure code generation bug for the PARISC targets. 15764 + Define __STDC_EXT__ for HPUX configurations. 15765 + Various POWERPC64 code generation bugfixes. 15766 + Fix abort for PPC targets using ELF (ex GNU/Linux). 15767 + Fix collect2 problems for AIX targets. 15768 + Correct handling of .file directive for PPC targets. 15769 + Fix bug in fix_trunc x86 patterns. 15770 + Fix x86 port to correctly pop the FP stack for functions that 15771 return structures in memory. 15772 + Fix minor bug in strlen x86 pattern. 15773 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 15774 + Fix template repository code to handle leading underscore in 15775 mangled names. 15776 + Fix weak/weak alias support for OpenBSD. 15777 + GNU/Linux for the ARM has C++ compatible include files. 15778 * Language & Runtime specific fixes. 15779 + Fix handling of constructor attribute in the C front-end which 15780 caused problems building the Chill runtime library on some 15781 targets. 15782 + Fix minor problem merging type qualifiers in the C front-end. 15783 + Fix aliasing bug for pointers and references (C/C++). 15784 + Fix incorrect "non-constant initializer bug" when -traditional 15785 or -fwritable-strings is enabled. 15786 + Fix build error for Chill front-end on SunOS. 15787 + Do not complain about duplicate instantiations when using 15788 -frepo (C++). 15789 + Fix array bounds handling in C++ front-end which caused 15790 problems with dwarf debugging information in some 15791 circumstances. 15792 + Fix minor namespace problem. 15793 + Fix problem linking java programs. 15794 15795 Additional Changes in GCC 2.95.3 15796 15797 * Generic bugfixes and improvements 15798 + Fix numerous problems that caused incorrect optimization in 15799 the register reloading code. 15800 + Fix numerous problems that caused incorrect optimization in 15801 the loop optimizer. 15802 + Fix aborts in the functions build_insn_chain and scan_loops 15803 under some circumstances. 15804 + Fix an alias analysis bug. 15805 + Fix an infinite compilation bug in the combiner. 15806 + A few problems with complex number support have been fixed. 15807 + It is no longer possible for gcc to act as a fork bomb when 15808 installed incorrectly. 15809 + The -fpack-struct option should be recognized now. 15810 + Fixed a bug that caused incorrect code to be generated due to 15811 a lost stack adjustment. 15812 * Platform specific bugfixes and improvements 15813 + Support building ARM toolchains hosted on Windows. 15814 + Fix attribute calculations in ARM toolchains. 15815 + arm-linux support has been improved. 15816 + Fix a PIC failure on sparc targets. 15817 + On ix86 targets, the regparm attribute should now work 15818 reliably. 15819 + Several updates for the h8300 port. 15820 + Fix problem building libio with glibc 2.2. 15821 15822 15823 For questions related to the use of GCC, please consult these web 15824 pages and the [17]GCC manuals. If that fails, the 15825 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15826 web pages and the development of GCC are welcome on our developer 15827 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 15828 archives. 15829 15830 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 15831 distribution of this entire article is permitted in any medium, 15832 provided this notice is preserved. 15833 15834 These pages are [22]maintained by the GCC team. Last modified 15835 2018-09-30[23]. 15836 15837 References 15838 15839 1. http://gcc.gnu.org/news/spill.html 15840 2. http://gcc.gnu.org/news/lcm.html 15841 3. http://gcc.gnu.org/news/cprop.html 15842 4. http://gcc.gnu.org/news/cfg.html 15843 5. http://gcc.gnu.org/news/dse.html 15844 6. http://gcc.gnu.org/news/hoist.html 15845 7. http://gcc.gnu.org/news/alias.html 15846 8. http://gcc.gnu.org/gcc-2.95/c++features.html 15847 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 15848 10. http://gcc.gnu.org/news/gcj-announce.txt 15849 11. http://gcc.gnu.org/news/javaannounce.html 15850 12. http://gcc.gnu.org/c99status.html 15851 13. http://gcc.gnu.org/news/chill.html 15852 14. http://gcc.gnu.org/news/sparc.html 15853 15. http://gcc.gnu.org/news/egcs-vcg.html 15854 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html 15855 17. https://gcc.gnu.org/onlinedocs/ 15856 18. mailto:gcc-help (a] gcc.gnu.org 15857 19. mailto:gcc (a] gcc.gnu.org 15858 20. https://gcc.gnu.org/lists.html 15859 21. https://www.fsf.org/ 15860 22. https://gcc.gnu.org/about.html 15861 23. http://validator.w3.org/check/referer 15862 ====================================================================== 15863 http://gcc.gnu.org/gcc-2.95/caveats.html 15864 GCC 2.95 Caveats 15865 15866 * GCC 2.95 will issue an error for invalid asm statements that had 15867 been silently accepted by earlier versions of the compiler. This is 15868 particularly noticeable when compiling older versions of the Linux 15869 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 15870 for more information on this issue. 15871 * GCC 2.95 implements type based alias analysis to disambiguate 15872 memory references. Some programs, particularly the Linux kernel 15873 violate ANSI/ISO aliasing rules and therefore may not operate 15874 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 15875 shipped with GCC 2.95) for more information on this issue. 