NEWS revision 1.12
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-8/index.html 9 GCC 8 Release Series 10 11 Mar 4, 2020 12 13 The [1]GNU project and the GCC developers are pleased to announce the 14 release of GCC 8.4. 15 16 This release is a bug-fix release, containing fixes for regressions in 17 GCC 8.3 relative to previous releases of GCC. 18 19 Release History 20 21 GCC 8.4 22 Mar 4, 2020 ([2]changes, [3]documentation) 23 24 GCC 8.3 25 Feb 22, 2019 ([4]changes, [5]documentation) 26 27 GCC 8.2 28 Jul 26, 2018 ([6]changes, [7]documentation) 29 30 GCC 8.1 31 May 2, 2018 ([8]changes, [9]documentation) 32 33 References and Acknowledgements 34 35 GCC used to stand for the GNU C Compiler, but since the compiler 36 supports several other languages aside from C, it now stands for the 37 GNU Compiler Collection. 38 39 A list of [10]successful builds is updated as new information becomes 40 available. 41 42 The GCC developers would like to thank the numerous people that have 43 contributed new features, improvements, bug fixes, and other changes as 44 well as test results to GCC. This [11]amazing group of volunteers is 45 what makes GCC successful. 46 47 For additional information about GCC please refer to the [12]GCC 48 project web site or contact the [13]GCC development mailing list. 49 50 To obtain GCC please use [14]our mirror sites or [15]our version 51 control system. 52 53 54 For questions related to the use of GCC, please consult these web 55 pages and the [16]GCC manuals. If that fails, the 56 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 57 web pages and the development of GCC are welcome on our developer 58 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 59 archives. 60 61 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 62 distribution of this entire article is permitted in any medium, 63 provided this notice is preserved. 64 65 These pages are [21]maintained by the GCC team. Last modified 66 2020-03-04[22]. 67 68 References 69 70 1. http://www.gnu.org/ 71 2. http://gcc.gnu.org/gcc-8/changes.html 72 3. http://gcc.gnu.org/onlinedocs/8.4.0/ 73 4. http://gcc.gnu.org/gcc-8/changes.html 74 5. http://gcc.gnu.org/onlinedocs/8.3.0/ 75 6. http://gcc.gnu.org/gcc-8/changes.html 76 7. http://gcc.gnu.org/onlinedocs/8.2.0/ 77 8. http://gcc.gnu.org/gcc-8/changes.html 78 9. http://gcc.gnu.org/onlinedocs/8.1.0/ 79 10. http://gcc.gnu.org/gcc-8/buildstat.html 80 11. http://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html 81 12. http://gcc.gnu.org/index.html 82 13. mailto:gcc (a] gcc.gnu.org 83 14. http://gcc.gnu.org/mirrors.html 84 15. http://gcc.gnu.org/git.html 85 16. https://gcc.gnu.org/onlinedocs/ 86 17. mailto:gcc-help (a] gcc.gnu.org 87 18. mailto:gcc (a] gcc.gnu.org 88 19. https://gcc.gnu.org/lists.html 89 20. https://www.fsf.org/ 90 21. https://gcc.gnu.org/about.html 91 22. http://validator.w3.org/check/referer 92 ====================================================================== 93 http://gcc.gnu.org/gcc-8/changes.html 94 GCC 8 Release Series 95 Changes, New Features, and Fixes 96 97 This page is a "brief" summary of some of the huge number of 98 improvements in GCC 8. You may also want to check out our [1]Porting to 99 GCC 8 page and the [2]full GCC documentation. 100 101 Caveats 102 103 * Support for the obsolete SDB/coff debug info format has been 104 removed. The option -gcoff no longer does anything. 105 * The Cilk+ extensions to the C and C++ languages have been removed. 106 * The MPX extensions to the C and C++ languages have been deprecated 107 and will be removed in a future release. 108 * The extension allowing arithmetic on std::atomic<void*> and types 109 like std::atomic<R(*)()> has been deprecated. 110 * The non-standard C++0x std::copy_exception function was removed. 111 std::make_exception_ptr should be used instead. 112 * Support for the powerpc*-*-*spe* target ports which have been 113 recently unmaintained and untested in GCC has been declared 114 obsolete in GCC 8 as [3]announced. Unless there is activity to 115 revive them, the next release of GCC will have their sources 116 permanently removed. 117 118 General Improvements 119 120 * Inter-procedural optimization improvements: 121 + Reworked run-time estimation metrics leading to more realistic 122 guesses driving inliner and cloning heuristics. 123 + The ipa-pure-const pass is extended to propagate the malloc 124 attribute, and the corresponding warning option 125 -Wsuggest-attribute=malloc emits a diagnostic for functions 126 which can be annotated with the malloc attribute. 127 * Profile driven optimization improvements: 128 + New infrastructure for representing profiles (both statically 129 guessed and profile feedback) which allows propagation of 130 additional information about the reliability of the profile. 131 + A number of improvements in the profile updating code solving 132 problems found by new verification code. 133 + Static detection of code which is not executed in a valid run 134 of the program. This includes paths which trigger undefined 135 behavior as well as calls to functions declared with the cold 136 attribute. Newly the noreturn attribute does not imply all 137 effects of cold to differentiate between exit (which is 138 noreturn) and abort (which is in addition not executed in 139 valid runs). 140 + -freorder-blocks-and-partition, a pass splitting function 141 bodies into hot and cold regions, is now enabled by default at 142 -O2 and higher for x86 and x86-64. 143 * Link-time optimization improvements: 144 + We have significantly improved debug information on ELF 145 targets using DWARF by properly preserving language-specific 146 information. This allows for example the libstdc++ 147 pretty-printers to work with LTO optimized executables. 148 * A new option -fcf-protection=[full|branch|return|none] is 149 introduced to perform code instrumentation to increase program 150 security by checking that target addresses of control-flow transfer 151 instructions (such as indirect function call, function return, 152 indirect jump) are valid. Currently the instrumentation is 153 supported on x86 GNU/Linux targets only. See the user guide for 154 further information about the option syntax and section "New 155 Targets and Target Specific Improvements" for IA-32/x86-64 for more 156 details. 157 * The -gcolumn-info option is now enabled by default. It includes 158 column information in addition to just filenames and line numbers 159 in DWARF debugging information. 160 * The polyhedral-based loop nest optimization pass 161 -floop-nest-optimize has been overhauled. It's still considered 162 experimental and may not result in any runtime improvements. 163 * Two new classical loop nest optimization passes have been added. 164 -floop-unroll-and-jam performs outer loop unrolling and fusing of 165 the inner loop copies. -floop-interchange exchanges loops in a loop 166 nest to improve data locality. Both passes are enabled by default 167 at -O3 and above. 168 * The classic loop nest optimization pass -ftree-loop-distribution 169 has been improved and enabled by default at -O3 and above. It 170 supports loop nest distribution in some restricted scenarios; it 171 also supports cancellable innermost loop distribution with loop 172 versioning under run-time alias checks. 173 * The new option -fstack-clash-protection causes the compiler to 174 insert probes whenever stack space is allocated statically or 175 dynamically to reliably detect stack overflows and thus mitigate 176 the attack vector that relies on jumping over a stack guard page as 177 provided by the operating system. 178 * A new pragma GCC unroll has been implemented in the C family of 179 languages, as well as Fortran and Ada, so as to make it possible 180 for the user to have a finer-grained control over the loop 181 unrolling optimization. 182 * GCC has been enhanced to detect more instances of meaningless or 183 mutually exclusive attribute specifications and handle such 184 conflicts more consistently. Mutually exclusive attribute 185 specifications are ignored with a warning regardless of whether 186 they appear on the same declaration or on distinct declarations of 187 the same entity. For example, because the noreturn attribute on the 188 second declaration below is mutually exclusive with the malloc 189 attribute on the first, it is ignored and a warning is issued. 190 > 191 void* __attribute__ ((malloc)) f (unsigned); 192 void* __attribute__ ((noreturn)) f (unsigned); 193 194 warning: ignoring attribute 'noreturn' because it conflicts with attribute 195 'malloc' [-Wattributes] 196 * The gcov tool can distinguish functions that begin on a same line 197 in a source file. This can be a different template instantiation or 198 a class constructor: 199 200 File 'ins.C' 201 Lines executed:100.00% of 8 202 Creating 'ins.C.gcov' 203 204 -: 0:Source:ins.C 205 -: 0:Graph:ins.gcno 206 -: 0:Data:ins.gcda 207 -: 0:Runs:1 208 -: 0:Programs:1 209 -: 1:template<class T> 210 -: 2:class Foo 211 -: 3:{ 212 -: 4: public: 213 2: 5: Foo(): b (1000) {} 214 ------------------ 215 Foo<char>::Foo(): 216 1: 5: Foo(): b (1000) {} 217 ------------------ 218 Foo<int>::Foo(): 219 1: 5: Foo(): b (1000) {} 220 ------------------ 221 2: 6: void inc () { b++; } 222 ------------------ 223 Foo<char>::inc(): 224 1: 6: void inc () { b++; } 225 ------------------ 226 Foo<int>::inc(): 227 1: 6: void inc () { b++; } 228 ------------------ 229 -: 7: 230 -: 8: private: 231 -: 9: int b; 232 -: 10:}; 233 -: 11: 234 1: 12:int main(int argc, char **argv) 235 -: 13:{ 236 1: 14: Foo<int> a; 237 1: 15: Foo<char> b; 238 -: 16: 239 1: 17: a.inc (); 240 1: 18: b.inc (); 241 1: 19:} 242 243 * The gcov tool has more accurate numbers for execution of lines in a 244 source file. 245 * The gcov tool can use TERM colors to provide more readable output. 246 * AddressSanitizer gained a new pair of sanitization options, 247 -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which 248 warn about subtraction (or comparison) of pointers that point to a 249 different memory object: 250 251 int 252 main () 253 { 254 /* Heap allocated memory. */ 255 char *heap1 = (char *)__builtin_malloc (42); 256 char *heap2 = (char *)__builtin_malloc (42); 257 if (heap1 > heap2) 258 return 1; 259 260 return 0; 261 } 262 263 ==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000 264 00050 265 #0 0x40070f in main /tmp/pointer-compare.c:7 266 #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 267 #2 0x400629 in _start (/tmp/a.out+0x400629) 268 269 0x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604 270 00000003a) 271 allocated by thread T0 here: 272 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 273 _malloc_linux.cc:86 274 #1 0x4006ea in main /tmp/pointer-compare.c:5 275 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 276 277 0x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604 278 00000007a) 279 allocated by thread T0 here: 280 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 281 _malloc_linux.cc:86 282 #1 0x4006f8 in main /tmp/pointer-compare.c:6 283 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 284 285 SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main 286 287 * The store merging pass has been enhanced to handle bit-fields and 288 not just constant stores, but also data copying from adjacent 289 memory locations into other adjacent memory locations, including 290 bitwise logical operations on the data. The pass can also handle 291 byte swapping into memory locations. 292 * The undefined behavior sanitizer gained two new options included in 293 -fsanitize=undefined: -fsanitize=builtin which diagnoses at run 294 time invalid arguments to __builtin_clz or __builtin_ctz prefixed 295 builtins, and -fsanitize=pointer-overflow which performs cheap run 296 time tests for pointer wrapping. 297 * A new attribute no_sanitize can be applied to functions to instruct 298 the compiler not to do sanitization of the options provided as 299 arguments to the attribute. Acceptable values for no_sanitize match 300 those acceptable by the -fsanitize command-line option. 301 302 void __attribute__ ((no_sanitize ("alignment", "object-size"))) 303 f () { /* Do something. */; } 304 305 New Languages and Language specific improvements 306 307 Ada 308 309 * For its internal exception handling used on the host for error 310 recovery in the front-end, the compiler now relies on the native 311 exception handling mechanism of the host platform, which should be 312 more efficient than the former mechanism. 313 314 BRIG (HSAIL) 315 316 In this release cycle, the focus for the BRIGFE was on stabilization 317 and performance improvements. Also a couple of completely new features 318 were added. 319 * Improved support for function and module scope group segment 320 variables. PRM specs define function and module scope group segment 321 variables as an experimental feature. However, PRM test suite uses 322 them. Now group segment is handled by separate book keeping of 323 module scope and function (kernel) offsets. Each function has a 324 "frame" in the group segment offset to which is given as an 325 argument, similar to traditional call stack frame handling. 326 * Reduce the number of type conversions due to the untyped HSAIL 327 registers. Instead of always representing the HSAIL's untyped 328 registers as unsigned int, the gccbrig now pre-analyzes the BRIG 329 code and builds the register variables as a type used the most when 330 storing or reading data to/from each register. This reduces the 331 number of total casts which cannot be always optimized away. 332 * Support for BRIG_KIND_NONE directives. 333 * Made -O3 the default optimization level for BRIGFE. 334 * Fixed illegal addresses generated from address expressions which 335 refer only to offset 0. 336 * Fixed a bug with reg+offset addressing on 32b segments. In 'large' 337 mode, the offset is treated as 32bits unless it's in global, 338 read-only or kernarg address space. 339 * Fixed a crash caused sometimes by calls with more than 4 arguments. 340 * Fixed a mis-execution issue with kernels that have both unexpanded 341 ID functions and calls to subfunctions. 342 * Treat HSAIL barrier builtins as setjmp/longjump style functions to 343 avoid illegal optimizations. 344 * Ensure per WI copies of private variables are aligned correctly. 345 * libhsail-rt: Assume the host runtime allocates the work group 346 memory. 347 348 C family 349 350 * New command-line options have been added for the C and C++ 351 compilers: 352 + [4]-Wmultistatement-macros warns about unsafe macros expanding 353 to multiple statements used as a body of a statement such as 354 if, else, while, switch, or for. 355 + [5]-Wstringop-truncation warns for calls to bounded string 356 manipulation functions such as strncat, strncpy, and stpncpy 357 that might either truncate the copied string or leave the 358 destination unchanged. For example, the following call to 359 strncat is diagnosed because it appends just three of the four 360 characters from the source string. 361 void append (char *buf, size_t bufsize) 362 { 363 strncat (buf, ".txt", 3); 364 } 365 warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [- 366 Wstringop-truncation] 367 Similarly, in the following example, the call to strncpy 368 specifies the size of the destination buffer as the bound. If 369 the length of the source string is equal to or greater than 370 this size the result of the copy will not be NUL-terminated. 371 Therefore, the call is also diagnosed. To avoid the warning, 372 specify sizeof buf - 1 as the bound and set the last element 373 of the buffer to NUL. 374 void copy (const char *s) 375 { 376 char buf[80]; 377 strncpy (buf, s, sizeof buf); 378 379 } 380 warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca 381 tion] 382 The -Wstringop-truncation option is included in -Wall. 383 Note that due to GCC bug [6]82944, defining strncat, strncpy, 384 or stpncpy as a macro in a system header as some 385 implementations do, suppresses the warning. 386 + [7]-Wif-not-aligned controls warnings issued in response to 387 invalid uses of objects declared with attribute 388 [8]warn_if_not_aligned. 389 The -Wif-not-aligned option is included in -Wall. 390 + [9]-Wmissing-attributes warns when a declaration of a function 391 is missing one or more attributes that a related function is 392 declared with and whose absence may adversely affect the 393 correctness or efficiency of generated code. For example, in 394 C++, the warning is issued when an explicit specialization of 395 a primary template declared with attribute alloc_align, 396 alloc_size, assume_aligned, format, format_arg, malloc, or 397 nonnull is declared without it. Attributes deprecated, error, 398 and warning suppress the warning. 399 The -Wmissing-attributes option is included in -Wall. 400 + [10]-Wpacked-not-aligned warns when a struct or union declared 401 with attribute packed defines a member with an explicitly 402 specified alignment greater than 1. Such a member will wind up 403 under-aligned. For example, a warning will be issued for the 404 definition of struct A in the following: 405 struct __attribute__ ((aligned (8))) 406 S8 { char a[8]; }; 407 408 struct __attribute__ ((packed)) A 409 { 410 struct S8 s8; 411 }; 412 warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned] 413 The -Wpacked-not-aligned option is included in -Wall. 414 + -Wcast-function-type warns when a function pointer is cast to 415 an incompatible function pointer. This warning is enabled by 416 -Wextra. 417 + -Wsizeof-pointer-div warns for suspicious divisions of the 418 size of a pointer by the size of the elements it points to, 419 which looks like the usual way to compute the array size but 420 won't work out correctly with pointers. This warning is 421 enabled by -Wall. 422 + -Wcast-align=strict warns whenever a pointer is cast such that 423 the required alignment of the target is increased. For 424 example, warn if a char * is cast to an int * regardless of 425 the target machine. 426 + -fprofile-abs-path creates absolute path names in the .gcno 427 files. This allows gcov to find the correct sources in 428 projects where compilations occur with different working 429 directories. 430 * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and 431 signed integer overflow is now undefined by default at all 432 optimization levels. Using -fsanitize=signed-integer-overflow is 433 now the preferred way to audit code, -Wstrict-overflow is 434 deprecated. 435 * The [11]-Warray-bounds option has been improved to detect more 436 instances of out-of-bounds array indices and pointer offsets. For 437 example, negative or excessive indices into flexible array members 438 and string literals are detected. 439 * The [12]-Wrestrict option introduced in GCC 7 has been enhanced to 440 detect many more instances of overlapping accesses to objects via 441 restrict-qualified arguments to standard memory and string 442 manipulation functions such as memcpy and strcpy. For example, the 443 strcpy call in the function below attempts to truncate the string 444 by replacing its initial characters with the last four. However, 445 because the function writes the terminating NUL into a[4], the 446 copies overlap and the call is diagnosed. 447 void f (void) 448 { 449 char a[] = "abcd1234"; 450 strcpy (a, a + 4); 451 452 } 453 warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset 454 4 [-Wrestrict] 455 The -Wrestrict option is included in -Wall. 456 * Several optimizer enhancements have enabled improvements to the 457 [13]-Wformat-overflow and [14]-Wformat-truncation options. The 458 warnings detect more instances of buffer overflow and truncation 459 than in GCC 7 and are better at avoiding certain kinds of false 460 positives. 461 * When reporting mismatching argument types at a function call, the C 462 and C++ compilers now underline both the argument and the pertinent 463 parameter in the declaration. 464 $ gcc arg-type-mismatch.cc 465 arg-type-mismatch.cc: In function 'int caller(int, int, float)': 466 arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*' 467 [-fpermissive] 468 return callee(first, second, third); 469 ^~~~~~ 470 arg-type-mismatch.cc:1:40: note: initializing argument 2 of 'int callee(int, c 471 onst char*, float)' 472 extern int callee(int one, const char *two, float three); 473 ~~~~~~~~~~~~^~~ 474 475 * When reporting on unrecognized identifiers, the C and C++ compilers 476 will now emit fix-it hints suggesting #include directives for 477 various headers in the C and C++ standard libraries. 478 $ gcc incomplete.c 479 incomplete.c: In function 'test': 480 incomplete.c:3:10: error: 'NULL' undeclared (first use in this function) 481 return NULL; 482 ^~~~ 483 incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge 484 t to '#include <stddef.h>'? 485 incomplete.c:1:1: 486 +#include <stddef.h> 487 const char *test(void) 488 incomplete.c:3:10: 489 return NULL; 490 ^~~~ 491 incomplete.c:3:10: note: each undeclared identifier is reported only once for ea 492 ch function it appears in 493 494 $ gcc incomplete.cc 495 incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type 496 std::string s("hello world"); 497 ^~~~~~ 498 incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you 499 forget to '#include <string>'? 500 +#include <string> 501 std::string s("hello world"); 502 ^~~ 503 504 * The C and C++ compilers now use more intuitive locations when 505 reporting on missing semicolons, and offer fix-it hints: 506 $ gcc t.c 507 t.c: In function 'test': 508 t.c:3:12: error: expected ';' before '}' token 509 return 42 510 ^ 511 ; 512 } 513 ~ 514 515 * When reporting on missing '}' and ')' tokens, the C and C++ 516 compilers will now highlight the corresponding '{' and '(' token, 517 issuing a 'note' if it's on a separate line: 518 $ gcc unclosed.c 519 unclosed.c: In function 'log_when_out_of_range': 520 unclosed.c:12:50: error: expected ')' before '{' token 521 && (temperature < MIN || temperature > MAX) { 522 ^~ 523 ) 524 unclosed.c:11:6: note: to match this '(' 525 if (logging_enabled && check_range () 526 ^ 527 528 or highlighting it directly if it's on the same line: 529 $ gcc unclosed-2.c 530 unclosed-2.c: In function 'test': 531 unclosed-2.c:8:45: error: expected ')' before '{' token 532 if (temperature < MIN || temperature > MAX { 533 ~ ^~ 534 ) 535 536 They will also emit fix-it hints. 537 538 C++ 539 540 * GCC 8 (-fabi-version=12) has a couple of corrections to the calling 541 convention, which changes the ABI for some uncommon code: 542 + Passing an empty class as an argument now takes up no space on 543 x86_64, as required by the psABI. 544 + Passing or returning a class with only deleted copy and move 545 constructors now uses the same calling convention as a class 546 with a non-trivial copy or move constructor. This only affects 547 C++17 mode, as in earlier standards passing or returning such 548 a class was impossible. 549 + WARNING: In GCC 8.1 the second change mistakenly also affects 550 classes with a deleted copy constructor and defaulted trivial 551 move constructor (bug [15]c++/86094). This issue is fixed in 552 GCC 8.2 (-fabi-version=13). 553 You can test whether these changes affect your code with -Wabi=11 554 (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are 555 problematic for your project, the GCC 7 ABI can be selected with 556 -fabi-version=11. 557 * The value of the C++11 alignof operator has been corrected to match 558 C _Alignof (minimum alignment) rather than GNU __alignof__ 559 (preferred alignment); on ia32 targets this means that 560 alignof(double) is now 4 rather than 8. Code that wants the 561 preferred alignment should use __alignof__ instead. 562 * New command-line options have been added for the C++ compiler to 563 control warnings: 564 + [16]-Wclass-memaccess warns when objects of non-trivial class 565 types are manipulated in potentially unsafe ways by raw memory 566 functions such as memcpy, or realloc. The warning helps detect 567 calls that bypass user-defined constructors or copy-assignment 568 operators, corrupt virtual table pointers, data members of 569 const-qualified types or references, or member pointers. The 570 warning also detects calls that would bypass access controls 571 to data members. For example, a call such as: 572 memcpy (&std::cout, &std::cerr, sizeof std::cout); 573 results in 574 warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t 575 o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n 576 o trivial copy-assignment [-Wclass-memaccess] 577 The -Wclass-memaccess option is included in -Wall. 578 * The C++ front end has experimental support for some of the upcoming 579 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 580 including designated initializers, default member initializers for 581 bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported), 582 lambda [=, this] captures, etc. For a full list of new features, 583 see [17]the C++ status page. 584 * When reporting on attempts to access private fields of a class or 585 struct, the C++ compiler will now offer fix-it hints showing how to 586 use an accessor function to get at the field in question, if one 587 exists. 