15876 * GCC 2.95 has a known bug in its handling of complex variables for 15877 64bit targets. Instead of silently generating incorrect code, GCC 15878 2.95 will issue a fatal error for situations it can not handle. 15879 This primarily affects the Fortran community as Fortran makes more 15880 use of complex variables than C or C++. 15881 * GCC 2.95 has an integrated libstdc++, but does not have an 15882 integrated libg++. Furthermore old libg++ releases will not work 15883 with GCC 2.95. You can retrieve a recent copy of libg++ from the 15884 [1]GCC ftp server. 15885 Note most C++ programs only need libstdc++. 15886 * Exception handling may not work with shared libraries, particularly 15887 on alphas, hppas, rs6000/powerpc and mips based platforms. 15888 Exception handling is known to work on x86 GNU/Linux platforms with 15889 shared libraries. 15890 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 15891 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 15892 or EGCS 1.1. As a result it may be necessary to fix C++ code before 15893 it will compile with GCC 2.95. 15894 * G++ is also converting toward the ISO C++ standard; as a result 15895 code which was previously valid (and thus accepted by other 15896 compilers and older versions of g++) may no longer be accepted. The 15897 flag -fpermissive may allow some non-conforming code to compile 15898 with GCC 2.95. 15899 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 15900 1.1.x, EGCS 1.0.x or GCC 2.8.x. 15901 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 15902 between Sept 30, 1998 and April 30, 1999 (the official end of the 15903 GCC 2.8 project). Future GCC releases will include all the changes 15904 from the defunct GCC 2.8 sources. 15905 15906 15907 For questions related to the use of GCC, please consult these web 15908 pages and the [2]GCC manuals. If that fails, the 15909 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15910 web pages and the development of GCC are welcome on our developer 15911 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 15912 15913 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 15914 distribution of this entire article is permitted in any medium, 15915 provided this notice is preserved. 15916 15917 These pages are [7]maintained by the GCC team. Last modified 15918 2018-09-30[8]. 15919 15920 References 15921 15922 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 15923 2. https://gcc.gnu.org/onlinedocs/ 15924 3. mailto:gcc-help (a] gcc.gnu.org 15925 4. mailto:gcc (a] gcc.gnu.org 15926 5. https://gcc.gnu.org/lists.html 15927 6. https://www.fsf.org/ 15928 7. https://gcc.gnu.org/about.html 15929 8. http://validator.w3.org/check/referer 15930 ====================================================================== 15931 http://gcc.gnu.org/egcs-1.1/index.html 15932 EGCS 1.1 15933 15934 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 15935 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 15936 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 15937 15938 EGCS is a free software project to further the development of the GNU 15939 compilers using an open development environment. 15940 15941 EGCS 1.1 is a major new release of the EGCS compiler system. It has 15942 been [1]extensively tested and is believed to be stable and suitable 15943 for widespread use. 15944 15945 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 15946 development sources; it contains all of the new features found in GCC 15947 2.8.1 as well as all new development from GCC up to June 6, 1998. 15948 15949 EGCS 1.1 also contains many improvements and features not found in GCC 15950 or in older versions of EGCS: 15951 * Global common subexpression elimination and global constant/copy 15952 propagation (aka [2]gcse) 15953 * Ongoing improvements to the [3]alias analysis support to allow for 15954 better optimizations throughout the compiler. 15955 * Vastly improved [4]C++ compiler and integrated C++ runtime 15956 libraries. 15957 * Fixes for the /tmp symlink race security problems. 15958 * New targets including mips16, arm-thumb and 64 bit PowerPC. 15959 * Improvements to GNU Fortran (g77) compiler and runtime library made 15960 since g77 version 0.5.23. 15961 15962 See the [5]new features page for a more complete list of new features 15963 found in EGCS 1.1 releases. 15964 15965 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 15966 1.1: 15967 * General improvements and fixes 15968 + Avoid some stack overflows when compiling large functions. 15969 + Avoid incorrect loop invariant code motions. 15970 + Fix some core dumps on Linux kernel code. 15971 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 15972 + Fix code generation problem in gcse. 15973 + Various documentation related fixes. 15974 * g++/libstdc++ improvements and fixes 15975 + MT safe EH fix for setjmp/longjmp based exception handling. 15976 + Fix a few bad interactions between optimization and exception 15977 handling. 15978 + Fixes for demangling of template names starting with "__". 15979 + Fix a bug that would fail to run destructors in some cases 15980 with -O2. 15981 + Fix 'new' of classes with virtual bases. 15982 + Fix crash building Qt on the Alpha. 15983 + Fix failure compiling WIFEXITED macro on GNU/Linux. 15984 + Fix some -frepo failures. 15985 * g77 and libf2c improvements and fixes 15986 + Various documentation fixes. 15987 + Avoid compiler crash on RAND intrinsic. 15988 + Fix minor bugs in makefiles exposed by BSD make programs. 15989 + Define _XOPEN_SOURCE for libI77 build to avoid potential 15990 problems on some 64-bit systems. 15991 + Fix problem with implicit endfile on rewind. 15992 + Fix spurious recursive I/O errors. 15993 * platform specific improvements and fixes 15994 + Match all versions of UnixWare7. 15995 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 15996 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 15997 from unsigned ints to double precision floats. 15998 + Fix ARM ABI issue with NetBSD. 15999 + Fix a few arm code generation bugs. 16000 + Fixincludes will fix additional broken SCO OpenServer header 16001 files. 16002 + Fix a m68k backend bug which caused invalid offsets in reg+d 16003 addresses. 16004 + Fix problems with 64bit AIX 4.3 support. 16005 + Fix handling of long longs for varargs/stdarg functions on the 16006 ppc. 16007 + Minor fixes to CPP predefines for Windows. 16008 + Fix code generation problems with gpr<->fpr copies for 64bit 16009 ppc. 16010 + Fix a few coldfire code generation bugs. 16011 + Fix some more header file problems on SunOS 4.x. 16012 + Fix assert.h handling for RTEMS. 16013 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 16014 + Fix x86 compiler abort in reg-stack pass. 16015 + Fix cygwin/windows problem with section attributes. 16016 + Fix Alpha code generation problem exposed by SMP Linux 16017 kernels. 16018 + Fix typo in m68k 32->64bit integer conversion. 16019 + Make sure target libraries build with -fPIC for PPC & Alpha 16020 targets. 16021 16022 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 16023 1.1.1: 16024 * General improvements and fixes 16025 + Fix bug in loop optimizer which caused the SPARC (and 16026 potentially other) ports to segfault. 16027 + Fix infinite recursion in alias analysis and combiner code. 16028 + Fix bug in regclass preferencing. 16029 + Fix incorrect loop reversal which caused incorrect code to be 16030 generated for several targets. 16031 + Fix return value for builtin memcpy. 16032 + Reduce compile time for certain loops which exposed quadratic 16033 behavior in the loop optimizer. 16034 + Fix bug which caused volatile memory to be written multiple 16035 times when only one write was needed/desired. 16036 + Fix compiler abort in caller-save.c 16037 + Fix combiner bug which caused incorrect code generation for 16038 certain division by constant operations. 16039 + Fix incorrect code generation due to a bug in range check 16040 optimizations. 16041 + Fix incorrect code generation due to mis-handling of clobbered 16042 values in CSE. 16043 + Fix compiler abort/segfault due to incorrect register 16044 splitting when unrolling loops. 16045 + Fix code generation involving autoincremented addresses with 16046 ternary operators. 16047 + Work around bug in the scheduler which caused qt to be 16048 mis-compiled on some platforms. 16049 + Fix code generation problems with -fshort-enums. 16050 + Tighten security for temporary files. 16051 + Improve compile time for codes which make heavy use of 16052 overloaded functions. 16053 + Fix multiply defined constructor/destructor symbol problems. 16054 + Avoid setting bogus RPATH environment variable during 16055 bootstrap. 16056 + Avoid GNU-make dependencies in the texinfo subdir. 16057 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 16058 --enable-cpp=<dirname> can be used to specify an additional 16059 install directory for the cpp wrapper script. 16060 + Fix CSE bug which caused incorrect label-label refs to appear 16061 on some platforms. 16062 + Avoid linking in EH routines from libgcc if they are not 16063 needed. 16064 + Avoid obscure bug in aliasing code. 16065 + Fix bug in weak symbol handling. 16066 * Platform-specific improvements and fixes 16067 + Fix detection of PPro/PII on Unixware 7. 16068 + Fix compiler segfault when building spec99 and other programs 16069 for SPARC targets. 16070 + Fix code-generation bugs for integer and floating point 16071 conditional move instructions on the PPro/PII. 16072 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 16073 + Fix build failure for the arc port. 16074 + Fix floating point format configuration for i?86-gnu port. 16075 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 16076 threads are enabled. 16077 + Fix coldfire code generation bugs. 16078 + Fix "unrecognized insn" problems for Alpha and PPC ports. 16079 + Fix h8/300 code generation problem with floating point values 16080 in memory. 16081 + Fix unrecognized insn problems for the m68k port. 16082 + Fix namespace-pollution problem for the x86 port. 16083 + Fix problems with old assembler on x86 NeXT systems. 16084 + Fix PIC code-generation problems for the SPARC port. 16085 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 16086 + Fix minor ISO namespace violation in Alpha varargs/stdarg 16087 support. 