588 $ gcc accessor.cc 589 accessor.cc: In function 'void test(foo*)': 590 accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context 591 if (ptr->m_ratio >= 0.5) 592 ^~~~~~~ 593 accessor.cc:7:10: note: declared private here 594 double m_ratio; 595 ^~~~~~~ 596 accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double 597 foo::get_ratio() const' 598 if (ptr->m_ratio >= 0.5) 599 ^~~~~~~ 600 get_ratio() 601 602 * The C++ compiler can now give you a hint if you use a macro before 603 it was defined (e.g. if you mess up the order of your #include 604 directives): 605 $ gcc ordering.cc 606 ordering.cc:2:24: error: expected ';' at end of member declaration 607 virtual void clone() const OVERRIDE { } 608 ^~~~~ 609 ; 610 ordering.cc:2:30: error: 'OVERRIDE' does not name a type 611 virtual void clone() const OVERRIDE { } 612 ^~~~~~~~ 613 ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined 614 In file included from ordering.cc:5: 615 c++11-compat.h:2: note: it was later defined here 616 #define OVERRIDE override 617 618 619 * The -Wold-style-cast diagnostic can now emit fix-it hints telling 620 you when you can use a static_cast, const_cast, or 621 reinterpret_cast. 622 $ gcc -c old-style-cast-fixits.cc -Wold-style-cast 623 old-style-cast-fixits.cc: In function 'void test(void*)': 624 old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [ 625 -Wold-style-cast] 626 foo *f = (foo *)ptr; 627 ^~~ 628 ---------- 629 static_cast<foo *> (ptr) 630 631 * When reporting on problems within extern "C" linkage 632 specifications, the C++ compiler will now display the location of 633 the start of the extern "C". 634 $ gcc -c extern-c.cc 635 extern-c.cc:3:1: error: template with C linkage 636 template <typename T> void test (void); 637 ^~~~~~~~ 638 In file included from extern-c.cc:1: 639 unclosed.h:1:1: note: 'extern "C"' linkage started here 640 extern "C" { 641 ^~~~~~~~~~ 642 extern-c.cc:3:39: error: expected '}' at end of input 643 template <typename T> void test (void); 644 ^ 645 In file included from extern-c.cc:1: 646 unclosed.h:1:12: note: to match this '{' 647 extern "C" { 648 ^ 649 650 * When reporting on mismatching template types, the C++ compiler will 651 now use color to highlight the mismatching parts of the template, 652 and will elide the parameters that are common between two 653 mismatching templates, printing [...] instead: 654 $ gcc templates.cc 655 templates.cc: In function 'void test()': 656 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 657 e>' to 'vector<int>' 658 fn_1(vector<double> ()); 659 ^~~~~~~~~~~~~~~~~ 660 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...] 661 ,double>' to 'map<[...],int>' 662 fn_2(map<int, double>()); 663 ^~~~~~~~~~~~~~~~~~ 664 665 Those [...] elided parameters can be seen using -fno-elide-type: 666 $ gcc templates.cc -fno-elide-type 667 templates.cc: In function 'void test()': 668 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 669 e>' to 'vector<int>' 670 fn_1(vector<double> ()); 671 ^~~~~~~~~~~~~~~~~ 672 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d 673 ouble>' to 'map<int,int>' 674 fn_2(map<int, double>()); 675 ^~~~~~~~~~~~~~~~~~ 676 677 The C++ compiler has also gained an option 678 -fdiagnostics-show-template-tree which visualizes such mismatching 679 templates in a hierarchical form: 680 $ gcc templates-2.cc -fdiagnostics-show-template-tree 681 templates-2.cc: In function 'void test()': 682 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 683 ble>' to 'vector<int>' 684 vector< 685 [double != int]> 686 fn_1(vector<double> ()); 687 ^~~~~~~~~~~~~~~~~ 688 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 689 ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m 690 ap<[...],vector<float>>,vector<float>>' 691 map< 692 map< 693 [...], 694 vector< 695 [double != float]>>, 696 vector< 697 [double != float]>> 698 fn_2(map<map<int, vector<double>>, vector<double>> ()); 699 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 700 701 which again works with -fno-elide-type: 702 $ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type 703 templates-2.cc: In function 'void test()': 704 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 705 ble>' to 'vector<int>' 706 vector< 707 [double != int]> 708 fn_1(vector<double> ()); 709 ^~~~~~~~~~~~~~~~~ 710 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 711 ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map 712 <int,vector<float>>,vector<float>>' 713 map< 714 map< 715 int, 716 vector< 717 [double != float]>>, 718 vector< 719 [double != float]>> 720 fn_2(map<map<int, vector<double>>, vector<double>> ()); 721 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 722 723 * Flowing off the end of a non-void function is considered 724 unreachable and may be subject to optimization on that basis. As a 725 result of this change, -Wreturn-type warnings are enabled by 726 default for C++. 727 728 Runtime Library (libstdc++) 729 730 * Improved experimental support for C++17, including the following 731 features: 732 + Deduction guides to support class template argument deduction. 733 + std::filesystem implementation. 734 + std::char_traits<char> and std::char_traits<wchar_t> are 735 usable in constant expressions. 736 + std::to_chars and std::from_chars (for integers only, not for 737 floating point types). 738 * Experimental support for C++2a: std::to_address (thanks to Glen 739 Fernandes) and std::endian. 740 * On GNU/Linux, std::random_device::entropy() accesses the kernel's 741 entropy count for the random device, if known (thanks to Xi 742 Ruoyao). 743 * Support for std::experimental::source_location. 744 * AddressSanitizer integration for std::vector, detecting 745 out-of-range accesses to the unused capacity of a vector. 746 * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the 747 Mathematical Special Functions. 748 749 Fortran 750 751 * The main version of libfortran has been changed to 5. 752 * Parameterized derived types, a major feature of Fortran 2003, have 753 been implemented. 754 * Partial support is provided for Fortran 2018 teams, which are 755 hierarchical subsets of images that execute independently of other 756 image subsets. 757 * The maximum rank for arrays has been increased to 15, conforming to 758 the Fortran 2008 standard. 759 * Transformational intrinsics are now fully supported in 760 initialization expressions. 761 * New flag -fc-prototypes to write C prototypes for BIND(C) 762 procedures and variables. 763 * If -fmax-stack-var-size is honored if given together with -Ofast, 764 -fstack-arrays is no longer set in that case. 765 * New options -fdefault-real-16 and -fdefault-real-10 to control the 766 default kind of REAL variables. 767 * A warning is now issued if an array subscript inside a DO loop 768 could lead to an out-of-bounds-access. The new option 769 -Wdo-subscript, enabled by -Wextra, warns about this even if the 770 compiler can not prove that the code will be executed. 771 * The Fortran front end now attempts to interchange loops if it is 772 deemed profitable. So far, this is restricted to FORALL and DO 773 CONCURRENT statements with multiple indices. This behavior be 774 controlled with the new flag -ffrontend-loop-interchange, which is 775 enabled with optimization by default. The 776 -Wfrontend-loop-interchange option warns about such occurrences. 777 * When an actual argument contains too few elements for a dummy 778 argument, an error is now issued. The -std=legacy option can be 779 used to still compile such code. 780 * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit 781 integers, making records larger than 2GiB possible. 782 * The GFORTRAN_DEFAULT_RECL environment variable no longer has any 783 effect. The record length for preconnected units is now larger than 784 any practical limit, same as for sequential access units opened 785 without an explicit RECL= specifier. 786 * Character variables longer than HUGE(0) elements are now possible 787 on 64-bit targets. Note that this changes the procedure call ABI 788 for all procedures with character arguments on 64-bit targets, as 789 the type of the hidden character length argument has changed. The 790 hidden character length argument is now of type INTEGER(C_SIZE_T). 791 * Partial support is provided for Fortran 2018 teams, which are 792 hierarchical subsets of images that execute independently of other 793 image subsets. 794 795 Go 796 797 * GCC 8 provides a complete implementation of the Go 1.10.1 user 798 packages. 799 * The garbage collector is now fully concurrent. As before, values 800 stored on the stack are scanned conservatively, but value stored in 801 the heap are scanned precisely. 802 * Escape analysis is fully implemented and enabled by default in the 803 Go frontend. This significantly reduces the number of heap 804 allocations by allocating values on the stack instead. 805 806 libgccjit 807 808 The libgccjit API gained four new entry points: 809 * [18]gcc_jit_type_get_vector and 810 * [19]gcc_jit_context_new_rvalue_from_vector for working with 811 vectors, 812 * [20]gcc_jit_type_get_aligned 813 * [21]gcc_jit_function_get_address 814 815 The C code generated by [22]gcc_jit_context_dump_reproducer_to_file is 816 now easier-to-read. 817 818 New Targets and Target Specific Improvements 819 820 AArch64 821 822 * The Armv8.4-A architecture is now supported. It can be used by 823 specifying the -march=armv8.4-a option. 824 * The Dot Product instructions are now supported as an optional 825 extension to the Armv8.2-A architecture and newer and are mandatory 826 on Armv8.4-A. The extension can be used by specifying the +dotprod 827 architecture extension. E.g. -march=armv8.2-a+dotprod. 828 * The Armv8-A +crypto extension has now been split into two 829 extensions for finer grained control: 830 + +aes which contains the Armv8-A AES crytographic instructions. 831 + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic 832 instructions. 833 Using +crypto will now enable these two extensions. 834 * New Armv8.4-A FP16 Floating Point Multiplication Variant 835 instructions have been added. These instructions are mandatory in 836 Armv8.4-A but available as an optional extension to Armv8.2-A and 837 Armv8.3-A. The new extension can be used by specifying the +fp16fml 838 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 839 the instructions can be enabled by specifying +fp16. 840 * New cryptographic instructions have been added as optional 841 extensions to Armv8.2-A and newer. These instructions can be 842 enabled with: 843 + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This 844 implies +sha2. 845 + +sm4 New SM3 and SM4 instructions from Armv8.4-A. 846 * The Scalable Vector Extension (SVE) is now supported as an optional 847 extension to the Armv8.2-A architecture and newer. This support 848 includes automatic vectorization with SVE instructions, but it does 849 not yet include the SVE Arm C Language Extensions (ACLE). It can be 850 enabled by specifying the +sve architecture extension (for example, 851 -march=armv8.2-a+sve). By default, the generated code works with 852 all vector lengths, but it can be made specific to N-bit vectors 853 using -msve-vector-bits=N. 854 * Support has been added for the following processors (GCC 855 identifiers in parentheses): 856 + Arm Cortex-A75 (cortex-a75). 857 + Arm Cortex-A55 (cortex-a55). 858 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 859 (cortex-a75.cortex-a55). 860 The GCC identifiers can be used as arguments to the -mcpu or -mtune 861 options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as 862 arguments to the equivalent target attributes and pragmas. 863 864 ARC 865 866 * Added support for: 867 + Fast interrupts. 868 + Naked functions. 869 + aux variable attributes. 870 + uncached type qualifier. 871 + Secure functions via sjli instruction. 872 * New exception handling implementation. 873 * Revamped trampoline implementation. 874 * Refactored small data feature implementation, controlled via -G 875 command line option. 876 * New support for reduced register set ARC architecture 877 configurations, controlled via -mrf16 command line option. 878 * Refurbished and improved support for zero overhead loops. 879 Introduced -mlpc-width command line option to control the width of 880 lp_count register. 881 882 ARM 883 884 * The -mfpu option now takes a new option setting of -mfpu=auto. When 885 set to this the floating-point and SIMD settings are derived from 886 the settings of the -mcpu or -march options. The internal CPU 887 configurations have been updated with information about the 888 permitted floating-point configurations supported. See the user 889 guide for further information about the extended option syntax for 890 controlling architectural extensions via the -march option. 891 -mfpu=auto is now the default setting unless the compiler has been 892 configured with an explicit --with-fpu option. 893 * The -march and -mcpu options now accept optional extensions to the 894 architecture or CPU option, allowing the user to enable or disable 895 any such extensions supported by that architecture or CPU such as 896 (but not limited to) floating-point and AdvancedSIMD. For example: 897 the option -mcpu=cortex-a53+nofp will generate code for the 898 Cortex-A53 processor with no floating-point support. This, in 899 combination with the new -mfpu=auto option, provides a 900 straightforward way of specifying a valid build target through a 901 single -mcpu or -march option. The -mtune option accepts the same 902 arguments as -mcpu but only the CPU name has an effect on tuning. 903 The architecture extensions do not have any effect. For details of 904 what extensions a particular architecture or CPU option supports 905 please refer to the [23]documentation. 906 * The -mstructure-size-boundary option has been deprecated and will 907 be removed in a future release. 908 * The default link behavior for Armv6 and Armv7-R targets has been 909 changed to produce BE8 format when generating big-endian images. A 910 new flag -mbe32 can be used to force the linker to produce legacy 911 BE32 format images. There is no change of behavior for Armv6-M and 912 other Armv7 or later targets: these already defaulted to BE8 913 format. This change brings GCC into alignment with other compilers 914 for the ARM architecture. 915 * The Armv8-R architecture is now supported. It can be used by 916 specifying the -march=armv8-r option. 917 * The Armv8.3-A architecture is now supported. It can be used by 918 specifying the -march=armv8.3-a option. 919 * The Armv8.4-A architecture is now supported. It can be used by 920 specifying the -march=armv8.4-a option. 921 * The Dot Product instructions are now supported as an optional 922 extension to the Armv8.2-A architecture and newer and are mandatory 923 on Armv8.4-A. The extension can be used by specifying the +dotprod 924 architecture extension. E.g. -march=armv8.2-a+dotprod. 925 * Support for setting extensions and architectures using the GCC 926 target pragma and attribute has been added. It can be used by 927 specifying #pragma GCC target ("arch=..."), #pragma GCC target 928 ("+extension"), __attribute__((target("arch=..."))) or 929 __attribute__((target("+extension"))). 930 * New Armv8.4-A FP16 Floating Point Multiplication Variant 931 instructions have been added. These instructions are mandatory in 932 Armv8.4-A but available as an optional extension to Armv8.2-A and 933 Armv8.3-A. The new extension can be used by specifying the +fp16fml 934 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 935 the instructions can be enabled by specifying +fp16. 936 * Support has been added for the following processors (GCC 937 identifiers in parentheses): 938 + Arm Cortex-A75 (cortex-a75). 939 + Arm Cortex-A55 (cortex-a55). 940 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 941 (cortex-a75.cortex-a55). 942 + Arm Cortex-R52 for Armv8-R (cortex-r52). 943 The GCC identifiers can be used as arguments to the -mcpu or -mtune 944 options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as 945 arguments to the equivalent target attributes and pragmas. 946 947 AVR 948 949 * The AVR port now supports the following XMEGA-like devices: 950 951 ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417, 952 ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617, 953 ATtiny3214, ATtiny3216, ATtiny3217 954 The new devices are listed under [24]-mmcu=avrxmega3. 955 + These devices see flash memory in the RAM address space, so 956 that features like PROGMEM and __flash are not needed any more 957 (as opposed to other AVR families for which read-only data 958 will be located in RAM except special, non-standard features 959 are used to locate and access such data). This requires that 960 the compiler is used with Binutils 2.29 or newer so that 961 [25]read-only data will be located in flash memory. 962 + A new command-line option -mshort-calls is supported. This 963 option is used internally for multilib selection of the 964 avrxmega3 variants. It is not an optimization option. Do not 965 set it by hand. 966 * The compiler now generates [26]efficient interrupt service routine 967 (ISR) prologues and epilogues. This is achieved by using the new 968 [27]AVR pseudo instruction __gcc_isr which is supported and 969 resolved by the GNU assembler. 970 + As the __gcc_isr pseudo-instruction will be resolved by the 971 assembler, inline assembly is transparent to the process. This 972 means that when inline assembly uses an instruction like INC 973 that clobbers the condition code, then the assembler will 974 detect this and generate an appropriate ISR prologue / 975 epilogue chunk to save / restore SREG as needed. 976 + A new command-line option -mno-gas-isr-prologues disables the 977 generation of the __gcc_isr pseudo instruction. Any non-naked 978 ISR will save and restore SREG, tmp_reg and zero_reg, no 979 matter whether the respective register is clobbered or used. 980 + The feature is turned on per default for all optimization 981 levels except for -O0 and -Og. It is explicitly enabled by 982 means of option -mgas-isr-prologues. 983 + Support has been added for a new [28]AVR function attribute 984 no_gccisr. It can be used to disable __gcc_isr pseudo 985 instruction generation for individual ISRs. 986 + This optimization is only available if GCC is configured with 987 GNU Binutils 2.29 or newer; or at least with a version of 988 Binutils that implements feature [29]PR21683. 989 * The compiler no more saves / restores registers in main; the effect 990 is the same as if attribute OS_task was specified for main. This 991 optimization can be switched off by the new command-line option 992 -mno-main-is-OS_task. 993 994 IA-32/x86-64 995 996 * The x86 port now supports the naked function attribute. 997 * Better tuning for znver1 and Intel Core based CPUs. 998 * Vectorization cost metrics has been reworked leading to significant 999 improvements on some benchmarks. 1000 * GCC now supports the Intel CPU named Cannonlake through 1001 -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA 1002 and SHA ISA extensions. 1003 * GCC now supports the Intel CPU named Icelake through 1004 -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES, 1005 AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ 1006 ISA extensions. 1007 * GCC now supports the Intel Control-flow Enforcement Technology 1008 (CET) extension through -fcf-protection option. 1009 1010 NDS32 1011 1012 * New command-line options -mext-perf, -mext-perf2, and -mext-string 1013 have been added for performance extension instructions. 1014 1015 Nios II 1016 1017 * The Nios II back end has been improved to generate better-optimized 1018 code. Changes include switching to LRA, more accurate cost models, 1019 and more compact code for addressing static variables. 1020 * New command-line options -mgprel-sec= and -mr0rel-sec= have been 1021 added. 1022 * The stack-smashing protection options are now enabled on Nios II. 1023 1024 PA-RISC 1025 1026 * The default call ABI on 32-bit linux has been changed from callee 1027 copies to caller copies. This affects objects larger than eight 1028 bytes passed by value. The goal is to improve compatibility with 1029 x86 and resolve issues with OpenMP. 1030 * Other PA-RISC targets are unchanged. 1031 1032 PowerPC / PowerPC64 / RS6000 1033 1034 * The PowerPC SPE support is split off to a separate powerpcspe port. 1035 The separate port is deprecated and might be removed in a future 1036 release. 1037 * The Paired Single support (as used on some PPC750 CPUs, -mpaired, 1038 powerpc*-*-linux*paired*) is deprecated and will be removed in a 1039 future release. 1040 * The Xilinx floating point support (-mxilinx-fpu, 1041 powerpc-xilinx-eabi*) is deprecated and will be removed in a future 1042 release. 1043 * Support for using big-endian AltiVec intrinsics on a little-endian 1044 target (-maltivec=be) is deprecated and will be removed in a future 1045 release. 1046 1047 Tile 1048 1049 * The TILE-Gx port is deprecated and will be removed in a future 1050 release. 1051 1052 Operating Systems 1053 1054 Windows 1055 1056 * GCC on Microsoft Windows can now be configured via 1057 --enable-mingw-wildcard or --disable-mingw-wildcard to force a 1058 specific behavior for GCC itself with regards to supporting the 1059 wildcard character. Prior versions of GCC would follow the 1060 configuration of the MinGW runtime. This behavior can still be 1061 obtained by not using the above options or by using 1062 --enable-mingw-wildcard=platform. 1063 1064 Improvements for plugin authors 1065 1066 * Plugins can now register a callback hook for when comments are 1067 encountered by the C and C++ compilers, e.g. allowing for plugins 1068 to handle documentation markup in code comments. 1069 * The gdbinit support script for debugging GCC now has a 1070 break-on-diagnostic command, providing an easy way to trigger a 1071 breakpoint whenever a diagnostic is emitted. 1072 * The API for creating fix-it hints now supports newlines, and for 1073 emitting mutually incompatible fix-it hints for one diagnostic. 1074 1075 GCC 8.1 1076 1077 This is the [30]list of problem reports (PRs) from GCC's bug tracking 1078 system that are known to be fixed in the 8.1 release. This list might 1079 not be complete (that is, it is possible that some PRs that have been 1080 fixed are not listed here). 1081 1082 GCC 8.2 1083 1084 This is the [31]list of problem reports (PRs) from GCC's bug tracking 1085 system that are known to be fixed in the 8.2 release. This list might 1086 not be complete (that is, it is possible that some PRs that have been 1087 fixed are not listed here). 1088 1089 General Improvements 1090 1091 * Fixed LTO link-time performance problems caused by an overflow in 1092 the partitioning algorithm while building large binaries. 1093 1094 Language Specific Changes 1095 1096 C++ 1097 1098 GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or 1099 returning of classes with a deleted copy constructor and defaulted 1100 trivial move constructor (bug [32]c++/86094). GCC 8.2 introduces 1101 -fabi-version=13 and makes it the default, ABI incompatibilities 1102 between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [33]C++ 1103 changes for more details. 1104 1105 Target Specific Changes 1106 1107 IA-32/x86-64 1108 1109 * -mtune=native performance regression [34]PR84413 on Intel Skylake 1110 processors has been fixed. 1111 1112 GCC 8.3 1113 1114 This is the [35]list of problem reports (PRs) from GCC's bug tracking 1115 system that are known to be fixed in the 8.3 release. This list might 1116 not be complete (that is, it is possible that some PRs that have been 1117 fixed are not listed here). 1118 1119 Windows 1120 1121 * A C++ Microsoft ABI bitfield layout bug, [36]PR87137 has been 1122 fixed. A non-field declaration could cause the current bitfield 1123 allocation unit to be completed, incorrectly placing a following 1124 bitfield into a new allocation unit. The Microsoft ABI is selected 1125 for: 1126 + Mingw targets 1127 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 1128 option is specified, or __attribute__((ms_struct)) is used 1129 + SuperH targets when the -mhitachi option is specified, or 1130 __attribute__((renesas)) is used 1131 GCC 8 introduced additional cases of this defect, but rather than 1132 resolve only those regressions, we decided to resolve all the cases 1133 of this defect in single change. 1134 1135 GCC 8.4 1136 1137 This is the [37]list of problem reports (PRs) from GCC's bug tracking 1138 system that are known to be fixed in the 8.4 release. This list might 1139 not be complete (that is, it is possible that some PRs that have been 1140 fixed are not listed here). 1141 1142 1143 For questions related to the use of GCC, please consult these web 1144 pages and the [38]GCC manuals. If that fails, the 1145 [39]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1146 web pages and the development of GCC are welcome on our developer 1147 list at [40]gcc (a] gcc.gnu.org. All of [41]our lists have public 1148 archives. 1149 1150 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and 1151 distribution of this entire article is permitted in any medium, 1152 provided this notice is preserved. 1153 1154 These pages are [43]maintained by the GCC team. Last modified 1155 2020-03-04[44]. 1156 1157 References 1158 1159 1. http://gcc.gnu.org/gcc-8/porting_to.html 1160 2. http://gcc.gnu.org/onlinedocs/index.html#current 1161 3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html 1162 4. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros 1163 5. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 1164 6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944 1165 7. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned 1166 8. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute 1167 9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 1168 10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned 1169 11. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds 1170 12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict 1171 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow 1172 14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation 1173 15. https://gcc.gnu.org/PR86094 1174 16. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess 1175 17. http://gcc.gnu.org/projects/cxx-status.html#cxx2a 1176 18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector 1177 19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector 1178 20. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned 1179 21. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address 1180 22. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file 1181 23. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options 1182 24. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html 1183 25. https://sourceware.org/PR21472 1184 26. https://gcc.gnu.org/PR20296 1185 27. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html 1186 28. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html 1187 29. https://sourceware.org/PR21683 1188 30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0 1189 31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2 1190 32. https://gcc.gnu.org/PR86094 1191 33. http://gcc.gnu.org/gcc-8/changes.html#cxx 1192 34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413 1193 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3 1194 36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 1195 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.4 1196 38. https://gcc.gnu.org/onlinedocs/ 1197 39. mailto:gcc-help (a] gcc.gnu.org 1198 40. mailto:gcc (a] gcc.gnu.org 1199 41. https://gcc.gnu.org/lists.html 1200 42. https://www.fsf.org/ 1201 43. https://gcc.gnu.org/about.html 1202 44. http://validator.w3.org/check/referer 1203 ====================================================================== 1204 http://gcc.gnu.org/gcc-7/index.html 1205 GCC 7 Release Series 1206 1207 Nov 14, 2019 1208 1209 The [1]GNU project and the GCC developers are pleased to announce the 1210 release of GCC 7.5. 1211 1212 This release is a bug-fix release, containing fixes for regressions in 1213 GCC 7.4 relative to previous releases of GCC. 1214 1215 Release History 1216 1217 GCC 7.5 1218 Nov 14, 2019 ([2]changes, [3]documentation) 1219 1220 GCC 7.4 1221 Dec 6, 2018 ([4]changes, [5]documentation) 1222 1223 GCC 7.3 1224 Jan 25, 2018 ([6]changes, [7]documentation) 1225 1226 GCC 7.2 1227 Aug 14, 2017 ([8]changes, [9]documentation) 1228 1229 GCC 7.1 1230 May 2, 2017 ([10]changes, [11]documentation) 1231 1232 References and Acknowledgements 1233 1234 GCC used to stand for the GNU C Compiler, but since the compiler 1235 supports several other languages aside from C, it now stands for the 1236 GNU Compiler Collection. 1237 1238 A list of [12]successful builds is updated as new information becomes 1239 available. 1240 1241 The GCC developers would like to thank the numerous people that have 1242 contributed new features, improvements, bug fixes, and other changes as 1243 well as test results to GCC. This [13]amazing group of volunteers is 1244 what makes GCC successful. 1245 1246 For additional information about GCC please refer to the [14]GCC 1247 project web site or contact the [15]GCC development mailing list. 1248 1249 To obtain GCC please use [16]our mirror sites or [17]our version 1250 control system. 1251 1252 1253 For questions related to the use of GCC, please consult these web 1254 pages and the [18]GCC manuals. If that fails, the 1255 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1256 web pages and the development of GCC are welcome on our developer 1257 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 1258 archives. 1259 1260 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 1261 distribution of this entire article is permitted in any medium, 1262 provided this notice is preserved. 1263 1264 These pages are [23]maintained by the GCC team. Last modified 1265 2020-01-14[24]. 1266 1267 References 1268 1269 1. http://www.gnu.org/ 1270 2. http://gcc.gnu.org/gcc-7/changes.html 1271 3. http://gcc.gnu.org/onlinedocs/7.5.0/ 1272 4. http://gcc.gnu.org/gcc-7/changes.html 1273 5. http://gcc.gnu.org/onlinedocs/7.4.0/ 1274 6. http://gcc.gnu.org/gcc-7/changes.html 1275 7. http://gcc.gnu.org/onlinedocs/7.3.0/ 1276 8. http://gcc.gnu.org/gcc-7/changes.html 1277 9. http://gcc.gnu.org/onlinedocs/7.2.0/ 1278 10. http://gcc.gnu.org/gcc-7/changes.html 1279 11. http://gcc.gnu.org/onlinedocs/7.1.0/ 1280 12. http://gcc.gnu.org/gcc-7/buildstat.html 1281 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 1282 14. http://gcc.gnu.org/index.html 1283 15. mailto:gcc (a] gcc.gnu.org 1284 16. http://gcc.gnu.org/mirrors.html 1285 17. http://gcc.gnu.org/git.html 1286 18. https://gcc.gnu.org/onlinedocs/ 1287 19. mailto:gcc-help (a] gcc.gnu.org 1288 20. mailto:gcc (a] gcc.gnu.org 1289 21. https://gcc.gnu.org/lists.html 1290 22. https://www.fsf.org/ 1291 23. https://gcc.gnu.org/about.html 1292 24. http://validator.w3.org/check/referer 1293 ====================================================================== 1294 http://gcc.gnu.org/gcc-7/changes.html 1295 GCC 7 Release Series 1296 Changes, New Features, and Fixes 1297 1298 This page is a brief summary of some of the huge number of improvements 1299 in GCC 7. For more information, see the [1]Porting to GCC 7 page and 1300 the [2]full GCC documentation. 1301 1302 Caveats 1303 1304 * GCC now uses [3]LRA (a new local register allocator) by default for 1305 new targets. 1306 * The non-standard C++0x type traits has_trivial_default_constructor, 1307 has_trivial_copy_constructor and has_trivial_copy_assign have been 1308 removed. 1309 * The libstdc++ [4]Profile Mode has been deprecated and will be 1310 removed in a future version. 1311 * The Cilk+ extensions to the C and C++ languages have been 1312 deprecated. 1313 * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that 1314 affects conformance to the procedure call standard (AAPCS) has been 1315 fixed. The bug affects some C++ code where class objects are passed 1316 by value to functions and could result in incorrect or inconsistent 1317 code being generated. This is an ABI change. If the option -Wpsabi 1318 is enabled (on by default) the compiler will emit a diagnostic note 1319 for code that might be affected. 1320 1321 General Optimizer Improvements 1322 1323 * GCC 7 can determine the return value or range of return values of 1324 some calls to the sprintf family of functions and make it available 1325 to other optimization passes. Some calls to the snprintf function 1326 with a zero size argument can be folded into constants. This 1327 optimization is included in -O1 and can be selectively controlled 1328 by the -fprintf-return-value option. 1329 * A new store merging pass has been added. It merges constant stores 1330 to adjacent memory locations into fewer, wider, stores. It is 1331 enabled by the -fstore-merging option and at the -O2 optimization 1332 level or higher (and -Os). 1333 * A new code hoisting optimization has been added to the partial 1334 redundancy elimination pass. It attempts to move evaluation of 1335 expressions executed on all paths to the function exit as early as 1336 possible. This primarily helps improve code size, but can improve 1337 the speed of the generated code as well. It is enabled by the 1338 -fcode-hoisting option and at the -O2 optimization level or higher 1339 (and -Os). 1340 * A new interprocedural bitwise constant propagation optimization has 1341 been added, which propagates knowledge about which bits of 1342 variables are known to be zero (including pointer alignment 1343 information) across the call graph. It is enabled by the 1344 -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled 1345 at the -O2 optimization level and higher (and -Os). This 1346 optimization supersedes interprocedural alignment propagation of 1347 GCC 6, and therefore the option -fipa-cp-alignment is now 1348 deprecated and ignored. 1349 * A new interprocedural value range propagation optimization has been 1350 added, which propagates integral range information across the call 1351 graph when variable values can be proven to be within those ranges. 1352 It is enabled by the -fipa-vrp option and at the -O2 optimization 1353 level and higher (and -Os). 1354 * A new loop splitting optimization pass has been added. Certain 1355 loops which contain a condition that is always true on one side of 1356 the iteration space and always false on the other are split into 1357 two loops, such that each of the two new loops iterates on just one 1358 side of the iteration space and the condition does not need to be 1359 checked inside of the loop. It is enabled by the -fsplit-loops 1360 option and at the -O3 optimization level or higher. 1361 * The shrink-wrapping optimization can now separate portions of 1362 prologues and epilogues to improve performance if some of the work 1363 done traditionally by prologues and epilogues is not needed on 1364 certain paths. This is controlled by the -fshrink-wrap-separate 1365 option, enabled by default. It requires target support, which is 1366 currently only implemented in the PowerPC and AArch64 ports. 1367 * AddressSanitizer gained a new sanitization option, 1368 -fsanitize-address-use-after-scope, which enables sanitization of 1369 variables whose address is taken and used after a scope where the 1370 variable is defined: 1371 1372 int 1373 main (int argc, char **argv) 1374 { 1375 char *ptr; 1376 { 1377 char my_char; 1378 ptr = &my_char; 1379 } 1380 1381 *ptr = 123; 1382 return *ptr; 1383 } 1384 1385 ==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99 1386 0 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958 1387 WRITE of size 1 at 0x7fffb8dba990 thread T0 1388 #0 0x4006d4 in main /tmp/use-after-scope-1.c:10 1389 #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290) 1390 #2 0x400739 in _start (/tmp/a.out+0x400739) 1391 1392 Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame 1393 #0 0x40067f in main /tmp/use-after-scope-1.c:3 1394 1395 This frame has 1 object(s): 1396 [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable 1397 1398 The option is enabled by default with -fsanitize=address and 1399 disabled by default with -fsanitize=kernel-address. Compared to the 1400 LLVM compiler, where the option already exists, the implementation 1401 in the GCC compiler has some improvements and advantages: 1402 + Complex uses of gotos and case labels are properly handled and 1403 should not report any false positive or false negatives. 1404 + C++ temporaries are sanitized. 1405 + Sanitization can handle invalid memory stores that are 1406 optimized out by the LLVM compiler when optimization is 1407 enabled. 1408 * The -fsanitize=signed-integer-overflow suboption of the 1409 UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even 1410 on arithmetic operations with generic vectors. 1411 * Version 5 of the [6]DWARF debugging information standard is 1412 supported through the -gdwarf-5 option. The DWARF version 4 1413 debugging information remains the default until consumers of 1414 debugging information are adjusted. 1415 1416 New Languages and Language specific improvements 1417 1418 OpenACC support in C, C++, and Fortran continues to be maintained and 1419 improved. See the [7]OpenACC and [8]Offloading wiki pages for further 1420 information. 1421 1422 Ada 1423 1424 * On mainstream native platforms, Ada programs no longer require the 1425 stack to be made executable in order to run properly. 1426 1427 BRIG (HSAIL) 1428 1429 Support for processing BRIG 1.0 files was added in this release. BRIG 1430 is a binary format for HSAIL (Heterogeneous System Architecture 1431 Intermediate Language). The BRIG front end can be used for implementing 1432 HSAIL "finalizers" (compilation of HSAIL to a native ISA) for 1433 GCC-supported targets. An implementation of an HSAIL runtime library, 1434 libhsail-rt is also included. 1435 1436 C family 1437 1438 * New command-line options have been added for the C and C++ 1439 compilers: 1440 + -Wimplicit-fallthrough warns when a switch case falls through. 1441 This warning has five different levels. The compiler is able 1442 to parse a wide range of fallthrough comments, depending on 1443 the level. It also handles control-flow statements, such as 1444 ifs. It's possible to suppress the warning by either adding a 1445 fallthrough comment, or by using a null statement: 1446 __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]]; 1447 (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning 1448 is enabled by -Wextra. 1449 + -Wpointer-compare warns when a pointer is compared with a zero 1450 character constant. Such code is now invalid in C++11 and GCC 1451 rejects it. This warning is enabled by default. 1452 + -Wduplicated-branches warns when an if-else has identical 1453 branches. 1454 + -Wrestrict warns when an argument passed to a 1455 restrict-qualified parameter aliases with another argument. 1456 + -Wmemset-elt-size warns for memset calls, when the first 1457 argument references an array, and the third argument is a 1458 number equal to the number of elements of the array, but not 1459 the size of the array. This warning is enabled by -Wall. 1460 + -Wint-in-bool-context warns about suspicious uses of integer 1461 values where boolean values are expected. This warning is 1462 enabled by -Wall. 1463 + -Wswitch-unreachable warns when a switch statement has 1464 statements between the controlling expression and the first 1465 case label which will never be executed. This warning is 1466 enabled by default. 1467 + -Wexpansion-to-defined warns when defined is used outside #if. 1468 This warning is enabled by -Wextra or -Wpedantic. 1469 + -Wregister warns about uses of the register storage specifier. 1470 In C++17 this keyword has been removed and for C++17 this is a 1471 pedantic warning enabled by default. The warning is not 1472 emitted for the GNU Explicit Register Variables extension. 1473 + -Wvla-larger-than=N warns about unbounded uses of 1474 variable-length arrays, and about bounded uses of 1475 variable-length arrays whose bound can be larger than N bytes. 1476 + -Wduplicate-decl-specifier warns when a declaration has 1477 duplicate const, volatile, restrict or _Atomic specifier. This 1478 warning is enabled by -Wall. 1479 * GCC 6's C and C++ front ends were able to offer suggestions for 1480 misspelled field names: 1481 1482 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 1483 you mean 'color'? 1484 return ptr->colour; 1485 ^~~~~~ 1486 1487 GCC 7 greatly expands the scope of these suggestions. Firstly, it 1488 adds fix-it hints to such suggestions: 1489 1490 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 1491 you mean 'color'? 1492 return ptr->colour; 1493 ^~~~~~ 1494 color 1495 1496 The suggestions now cover many other things, such as misspelled 1497 function names: 1498 1499 spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi 1500 dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat 1501 ion] 1502 gtk_widget_showall (w); 1503 ^~~~~~~~~~~~~~~~~~ 1504 gtk_widget_show_all 1505 1506 misspelled macro names and enum values: 1507 1508 spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun 1509 ction); did you mean 'MAX_ITEMS'? 1510 int array[MAX_ITEM]; 1511 ^~~~~~~~ 1512 MAX_ITEMS 1513 1514 misspelled type names: 1515 1516 spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si 1517 gned'? 1518 void test (singed char e); 1519 ^~~~~~ 1520 signed 1521 1522 and, in the C front end, named initializers: 1523 1524 test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color 1525 '? 1526 struct s test = { .colour = 3 }; 1527 ^~~~~~ 1528 color 1529 1530 * The preprocessor can now offer suggestions for misspelled 1531 directives, e.g.: 1532 1533 test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif? 1534 #endfi 1535 ^~~~~ 1536 endif 1537 1538 * Warnings about format strings now underline the pertinent part of 1539 the string, and can offer suggested fixes. In some cases, the 1540 pertinent argument is underlined. 1541 1542 test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume 1543 nt 3 has type 'int' [-Wformat=] 1544 printf ("foo: %d bar: %s baz: %d", 100, i + j, 102); 1545 ~^ ~~~~~ 1546 %d 1547 1548 * The new -Wdangling-else command-line option has been split out of 1549 -Wparentheses and warns about dangling else. 1550 * The -Wshadow warning has been split into three variants. 1551 -Wshadow=global warns for any shadowing. This is the default when 1552 using -Wshadow without any argument. -Wshadow=local only warns for 1553 a local variable shadowing another local variable or parameter. 1554 -Wshadow=compatible-local only warns for a local variable shadowing 1555 another local variable or parameter whose type is compatible (in 1556 C++ compatible means that the type of the shadowing variable can be 1557 converted to that of the shadowed variable). 1558 The following example shows the different kinds of shadow warnings: 1559 1560 enum operation { add, count }; 1561 struct container { int nr; }; 1562 1563 int 1564 container_count (struct container c, int count) 1565 { 1566 int r = 0; 1567 for (int count = 0; count > 0; count--) 1568 { 1569 struct container count = c; 1570 r += count.nr; 1571 } 1572 return r; 1573 } 1574 1575 -Wshadow=compatible-local will warn for the parameter being 1576 shadowed with the same type: 1577 1578 warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow= 1579 compatible-local] 1580 for (int count = 0; count > 0; count--) 1581 ^~~~~ 1582 warn-test.c:5:42: note: shadowed declaration is here 1583 container_count (struct container c, int count) 1584 ^~~~~ 1585 1586 -Wshadow=local will warn for the above and for the shadowed 1587 declaration with incompatible type: 1588 1589 warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws 1590 hadow=local] 1591 struct container count = c; 1592 ^~~~~ 1593 warn-test.c:8:12: note: shadowed declaration is here 1594 for (int count = 0; count > 0; count--) 1595 ^~~~~ 1596 1597 -Wshadow=global will warn for all of the above and the shadowing of 1598 the global declaration: 1599 1600 warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [ 1601 -Wshadow] 1602 container_count (struct container c, int count) 1603 ^~~~~ 1604 warn-test.c:1:23: note: shadowed declaration is here 1605 enum operation { add, count }; 1606 ^~~~~ 1607 1608 * GCC 7 contains a number of enhancements that help detect buffer 1609 overflow and other forms of invalid memory accesses. 1610 + The -Walloc-size-larger-than=size option detects calls to 1611 standard and user-defined memory allocation functions 1612 decorated with attribute alloc_size whose argument exceeds the 1613 specified size (PTRDIFF_MAX by default). The option also 1614 detects arithmetic overflow in the computation of the size in 1615 two-argument allocation functions like calloc where the total 1616 size is the product of the two arguments. Since calls with an 1617 excessive size cannot succeed they are typically the result of 1618 programming errors. Such bugs have been known to be the source 1619 of security vulnerabilities and a target of exploits. 1620 -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall. 1621 For example, the following call to malloc incorrectly tries to 1622 avoid passing a negative argument to the function and instead 1623 ends up unconditionally invoking it with an argument less than 1624 or equal to zero. Since after conversion to the type of the 1625 argument of the function (size_t) a negative argument results 1626 in a value in excess of the maximum PTRDIFF_MAX the call is 1627 diagnosed. 1628 1629 void* f (int n) 1630 { 1631 return malloc (n > 0 ? 0 : n); 1632 } 1633 1634 warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2 1635 147483647 [-Walloc-size-larger-than=] 1636 1637 + The -Walloc-zero option detects calls to standard and 1638 user-defined memory allocation functions decorated with 1639 attribute alloc_size with a zero argument. -Walloc-zero is not 1640 included in either -Wall or -Wextra and must be explicitly 1641 enabled. 1642 + The -Walloca option detects all calls to the alloca function 1643 in the program. -Walloca is not included in either -Wall or 1644 -Wextra and must be explicitly enabled. 1645 + The -Walloca-larger-than=size option detects calls to the 1646 alloca function whose argument either may exceed the specified 1647 size, or that is not known to be sufficiently constrained to 1648 avoid exceeding it. -Walloca-larger-than is not included in 1649 either -Wall or -Wextra and must be explicitly enabled. 1650 For example, compiling the following snippet with 1651 -Walloca-larger-than=1024 results in a warning because even 1652 though the code appears to call alloca only with sizes of 1kb 1653 and less, since n is signed, a negative value would result in 1654 a call to the function well in excess of the limit. 1655 1656 void f (int n) 1657 { 1658 char *d; 1659 if (n < 1025) 1660 d = alloca (n); 1661 else 1662 d = malloc (n); 1663 1664 } 1665 1666 warning: argument to 'alloca may be too large due to conversion from 'int' to 'l 1667 ong unsigned int' [-Walloca-larger-than=] 1668 1669 In contrast, a call to alloca that isn't bounded at all such 1670 as in the following function will elicit the warning below 1671 regardless of the size argument to the option. 1672 1673 void f (size_t n) 1674 { 1675 char *d = alloca (n); 1676 1677 } 1678 1679 warning: unbounded use of 'alloca' [-Walloca-larger-than=] 1680 1681 + The -Wformat-overflow=level option detects certain and likely 1682 buffer overflow in calls to the sprintf family of formatted 1683 output functions. Although the option is enabled even without 1684 optimization it works best with -O2 and higher. 1685 For example, in the following snippet the call to sprintf is 1686 diagnosed because even though its output has been constrained 1687 using the modulo operation it could result in as many as three 1688 bytes if mday were negative. The solution is to either 1689 allocate a larger buffer or make sure the argument is not 1690 negative, for example by changing mday's type to unsigned or 1691 by making the type of the second operand of the modulo 1692 expression unsigned: 100U. 1693 1694 void* f (int mday) 1695 { 1696 char *buf = malloc (3); 1697 sprintf (buf, "%02i", mday % 100); 1698 return buf; 1699 } 1700 1701 warning: 'sprintf may write a terminating nul past the end of the destination [- 1702 Wformat-overflow=] 1703 note: 'sprintf' output between 3 and 4 bytes into a destination of size 3 1704 1705 + The -Wformat-truncation=level option detects certain and 1706 likely output truncation in calls to the snprintf family of 1707 formatted output functions. -Wformat-truncation=1 is included 1708 in -Wall and enabled without optimization but works best with 1709 -O2 and higher. 