16088 + Fix incorrect "braf" instruction usage for the SH port. 16089 + Fix minor bug in va-sh which prevented its use with -ansi. 16090 + Fix problems recognizing and supporting FreeBSD. 16091 + Handle OpenBSD systems correctly. 16092 + Minor fixincludes fix for Digital UNIX 4.0B. 16093 + Fix problems with ctors/dtors in SCO shared libraries. 16094 + Abort instead of generating incorrect code for PPro/PII 16095 floating point conditional moves. 16096 + Avoid multiply defined symbols on GNU/Linux systems using 16097 libc-5.4.xx. 16098 + Fix abort in alpha compiler. 16099 * Fortran-specific fixes 16100 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 16101 is in the documented, non-Y2K-compliant range of 0-99, instead 16102 of being returned as 100 in the year 2000. 16103 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 16104 milliseconds value properly in Values(8). 16105 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 16106 information properly in SArray(7). 16107 16108 Each release includes installation instructions in both HTML and 16109 plaintext forms (see the INSTALL directory in the toplevel directory of 16110 the distribution). However, we also keep the most up to date 16111 installation instructions and [6]build/test status on our web page. We 16112 will update those pages as new information becomes available. 16113 16114 The EGCS project would like to thank the numerous people that have 16115 contributed new features, test results, bugfixes, etc. This [7]amazing 16116 group of volunteers is what makes EGCS successful. 16117 16118 And finally, we can't in good conscience fail to mention some 16119 [8]caveats to using EGCS 1.1. 16120 16121 Download EGCS from egcs.cygnus.com (USA California). 16122 16123 The EGCS 1.1 release is also available on many mirror sites. 16124 [9]Goto mirror list to find a closer site. 16125 16126 16127 For questions related to the use of GCC, please consult these web 16128 pages and the [10]GCC manuals. If that fails, the 16129 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16130 web pages and the development of GCC are welcome on our developer 16131 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 16132 archives. 16133 16134 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 16135 distribution of this entire article is permitted in any medium, 16136 provided this notice is preserved. 16137 16138 These pages are [15]maintained by the GCC team. Last modified 16139 2018-09-30[16]. 16140 16141 References 16142 16143 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 16144 2. http://gcc.gnu.org/news/gcse.html 16145 3. http://gcc.gnu.org/news/alias.html 16146 4. http://gcc.gnu.org/egcs-1.1/c++features.html 16147 5. http://gcc.gnu.org/egcs-1.1/features.html 16148 6. http://gcc.gnu.org/egcs-1.1/buildstat.html 16149 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16150 8. http://gcc.gnu.org/egcs-1.1/caveats.html 16151 9. http://gcc.gnu.org/mirrors.html 16152 10. https://gcc.gnu.org/onlinedocs/ 16153 11. mailto:gcc-help (a] gcc.gnu.org 16154 12. mailto:gcc (a] gcc.gnu.org 16155 13. https://gcc.gnu.org/lists.html 16156 14. https://www.fsf.org/ 16157 15. https://gcc.gnu.org/about.html 16158 16. http://validator.w3.org/check/referer 16159 ====================================================================== 16160 http://gcc.gnu.org/egcs-1.1/features.html 16161 EGCS 1.1 new features 16162 16163 * Integrated GNU Fortran (g77) compiler and runtime library with 16164 improvements, based on g77 version 0.5.23. 16165 * Vast improvements in the C++ compiler; so many they have [1]page of 16166 their own! 16167 * Compiler implements [2]global common subexpression elimination and 16168 global copy/constant propagation. 16169 * More major improvements in the [3]alias analysis code. 16170 * More major improvements in the exception handling code to improve 16171 performance, lower static overhead and provide the infrastructure 16172 for future improvements. 16173 * The infamous /tmp symlink race security problems have been fixed. 16174 * The regmove optimization pass has been nearly completely rewritten 16175 to improve performance of generated code. 16176 * The compiler now recomputes register usage information before local 16177 register allocation. By providing more accurate information to the 16178 priority based allocator, we get better register allocation. 16179 * The register reloading phase of the compiler optimizes spill code 16180 much better than in previous releases. 16181 * Some bad interactions between the register allocator and 16182 instruction scheduler have been fixed, resulting in much better 16183 code for certain programs. Additionally, we have tuned the 16184 scheduler in various ways to improve performance of generated code 16185 for some architectures. 16186 * The compiler's branch shortening algorithms have been significantly 16187 improved to work better on targets which align jump targets. 16188 * The compiler now supports -Os to prefer optimizing for code space 16189 over optimizing for code speed. 16190 * The compiler will now totally eliminate library calls which compute 16191 constant values. This primarily helps targets with no integer 16192 div/mul support and targets without floating point support. 16193 * The compiler now supports an extensive "--help" option. 