1710 For example, the following function attempts to format an 1711 integer between 0 and 255 in hexadecimal, including the 0x 1712 prefix, into a buffer of four characters. But since the 1713 function must always terminate output by the null character 1714 ('\0') such a buffer is only big enough to fit just one digit 1715 plus the prefix. Therefore the snprintf call is diagnosed. To 1716 avoid the warning either use a bigger buffer or handle the 1717 function's return value which indicates whether or not its 1718 output has been truncated. 1719 1720 void f (unsigned x) 1721 { 1722 char d[4]; 1723 snprintf (d, sizeof d, "%#02x", x & 0xff); 1724 1725 } 1726 1727 warning: 'snprintf' output may be truncated before the last format character [-W 1728 format-truncation=] 1729 note: 'snprintf' output between 3 and 5 bytes into a destination of size 4 1730 1731 + The -Wnonnull option has been enhanced to detect a broader set 1732 of cases of passing null pointers to functions that expect a 1733 non-null argument (those decorated with attribute nonnull). By 1734 taking advantage of optimizations the option can detect many 1735 more cases of the problem than in prior GCC versions. 1736 + The -Wstringop-overflow=type option detects buffer overflow in 1737 calls to string handling functions like memcpy and strcpy. The 1738 option relies on [9]Object Size Checking and has an effect 1739 similar to defining the _FORTIFY_SOURCE macro. 1740 -Wstringop-overflow=2 is enabled by default. 1741 For example, in the following snippet, because the call to 1742 strncat specifies a maximum that allows the function to write 1743 past the end of the destination, it is diagnosed. To correct 1744 the problem and avoid the overflow the function should be 1745 called with a size of at most sizeof d - strlen(d) - 1. 1746 1747 void f (const char *fname) 1748 { 1749 char d[8]; 1750 strncpy (d, "/tmp/", sizeof d); 1751 strncat (d, fname, sizeof d); 1752 1753 } 1754 1755 warning: specified bound 8 equals the size of the destination [-Wstringop-overfl 1756 ow=] 1757 1758 * The <limits.h> header provided by GCC defines macros such as 1759 INT_WIDTH for the width in bits of integer types, if 1760 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 1761 included. The <stdint.h> header defines such macros as SIZE_WIDTH 1762 and INTMAX_WIDTH for the width of some standard typedef names for 1763 integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined 1764 before the header is included; note that GCC's implementation of 1765 this header is only used for freestanding compilations, not hosted 1766 compilations, on most systems. These macros come from ISO/IEC TS 1767 18661-1:2014. 1768 * The <float.h> header provided by GCC defines the macro 1769 CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if 1770 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 1771 included. This represents the number of decimal digits for which 1772 conversions between decimal character strings and binary formats, 1773 in both directions, are correctly rounded, and currently has the 1774 value of UINTMAX_MAX on all systems, reflecting that GCC's 1775 compile-time conversions are correctly rounded for any number of 1776 digits. 1777 * New __builtin_add_overflow_p, __builtin_sub_overflow_p, 1778 __builtin_mul_overflow_p built-in functions have been added. These 1779 work similarly to their siblings without the _p suffix, but do not 1780 actually store the result of the arithmetics anywhere, just return 1781 whether the operation would overflow. Calls to these built-ins with 1782 integer constant arguments evaluate to integer constants 1783 expressions. 1784 For example, in the following, c is assigned the result of a * b 1785 only if the multiplication does not overflow, otherwise it is 1786 assigned the value zero. The multiplication is performed at 1787 compile-time and without triggering a -Woverflow warning. 1788 1789 enum { 1790 a = 12345678, 1791 b = 87654321, 1792 c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b 1793 }; 1794 1795 C 1796 1797 * The C front end now supports type names _FloatN for floating-point 1798 types with IEEE interchange formats and _FloatNx for floating-point 1799 types with IEEE extended formats. These type names come from 1800 ISO/IEC TS 18661-3:2015. 1801 The set of types supported depends on the target for which GCC is 1802 configured. Most targets support _Float32, _Float32x and _Float64. 1803 _Float128 is supported on targets where IEEE binary128 encoding was 1804 already supported as long double or __float128. _Float64x is 1805 supported on targets where a type with either binary128 or Intel 1806 extended precision format is available. 1807 Constants with these types are supported using suffixes fN, FN, fNx 1808 and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are 1809 defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is 1810 defined before it is included. 1811 These new types are always distinct from each other and from float, 1812 double and long double, even if they have the same encoding. 1813 Complex types such as _Complex _Float128 are also supported. 1814 Type-generic built-in functions such as __builtin_isinf support the 1815 new types, and the following type-specific built-in functions have 1816 versions (suffixed fN or fNx) for the new types: 1817 __builtin_copysign, __builtin_fabs, __builtin_huge_val, 1818 __builtin_inf, __builtin_nan, __builtin_nans. 1819 * Compilation with -fopenmp is now compatible with the C11 _Atomic 1820 keyword. 1821 1822 C++ 1823 1824 * The C++ front end has experimental support for all of the current 1825 C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if 1826 constexpr, class template argument deduction, auto template 1827 parameters, and structured bindings. For a full list of new 1828 features, see [10]the C++ status page. 1829 * C++17 support for new of over-aligned types can be enabled in other 1830 modes with the -faligned-new flag. 1831 * The C++17 evaluation order requirements can be selected in other 1832 modes with the -fstrong-eval-order flag, or disabled in C++17 mode 1833 with -fno-strong-eval-order. 1834 * The default semantics of inherited constructors has changed in all 1835 modes, following [11]P0136. Essentially, overload resolution 1836 happens as if calling the inherited constructor directly, and the 1837 compiler fills in construction of the other bases and members as 1838 needed. Most uses should not need any changes. The old behavior can 1839 be restored with -fno-new-inheriting-ctors, or -fabi-version less 1840 than 11. 1841 * The resolution of DR 150 on matching of template template 1842 parameters, allowing default template arguments to make a template 1843 match a parameter, is currently enabled by default in C++17 mode 1844 only. The default can be overridden with -f{no-,}new-ttp-matching. 1845 * The C++ front end will now provide fix-it hints for some missing 1846 semicolons, allowing for automatic fixes by IDEs: 1847 1848 test.cc:4:11: error: expected ';' after class definition 1849 class a {} 1850 ^ 1851 ; 1852 1853 * -Waligned-new has been added to the C++ front end. It warns about 1854 new of type with extended alignment without -faligned-new. 1855 1856 Runtime Library (libstdc++) 1857 1858 * The type of exception thrown by iostreams, std::ios_base::failure, 1859 now uses the [12]cxx11 ABI. 1860 * Experimental support for C++17, including the following new 1861 features: 1862 + std::string_view; 1863 + std::any, std::optional, and std::variant; 1864 + std::invoke, std::is_invocable, std::is_nothrow_invocable, and 1865 invoke_result; 1866 + std::is_swappable, and std::is_nothrow_swappable; 1867 + std::apply, and std::make_from_tuple; 1868 + std::void_t, std::bool_constant, std::conjunction, 1869 std::disjunction, and std::negation; 1870 + Variable templates for type traits; 1871 + Mathematical Special Functions; 1872 + std::chrono::floor, std::chrono::ceil, std::chrono::round, and 1873 std::chrono::abs; 1874 + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot; 1875 + std::scoped_lock, std::shared_mutex, 1876 std::atomic<T>::is_always_lock_free; 1877 + std::sample, std::default_searcher, std::boyer_moore_searcher 1878 and std::boyer_moore_horspool_searcher; 1879 + Extraction and re-insertion of map and set nodes, try_emplace 1880 members for maps, and functions for accessing containers 1881 std::size, std::empty, and std::data; 1882 + std::shared_ptr support for arrays, 1883 std::shared_ptr<T>::weak_type, 1884 std::enable_shared_from_this<T>::weak_from_this(), and 1885 std::owner_less<void>; 1886 + std::byte; 1887 + std::as_const, std::not_fn, 1888 std::has_unique_object_representations, constexpr 1889 std::addressof. 1890 Thanks to Daniel Krgler, Tim Shen, Edward Smith-Rowland, and Ville 1891 Voutilainen for work on the C++17 support. 1892 * A new power-of-two rehashing policy for use with the _Hashtable 1893 internals, thanks to Franois Dumont. 1894 1895 Fortran 1896 1897 * Support for a number of extensions for compatibility with legacy 1898 code with new flags: 1899 + -fdec-structure Support for DEC STRUCTURE and UNION 1900 + -fdec-intrinsic-ints Support for new integer intrinsics with 1901 B/I/J/K prefixes such as BABS, JIAND... 1902 + -fdec-math Support for additional math intrinsics, including 1903 COTAN and degree-valued trigonometric functions such as TAND, 1904 ASIND... 1905 + -fdec Enable the -fdec-* family of extensions. 1906 * New flag -finit-derived to allow default initialization of 1907 derived-type variables. 1908 * Improved DO loops with step equal to 1 or -1, generates faster code 1909 without a loop preheader. A new warning, -Wundefined-do-loop, warns 1910 when a loop iterates either to HUGE(i) (with step equal to 1), or 1911 to -HUGE(i) (with step equal to -1). Invalid behavior can be caught 1912 at run time with -fcheck=do enabled: 1913 1914 program test 1915 implicit none 1916 integer(1) :: i 1917 do i = -HUGE(i)+10, -HUGE(i)-1, -1 1918 print *, i 1919 end do 1920 end program test 1921 1922 At line 8 of file do_check_12.f90 1923 Fortran runtime error: Loop iterates infinitely 1924 1925 * Version 4.5 of the [13]OpenMP specification is now partially 1926 supported in the Fortran compiler; the largest missing item is 1927 structure element mapping. 1928 * User-defined derived-type input/output (UDTIO) is added. 1929 * Derived type coarrays with allocatable and pointer components are 1930 partially supported. 1931 * Non-constant stop codes and error stop codes (Fortran 2015 1932 feature). 1933 * Derived types with allocatable components of recursive type. 1934 * Intrinsic assignment to polymorphic variables. 1935 * Improved submodule support. 1936 * Improved diagnostics (polymorphic results in pure functions). 1937 * Coarray: Support for failed images (Fortan 2015 feature). 1938 1939 Go 1940 1941 * GCC 7 provides a complete implementation of the Go 1.8.1 user 1942 packages. 1943 * Compared to the Go 1.8.1 toolchain, the garbage collector is more 1944 conservative and less concurrent. 1945 * Escape analysis is available for experimental use via the 1946 -fgo-optimize-allocs option. The -fgo-debug-escape prints 1947 information useful for debugging escape analysis choices. 1948 1949 Java (GCJ) 1950 1951 The GCC Java front end and associated libjava runtime library have been 1952 removed from GCC. 1953 1954 libgccjit 1955 1956 The libgccjit API gained support for marking calls as requiring 1957 tail-call optimization via a new entry point: 1958 [14]gcc_jit_rvalue_set_bool_require_tail_call. 1959 1960 libgccjit performs numerous checks at the API boundary, but if these 1961 succeed, it previously ignored errors and other diagnostics emitted 1962 within the core of GCC, and treated the compile of a gcc_jit_context as 1963 having succeeded. As of GCC 7 it now ensures that if any diagnostics 1964 are emitted, they are visible from the libgccjit API, and that the the 1965 context is flagged as having failed. 1966 1967 New Targets and Target Specific Improvements 1968 1969 AArch64 1970 1971 * GCC has been updated to the latest revision of the procedure call 1972 standard (AAPCS64) to provide support for parameter passing when 1973 data types have been over-aligned. 1974 * The ARMv8.3-A architecture is now supported. It can be used by 1975 specifying the -march=armv8.3-a option. 1976 * The option -msign-return-address= is supported to enable return 1977 address protection using ARMv8.3-A Pointer Authentication 1978 Extensions. For more information on the arguments accepted by this 1979 option, please refer to [15]AArch64-Options. 1980 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 1981 Extensions are now supported. They can be used by specifying the 1982 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 1983 Floating-Point Extensions introduce new half-precision data 1984 processing floating-point instructions. 1985 * Support has been added for the following processors (GCC 1986 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom 1987 Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium 1988 ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx 1989 (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1), 1990 Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor). 1991 The GCC identifiers can be used as arguments to the -mcpu or -mtune 1992 options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as 1993 arguments to the equivalent target attributes and pragmas. 1994 1995 ARC 1996 1997 * Added support for ARC HS and ARC EM processors. 1998 * Added support for ARC EM variation found in Intel QuarkSE SoCs. 1999 * Added support for NPS400 ARC700 based CPUs. 2000 * Thread Local Storage is now supported by ARC CPUs. 2001 * Fixed errors for ARC600 when using 32x16 multiplier option. 2002 * Fixed PIE for ARC CPUs. 2003 * New CPU templates are supported via multilib. 2004 2005 ARM 2006 2007 * Support for the ARMv5 and ARMv5E architectures has been deprecated 2008 (which have no known implementations) and will be removed in a 2009 future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ 2010 architectures remain supported. The values armv5 and armv5e of 2011 -march are thus deprecated. 2012 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 2013 Extensions are now supported. They can be used by specifying the 2014 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 2015 Floating-Point Extensions introduce new half-precision data 2016 processing floating-point instructions. 2017 * The ARMv8-M architecture is now supported in its two architecture 2018 profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and 2019 Floating-Point Extensions. They can be used by specifying the 2020 -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options. 2021 * Support has been added for the following processors (GCC 2022 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM 2023 Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC 2024 identifiers can be used as arguments to the -mcpu or -mtune 2025 options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33. 2026 * A new command-line option -mpure-code has been added. It does not 2027 allow constant data to be placed in code sections. This option is 2028 only available when generating non-PIC code for ARMv7-M targets. 2029 * Support for the ACLE Coprocessor Intrinsics has been added. This 2030 enables the generation of coprocessor instructions through the use 2031 of intrinsics such as cdp, ldc, and others. 2032 * The configure option --with-multilib-list now accepts the value 2033 rmprofile to build multilib libraries for a range of embedded 2034 targets. See our [16]installation instructions for details. 2035 2036 AVR 2037 2038 * On the reduced Tiny cores, the progmem [17]variable attribute is 2039 now properly supported. Respective read-only variables are located 2040 in flash memory in section .progmem.data. No special code is needed 2041 to access such variables; the compiler automatically adds an offset 2042 of 0x4000 to all addresses, which is needed to access variables in 2043 flash memory. As opposed to ordinary cores where it is sufficient 2044 to specify the progmem attribute with definitions, on the reduced 2045 Tiny cores the attribute also has to be specified with (external) 2046 declarations: 2047 2048 extern const int array[] __attribute__((__progmem__)); 2049 2050 int get_value2 (void) 2051 { 2052 /* Access via addresses array + 0x4004 and array + 0x4005. */ 2053 return array[2]; 2054 } 2055 2056 const int* get_address (unsigned idx) 2057 { 2058 /* Returns array + 0x4000 + 2 * idx. */ 2059 return &array[idx]; 2060 } 2061 2062 * A new command-line option -Wmisspelled-isr has been added. It turns 2063 off or turns into errors warnings that are reported for 2064 interrupt service routines (ISRs) which don't follow AVR-LibC's 2065 naming convention of prefixing ISR names with __vector. 2066 * __builtin_avr_nops(n) is a new [18]built-in function that inserts n 2067 NOP instructions into the instruction stream. n must be a value 2068 known at compile time. 2069 2070 IA-32/x86-64 2071 2072 * Support for the AVX-512 Fused Multiply Accumulation Packed Single 2073 precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word 2074 variable precision (4VNNIW), AVX-512 Vector Population Count 2075 (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has 2076 been added. 2077 2078 NVPTX 2079 2080 * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs. 2081 See the [19]Offloading Wiki on how to configure it. 2082 2083 PowerPC / PowerPC64 / RS6000 2084 2085 * The PowerPC port now uses LRA by default. 2086 * GCC now diagnoses inline assembly that clobbers register r2. This 2087 has always been invalid code, and is no longer quietly tolerated. 2088 * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been 2089 enhanced to generate more of the new instructions by default, and 2090 to provide more built-in functions to generate code for other new 2091 instructions. 2092 * The configuration option --enable-gnu-indirect-function is now 2093 enabled by default on PowerPC GNU/Linux builds. 2094 * The PowerPC port will now allow 64-bit and 32-bit integer types to 2095 be allocated to the VSX vector registers (ISA 2.06 and above). In 2096 addition, on ISA 3.0, 16-bit and 8-bit integer types can be 2097 allocated in the vector registers. Previously, only 64-bit integer 2098 types were allowed in the traditional floating point registers. 2099 * New options -mstack-protector-guard=global, 2100 -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and 2101 -mstack-protector-guard-offset= change how the stack protector gets 2102 the value to use as canary. 2103 2104 S/390, System z, IBM z Systems, IBM Z 2105 2106 * Support for the IBM z14 processor has been added. When using the 2107 -march=z14 option, the compiler will generate code making use of 2108 the new instructions introduced with the vector enhancement 2109 facility and the miscellaneous instruction extension facility 2. 2110 The -mtune=z14 option enables z14 specific instruction scheduling 2111 without making use of new instructions. 2112 * Builtins for the new vector instructions have been added and can be 2113 enabled using the -mzvector option. 2114 2115 RISC-V 2116 2117 * Support for the RISC-V instruction set has been added. 2118 2119 RX 2120 2121 Basic support for atomic built-in function has been added. It is 2122 currently implemented by flipping interrupts off and on as needed. 2123 2124 SH 2125 2126 * Support for SH5/SH64 has been removed. 2127 * Improved utilization of delay slots on SH2A. 2128 * Improved utilization of zero-displacement conditional branches. 2129 * The following deprecated options have been removed 2130 + -mcbranchdi 2131 + -mcmpeqdi 2132 + -minvalid-symbols 2133 + -msoft-atomic 2134 + -mspace 2135 + -madjust-unroll 2136 * Support for the following SH2A instructions has been added 2137 + mov.b @-Rm,R0 2138 + mov.w @-Rm,R0 2139 + mov.l @-Rm,R0 2140 + mov.b R0,@Rn+ 2141 + mov.w R0,@Rn+ 2142 + mov.l R0,@Rn+ 2143 2144 SPARC 2145 2146 * The SPARC port now uses LRA by default. 2147 * Support for the new Subtract-Extended-with-Carry instruction 2148 available in SPARC M7 (Niagara 7) has been added. 2149 2150 Operating Systems 2151 2152 AIX 2153 2154 * Visibility support has been enabled for AIX 7.1 and above. 2155 2156 Fuchsia 2157 2158 * Support has been added for the [20]Fuchsia OS. 2159 2160 RTEMS 2161 2162 * The ABI changes on ARM so that no short enums are used by default. 2163 2164 Other significant improvements 2165 2166 * -fverbose-asm previously emitted information on the meanings of 2167 assembly expressions. This has been extended so that it now also 2168 prints comments showing the source lines that correspond to the 2169 assembly, making it easier to read the generated assembly 2170 (especially with larger functions). For example, given this C 2171 source file: 2172 2173 int test (int n) 2174 { 2175 int i; 2176 int total = 0; 2177 2178 for (i = 0; i < n; i++) 2179 total += i * i; 2180 return total; 2181 } 2182 2183 -fverbose-asm now gives output similar to this for the function 2184 body (when compiling for x86_64, with -Os): 2185 2186 .text 2187 .globl test 2188 .type test, @@function 2189 test: 2190 .LFB0: 2191 .cfi_startproc 2192 # example.c:4: int total = 0; 2193 xorl %eax, %eax # <retval> 2194 # example.c:6: for (i = 0; i < n; i++) 2195 xorl %edx, %edx # i 2196 .L2: 2197 # example.c:6: for (i = 0; i < n; i++) 2198 cmpl %edi, %edx # n, i 2199 jge .L5 #, 2200 # example.c:7: total += i * i; 2201 movl %edx, %ecx # i, tmp92 2202 imull %edx, %ecx # i, tmp92 2203 # example.c:6: for (i = 0; i < n; i++) 2204 incl %edx # i 2205 # example.c:7: total += i * i; 2206 addl %ecx, %eax # tmp92, <retval> 2207 jmp .L2 # 2208 .L5: 2209 # example.c:10: } 2210 ret 2211 .cfi_endproc 2212 2213 * Two new options have been added for printing fix-it hints: 2214 + -fdiagnostics-parseable-fixits allows for fix-it hints to be 2215 emitted in a machine-readable form, suitable for consumption 2216 by IDEs. For example, given: 2217 2218 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 2219 you mean 'color'? 2220 return ptr->colour; 2221 ^~~~~~ 2222 color 2223 2224 it will emit: 2225 2226 fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color" 2227 2228 + -fdiagnostics-generate-patch will print a patch in "unified" 2229 format after any diagnostics are printed, showing the result 2230 of applying all fix-it hints. For the above example it would 2231 emit: 2232 2233 --- spellcheck-fields.cc 2234 +++ spellcheck-fields.cc 2235 @@ -49,5 +49,5 @@ 2236 2237 color get_color(struct s *ptr) 2238 { 2239 - return ptr->colour; 2240 + return ptr->color; 2241 } 2242 2243 * The gcc and g++ driver programs will now provide suggestions for 2244 misspelled arguments to command-line options. 2245 2246 $ gcc -c test.c -ftls-model=global-dinamic 2247 gcc: error: unknown TLS model 'global-dinamic' 2248 gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo 2249 cal-dynamic local-exec; did you mean 'global-dynamic'? 2250 2251 * The compiler will now provide suggestions for misspelled 2252 parameters. 2253 2254 $ gcc -c test.c --param max-early-inliner-iteration=3 2255 cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma 2256 x-early-inliner-iterations'? 2257 2258 * Profile-guided optimization (PGO) instrumentation, as well as test 2259 coverage (GCOV), can newly instrument constructors (functions marks 2260 with __attribute__((constructor))), destructors and C++ 2261 constructors (and destructors) of classes that are used as the type 2262 of a global variable. 2263 * A new option -fprofile-update=atomic prevents creation of corrupted 2264 profiles created during an instrumentation run (-fprofile=generate) 2265 of an application. The downside of the option is a speed penalty. 2266 Providing -pthread on the command line selects atomic profile 2267 updating (when supported by the target). 2268 * GCC's already extensive testsuite has gained some new capabilities, 2269 to further improve the reliability of the compiler: 2270 + GCC now has an internal unit-testing API and a suite of tests 2271 for programmatic self-testing of subsystems. 2272 + GCC's C front end has been extended so that it can parse dumps 2273 of GCC's internal representations, allowing for DejaGnu tests 2274 that more directly exercise specific optimization passes. This 2275 covers both the [21]GIMPLE representation (for testing 2276 higher-level optimizations) and the [22]RTL representation, 2277 allowing for more direct testing of lower-level details, such 2278 as register allocation and instruction selection. 2279 2280 GCC 7.1 2281 2282 This is the [23]list of problem reports (PRs) from GCC's bug tracking 2283 system that are known to be fixed in the 7.1 release. This list might 2284 not be complete (that is, it is possible that some PRs that have been 2285 fixed are not listed here). 2286 2287 GCC 7.2 2288 2289 This is the [24]list of problem reports (PRs) from GCC's bug tracking 2290 system that are known to be fixed in the 7.2 release. This list might 2291 not be complete (that is, it is possible that some PRs that have been 2292 fixed are not listed here). 2293 2294 Target Specific Changes 2295 2296 SPARC 2297 2298 * Support for the SPARC M8 processor has been added. 2299 * The switches -mfix-ut700 and -mfix-gr712rc have been added to work 2300 around an erratum in LEON3FT processors. 2301 * Use of the Floating-point Multiply Single to Double (FsMULd) 2302 instruction can now be controlled by the -mfsmuld and -fno-fsmuld 2303 options. 2304 2305 Operating Systems 2306 2307 RTEMS 2308 2309 * The Ada run-time support uses now thread-local storage (TLS). 2310 * Support for RISC-V has been added. 2311 * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long 2312 double has been added. 2313 2314 GCC 7.3 2315 2316 This is the [25]list of problem reports (PRs) from GCC's bug tracking 2317 system that are known to be fixed in the 7.3 release. This list might 2318 not be complete (that is, it is possible that some PRs that have been 2319 fixed are not listed here). 2320 2321 Target Specific Changes 2322 2323 SPARC 2324 2325 * Workarounds for the four [26]LEON3FT errata GRLIB-TN-0010..0013 2326 have been added. Relevant errata are activated by the target 2327 specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches. 2328 2329 Operating Systems 2330 2331 RTEMS 2332 2333 * Support has been added for Epiphany target. 2334 2335 GCC 7.4 2336 2337 This is the [27]list of problem reports (PRs) from GCC's bug tracking 2338 system that are known to be fixed in the 7.4 release. This list might 2339 not be complete (that is, it is possible that some PRs that have been 2340 fixed are not listed here). 2341 2342 GCC 7.5 2343 2344 This is the [28]list of problem reports (PRs) from GCC's bug tracking 2345 system that are known to be fixed in the 7.5 release. This list might 2346 not be complete (that is, it is possible that some PRs that have been 2347 fixed are not listed here). 2348 2349 2350 For questions related to the use of GCC, please consult these web 2351 pages and the [29]GCC manuals. If that fails, the 2352 [30]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2353 web pages and the development of GCC are welcome on our developer 2354 list at [31]gcc (a] gcc.gnu.org. All of [32]our lists have public 2355 archives. 2356 2357 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 2358 distribution of this entire article is permitted in any medium, 2359 provided this notice is preserved. 2360 2361 These pages are [34]maintained by the GCC team. Last modified 2362 2019-11-28[35]. 2363 2364 References 2365 2366 1. http://gcc.gnu.org/gcc-7/porting_to.html 2367 2. http://gcc.gnu.org/onlinedocs/index.html#current 2368 3. https://gcc.gnu.org/wiki/LRAIsDefault 2369 4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html 2370 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728 2371 6. http://www.dwarfstd.org/Download.php 2372 7. https://gcc.gnu.org/wiki/OpenACC 2373 8. https://gcc.gnu.org/wiki/Offloading 2374 9. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html 2375 10. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 2376 11. http://wg21.link/p0136 2377 12. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html 2378 13. https://www.openmp.org/specifications/ 2379 14. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call 2380 15. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options 2381 16. https://gcc.gnu.org/install/configure.html 2382 17. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html 2383 18. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html 2384 19. https://gcc.gnu.org/wiki/Offloading 2385 20. https://fuchsia.googlesource.com/ 2386 21. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html 2387 22. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html 2388 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0 2389 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2 2390 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3 2391 26. https://www.gaisler.com/index.php/information/app-tech-notes 2392 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4 2393 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.5 2394 29. https://gcc.gnu.org/onlinedocs/ 2395 30. mailto:gcc-help (a] gcc.gnu.org 2396 31. mailto:gcc (a] gcc.gnu.org 2397 32. https://gcc.gnu.org/lists.html 2398 33. https://www.fsf.org/ 2399 34. https://gcc.gnu.org/about.html 2400 35. http://validator.w3.org/check/referer 2401 ====================================================================== 2402 http://gcc.gnu.org/gcc-6/index.html 2403 GCC 6 Release Series 2404 2405 (This release series is no longer supported.) 2406 2407 October 26, 2018 2408 2409 The [1]GNU project and the GCC developers are pleased to announce the 2410 release of GCC 6.5. 2411 2412 This release is a bug-fix release, containing fixes for regressions in 2413 GCC 6.4 relative to previous releases of GCC. 2414 2415 Release History 2416 2417 GCC 6.5 2418 October 26, 2018 ([2]changes, [3]documentation) 2419 2420 GCC 6.4 2421 July 4, 2017 ([4]changes, [5]documentation) 2422 2423 GCC 6.3 2424 December 21, 2016 ([6]changes, [7]documentation) 2425 2426 GCC 6.2 2427 August 22, 2016 ([8]changes, [9]documentation) 2428 2429 GCC 6.1 2430 April 27, 2016 ([10]changes, [11]documentation) 2431 2432 References and Acknowledgements 2433 2434 GCC used to stand for the GNU C Compiler, but since the compiler 2435 supports several other languages aside from C, it now stands for the 2436 GNU Compiler Collection. 2437 2438 A list of [12]successful builds is updated as new information becomes 2439 available. 2440 2441 The GCC developers would like to thank the numerous people that have 2442 contributed new features, improvements, bug fixes, and other changes as 2443 well as test results to GCC. This [13]amazing group of volunteers is 2444 what makes GCC successful. 2445 2446 For additional information about GCC please refer to the [14]GCC 2447 project web site or contact the [15]GCC development mailing list. 2448 2449 To obtain GCC please use [16]our mirror sites or [17]our version 2450 control system. 2451 2452 2453 For questions related to the use of GCC, please consult these web 2454 pages and the [18]GCC manuals. If that fails, the 2455 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2456 web pages and the development of GCC are welcome on our developer 2457 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 2458 archives. 2459 2460 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 2461 distribution of this entire article is permitted in any medium, 2462 provided this notice is preserved. 2463 2464 These pages are [23]maintained by the GCC team. Last modified 2465 2020-01-14[24]. 2466 2467 References 2468 2469 1. http://www.gnu.org/ 2470 2. http://gcc.gnu.org/gcc-6/changes.html 2471 3. http://gcc.gnu.org/onlinedocs/6.5.0/ 2472 4. http://gcc.gnu.org/gcc-6/changes.html 2473 5. http://gcc.gnu.org/onlinedocs/6.4.0/ 2474 6. http://gcc.gnu.org/gcc-6/changes.html 2475 7. http://gcc.gnu.org/onlinedocs/6.3.0/ 2476 8. http://gcc.gnu.org/gcc-6/changes.html 2477 9. http://gcc.gnu.org/onlinedocs/6.2.0/ 2478 10. http://gcc.gnu.org/gcc-6/changes.html 2479 11. http://gcc.gnu.org/onlinedocs/6.1.0/ 2480 12. http://gcc.gnu.org/gcc-6/buildstat.html 2481 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 2482 14. http://gcc.gnu.org/index.html 2483 15. mailto:gcc (a] gcc.gnu.org 2484 16. http://gcc.gnu.org/mirrors.html 2485 17. http://gcc.gnu.org/git.html 2486 18. https://gcc.gnu.org/onlinedocs/ 2487 19. mailto:gcc-help (a] gcc.gnu.org 2488 20. mailto:gcc (a] gcc.gnu.org 2489 21. https://gcc.gnu.org/lists.html 2490 22. https://www.fsf.org/ 2491 23. https://gcc.gnu.org/about.html 2492 24. http://validator.w3.org/check/referer 2493 ====================================================================== 2494 http://gcc.gnu.org/gcc-6/changes.html 2495 GCC 6 Release Series 2496 Changes, New Features, and Fixes 2497 2498 This page is a brief summary of some of the huge number of improvements 2499 in GCC 6. For more information, see the [1]Porting to GCC 6 page and 2500 the [2]full GCC documentation. 2501 2502 Caveats 2503 2504 * The default mode for C++ is now -std=gnu++14 instead of 2505 -std=gnu++98. 2506 * Support for a number of older systems and recently unmaintained or 2507 untested target ports of GCC has been declared obsolete in GCC 6. 2508 Unless there is activity to revive them, the next release of GCC 2509 will have their sources permanently removed. 2510 The following ports for individual systems on particular 2511 architectures have been obsoleted: 2512 + SH5 / SH64 (sh64-*-*) as announced [3]here. 2513 * The AVR port requires binutils version 2.26.1 or later for the fix 2514 for [4]PR71151 to work. 2515 * The GCC 6.5 release has an accidental ABI incompatibility for 2516 nested std::pair objects, for more details see [5]PR 87822. The bug 2517 causes a layout change for pairs where the first member is also a 2518 pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series 2519 is closed so the bug in GCC 6.5 will not be fixed upstream, but 2520 there is a patch in the bug report to allow it to be fixed by 2521 anybody packaging GCC 6.5 or installing it themselves. 2522 2523 General Optimizer Improvements 2524 2525 * UndefinedBehaviorSanitizer gained a new sanitization option, 2526 -fsanitize=bounds-strict, which enables strict checking of array 2527 bounds. In particular, it enables -fsanitize=bounds as well as 2528 instrumentation of flexible array member-like arrays. 2529 * Type-based alias analysis now disambiguates accesses to different 2530 pointers. This improves precision of the alias oracle by about 2531 20-30% on higher-level C++ programs. Programs doing invalid type 2532 punning of pointer types may now need -fno-strict-aliasing to work 2533 correctly. 2534 * Alias analysis now correctly supports the weakref and alias 2535 attributes. This allows accessing both a variable and its alias in 2536 one translation unit which is common with link-time optimization. 2537 * Value range propagation now assumes that the this pointer in C++ 2538 member functions is non-null. This eliminates common null pointer 2539 checks but also breaks some non-conforming code-bases (such as 2540 Qt-5, Chromium, KDevelop). As a temporary work-around 2541 -fno-delete-null-pointer-checks can be used. Wrong code can be 2542 identified by using -fsanitize=undefined. 2543 * Link-time optimization improvements: 2544 + warning and error attributes are now correctly preserved by 2545 declaration linking and thus -D_FORTIFY_SOURCE=2 is now 2546 supported with -flto. 2547 + Type merging was fixed to handle C and Fortran 2548 interoperability rules as defined by the Fortran 2008 language 2549 standard. 2550 As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable 2551 with char in all cases because it is an array while char is 2552 scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In 2553 general, this inter-operability cannot be implemented, for 2554 example on targets where the argument passing convention for 2555 arrays differs from scalars. 2556 + More type information is now preserved at link time, reducing 2557 the loss of accuracy of the type-based alias analysis compared 2558 to builds without link-time optimization. 2559 + Invalid type punning on global variables and declarations is 2560 now reported with -Wodr-type-mismatch. 2561 + The size of LTO object files was reduced by about 11% 2562 (measured by compiling Firefox 46.0). 2563 + Link-time parallelization (enabled using -flto=n) was 2564 significantly improved by decreasing the size of streamed data 2565 when partitioning programs. The size of streamed IL while 2566 compiling Firefox 46.0 was reduced by 66%. 2567 + The linker plugin was extended to pass information about the 2568 type of binary produced to the GCC back end. (That can also be 2569 controlled manually by -flinker-output.) This makes it 2570 possible to properly configure the code generator and support 2571 incremental linking. Incremental linking of LTO objects by gcc 2572 -r is now supported for plugin-enabled setups. 2573 There are two ways to perform incremental linking: 2574 1. Linking by ld -r will result in an object file with all 2575 sections from individual object files mechanically 2576 merged. This delays the actual link-time optimization to 2577 the final linking step and thus permits whole program 2578 optimization. Linking the final binary with such object 2579 files is however slower. 2580 2. Linking by gcc -r will lead to link-time optimization and 2581 emit the final binary into the object file. Linking such 2582 an object file is fast but avoids any benefits from whole 2583 program optimization. 2584 GCC 7 will support incremental link-time optimization with gcc 2585 -r. 2586 * Inter-procedural optimization improvements: 2587 + Basic jump threading is now performed before profile 2588 construction and inline analysis, resulting in more realistic 2589 size and time estimates that drive the heuristics of the 2590 inliner and function cloning passes. 2591 + Function cloning now more aggressively eliminates unused 2592 function parameters. 2593 2594 New Languages and Language specific improvements 2595 2596 Compared to GCC 5, the GCC 6 release series includes a much improved 2597 implementation of the [6]OpenACC 2.0a specification. Highlights are: 2598 * In addition to single-threaded host-fallback execution, offloading 2599 is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit 2600 little-endian GNU/Linux host systems. For nvptx offloading, with 2601 the OpenACC parallel construct, the execution model allows for an 2602 arbitrary number of gangs, up to 32 workers, and 32 vectors. 2603 * Initial support for parallelized execution of OpenACC kernels 2604 constructs: 2605 + Parallelization of a kernels region is switched on by 2606 -fopenacc combined with -O2 or higher. 2607 + Code is offloaded onto multiple gangs, but executes with just 2608 one worker, and a vector length of 1. 2609 + Directives inside a kernels region are not supported. 2610 + Loops with reductions can be parallelized. 2611 + Only kernels regions with one loop nest are parallelized. 2612 + Only the outer-most loop of a loop nest can be parallelized. 2613 + Loop nests containing sibling loops are not parallelized. 2614 Typically, using the OpenACC parallel construct gives much better 2615 performance, compared to the initial support of the OpenACC kernels 2616 construct. 2617 * The device_type clause is not supported. The bind and nohost 2618 clauses are not supported. The host_data directive is not supported 2619 in Fortran. 2620 * Nested parallelism (cf. CUDA dynamic parallelism) is not supported. 2621 * Usage of OpenACC constructs inside multithreaded contexts (such as 2622 created by OpenMP, or pthread programming) is not supported. 2623 * If a call to the acc_on_device function has a compile-time constant 2624 argument, the function call evaluates to a compile-time constant 2625 value only for C and C++ but not for Fortran. 2626 2627 See the [7]OpenACC and [8]Offloading wiki pages for further 2628 information. 2629 2630 C family 2631 2632 * Version 4.5 of the [9]OpenMP specification is now supported in the 2633 C and C++ compilers. 2634 * The C and C++ compilers now support attributes on enumerators. For 2635 instance, it is now possible to mark enumerators as deprecated: 2636 2637 enum { 2638 newval, 2639 oldval __attribute__ ((deprecated ("too old"))) 2640 }; 2641 2642 * Source locations for the C and C++ compilers are now tracked as 2643 ranges, rather than just points, making it easier to identify the 2644 subexpression of interest within a complicated expression. For 2645 example: 2646 2647 test.cc: In function 'int test(int, int, foo, int, int)': 2648 test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo' 2649 ) 2650 return p + q * r * s + t; 2651 ~~^~~ 2652 2653 In addition, there is now initial support for precise diagnostic 2654 locations within strings: 2655 2656 format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in 2657 t' argument [-Wformat=] 2658 printf("%*d"); 2659 ^ 2660 2661 * Diagnostics can now contain "fix-it hints", which are displayed in 2662 context underneath the relevant source code. For example: 2663 2664 fixits.c: In function 'bad_deref': 2665 fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'? 2666 return ptr.x; 2667 ^ 2668 -> 2669 2670 * The C and C++ compilers now offer suggestions for misspelled field 2671 names: 2672 2673 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 2674 you mean 'color'? 2675 return ptr->colour; 2676 ^~~~~~ 2677 2678 * New command-line options have been added for the C and C++ 2679 compilers: 2680 + -Wshift-negative-value warns about left shifting a negative 2681 value. 2682 + -Wshift-overflow warns about left shift overflows. This 2683 warning is enabled by default. -Wshift-overflow=2 also warns 2684 about left-shifting 1 into the sign bit. 2685 + -Wtautological-compare warns if a self-comparison always 2686 evaluates to true or false. This warning is enabled by -Wall. 2687 + -Wnull-dereference warns if the compiler detects paths that 2688 trigger erroneous or undefined behavior due to dereferencing a 2689 null pointer. This option is only active when 2690 -fdelete-null-pointer-checks is active, which is enabled by 2691 optimizations in most targets. The precision of the warnings 2692 depends on the optimization options used. 2693 + -Wduplicated-cond warns about duplicated conditions in an 2694 if-else-if chain. 2695 + -Wmisleading-indentation warns about places where the 2696 indentation of the code gives a misleading idea of the block 2697 structure of the code to a human reader. For example, given 2698 [10]CVE-2014-1266: 2699 2700 sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange': 2701 sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin 2702 g-indentation] 2703 if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0) 2704 ^~ 2705 sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly 2706 indented as if it is guarded by the 'if' 2707 goto fail; 2708 ^~~~ 2709 2710 This warning is enabled by -Wall. 2711 * The C and C++ compilers now emit saner error messages if 2712 merge-conflict markers are present in a source file. 2713 2714 test.c:3:1: error: version control conflict marker in file 2715 <<<<<<< HEAD 2716 ^~~~~~~ 2717 2718 C 2719 2720 * It is possible to disable warnings when an initialized field of a 2721 structure or a union with side effects is being overridden when 2722 using designated initializers via a new warning option 2723 -Woverride-init-side-effects. 2724 * A new type attribute scalar_storage_order applying to structures 2725 and unions has been introduced. It specifies the storage order (aka 2726 endianness) in memory of scalar fields in structures or unions. 2727 2728 C++ 2729 2730 * The default mode has been changed to -std=gnu++14. 2731 * [11]C++ Concepts are now supported when compiling with -fconcepts. 2732 * -flifetime-dse is more aggressive in dead-store elimination in 2733 situations where a memory store to a location precedes a 2734 constructor to that memory location. 2735 * G++ now supports [12]C++17 fold expressions, u8 character literals, 2736 extended static_assert, and nested namespace definitions. 2737 * G++ now allows constant evaluation for all non-type template 2738 arguments. 2739 * G++ now supports C++ Transactional Memory when compiling with 2740 -fgnu-tm. 2741 2742 Runtime Library (libstdc++) 2743 2744 * Extensions to the C++ Library to support mathematical special 2745 functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland. 2746 * Experimental support for C++17, including the following new 2747 features: 2748 + std::uncaught_exceptions function (this is also available for 2749 -std=gnu++NN modes); 2750 + new member functions try_emplace and insert_or_assign for 2751 unique_key maps; 2752 + non-member functions std::size, std::empty, and std::data for 2753 accessing containers and arrays; 2754 + std::invoke; 2755 + std::shared_mutex; 2756 + std::void_t and std::bool_constant metaprogramming utilities. 2757 Thanks to Ville Voutilainen for contributing many of the C++17 2758 features. 2759 * An experimental implementation of the File System TS. 2760 * Experimental support for most features of the second version of the 2761 Library Fundamentals TS. This includes polymorphic memory resources 2762 and array support in shared_ptr, thanks to Fan You. 2763 * Some assertions checked by Debug Mode can now also be enabled by 2764 _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro 2765 have less run-time overhead than the full _GLIBCXX_DEBUG checks and 2766 don't affect the library ABI, so can be enabled per-translation 2767 unit. 2768 * Timed mutex types are supported on more targets, including Darwin. 2769 * Improved std::locale support for DragonFly and FreeBSD, thanks to 2770 John Marino and Andreas Tobler. 2771 2772 Fortran 2773 2774 * Fortran 2008 SUBMODULE support. 2775 * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY 2776 support. 2777 * Improved support for Fortran 2003 deferred-length character 2778 variables. 2779 * Improved support for OpenMP and OpenACC. 2780 * The MATMUL intrinsic is now inlined for straightforward cases if 2781 front-end optimization is active. The maximum size for inlining can 2782 be set to n with the -finline-matmul-limit=n option and turned off 2783 with -finline-matmul-limit=0. 2784 * The -Wconversion-extra option will warn about REAL constants which 2785 have excess precision for their kind. 2786 * The -Winteger-division option has been added, which warns about 2787 divisions of integer constants which are truncated. This option is 2788 included in -Wall by default. 2789 2790 libgccjit 2791 2792 * The driver code is now run in-process within libgccjit, providing a 2793 small speed-up of the compilation process. 2794 * The API has gained entrypoints for 2795 + [13]timing how long was spent in different parts of code, 2796 + [14]creating switch statements, 2797 + [15]allowing unreachable basic blocks in a function, and 2798 + [16]adding arbitrary command-line options to a compilation. 2799 2800 New Targets and Target Specific Improvements 2801 2802 AArch64 2803 2804 * A number of AArch64-specific options have been added. The most 2805 important ones are summarised in this section; for more detailed 2806 information please refer to the documentation. 2807 * The command-line options -march=native, -mcpu=native and 2808 -mtune=native are now available on native AArch64 GNU/Linux 2809 systems. Specifying these options causes GCC to auto-detect the 2810 host CPU and choose the optimal setting for that system. 2811 * -fpic is now supported when generating code for the small code 2812 model (-mcmodel=small). The size of the global offset table (GOT) 2813 is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the 2814 ILP32 SysV ABI. 2815 * The AArch64 port now supports target attributes and pragmas. Please 2816 refer to the [17]documentation for details of available attributes 2817 and pragmas as well as usage instructions. 2818 * Link-time optimization across translation units with different 2819 target-specific options is now supported. 2820 * The option -mtls-size= is now supported. It can be used to specify 2821 the bit size of TLS offsets, allowing GCC to generate better TLS 2822 instruction sequences. 2823 * The option -fno-plt is now fully functional. 2824 * The ARMv8.1-A architecture and the Large System Extensions are now 2825 supported. They can be used by specifying the -march=armv8.1-a 2826 option. Additionally, the +lse option extension can be used in a 2827 similar fashion to other option extensions. The Large System 2828 Extensions introduce new instructions that are used in the 2829 implementation of atomic operations. 2830 * The ACLE half-precision floating-point type __fp16 is now supported 2831 in the C and C++ languages. 2832 * The ARM Cortex-A35 processor is now supported via the 2833 -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the 2834 equivalent target attributes and pragmas. 