16194 * cpplib has been greatly improved and may be suitable for limited 16195 use. 16196 * Memory footprint for the compiler has been significantly reduced 16197 for some pathological cases. 16198 * The time to build EGCS has been improved for certain targets 16199 (particularly the alpha and mips platforms). 16200 * Many infrastructure improvements throughout the compiler, plus the 16201 usual mountain of bugfixes and minor improvements. 16202 * Target dependent improvements: 16203 + SPARC port now includes V8 plus and V9 support as well as 16204 performance tuning for Ultra class machines. The SPARC port 16205 now uses the Haifa scheduler. 16206 + Alpha port has been tuned for the EV6 processor and has an 16207 optimized expansion of memcpy/bzero. The Alpha port now uses 16208 the Haifa scheduler. 16209 + RS6000/PowerPC: support for the Power64 architecture and AIX 16210 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 16211 + x86: Alignment of static store data and jump targets is per 16212 Intel recommendations now. Various improvements throughout the 16213 x86 port to improve performance on Pentium processors 16214 (including improved epilogue sequences for Pentium chips and 16215 backend improvements which should help register allocation on 16216 all x86 variants. Conditional move support has been fixed and 16217 enabled for PPro processors. The x86 port also better supports 16218 64bit operations now. Unixware 7, a System V Release 5 target, 16219 is now supported and SCO OpenServer targets can support GAS. 16220 + MIPS has improved multiply/multiply-add support and now 16221 includes mips16 ISA support. 16222 + M68k has many micro-optimizations and Coldfire fixes. 16223 * Core compiler is based on the GCC development tree from June 9, 16224 1998, so we have all of the [4]features found in GCC 2.8. 16225 16226 16227 For questions related to the use of GCC, please consult these web 16228 pages and the [5]GCC manuals. If that fails, the 16229 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16230 web pages and the development of GCC are welcome on our developer 16231 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 16232 16233 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 16234 distribution of this entire article is permitted in any medium, 16235 provided this notice is preserved. 16236 16237 These pages are [10]maintained by the GCC team. Last modified 16238 2018-09-30[11]. 16239 16240 References 16241 16242 1. http://gcc.gnu.org/egcs-1.1/c++features.html 16243 2. http://gcc.gnu.org/news/gcse.html 16244 3. http://gcc.gnu.org/news/alias.html 16245 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html 16246 5. https://gcc.gnu.org/onlinedocs/ 16247 6. mailto:gcc-help (a] gcc.gnu.org 16248 7. mailto:gcc (a] gcc.gnu.org 16249 8. https://gcc.gnu.org/lists.html 16250 9. https://www.fsf.org/ 16251 10. https://gcc.gnu.org/about.html 16252 11. http://validator.w3.org/check/referer 16253 ====================================================================== 16254 http://gcc.gnu.org/egcs-1.1/caveats.html 16255 EGCS 1.1 Caveats 16256 16257 * EGCS has an integrated libstdc++, but does not have an integrated 16258 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 16259 Lu has made a libg++-2.8.1.2 snapshot available which may work with 16260 EGCS. 16261 Note most C++ programs only need libstdc++. 16262 * Exception handling may not work with shared libraries, particularly 16263 on alphas, hppas, rs6000/powerpc and mips based platforms. 16264 Exception handling is known to work on x86-linux platforms with 16265 shared libraries. 16266 * Some versions of the Linux kernel have bugs which prevent them from 16267 being compiled or from running when compiled by EGCS. See the FAQ 16268 (as shipped with EGCS 1.1) for additional information. 16269 * In general, EGCS is more rigorous about rejecting invalid C++ code 16270 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 16271 a result it may be necessary to fix C++ code before it will compile 16272 with EGCS. 16273 * G++ is also converting toward the ISO C++ standard; as a result 16274 code which was previously valid (and thus accepted by other 16275 compilers and older versions of g++) may no longer be accepted. 16276 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 16277 or GCC 2.8.x due to changes necessary to support thread safe 16278 exception handling. 16279 16280 16281 For questions related to the use of GCC, please consult these web 16282 pages and the [1]GCC manuals. If that fails, the 16283 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16284 web pages and the development of GCC are welcome on our developer 16285 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 16286 16287 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 16288 distribution of this entire article is permitted in any medium, 16289 provided this notice is preserved. 16290 16291 These pages are [6]maintained by the GCC team. Last modified 16292 2018-09-30[7]. 