2835 * The Qualcomm QDF24xx processor is now supported via the 2836 -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent 2837 target attributes and pragmas. 2838 * Code generation for the ARM Cortex-A57 processor is improved. Among 2839 general code generation improvements, a better algorithm is added 2840 for allocating registers to floating-point multiply-accumulate 2841 instructions offering increased performance when compiling with 2842 -mcpu=cortex-a57 or -mtune=cortex-a57. 2843 * Code generation for the ARM Cortex-A53 processor is improved. A 2844 more accurate instruction scheduling model for the processor is now 2845 used, and a number of compiler tuning parameters have been set to 2846 offer increased performance when compiling with -mcpu=cortex-a53 or 2847 -mtune=cortex-a53. 2848 * Code generation for the Samsung Exynos M1 processor is improved. A 2849 more accurate instruction scheduling model for the processor is now 2850 used, and a number of compiler tuning parameters have been set to 2851 offer increased performance when compiling with -mcpu=exynos-m1 or 2852 -mtune=exynos-m1. 2853 * Improvements in the generation of conditional branches and literal 2854 pools allow the compiler to compile functions of a large size. 2855 Constant pools are now placed into separate rodata sections. The 2856 new option -mpc-relative-literal-loads generates per-function 2857 literal pools, limiting the maximum size of functions to 1MiB. 2858 * Several correctness issues generating Advanced SIMD instructions 2859 for big-endian targets have been fixed resulting in improved code 2860 generation for ACLE intrinsics with -mbig-endian. 2861 2862 ARM 2863 2864 * Support for revisions of the ARM architecture prior to ARMv4t has 2865 been deprecated and will be removed in a future GCC release. The 2866 -mcpu and -mtune values that are deprecated are: arm2, arm250, 2867 arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, 2868 arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, 2869 arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm, 2870 strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value 2871 arm7tdmi is still supported. The values of -march that are 2872 deprecated are: armv2,armv2a,armv3,armv3m,armv4. 2873 * The ARM port now supports target attributes and pragmas. Please 2874 refer to the [18]documentation for details of available attributes 2875 and pragmas as well as usage instructions. 2876 * Support has been added for the following processors (GCC 2877 identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM 2878 Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC 2879 identifiers can be used as arguments to the -mcpu or -mtune 2880 options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35. 2881 2882 Heterogeneous Systems Architecture 2883 2884 * GCC can now generate HSAIL (Heterogeneous System Architecture 2885 Intermediate Language) for simple OpenMP device constructs if 2886 configured with --enable-offload-targets=hsa. A new libgomp plugin 2887 then runs the HSA GPU kernels implementing these constructs on HSA 2888 capable GPUs via a standard HSA run time. 2889 If the HSA compilation back end determines it cannot output HSAIL 2890 for a particular input, it gives a warning by default. These 2891 warnings can be suppressed with -Wno-hsa. To give a few examples, 2892 the HSA back end does not implement compilation of code using 2893 function pointers, automatic allocation of variable sized arrays, 2894 functions with variadic arguments as well as a number of other less 2895 common programming constructs. 2896 When compilation for HSA is enabled, the compiler attempts to 2897 compile composite OpenMP constructs 2898 2899 #pragma omp target teams distribute parallel for 2900 2901 into parallel HSA GPU kernels. 2902 2903 IA-32/x86-64 2904 2905 * GCC now supports the Intel CPU named Skylake with AVX-512 2906 extensions through -march=skylake-avx512. The switch enables the 2907 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 2908 AVX-512DQ. 2909 * Support for new AMD instructions monitorx and mwaitx has been 2910 added. This includes new intrinsic and built-in support. It is 2911 enabled through option -mmwaitx. The instructions monitorx and 2912 mwaitx implement the same functionality as the old monitor and 2913 mwait instructions. In addition mwaitx adds a configurable timer. 2914 The timer value is received as third argument and stored in 2915 register %ebx. 2916 * x86-64 targets now allow stack realignment from a word-aligned 2917 stack pointer using the command-line option -mstackrealign or 2918 __attribute__ ((force_align_arg_pointer)). This allows functions 2919 compiled with a vector-aligned stack to be invoked from objects 2920 that keep only word-alignment. 2921 * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These 2922 can be used to access data via the %fs and %gs segments without 2923 having to resort to inline assembly. Please refer to the 2924 [19]documentation for usage instructions. 2925 * Support for AMD Zen (family 17h) processors is now available 2926 through the -march=znver1 and -mtune=znver1 options. 2927 2928 MeP 2929 2930 * Support for the MeP (mep-elf) architecture has been deprecated and 2931 will be removed in a future GCC release. 2932 2933 MSP430 2934 2935 * The MSP430 compiler now has the ability to automatically distribute 2936 code and data between low memory (addresses below 64K) and high 2937 memory. This only applies to parts that actually have both memory 2938 regions and only if the linker script for the part has been 2939 specifically set up to support this feature. 2940 A new attribute of either can be applied to both functions and 2941 data, and this tells the compiler to place the object into low 2942 memory if there is room and into high memory otherwise. Two other 2943 new attributes - lower and upper - can be used to explicitly state 2944 that an object should be placed in the specified memory region. If 2945 there is not enough left in that region the compilation will fail. 2946 Two new command-line options - -mcode-region=[lower|upper|either] 2947 and -mdata-region=[lower|upper|either] - can be used to tell the 2948 compiler what to do with objects that do not have one of these new 2949 attributes. 2950 2951 PowerPC / PowerPC64 / RS6000 2952 2953 * PowerPC64 now supports IEEE 128-bit floating-point using the 2954 __float128 data type. In GCC 6, this is not enabled by default, but 2955 you can enable it with -mfloat128. The IEEE 128-bit floating-point 2956 support requires the use of the VSX instruction set. IEEE 128-bit 2957 floating-point values are passed and returned as a single vector 2958 value. The software emulator for IEEE 128-bit floating-point 2959 support is only built on PowerPC GNU/Linux systems where the 2960 default CPU is at least power7. On future ISA 3.0 systems (POWER 9 2961 and later), you will be able to use the -mfloat128-hardware option 2962 to use the ISA 3.0 instructions that support IEEE 128-bit 2963 floating-point. An additional type (__ibm128) has been added to 2964 refer to the IBM extended double type that normally implements long 2965 double. This will allow for a future transition to implementing 2966 long double with IEEE 128-bit floating-point. 2967 * Basic support has been added for POWER9 hardware that will use the 2968 recently published OpenPOWER ISA 3.0 instructions. The following 2969 new switches are available: 2970 + -mcpu=power9: Implement all of the ISA 3.0 instructions 2971 supported by the compiler. 2972 + -mtune=power9: In the future, apply tuning for POWER9 systems. 2973 Currently, POWER8 tunings are used. 2974 + -mmodulo: Generate code using the ISA 3.0 integer instructions 2975 (modulus, count trailing zeros, array index support, integer 2976 multiply/add). 2977 + -mpower9-fusion: Generate code to suitably fuse instruction 2978 sequences for a POWER9 system. 2979 + -mpower9-dform: Generate code to use the new D-form 2980 (register+offset) memory instructions for the vector 2981 registers. 2982 + -mpower9-vector: Generate code using the new ISA 3.0 vector 2983 (VSX or Altivec) instructions. 2984 + -mpower9-minmax: Reserved for future development. 2985 + -mtoc-fusion: Keep TOC entries together to provide more fusion 2986 opportunities. 2987 * New constraints have been added to support IEEE 128-bit 2988 floating-point and ISA 3.0 instructions: 2989 + wb: Altivec register if -mpower9-dform is enabled. 2990 + we: VSX register if -mpower9-vector is enabled for 64-bit code 2991 generation. 2992 + wo: VSX register if -mpower9-vector is enabled. 2993 + wp: Reserved for future use if long double is implemented with 2994 IEEE 128-bit floating-point instead of IBM extended double. 2995 + wq: VSX register if -mfloat128 is enabled. 2996 + wF: Memory operand suitable for POWER9 fusion load/store. 2997 + wG: Memory operand suitable for TOC fusion memory references. 2998 + wL: Integer constant identifying the element number mfvsrld 2999 accesses within a vector. 3000 * Support has been added for __builtin_cpu_is() and 3001 __builtin_cpu_supports(), allowing for very fast access to 3002 AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of 3003 glibc 2.23 or later. 3004 * All hardware transactional memory builtins now correctly behave as 3005 memory barriers. Programmers can use #ifdef __TM_FENCE__ to 3006 determine whether their "old" compiler treats the builtins as 3007 barriers. 3008 * Split-stack support has been added for gccgo on PowerPC64 for both 3009 big- and little-endian (but not for 32-bit). The gold linker from 3010 at least binutils 2.25.1 must be available in the PATH when 3011 configuring and building gccgo to enable split stack. (The 3012 requirement for binutils 2.25.1 applies to PowerPC64 only.) The 3013 split-stack feature allows a small initial stack size to be 3014 allocated for each goroutine, which increases as needed. 3015 * GCC on PowerPC now supports the standard lround function. 3016 * A new configuration option ---with-advance-toolchain=at was added 3017 for PowerPC 64-bit GNU/Linux systems to use the header files, 3018 library files, and the dynamic linker from a specific Advance 3019 Toolchain release instead of the default versions that are provided 3020 by the GNU/Linux distribution. In general, this option is intended 3021 for the developers of GCC, and it is not intended for general use. 3022 * The "q", "S", "T", and "t" asm-constraints have been removed. 3023 * The "b", "B", "m", "M", and "W" format modifiers have been removed. 3024 3025 S/390, System z, IBM z Systems 3026 3027 * Support for the IBM z13 processor has been added. When using the 3028 -march=z13 option, the compiler will generate code making use of 3029 the new instructions and registers introduced with the vector 3030 extension facility. The -mtune=z13 option enables z13 specific 3031 instruction scheduling without making use of new instructions. 3032 Compiling code with -march=z13 reduces the default alignment of 3033 vector types bigger than 8 bytes to 8. This is an ABI change and 3034 care must be taken when linking modules compiled with different 3035 arch levels which interchange variables containing vector type 3036 values. For newly compiled code the GNU linker will emit a warning. 3037 * The -mzvector option enables a C/C++ language extension. This 3038 extension provides a new keyword vector which can be used to define 3039 vector type variables. (Note: This is not available when enforcing 3040 strict standard compliance e.g. with -std=c99. Either enable GNU 3041 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 3042 Additionally a set of overloaded builtins is provided which is 3043 partially compatible to the PowerPC Altivec builtins. In order to 3044 make use of these builtins the vecintrin.h header file needs to be 3045 included. 3046 * The new command-line options -march=native, and -mtune=native are 3047 now available on native IBM z Systems. Specifying these options 3048 causes GCC to auto-detect the host CPU and choose the optimal 3049 setting for that system. If GCC is unable to detect the host CPU 3050 these options have no effect. 3051 * The IBM z Systems port now supports target attributes and pragmas. 3052 Please refer to the [20]documentation for details of available 3053 attributes and pragmas as well as usage instructions. 3054 * -fsplit-stack is now supported as part of the IBM z Systems port. 3055 This feature requires a recent gold linker to be used. 3056 * Support for the g5 and g6 -march=/-mtune= CPU level switches has 3057 been deprecated and will be removed in a future GCC release. -m31 3058 from now on defaults to -march=z900 if not specified otherwise. 3059 -march=native on a g5/g6 machine will default to -march=z900. 3060 3061 SH 3062 3063 * Support for SH5 / SH64 has been declared obsolete and will be 3064 removed in future releases. 3065 * Support for the FDPIC ABI has been added. It can be enabled using 3066 the new -mfdpic target option and --enable-fdpic configure option. 3067 3068 SPARC 3069 3070 * An ABI bug has been fixed in 64-bit mode. Unfortunately, this 3071 change will break binary compatibility with earlier releases for 3072 code it affects, but this should be pretty rare in practice. The 3073 conditions are: a 16-byte structure containing a double or a 8-byte 3074 vector in the second half is passed to a subprogram in slot #15, 3075 for example as 16th parameter if the first 15 ones have at most 8 3076 bytes. The double or vector was wrongly passed in floating-point 3077 register %d32 in lieu of on the stack as per the SPARC calling 3078 conventions. 3079 3080 Operating Systems 3081 3082 AIX 3083 3084 * DWARF debugging support for AIX 7.1 has been enabled as an optional 3085 debugging format. A more recent Technology Level (TL) and GCC built 3086 with that level are required for full exploitation of DWARF 3087 debugging capabilities. 3088 3089 Linux 3090 3091 * Support for the [21]musl C library was added for the AArch64, ARM, 3092 MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and 3093 x86_64 targets. It can be selected using the new -mmusl option in 3094 case musl is not the default libc. GCC defaults to musl libc if it 3095 is built with a target triplet matching the *-linux-musl* pattern. 3096 3097 RTEMS 3098 3099 * The RTEMS thread model implementation changed. Mutexes now use 3100 self-contained objects defined in Newlib <sys/lock.h> instead of 3101 Classic API semaphores. The keys for thread specific data and the 3102 once function are directly defined via <pthread.h>. Self-contained 3103 condition variables are provided via Newlib <sys/lock.h>. The RTEMS 3104 thread model also supports C++11 threads. 3105 * OpenMP support now uses self-contained objects provided by Newlib 3106 <sys/lock.h> and offers a significantly better performance compared 3107 to the POSIX configuration of libgomp. It is possible to configure 3108 thread pools for each scheduler instance via the environment 3109 variable GOMP_RTEMS_THREAD_POOLS. 3110 3111 Solaris 3112 3113 * Solaris 12 is now fully supported. Minimal support had already been 3114 present in GCC 5.3. 3115 * Solaris 12 provides a full set of startup files (crt1.o, crti.o, 3116 crtn.o), which GCC now prefers over its own ones. 3117 * Position independent executables (PIE) are now supported on Solaris 3118 12. 3119 * Constructor priority is now supported on Solaris 12 with the system 3120 linker. 3121 * libvtv has been ported to Solaris 11 and up. 3122 3123 Windows 3124 3125 * The option -mstackrealign is now automatically activated in 32-bit 3126 mode whenever the use of SSE instructions is requested. 3127 3128 Other significant improvements 3129 3130 * The gcc and g++ driver programs will now provide suggestions for 3131 misspelled command-line options. 3132 3133 $ gcc -static-libfortran test.f95 3134 gcc: error: unrecognized command line option '-static-libfortran'; did you mean 3135 '-static-libgfortran'? 3136 3137 * The --enable-default-pie configure option enables generation of PIE 3138 by default. 3139 3140 GCC 6.2 3141 3142 This is the [22]list of problem reports (PRs) from GCC's bug tracking 3143 system that are known to be fixed in the 6.2 release. This list might 3144 not be complete (that is, it is possible that some PRs that have been 3145 fixed are not listed here). 3146 3147 Target Specific Changes 3148 3149 SPARC 3150 3151 * Support for --with-cpu-32 and --with-cpu-64 configure options has 3152 been added on bi-architecture platforms. 3153 * Support for the SPARC M7 (Niagara 7) processor has been added. 3154 * Support for the VIS 4.0 instruction set has been added. 3155 3156 GCC 6.3 3157 3158 This is the [23]list of problem reports (PRs) from GCC's bug tracking 3159 system that are known to be fixed in the 6.3 release. This list might 3160 not be complete (that is, it is possible that some PRs that have been 3161 fixed are not listed here). 3162 3163 Target Specific Changes 3164 3165 IA-32/x86-64 3166 3167 * Support for the [24]deprecated pcommit instruction has been 3168 removed. 3169 3170 GCC 6.4 3171 3172 This is the [25]list of problem reports (PRs) from GCC's bug tracking 3173 system that are known to be fixed in the 6.4 release. This list might 3174 not be complete (that is, it is possible that some PRs that have been 3175 fixed are not listed here). 3176 3177 Operating Systems 3178 3179 RTEMS 3180 3181 * The ABI changes on ARM so that no short enums are used by default. 3182 3183 GCC 6.5 3184 3185 This is the [26]list of problem reports (PRs) from GCC's bug tracking 3186 system that are known to be fixed in the 6.5 release. This list might 3187 not be complete (that is, it is possible that some PRs that have been 3188 fixed are not listed here). 3189 3190 3191 For questions related to the use of GCC, please consult these web 3192 pages and the [27]GCC manuals. If that fails, the 3193 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3194 web pages and the development of GCC are welcome on our developer 3195 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 3196 archives. 3197 3198 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 3199 distribution of this entire article is permitted in any medium, 3200 provided this notice is preserved. 3201 3202 These pages are [32]maintained by the GCC team. Last modified 3203 2019-11-28[33]. 3204 3205 References 3206 3207 1. http://gcc.gnu.org/gcc-6/porting_to.html 3208 2. http://gcc.gnu.org/onlinedocs/index.html#current 3209 3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html 3210 4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151 3211 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822 3212 6. https://www.openacc.org/ 3213 7. https://gcc.gnu.org/wiki/OpenACC 3214 8. https://gcc.gnu.org/wiki/Offloading 3215 9. https://www.openmp.org/specifications/ 3216 10. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266 3217 11. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf 3218 12. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 3219 13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html 3220 14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch 3221 15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks 3222 16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option 3223 17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes 3224 18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes 3225 19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces 3226 20. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes 3227 21. http://www.musl-libc.org/ 3228 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2 3229 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3 3230 24. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction 3231 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4 3232 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5 3233 27. https://gcc.gnu.org/onlinedocs/ 3234 28. mailto:gcc-help (a] gcc.gnu.org 3235 29. mailto:gcc (a] gcc.gnu.org 3236 30. https://gcc.gnu.org/lists.html 3237 31. https://www.fsf.org/ 3238 32. https://gcc.gnu.org/about.html 3239 33. http://validator.w3.org/check/referer 3240 ====================================================================== 3241 http://gcc.gnu.org/gcc-5/index.html 3242 GCC 5 Release Series 3243 3244 (This release series is no longer supported.) 3245 3246 October 10, 2017 3247 3248 The [1]GNU project and the GCC developers are pleased to announce the 3249 release of GCC 5.5. 3250 3251 This release is a bug-fix release, containing fixes for regressions in 3252 GCC 5.4 relative to previous releases of GCC. 3253 3254 Release History 3255 3256 GCC 5.5 3257 October 10, 2017 ([2]changes, [3]documentation) 3258 3259 GCC 5.4 3260 June 3, 2016 ([4]changes, [5]documentation) 3261 3262 GCC 5.3 3263 December 4, 2015 ([6]changes, [7]documentation) 3264 3265 GCC 5.2 3266 July 16, 2015 ([8]changes, [9]documentation) 3267 3268 GCC 5.1 3269 April 22, 2015 ([10]changes, [11]documentation) 3270 3271 References and Acknowledgements 3272 3273 GCC used to stand for the GNU C Compiler, but since the compiler 3274 supports several other languages aside from C, it now stands for the 3275 GNU Compiler Collection. 3276 3277 A list of [12]successful builds is updated as new information becomes 3278 available. 3279 3280 The GCC developers would like to thank the numerous people that have 3281 contributed new features, improvements, bug fixes, and other changes as 3282 well as test results to GCC. This [13]amazing group of volunteers is 3283 what makes GCC successful. 3284 3285 For additional information about GCC please refer to the [14]GCC 3286 project web site or contact the [15]GCC development mailing list. 3287 3288 To obtain GCC please use [16]our mirror sites or [17]our version 3289 control system. 3290 3291 3292 For questions related to the use of GCC, please consult these web 3293 pages and the [18]GCC manuals. If that fails, the 3294 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3295 web pages and the development of GCC are welcome on our developer 3296 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 3297 archives. 3298 3299 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 3300 distribution of this entire article is permitted in any medium, 3301 provided this notice is preserved. 3302 3303 These pages are [23]maintained by the GCC team. Last modified 3304 2020-01-14[24]. 3305 3306 References 3307 3308 1. http://www.gnu.org/ 3309 2. http://gcc.gnu.org/gcc-5/changes.html 3310 3. http://gcc.gnu.org/onlinedocs/5.5.0/ 3311 4. http://gcc.gnu.org/gcc-5/changes.html 3312 5. http://gcc.gnu.org/onlinedocs/5.4.0/ 3313 6. http://gcc.gnu.org/gcc-5/changes.html 3314 7. http://gcc.gnu.org/onlinedocs/5.3.0/ 3315 8. http://gcc.gnu.org/gcc-5/changes.html 3316 9. http://gcc.gnu.org/onlinedocs/5.2.0/ 3317 10. http://gcc.gnu.org/gcc-5/changes.html 3318 11. http://gcc.gnu.org/onlinedocs/5.1.0/ 3319 12. http://gcc.gnu.org/gcc-5/buildstat.html 3320 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 3321 14. http://gcc.gnu.org/index.html 3322 15. mailto:gcc (a] gcc.gnu.org 3323 16. http://gcc.gnu.org/mirrors.html 3324 17. http://gcc.gnu.org/git.html 3325 18. https://gcc.gnu.org/onlinedocs/ 3326 19. mailto:gcc-help (a] gcc.gnu.org 3327 20. mailto:gcc (a] gcc.gnu.org 3328 21. https://gcc.gnu.org/lists.html 3329 22. https://www.fsf.org/ 3330 23. https://gcc.gnu.org/about.html 3331 24. http://validator.w3.org/check/referer 3332 ====================================================================== 3333 http://gcc.gnu.org/gcc-5/changes.html 3334 GCC 5 Release Series 3335 Changes, New Features, and Fixes 3336 3337 Caveats 3338 3339 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 3340 * The C++ runtime library (libstdc++) uses a new ABI by default (see 3341 [1]below). 3342 * The Graphite framework for loop optimizations no longer requires 3343 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 3344 The installation manual contains more information about 3345 requirements to build GCC. 3346 * The non-standard C++0x type traits has_trivial_default_constructor, 3347 has_trivial_copy_constructor and has_trivial_copy_assign have been 3348 deprecated and will be removed in a future version. The standard 3349 C++11 traits is_trivially_default_constructible, 3350 is_trivially_copy_constructible and is_trivially_copy_assignable 3351 should be used instead. 3352 * On AVR, support has been added for the devices 3353 ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer. 3354 * The AVR port uses a new scheme to describe supported devices: For 3355 each supported device the compiler provides a device-specific 3356 [2]spec file. If the compiler is used together with AVR-LibC, this 3357 requires at least GCC 5.2 and a version of AVR-LibC which 3358 implements [3]feature #44574. 3359 3360 General Optimizer Improvements 3361 3362 * Inter-procedural optimization improvements: 3363 + An Identical Code Folding (ICF) pass (controlled via 3364 -fipa-icf) has been added. Compared to the identical code 3365 folding performed by the Gold linker this pass does not 3366 require function sections. It also performs merging before 3367 inlining, so inter-procedural optimizations are aware of the 3368 code re-use. On the other hand not all unifications performed 3369 by a linker are doable by GCC which must honor aliasing 3370 information. During link-time optimization of Firefox, this 3371 pass unifies about 31000 functions, that is 14% overall. 3372 + The devirtualization pass was significantly improved by adding 3373 better support for speculative devirtualization and dynamic 3374 type detection. About 50% of virtual calls in Firefox are now 3375 speculatively devirtualized during link-time optimization. 3376 + A new comdat localization pass allows the linker to eliminate 3377 more dead code in presence of C++ inline functions. 3378 + Virtual tables are now optimized. Local aliases are used to 3379 reduce dynamic linking time of C++ virtual tables on ELF 3380 targets and data alignment has been reduced to limit data 3381 segment bloat. 3382 + A new -fno-semantic-interposition option can be used to 3383 improve code quality of shared libraries where interposition 3384 of exported symbols is not allowed. 3385 + Write-only variables are now detected and optimized out. 3386 + With profile feedback the function inliner can now bypass 3387 --param inline-insns-auto and --param inline-insns-single 3388 limits for hot calls. 3389 + The IPA reference pass was significantly sped up making it 3390 feasible to enable -fipa-reference with -fprofile-generate. 3391 This also solves a bottleneck seen when building Chromium with 3392 link-time optimization. 3393 + The symbol table and call-graph API was reworked to C++ and 3394 simplified. 3395 + The interprocedural propagation of constants now also 3396 propagates alignments of pointer parameters. This for example 3397 means that the vectorizer often does not need to generate loop 3398 prologues and epilogues to make up for potential 3399 misalignments. 3400 * Link-time optimization improvements: 3401 + One Definition Rule based merging of C++ types has been 3402 implemented. Type merging enables better devirtualization and 3403 alias analysis. Streaming extra information needed to merge 3404 types adds about 2-6% of memory size and object size increase. 3405 This can be controlled by -flto-odr-type-merging. 3406 + Command-line optimization and target options are now streamed 3407 on a per-function basis and honored by the link-time 3408 optimizer. This change makes link-time optimization a more 3409 transparent replacement of per-file optimizations. It is now 3410 possible to build projects that require different optimization 3411 settings for different translation units (such as -ffast-math, 3412 -mavx, or -finline). Contrary to earlier GCC releases, the 3413 optimization and target options passed on the link command 3414 line are ignored. 3415 Note that this applies only to those command-line options that 3416 can be passed to optimize and target attributes. Command-line 3417 options affecting global code generation (such as -fpic), 3418 warnings (such as -Wodr), optimizations affecting the way 3419 static variables are optimized (such as -fcommon), debug 3420 output (such as -g), and --param parameters can be applied 3421 only to the whole link-time optimization unit. In these cases, 3422 it is recommended to consistently use the same options at both 3423 compile time and link time. 3424 + GCC bootstrap now uses slim LTO object files. 3425 + Memory usage and link times were improved. Tree merging was 3426 sped up, memory usage of GIMPLE declarations and types was 3427 reduced, and, support for on-demand streaming of variable 3428 constructors was added. 3429 * Feedback directed optimization improvements: 3430 + A new auto-FDO mode uses profiles collected by low overhead 3431 profiling tools (perf) instead of more expensive program 3432 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 3433 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 3434 traditional feedback directed optimization. 3435 + Profile precision was improved in presence of C++ inline and 3436 extern inline functions. 3437 + The new gcov-tool utility allows manipulating profiles. 3438 + Profiles are now more tolerant to source file changes (this 3439 can be controlled by --param profile-func-internal-id). 3440 * Register allocation improvements: 3441 + A new local register allocator (LRA) sub-pass, controlled by 3442 -flra-remat, implements control-flow sensitive global register 3443 rematerialization. Instead of spilling and restoring a 3444 register value, it is recalculated if it is profitable. The 3445 sub-pass improved SPEC2000 generated code by 1% and 0.5% 3446 correspondingly on ARM and x86-64. 3447 + Reuse of the PIC hard register, instead of using a fixed 3448 register, was implemented on x86/x86-64 targets. This improves 3449 generated PIC code performance as more hard registers can be 3450 used. Shared libraries can significantly benefit from this 3451 optimization. Currently it is switched on only for x86/x86-64 3452 targets. As RA infrastructure is already implemented for PIC 3453 register reuse, other targets might follow this in the future. 3454 + A simple form of inter-procedural RA was implemented. When it 3455 is known that a called function does not use caller-saved 3456 registers, save/restore code is not generated around the call 3457 for such registers. This optimization can be controlled by 3458 -fipa-ra 3459 + LRA is now much more effective at generating spills of general 3460 registers into vector registers instead of memory on 3461 architectures (e.g., modern Intel processors) where this is 3462 profitable. 3463 * UndefinedBehaviorSanitizer gained a few new sanitization options: 3464 + -fsanitize=float-divide-by-zero: detect floating-point 3465 division by zero; 3466 + -fsanitize=float-cast-overflow: check that the result of 3467 floating-point type to integer conversions do not overflow; 3468 + -fsanitize=bounds: enable instrumentation of array bounds and 3469 detect out-of-bounds accesses; 3470 + -fsanitize=alignment: enable alignment checking, detect 3471 various misaligned objects; 3472 + -fsanitize=object-size: enable object size checking, detect 3473 various out-of-bounds accesses. 3474 + -fsanitize=vptr: enable checking of C++ member function calls, 3475 member accesses and some conversions between pointers to base 3476 and derived classes, detect if the referenced object does not 3477 have the correct dynamic type. 3478 * Pointer Bounds Checker, a bounds violation detector, has been added 3479 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 3480 instrumented with run-time checks of used pointers against their 3481 bounds to detect pointer bounds violations (overflows). The Pointer 3482 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 3483 new ISA extension Intel MPX support. See the Pointer Bounds Checker 3484 [4]Wiki page for more details. 3485 3486 New Languages and Language specific improvements 3487 3488 * [5]OpenMP 4.0 specification offloading features are now supported 3489 by the C, C++, and Fortran compilers. Generic changes: 3490 + Infrastructure (suitable for any vendor). 3491 + Testsuite which covers offloading from the [6]OpenMP 4.0 3492 Examples document. 3493 Specific for upcoming Intel Xeon Phi products: 3494 + Run-time library. 3495 + Card emulator. 3496 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 3497 specification. OpenACC is intended for programming accelerator 3498 devices such as GPUs. See [7]the OpenACC wiki page for more 3499 information. 3500 3501 C family 3502 3503 * The default setting of the -fdiagnostics-color= command-line option 3504 is now [8]configurable when building GCC using configuration option 3505 --with-diagnostics-color=. The possible values are: never, always, 3506 auto and auto-if-env. The new default auto uses color only when the 3507 standard error is a terminal. The default in GCC 4.9 was 3508 auto-if-env, which is equivalent to auto if there is a non-empty 3509 GCC_COLORS environment variable, and never otherwise. As in GCC 3510 4.9, an empty GCC_COLORS variable in the environment will always 3511 disable colors, no matter what the default is or what command-line 3512 options are used. 3513 * A new command-line option -Wswitch-bool has been added for the C 3514 and C++ compilers, which warns whenever a switch statement has an 3515 index of boolean type. 3516 * A new command-line option -Wlogical-not-parentheses has been added 3517 for the C and C++ compilers, which warns about "logical not" used 3518 on the left hand side operand of a comparison. 3519 * A new command-line option -Wsizeof-array-argument has been added 3520 for the C and C++ compilers, which warns when the sizeof operator 3521 is applied to a parameter that has been declared as an array in a 3522 function definition. 3523 * A new command-line option -Wbool-compare has been added for the C 3524 and C++ compilers, which warns about boolean expressions compared 3525 with an integer value different from true/false. 3526 * Full support for [9]Cilk Plus has been added to the GCC compiler. 3527 Cilk Plus is an extension to the C and C++ languages to support 3528 data and task parallelism. 3529 * A new attribute no_reorder prevents reordering of selected symbols 3530 against other such symbols or inline assembler. This enables to 3531 link-time optimize the Linux kernel without having to resort to 3532 -fno-toplevel-reorder that disables several optimizations. 3533 * New preprocessor constructs, __has_include and __has_include_next, 3534 to test the availability of headers have been added. 3535 This demonstrates a way to include the header <optional> only if it 3536 is available: 3537 3538 #ifdef __has_include 3539 # if __has_include(<optional>) 3540 # include <optional> 3541 # define have_optional 1 3542 # elif __has_include(<experimental/optional>) 3543 # include <experimental/optional> 3544 # define have_optional 1 3545 # define experimental_optional 3546 # else 3547 # define have_optional 0 3548 # endif 3549 #endif 3550 3551 The header search paths for __has_include and __has_include_next 3552 are equivalent to those of the standard directive #include and the 3553 extension #include_next respectively. 3554 * A new built-in function-like macro to determine the existence of an 3555 attribute, __has_attribute, has been added. The equivalent built-in 3556 macro __has_cpp_attribute was added to C++ to support 3557 [10]Feature-testing recommendations for C++. The macro 3558 __has_attribute is added to all C-like languages as an extension: 3559 3560 int 3561 #ifdef __has_attribute 3562 # if __has_attribute(__noinline__) 3563 __attribute__((__noinline__)) 3564 # endif 3565 #endif 3566 foo(int x); 3567 3568 If an attribute exists, a nonzero constant integer is returned. For 3569 standardized C++ attributes a date is returned, otherwise the 3570 constant returned is 1. Both __has_attribute and 3571 __has_cpp_attribute will add underscores to an attribute name if 3572 necessary to resolve the name. For C++11 and onwards the attribute 3573 may be scoped. 3574 * A new set of built-in functions for arithmetics with overflow 3575 checking has been added: __builtin_add_overflow, 3576 __builtin_sub_overflow and __builtin_mul_overflow and for 3577 compatibility with clang also other variants. These builtins have 3578 two integral arguments (which don't need to have the same type), 3579 the arguments are extended to infinite precision signed type, +, - 3580 or * is performed on those, and the result is stored in an integer 3581 variable pointed to by the last argument. If the stored value is 3582 equal to the infinite precision result, the built-in functions 3583 return false, otherwise true. The type of the integer variable that 3584 will hold the result can be different from the types of the first 3585 two arguments. The following snippet demonstrates how this can be 3586 used in computing the size for the calloc function: 3587 3588 void * 3589 calloc (size_t x, size_t y) 3590 { 3591 size_t sz; 3592 if (__builtin_mul_overflow (x, y, &sz)) 3593 return NULL; 3594 void *ret = malloc (sz); 3595 if (ret) memset (res, 0, sz); 3596 return ret; 3597 } 3598 3599 On e.g. i?86 or x86-64 the above will result in a mul instruction 3600 followed by a jump on overflow. 3601 * The option -fextended-identifiers is now enabled by default for 3602 C++, and for C99 and later C versions. Various bugs in the 3603 implementation of extended identifiers have been fixed. 3604 3605 C 3606 3607 * The default mode has been changed to -std=gnu11. 3608 * A new command-line option -Wc90-c99-compat has been added to warn 3609 about features not present in ISO C90, but present in ISO C99. 3610 * A new command-line option -Wc99-c11-compat has been added to warn 3611 about features not present in ISO C99, but present in ISO C11. 3612 * It is possible to disable warnings about conversions between 3613 pointers that have incompatible types via a new warning option 3614 -Wno-incompatible-pointer-types; warnings about implicit 3615 incompatible integer to pointer and pointer to integer conversions 3616 via a new warning option -Wno-int-conversion; and warnings about 3617 qualifiers on pointers being discarded via a new warning option 3618 -Wno-discarded-qualifiers. 3619 * To allow proper use of const qualifiers with multidimensional 3620 arrays, GCC will not warn about incompatible pointer types anymore 3621 for conversions between pointers to arrays with and without const 3622 qualifier (except when using -pedantic). Instead, a new warning is 3623 emitted only if the const qualifier is lost. This can be controlled 3624 with a new warning option -Wno-discarded-array-qualifiers. 3625 * The C front end now generates more precise caret diagnostics. 3626 * The -pg command-line option now only affects the current file in an 3627 LTO build. 3628 3629 C++ 3630 3631 * G++ now supports [11]C++14 variable templates. 3632 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 3633 * Excessive template instantiation depth is now a fatal error. This 3634 prevents excessive diagnostics that usually do not help to identify 3635 the problem. 3636 * G++ and libstdc++ now implement the feature-testing macros from 3637 [12]Feature-testing recommendations for C++. 3638 * G++ now allows typename in a template template parameter. 3639 3640 template<template<typename> typename X> struct D; // OK 3641 3642 * G++ now supports [13]C++14 aggregates with non-static data member 3643 initializers. 3644 3645 struct A { int i, j = i; }; 3646 A a = { 42 }; // a.j is also 42 3647 3648 * G++ now supports [14]C++14 extended constexpr. 3649 3650 constexpr int f (int i) 3651 { 3652 int j = 0; 3653 for (; i > 0; --i) 3654 ++j; 3655 return j; 3656 } 3657 3658 constexpr int i = f(42); // i is 42 3659 3660 * G++ now supports the [15]C++14 sized deallocation functions. 3661 3662 void operator delete (void *, std::size_t) noexcept; 3663 void operator delete[] (void *, std::size_t) noexcept; 3664 3665 * A new One Definition Rule violation warning (controlled by -Wodr) 3666 detects mismatches in type definitions and virtual table contents 3667 during link-time optimization. 3668 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 3669 developers to annotate programs with final specifiers (or anonymous 3670 namespaces) to improve code generation. These warnings can be used 3671 at compile time, but they are more useful in combination with 3672 link-time optimization. 3673 * G++ no longer supports [16]N3639 variable length arrays, as they 3674 were removed from the C++14 working paper prior to ratification. 3675 GNU VLAs are still supported, so VLA support is now the same in 3676 C++14 mode as in C++98 and C++11 modes. 3677 * G++ now allows passing a non-trivially-copyable class via C 3678 varargs, which is conditionally-supported with 3679 implementation-defined semantics in the standard. This uses the 3680 same calling convention as a normal value parameter. 3681 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 3682 various mangling bugs are fixed, but G++ will still emit aliases 3683 with the old, wrong mangling where feasible. -Wabi=2 will warn 3684 about differences between ABI version 2 and the current setting. 3685 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 3686 to be unaffected, but -Wabi=8 will warn about a non-static data 3687 member with type std::nullptr_t which changes position due to this 3688 change. 3689 3690 Runtime Library (libstdc++) 3691 3692 * A [17]Dual ABI is provided by the library. A new ABI is enabled by 3693 default. The old ABI is still supported and can be used by defining 3694 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 3695 standard library headers. 3696 * A new implementation of std::string is enabled by default, using 3697 the small string optimization instead of copy-on-write reference 3698 counting. 3699 * A new implementation of std::list is enabled by default, with an 3700 O(1) size() function; 3701 * [18]Full support for C++11, including the following new features: 3702 + std::deque and std::vector<bool> meet the allocator-aware 3703 container requirements; 3704 + movable and swappable iostream classes; 3705 + support for std::align and std::aligned_union; 3706 + type traits std::is_trivially_copyable, 3707 std::is_trivially_constructible, std::is_trivially_assignable 3708 etc.; 3709 + I/O manipulators std::put_time, std::get_time, std::hexfloat 3710 and std::defaultfloat; 3711 + generic locale-aware std::isblank; 3712 + locale facets for Unicode conversion; 3713 + atomic operations for std::shared_ptr; 3714 + std::notify_all_at_thread_exit() and functions for making 3715 futures ready at thread exit. 3716 * Support for the C++11 hexfloat manipulator changes how the num_put 3717 facet formats floating point types when 3718 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 3719 This change affects all language modes, even though the C++98 3720 standard gave no special meaning to that combination of flags. To 3721 prevent the use of hexadecimal notation for floating point types 3722 use str.unsetf(std::ios_base::floatfield) to clear the relevant 3723 bits in str.flags(). 3724 * [19]Full experimental support for C++14, including the following 3725 new features: 3726 + std::is_final type trait; 3727 + heterogeneous comparison lookup in associative containers. 3728 + global functions cbegin, cend, rbegin, rend, crbegin, and 3729 crend for range access to containers, arrays and initializer 3730 lists. 3731 * [20]Improved experimental support for the Library Fundamentals TS, 3732 including: 3733 + class std::experimental::any; 3734 + function template std::experimental::apply; 3735 + function template std::experimental::sample; 3736 + function template std::experimental::search and related 3737 searcher types; 3738 + variable templates for type traits; 3739 + function template std::experimental::not_fn. 3740 * New random number distributions logistic_distribution and 3741 uniform_on_sphere_distribution as extensions. 3742 * [21]GDB Xmethods for containers and std::unique_ptr. 3743 3744 Fortran 3745 3746 * Compatibility notice: 3747 + The version of the module files (.mod) has been incremented. 3748 + For free-form source files [22]-Werror=line-truncation is now 3749 enabled by default. Note that comments exceeding the line 3750 length are not diagnosed. (For fixed-form source code, the 3751 same warning is available but turned off by default, such that 3752 excess characters are ignored. -ffree-line-length-n and 3753 -ffixed-line-length-n can be used to modify the default line 3754 lengths of 132 and 72 columns, respectively.) 3755 + The -Wtabs option is now more sensible: with -Wtabs the 3756 compiler warns if it encounters tabs and with -Wno-tabs this 3757 warning is turned off. Before, -Wno-tabs warned and -Wtabs 3758 disabled the warning. As before, this warning is also enabled 3759 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 3760 options of -std=. 3761 * Incomplete support for colorizing diagnostics emitted by gfortran 3762 has been added. The option [23]-fdiagnostics-color controls when 3763 color is used in diagnostics. The default value of this option can 3764 be [24]configured when building GCC. The GCC_COLORS environment 3765 variable can be used to customize the colors or disable coloring 3766 completely. Sample diagnostics output: 3767 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 3768 test.f90:6:1: 3769 3770 0 continue 3771 1 3772 Error: Zero is not a valid statement label at (1) 3773 test.f90:9:6: 3774 3775 USE foo 3776 1 3777 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 3778 3779 * The -Wuse-without-only option has been added to warn when a USE 3780 statement has no ONLY qualifier and thus implicitly imports all 3781 public entities of the used module. 3782 * Formatted READ and WRITE statements now work correctly in 3783 locale-aware programs. For more information and potential caveats, 3784 see [25]Section 5.3 Thread-safety of the runtime library in the 3785 manual. 3786 * [26]Fortran 2003: 3787 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 3788 IEEE_ARITHMETIC) are now supported. 3789 * [27]Fortran 2008: 3790 + [28]Coarrays: Full experimental support of Fortran 2008's 3791 coarrays with -fcoarray=lib except for allocatable/pointer 3792 components of derived-type coarrays. GCC currently only ships 3793 with a single-image library (libcaf_single), but multi-image 3794 support based on MPI and GASNet is provided by the libraries 3795 of the [29]OpenCoarrays project. 3796 * TS18508 Additional Parallel Features in Fortran: 3797 + Support for the collective intrinsic subroutines CO_MAX, 3798 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 3799 including -fcoarray=lib support. 3800 + Support for the new atomic intrinsics has been added, 3801 including -fcoarray=lib support. 3802 * Fortran 2015: 3803 + Support for IMPLICIT NONE (external, type). 3804 + ERROR STOP is now permitted in pure procedures. 3805 3806 Go 3807 3808 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 3809 * Building GCC 5 with Go enabled will install two new programs: 3810 [30]go and [31]gofmt. 3811 3812 libgccjit 3813 3814 New in GCC 5 is the ability to build GCC as a shared library for 3815 embedding in other processes (such as interpreters), suitable for 3816 Just-In-Time compilation to machine code. 3817 3818 The shared library has a [32]C API and a [33]C++ wrapper API providing 3819 some "syntactic sugar". There are also bindings available from 3rd 3820 parties for [34]Python and for [35]D. 3821 3822 For example, this library can be used by interpreters for [36]compiling 3823 functions from bytecode to machine code. 3824 3825 The library can also be used for ahead-of-time compilation, enabling 3826 GCC to be plugged into a pre-existing front end. An example of using 3827 this to build a compiler for an esoteric language we'll refer to as 3828 "brainf" can be seen [37]here. 3829 3830 libgccjit is licensed under the GPLv3 (or at your option, any later 3831 version) 3832 3833 It should be regarded as experimental at this time. 3834 3835 New Targets and Target Specific Improvements 3836 3837 Reporting stack usage 3838 3839 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 3840 -fstack-usage option. 3841 3842 AArch64 3843 3844 * Code generation for the ARM Cortex-A57 processor has been improved. 3845 A more accurate instruction scheduling model for the processor is 3846 now used, and a number of compiler tuning parameters have been set 3847 to offer increased performance when compiling with -mcpu=cortex-a57 3848 or -mtune=cortex-a57. 3849 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 3850 and can be enabled by giving the -mfix-cortex-a53-835769 option. 3851 Alternatively it can be enabled by default by configuring GCC with 3852 the --enable-fix-cortex-a53-835769 option. 