16293 16294 References 16295 16296 1. https://gcc.gnu.org/onlinedocs/ 16297 2. mailto:gcc-help (a] gcc.gnu.org 16298 3. mailto:gcc (a] gcc.gnu.org 16299 4. https://gcc.gnu.org/lists.html 16300 5. https://www.fsf.org/ 16301 6. https://gcc.gnu.org/about.html 16302 7. http://validator.w3.org/check/referer 16303 ====================================================================== 16304 http://gcc.gnu.org/egcs-1.0/index.html 16305 EGCS 1.0 16306 16307 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 16308 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 16309 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 16310 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 16311 16312 EGCS is a collaborative effort involving several groups of hackers 16313 using an open development model to accelerate development and testing 16314 of GNU compilers and runtime libraries. 16315 16316 An important goal of EGCS is to allow wide scale testing of 16317 experimental features and optimizations; therefore, EGCS contains some 16318 features and optimizations which are still under development. However, 16319 EGCS has been carefully tested and should be comparable in quality to 16320 most GCC releases. 16321 16322 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 16323 development sources; it contains nearly all of the new features found 16324 in GCC 2.8. 16325 16326 EGCS 1.0 also contains many improvements and features not found in GCC 16327 2.7 and even the GCC 2.8 series (which was released after the original 16328 EGCS 1.0 release). 16329 * Integrated C++ runtime libraries, including support for most major 16330 GNU/Linux systems! 16331 * The integrated libstdc++ library includes a verbatim copy of SGI's 16332 STL release. 16333 * Integrated GNU Fortran compiler. 16334 * New instruction scheduler. 16335 * New alias analysis code. 16336 16337 See the [1]new features page for a more complete list of new features. 16338 16339 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 16340 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 16341 EGCS 1.0 release: 16342 * Add support for Red Hat 5.0 Linux and better support for Linux 16343 systems using glibc2. 16344 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 16345 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 16346 fix these problems. 16347 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 16348 handling interfaces. 16349 To avoid future compatibility problems, we strongly urge anyone who 16350 is planning on distributing shared libraries that contain C++ code 16351 to upgrade to EGCS 1.0.1 first. 16352 Soon after EGCS 1.0 was released, the GCC developers made some 16353 incompatible changes in libgcc's exception handling interfaces. 16354 These changes were needed to solve problems on some platforms. This 16355 means that GCC 2.8.0, when released, will not be seamlessly 16356 compatible with shared libraries built by EGCS 1.0. The reason is 16357 that the libgcc.a in GCC 2.8.0 will not contain a function needed 16358 by the old interface. 16359 The result of this is that there may be compatibility problems with 16360 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 16361 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 16362 and libgcc.a has the support routines for both the old and the new 16363 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 16364 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 16365 The maintainers of GCC 2.x have decided against including seamless 16366 support for the old interface in 2.8.0, since it was never 16367 "official", so to avoid future compatibility problems we recommend 16368 against distributing any shared libraries built by EGCS 1.0 that 16369 contain C++ code (upgrade to 1.0.1 and use that). 16370 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 16371 The x86 changes fix code generation errors exposed when building 16372 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 16373 The hppa change fixes a compiler abort when configured for use with 16374 RTEMS. 16375 The MIPS changes fix problems with the definition of LONG_MAX on 16376 newer systems, allow for command line selection of the target ABI, 16377 and fix one code generation problem. 16378 The rs6000/ppc change fixes some problems with passing structures 16379 to varargs/stdarg functions. 16380 * A few machine independent bugfixes, mostly to fix code generation 16381 errors when building Linux kernels or glibc. 16382 * Fix a few critical exception handling and template bugs in the C++ 16383 compiler. 16384 * Fix Fortran namelist bug on alphas. 16385 * Fix build problems on x86-solaris systems. 16386 16387 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 16388 serious problems in EGCS 1.0.1. 16389 * General improvements and fixes 16390 + Memory consumption significantly reduced, especially for 16391 templates and inline functions. 16392 + Fix various problems with glibc2.1. 16393 + Fix loop optimization bug exposed by rs6000/ppc port. 16394 + Fix to avoid potential code generation problems in jump.c. 16395 + Fix some undefined symbol problems in dwarf1 debug support. 16396 * g++/libstdc++ improvements and fixes 16397 + libstdc++ in the EGCS release has been updated and should be 16398 link compatible with libstdc++-2.8. 16399 + Various fixes in libio/libstdc++ to work better on GNU/Linux 16400 systems. 16401 + Fix problems with duplicate symbols on systems that do not 16402 support weak symbols. 16403 + Memory corruption bug and undefined symbols in bastring have 16404 been fixed. 16405 + Various exception handling fixes. 16406 + Fix compiler abort for very long thunk names. 16407 * g77 improvements and fixes 16408 + Fix compiler crash for omitted bound in Fortran CASE 16409 statement. 16410 + Add missing entries to g77 lang-options. 16411 + Fix problem with -fpedantic in the g77 compiler. 16412 + Fix "backspace" problem with g77 on alphas. 16413 + Fix x86 backend problem with Fortran literals and -fpic. 16414 + Fix some of the problems with negative subscripts for g77 on 16415 alphas. 16416 + Fixes for Fortran builds on cygwin32/mingw32. 16417 * platform specific improvements and fixes 16418 + Fix long double problems on x86 (exposed by glibc). 16419 + x86 ports define i386 again to keep imake happy. 16420 + Fix exception handling support on NetBSD ports. 16421 + Several changes to collect2 to fix many problems with AIX. 16422 + Define __ELF__ for GNU/Linux on rs6000. 16423 + Fix -mcall-linux problem on GNU/Linux on rs6000. 16424 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 16425 + Allow autoconf to select a proper install problem on AIX 3.1. 16426 + m68k port support includes -mcpu32 option as well as cpu32 16427 multilibs. 16428 + Fix stdarg bug for irix6. 16429 + Allow EGCS to build on irix5 without the gnu assembler. 16430 + Fix problem with static linking on sco5. 16431 + Fix bootstrap on sco5 with native compiler. 16432 + Fix for abort building newlib on H8 target. 16433 + Fix fixincludes handling of math.h on SunOS. 16434 + Minor fix for Motorola 3300 m68k systems. 16435 16436 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 16437 problems reported by Red Hat for builds of Red Hat 5.1. 16438 * Generic bugfixes: 16439 + Fix a typo in the libio library which resulted in incorrect 16440 behavior of istream::get. 16441 + Fix the Fortran negative array index problem. 16442 + Fix a major problem with the ObjC runtime thread support 16443 exposed by glibc2. 16444 + Reduce memory consumption of the Haifa scheduler. 16445 * Target specific bugfixes: 16446 + Fix one x86 floating point code generation bug exposed by 16447 glibc2 builds. 16448 + Fix one x86 internal compiler error exposed by glibc2 builds. 16449 + Fix profiling bugs on the Alpha. 16450 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 16451 + Fix rs6000/ppc bug when converting values from integer types 16452 to floating point types. 16453 16454 The EGCS 1.0 releases include installation instructions in both HTML 16455 and plaintext forms (see the INSTALL directory in the toplevel 16456 directory of the distribution). However, we also keep the most up to 16457 date installation instructions and [2]build/test status on our web 16458 page. We will update those pages as new information becomes available. 16459 16460 And, we can't in good conscience fail to mention some [3]caveats to 16461 using EGCS. 16462 16463 Update: Big thanks to Stanford for providing a high speed link for 16464 downloading EGCS (go.cygnus.com)! 16465 16466 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 16467 (USA California -- High speed link provided by Stanford). 16468 16469 The EGCS 1.0 release is also available many mirror sites. 16470 [4]Goto mirror list to find a closer site 16471 16472 We'd like to thank the numerous people that have contributed new 16473 features, test results, bugfixes, etc. Unfortunately, they're far too 16474 numerous to mention by name. 16475 16476 16477 For questions related to the use of GCC, please consult these web 16478 pages and the [5]GCC manuals. If that fails, the 16479 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16480 web pages and the development of GCC are welcome on our developer 16481 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 16482 16483 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 16484 distribution of this entire article is permitted in any medium, 16485 provided this notice is preserved. 16486 16487 These pages are [10]maintained by the GCC team. Last modified 16488 2018-09-30[11]. 16489 16490 References 16491 16492 1. http://gcc.gnu.org/egcs-1.0/features.html 16493 2. http://gcc.gnu.org/egcs-1.0/buildstat.html 16494 3. http://gcc.gnu.org/egcs-1.0/caveats.html 16495 4. http://gcc.gnu.org/mirrors.html 16496 5. https://gcc.gnu.org/onlinedocs/ 16497 6. mailto:gcc-help (a] gcc.gnu.org 16498 7. mailto:gcc (a] gcc.gnu.org 16499 8. https://gcc.gnu.org/lists.html 16500 9. https://www.fsf.org/ 16501 10. https://gcc.gnu.org/about.html 16502 11. http://validator.w3.org/check/referer 16503 ====================================================================== 16504 http://gcc.gnu.org/egcs-1.0/features.html 16505 EGCS 1.0 features 16506 16507 * Core compiler is based on the gcc2 development tree from Aug 2, 16508 1997, so we have most of the [1]features found in GCC 2.8. 16509 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 16510 * Vast improvements in the C++ compiler; so many they have [2]page of 16511 their own! 16512 * Integrated C++ runtime libraries, including support for most major 16513 GNU/Linux systems! 16514 * New instruction scheduler from IBM Haifa which includes support for 16515 function wide instruction scheduling as well as superscalar 16516 scheduling. 16517 * Significantly improved alias analysis code. 16518 * Improved register allocation for two address machines. 16519 * Significant code generation improvements for Fortran code on 16520 Alphas. 16521 * Various optimizations from the g77 project as well as improved loop 16522 optimizations. 16523 * Dwarf2 debug format support for some targets. 16524 * egcs libstdc++ includes the SGI STL implementation without changes. 16525 * As a result of these and other changes, egcs libstc++ is not binary 16526 compatible with previous releases of libstdc++. 16527 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 16528 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 16529 1.1), Support for RTEMS on several embedded targets, Support for 16530 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 16531 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 16532 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 16533 * RS6000/PowerPC ports generate code which can run on all 16534 RS6000/PowerPC variants by default. 16535 * -mcpu= and -march= switches for the x86 port to allow better 16536 control over how the x86 port generates code. 16537 * Includes the template repository patch (aka repo patch); note the 16538 new template code makes repo obsolete for ELF systems using gnu-ld 16539 such as GNU/Linux. 16540 * Plus the usual assortment of bugfixes and improvements. 16541 16542 16543 For questions related to the use of GCC, please consult these web 16544 pages and the [3]GCC manuals. If that fails, the 16545 [4]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16546 web pages and the development of GCC are welcome on our developer 16547 list at [5]gcc (a] gcc.gnu.org. All of [6]our lists have public archives. 16548 16549 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 16550 distribution of this entire article is permitted in any medium, 16551 provided this notice is preserved. 16552 16553 These pages are [8]maintained by the GCC team. Last modified 16554 2018-09-30[9]. 16555 16556 References 16557 16558 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html 16559 2. http://gcc.gnu.org/egcs-1.0/c++features.html 16560 3. https://gcc.gnu.org/onlinedocs/ 16561 4. mailto:gcc-help (a] gcc.gnu.org 16562 5. mailto:gcc (a] gcc.gnu.org 16563 6. https://gcc.gnu.org/lists.html 16564 7. https://www.fsf.org/ 16565 8. https://gcc.gnu.org/about.html 16566 9. http://validator.w3.org/check/referer 16567 ====================================================================== 16568 http://gcc.gnu.org/egcs-1.0/caveats.html 16569 EGCS 1.0 Caveats 16570 16571 * EGCS has an integrated libstdc++, but does not have an integrated 16572 libg++. Furthermore old libg++ releases will not work with egc; HJ 16573 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 16574 Note most C++ programs only need libstdc++. 16575 * Note that using -pedantic or -Wreturn-type can cause an explosion 16576 in the amount of memory needed for template-heavy C++ code, such as 16577 code that uses STL. Also note that -Wall includes -Wreturn-type, so 16578 if you use -Wall you will need to specify -Wno-return-type to turn 16579 it off. 16580 * Exception handling may not work with shared libraries, particularly 16581 on alphas, hppas, and mips based platforms. Exception handling is 16582 known to work on x86-linux platforms with shared libraries. 16583 * Some versions of the Linux kernel have bugs which prevent them from 16584 being compiled or from running when compiled by EGCS. See the FAQ 16585 (as shipped with EGCS 1.0) for additional information. 16586 * In general, EGCS is more rigorous about rejecting invalid C++ code 16587 or deprecated C++ constructs than G++ 2.7. As a result it may be 16588 necessary to fix C++ code before it will compile with EGCS. 16589 * G++ is also aggressively tracking the C++ standard; as a result 16590 code which was previously valid (and thus accepted by other 16591 compilers and older versions of G++) may no longer be accepted. 16592 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 16593 1.0.x and later releases should work with Red Hat Linux 5.0. 16594 16595 16596 For questions related to the use of GCC, please consult these web 16597 pages and the [1]GCC manuals. If that fails, the 16598 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16599 web pages and the development of GCC are welcome on our developer 16600 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 16601 16602 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 16603 distribution of this entire article is permitted in any medium, 16604 provided this notice is preserved. 16605 16606 These pages are [6]maintained by the GCC team. Last modified 16607 2018-09-30[7]. 16608 16609 References 16610 16611 1. https://gcc.gnu.org/onlinedocs/ 16612 2. mailto:gcc-help (a] gcc.gnu.org 16613 3. mailto:gcc (a] gcc.gnu.org 16614 4. https://gcc.gnu.org/lists.html 16615 5. https://www.fsf.org/ 16616 6. https://gcc.gnu.org/about.html 16617 7. http://validator.w3.org/check/referer 16618 ====================================================================== 16619
Indexes created Sat Jun 27 00:24:51 UTC 2026