3853 * The optional cryptographic extensions to the ARMv8-A architecture 3854 are no longer enabled by default when specifying the 3855 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 3856 options. To enable these extensions add +crypto to the value of 3857 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 3858 * Support has been added for the following processors (GCC 3859 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 3860 initial support for its big.LITTLE combination with the ARM 3861 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 3862 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 3863 The GCC identifiers can be used as arguments to the -mcpu or -mtune 3864 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 3865 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 3866 support for the Cortex-A72. 3867 * The transitional options -mlra and -mno-lra have been removed. The 3868 AArch64 backend now uses the local register allocator (LRA) only. 3869 3870 ARM 3871 3872 * Thumb-1 assembly code is now generated in unified syntax. The new 3873 option -masm-syntax-unified specifies whether inline assembly code 3874 is using unified syntax. By default the option is off which means 3875 non-unified syntax is used. However this is subject to change in 3876 future releases. Eventually the non-unified syntax will be 3877 deprecated. 3878 * It is now a configure-time error to use the --with-cpu configure 3879 option with either of --with-tune or --with-arch. 3880 * Code generation for the ARM Cortex-A57 processor has been improved. 3881 A more accurate instruction scheduling model for the processor is 3882 now used, and a number of compiler tuning parameters have been set 3883 to offer increased performance when compiling with -mcpu=cortex-a57 3884 or -mtune=cortex-a57. 3885 * Support has been added for the following processors (GCC 3886 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 3887 initial support for its big.LITTLE combination with the ARM 3888 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 3889 initial support for its big.LITTLE combination with the ARM 3890 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 3891 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 3892 The GCC identifiers can be used as arguments to the -mcpu or -mtune 3893 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 3894 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 3895 support for the Cortex-A72. 3896 * The deprecated option -mwords-little-endian has been removed. 3897 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 3898 -mtpcs-leaf-frame which are only applicable to the old ABI have 3899 been deprecated. 3900 * The transitional options -mlra and -mno-lra have been removed. The 3901 ARM backend now uses the local register allocator (LRA) only. 3902 3903 AVR 3904 3905 * The compiler no more supports individual devices like ATmega8. 3906 Specifying, say, -mmcu=atmega8 triggers the usage of the 3907 device-specific [38]spec file specs-atmega8 which is part of the 3908 installation and describes options for the sub-processes like 3909 compiler proper, assembler and linker. You can add support for a 3910 new device -mmcu=mydevice as follows: 3911 1. In an empty directory /someplace, create a new directory 3912 device-specs. 3913 2. Copy a device spec file from the installed device-specs 3914 folder, follow the comments in that file and then save it as 3915 /someplace/device-specs/specs-mydevice. 3916 3. Add -B /someplace -mmcu=mydevice to the compiler's 3917 command-line options. Notice that /someplace must specify an 3918 absolute path and that mydevice must not start with "avr". 3919 4. Provided you have a device-specific library libmydevice.a 3920 available, you can put it at /someplace, dito for a 3921 device-specific startup file crtmydevice.o. 3922 The contents of the device spec files depend on the compiler's 3923 configuration, in particular on --with-avrlibc=no and whether or 3924 not it is configured for RTEMS. 3925 * A new command-line option -nodevicelib has been added. It prevents 3926 the compiler from linking against AVR-LibC's device-specific 3927 library libdevice.a. 3928 * The following three command-line options have been added: 3929 3930 -mrmw 3931 Set if the device supports the read-modify-write 3932 instructions LAC, LAS, LAT and XCH. 3933 3934 -mn-flash=size 3935 Specify the flash size of the device in units of 64 KiB, 3936 rounded up to the next integer as needed. This option 3937 affects the availability of the [39]AVR address-spaces. 3938 3939 -mskip-bug 3940 Set if the device is affected by the respective silicon 3941 bug. 3942 3943 In general, you don't need to set these options by hand. The new 3944 device-specific spec file will set them as needed. 3945 3946 IA-32/x86-64 3947 3948 * New [40]ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of 3949 Intel's CPU codenamed Skylake Server was added to GCC. That 3950 includes inline assembly support, new intrinsics, and basic 3951 autovectorization. These new AVX-512 extensions are available via 3952 the following GCC switches: AVX-512 Vector Length EVEX feature: 3953 -mavx512vl, AVX-512 Byte and Word instructions: -mavx512bw, AVX-512 3954 Dword and Qword instructions: -mavx512dq, AVX-512 FMA-52 3955 instructions: -mavx512ifma and for AVX-512 Vector Bit Manipulation 3956 Instructions: -mavx512vbmi. 3957 * New ISA extensions support Intel MPX was added to GCC. This new 3958 extension is available via the -mmpx compiler switch. Intel MPX is 3959 a set of processor features which, with compiler, run-time library 3960 and OS support, brings increased robustness to software by run-time 3961 checking pointer references against their bounds. In GCC Intel MPX 3962 is supported by Pointer Bounds Checker and libmpx run-time 3963 libraries. 3964 * The new -mrecord-mcount option for -pg generates a Linux kernel 3965 style table of pointers to mcount or __fentry__ calls at the 3966 beginning of functions. The new -mnop-mcount option in addition 3967 also generates nops in place of the __fentry__ or mcount call, so 3968 that a call per function can be later patched in. This can be used 3969 for low overhead tracing or hot code patching. 3970 * The new -malign-data option controls how GCC aligns variables. 3971 -malign-data=compat uses increased alignment compatible with GCC 3972 4.8 and earlier, -malign-data=abi uses alignment as specified by 3973 the psABI, and -malign-data=cacheline uses increased alignment to 3974 match the cache line size. -malign-data=compat is the default. 3975 * The new -mskip-rax-setup option skips setting up the RAX register 3976 when SSE is disabled and there are no variable arguments passed in 3977 vector registers. This can be used to optimize the Linux kernel. 3978 3979 MIPS 3980 3981 * MIPS Releases 3 and 5 are now directly supported. Use the 3982 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 3983 to enable code-generation for these processors. 3984 * The Imagination P5600 processor is now supported using the 3985 -march=p5600 command-line option. 3986 * The Cavium Octeon3 processor is now supported using the 3987 -march=octeon3 command-line option. 3988 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 3989 command-line options. 3990 * The o32 ABI has been modified and extended. The o32 64-bit 3991 floating-point register support is now obsolete and has been 3992 removed. It has been replaced by three ABI extensions FPXX, FP64A, 3993 and FP64. The meaning of the -mfp64 command-line option has 3994 changed. It is now used to enable the FP64A and FP64 ABI 3995 extensions. 3996 + The FPXX extension requires that code generated to access 3997 double-precision values use even-numbered registers. Code that 3998 adheres to this extension is link-compatible with all other 3999 o32 double-precision ABI variants and will execute correctly 4000 in all hardware FPU modes. The command-line options -mabi=32 4001 -mfpxx can be used to enable this extension. MIPS II is the 4002 minimum processor required. 4003 + The o32 FP64A extension requires that floating-point registers 4004 be 64-bit and odd-numbered single-precision registers are not 4005 allowed. Code that adheres to the o32 FP64A variant is 4006 link-compatible with all other o32 double-precision ABI 4007 variants. The command-line options -mabi=32 -mfp64 4008 -mno-odd-spreg can be used to enable this extension. MIPS32R2 4009 is the minimum processor required. 4010 + The o32 FP64 extension also requires that floating-point 4011 registers be 64-bit, but permits the use of single-precision 4012 registers. Code that adheres to the o32 FP64 variant is 4013 link-compatible with o32 FPXX and o32 FP64A variants only, 4014 i.e. it is not compatible with the original o32 4015 double-precision ABI. The command-line options -mabi=32 -mfp64 4016 -modd-spreg can be used to enable this extension. MIPS32R2 is 4017 the minimum processor required. 4018 The new ABI variants can be enabled by default using the configure 4019 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 4020 It is strongly recommended that all vendors begin to set o32 FPXX 4021 as the default ABI. This will be required to run the generated code 4022 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 4023 MIPSR6 cores. 4024 * GCC will now pass all floating-point options to the assembler if 4025 GNU binutils 2.25 is used. As a result, any inline assembly code 4026 that uses hard-float instructions should be amended to include a 4027 .set directive to override the global assembler options when 4028 compiling for soft-float targets. 4029 4030 NDS32 4031 4032 * The variadic function ABI implementation is now compatible with 4033 past Andes toolchains where the caller uses registers to pass 4034 arguments and the callee is in charge of pushing them on stack. 4035 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 4036 been removed since they are not yet available in the nds32 port of 4037 GNU binutils. 4038 * A new option -mcmodel=[small|medium|large] supports varied code 4039 models on code generation. The -mgp-direct option became 4040 meaningless and can be discarded. 4041 4042 RX 4043 4044 * A new command line option -mno-allow-string-insns can be used to 4045 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 4046 SWHILE and RMPA instructions. An erratum released by Renesas shows 4047 that it is unsafe to use these instructions on addresses within the 4048 I/O space of the processor. The new option can be used when the 4049 programmer is concerned that the I/O space might be accessed. The 4050 default is still to enable these instructions. 4051 4052 SH 4053 4054 * The compiler will now pass the appropriate --isa= option to the 4055 assembler. 4056 * The default handling for the GBR has been changed from call 4057 clobbered to call preserved. The old behavior can be reinstated by 4058 specifying the option -fcall-used-gbr. 4059 * Support for the SH4A fpchg instruction has been added which will be 4060 utilized when switching between single and double precision FPU 4061 modes. 4062 * The compiler no longer uses the __fpscr_values array for switching 4063 between single and double FPU precision modes on non-SH4A targets. 4064 Instead mode switching will now be performed by storing, modifying 4065 and reloading the FPSCR, so that other FPSCR bits are preserved 4066 across mode switches. The __fpscr_values array that is defined in 4067 libgcc is still present for backwards compatibility, but it will 4068 not be referenced by compiler generated code anymore. 4069 * New builtin functions __builtin_sh_get_fpscr and 4070 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 4071 function will mask the specified bits in such a way that the SZ, PR 4072 and FR mode bits will be preserved, while changing the other bits. 4073 These new functions do not reference the __fpscr_values array. The 4074 old functions __set_fpscr and __get_fpscr in libgcc which access 4075 the __fpscr_values array are still present for backwards 4076 compatibility, but their usage is highly discouraged. 4077 * Some improvements to code generated for __atomic built-in 4078 functions. 4079 * When compiling for SH2E the compiler will no longer force the usage 4080 of delay slots for conditional branch instructions bt and bf. The 4081 old behavior can be reinstated (e.g. to work around a hardware bug 4082 in the original SH7055) by specifying the new option 4083 -mcbranch-force-delay-slot. 4084 4085 Operating Systems 4086 4087 AIX 4088 4089 * GCC now supports stabs debugging continuation lines to allow long 4090 stabs debug information without overflow that generates AIX linker 4091 errors. 4092 4093 DragonFly BSD 4094 4095 * GCC now supports the DragonFly BSD operating system. 4096 4097 FreeBSD 4098 4099 * GCC now supports the FreeBSD operating system for the arm port 4100 through the arm*-*-freebsd* target triplets. 4101 4102 VxWorks MILS 4103 4104 * GCC now supports the MILS (Multiple Independent Levels of Security) 4105 variant of WindRiver's VxWorks operating system for PowerPC 4106 targets. 4107 4108 Other significant improvements 4109 4110 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 4111 to set the compiler to use. 4112 4113 * When the new command-line option -freport-bug is used, GCC 4114 automatically generates a developer-friendly reproducer whenever an 4115 internal compiler error is encountered. 4116 4117 GCC 5.2 4118 4119 This is the [41]list of problem reports (PRs) from GCC's bug tracking 4120 system that are known to be fixed in the 5.2 release. This list might 4121 not be complete (that is, it is possible that some PRs that have been 4122 fixed are not listed here). 4123 4124 Target Specific Changes 4125 4126 IA-32/x86-64 4127 4128 * Support for new AMD instructions monitorx and mwaitx has been 4129 added. This includes new intrinsic and built-in support. It is 4130 enabled through option -mmwaitx. The instructions monitorx and 4131 mwaitx implement the same functionality as the old monitor and 4132 mwait instructions. In addition mwaitx adds a configurable timer. 4133 The timer value is received as third argument and stored in 4134 register %ebx. 4135 4136 S/390, System z, IBM z Systems 4137 4138 * Support for the IBM z13 processor has been added. When using the 4139 -march=z13 option, the compiler will generate code making use of 4140 the new instructions and registers introduced with the vector 4141 extension facility. The -mtune=z13 option enables z13 specific 4142 instruction scheduling without making use of new instructions. 4143 Compiling code with -march=z13 reduces the default alignment of 4144 vector types bigger than 8 bytes to 8. This is an ABI change and 4145 care must be taken when linking modules compiled with different 4146 arch levels which interchange variables containing vector type 4147 values. For newly compiled code the GNU linker will emit a warning. 4148 * The -mzvector option enables a C/C++ language extension. This 4149 extension provides a new keyword vector which can be used to define 4150 vector type variables. (Note: This is not available when enforcing 4151 strict standard compliance e.g. with -std=c99. Either enable GNU 4152 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 4153 Additionally a set of overloaded builtins is provided which is 4154 partially compatible to the PowerPC Altivec builtins. In order to 4155 make use of these builtins the vecintrin.h header file needs to be 4156 included. 4157 4158 GCC 5.3 4159 4160 This is the [42]list of problem reports (PRs) from GCC's bug tracking 4161 system that are known to be fixed in the 5.3 release. This list might 4162 not be complete (that is, it is possible that some PRs that have been 4163 fixed are not listed here). 4164 4165 Target Specific Changes 4166 4167 IA-32/x86-64 4168 4169 * GCC now supports the Intel CPU named Skylake with AVX-512 4170 extensions through -march=skylake-avx512. The switch enables the 4171 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 4172 AVX-512DQ. 4173 4174 S/390, System z, IBM z Systems 4175 4176 * With this version of GCC IBM z Systems support has been added to 4177 the GO runtime environment. GCC 5.3 has proven to be able to 4178 compile larger GO applications on IBM z Systems. 4179 4180 GCC 5.4 4181 4182 This is the [43]list of problem reports (PRs) from GCC's bug tracking 4183 system that are known to be fixed in the 5.4 release. This list might 4184 not be complete (that is, it is possible that some PRs that have been 4185 fixed are not listed here). 4186 4187 GCC 5.5 4188 4189 This is the [44]list of problem reports (PRs) from GCC's bug tracking 4190 system that are known to be fixed in the 5.5 release. This list might 4191 not be complete (that is, it is possible that some PRs that have been 4192 fixed are not listed here). 4193 4194 Target Specific Changes 4195 4196 IA-32/x86-64 4197 4198 * Support for the [45]deprecated pcommit instruction has been 4199 removed. 4200 4201 4202 For questions related to the use of GCC, please consult these web 4203 pages and the [46]GCC manuals. If that fails, the 4204 [47]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4205 web pages and the development of GCC are welcome on our developer 4206 list at [48]gcc (a] gcc.gnu.org. All of [49]our lists have public 4207 archives. 4208 4209 Copyright (C) [50]Free Software Foundation, Inc. Verbatim copying and 4210 distribution of this entire article is permitted in any medium, 4211 provided this notice is preserved. 4212 4213 These pages are [51]maintained by the GCC team. Last modified 4214 2019-11-28[52]. 4215 4216 References 4217 4218 1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx 4219 2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 4220 3. https://savannah.nongnu.org/bugs/?44574 4221 4. https://gcc.gnu.org/wiki/Intel MPX support in the GCC compiler 4222 5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf 4223 6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf 4224 7. https://gcc.gnu.org/wiki/OpenACC 4225 8. https://gcc.gnu.org/install/configure.html 4226 9. https://www.cilkplus.org/ 4227 10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 4228 11. http://gcc.gnu.org/projects/cxx1y.html 4229 12. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 4230 13. http://gcc.gnu.org/projects/cxx1y.html 4231 14. http://gcc.gnu.org/projects/cxx1y.html 4232 15. http://gcc.gnu.org/projects/cxx1y.html 4233 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 4234 17. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 4235 18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 4236 19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 4237 20. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 4238 21. https://sourceware.org/gdb/current/onlinedocs/gdb/Xmethods-In-Python.html 4239 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 4240 23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 4241 24. https://gcc.gnu.org/install/configure.html 4242 25. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 4243 26. https://gcc.gnu.org/wiki/Fortran2003Status 4244 27. https://gcc.gnu.org/wiki/Fortran2008Status 4245 28. https://gcc.gnu.org/wiki/Coarray 4246 29. http://www.opencoarrays.org/ 4247 30. https://golang.org/cmd/go/ 4248 31. https://golang.org/cmd/gofmt/ 4249 32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 4250 33. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 4251 34. https://github.com/davidmalcolm/pygccjit 4252 35. https://github.com/ibuclaw/gccjitd 4253 36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 4254 37. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 4255 38. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 4256 39. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 4257 40. https://software.intel.com/sites/default/files/managed/39/c5/325462-sdm-vol-1-2abcd-3abcd.pdf 4258 41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 4259 42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 4260 43. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 4261 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 4262 45. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction 4263 46. https://gcc.gnu.org/onlinedocs/ 4264 47. mailto:gcc-help (a] gcc.gnu.org 4265 48. mailto:gcc (a] gcc.gnu.org 4266 49. https://gcc.gnu.org/lists.html 4267 50. https://www.fsf.org/ 4268 51. https://gcc.gnu.org/about.html 4269 52. http://validator.w3.org/check/referer 4270 ====================================================================== 4271 http://gcc.gnu.org/gcc-4.9/index.html 4272 GCC 4.9 Release Series 4273 4274 (This release series is no longer supported.) 4275 4276 Aug 3, 2016 4277 4278 The [1]GNU project and the GCC developers are pleased to announce the 4279 release of GCC 4.9.4. 4280 4281 This release is a bug-fix release, containing fixes for regressions in 4282 GCC 4.9.3 relative to previous releases of GCC. 4283 4284 Release History 4285 4286 GCC 4.9.4 4287 Aug 3, 2016 ([2]changes, [3]documentation) 4288 4289 GCC 4.9.3 4290 June 26, 2015 ([4]changes, [5]documentation) 4291 4292 GCC 4.9.2 4293 October 30, 2014 ([6]changes, [7]documentation) 4294 4295 GCC 4.9.1 4296 July 16, 2014 ([8]changes, [9]documentation) 4297 4298 GCC 4.9.0 4299 April 22, 2014 ([10]changes, [11]documentation) 4300 4301 References and Acknowledgements 4302 4303 GCC used to stand for the GNU C Compiler, but since the compiler 4304 supports several other languages aside from C, it now stands for the 4305 GNU Compiler Collection. 4306 4307 A list of [12]successful builds is updated as new information becomes 4308 available. 4309 4310 The GCC developers would like to thank the numerous people that have 4311 contributed new features, improvements, bug fixes, and other changes as 4312 well as test results to GCC. This [13]amazing group of volunteers is 4313 what makes GCC successful. 4314 4315 For additional information about GCC please refer to the [14]GCC 4316 project web site or contact the [15]GCC development mailing list. 4317 4318 To obtain GCC please use [16]our mirror sites or [17]our version 4319 control system. 4320 4321 4322 For questions related to the use of GCC, please consult these web 4323 pages and the [18]GCC manuals. If that fails, the 4324 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4325 web pages and the development of GCC are welcome on our developer 4326 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 4327 archives. 4328 4329 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 4330 distribution of this entire article is permitted in any medium, 4331 provided this notice is preserved. 4332 4333 These pages are [23]maintained by the GCC team. Last modified 4334 2020-01-14[24]. 4335 4336 References 4337 4338 1. http://www.gnu.org/ 4339 2. http://gcc.gnu.org/gcc-4.9/changes.html 4340 3. http://gcc.gnu.org/onlinedocs/4.9.4/ 4341 4. http://gcc.gnu.org/gcc-4.9/changes.html 4342 5. http://gcc.gnu.org/onlinedocs/4.9.3/ 4343 6. http://gcc.gnu.org/gcc-4.9/changes.html 4344 7. http://gcc.gnu.org/onlinedocs/4.9.2/ 4345 8. http://gcc.gnu.org/gcc-4.9/changes.html 4346 9. http://gcc.gnu.org/onlinedocs/4.9.1/ 4347 10. http://gcc.gnu.org/gcc-4.9/changes.html 4348 11. https://gcc.gnu.org/onlinedocs/4.9.0/ 4349 12. http://gcc.gnu.org/gcc-4.9/buildstat.html 4350 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 4351 14. http://gcc.gnu.org/index.html 4352 15. mailto:gcc (a] gcc.gnu.org 4353 16. http://gcc.gnu.org/mirrors.html 4354 17. http://gcc.gnu.org/git.html 4355 18. https://gcc.gnu.org/onlinedocs/ 4356 19. mailto:gcc-help (a] gcc.gnu.org 4357 20. mailto:gcc (a] gcc.gnu.org 4358 21. https://gcc.gnu.org/lists.html 4359 22. https://www.fsf.org/ 4360 23. https://gcc.gnu.org/about.html 4361 24. http://validator.w3.org/check/referer 4362 ====================================================================== 4363 http://gcc.gnu.org/gcc-4.9/changes.html 4364 GCC 4.9 Release Series 4365 Changes, New Features, and Fixes 4366 4367 Caveats 4368 4369 * The mudflap run time checker has been removed. The mudflap options 4370 remain, but do nothing. 4371 * Support for a number of older systems and recently unmaintained or 4372 untested target ports of GCC has been declared obsolete in GCC 4.9. 4373 Unless there is activity to revive them, the next release of GCC 4374 will have their sources permanently removed. 4375 The following ports for individual systems on particular 4376 architectures have been obsoleted: 4377 + Solaris 9 (*-*-solaris2.9). Details can be found in the 4378 [1]announcement. 4379 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 4380 float64x1_t exported by arm_neon.h are defined to be the same as 4381 their base types. This results in incorrect application of 4382 parameter passing rules to arguments of types int64x1_t and 4383 uint64x1_t, with respect to the AAPCS64 ABI specification. In 4384 addition, names of C++ functions with parameters of these types 4385 (including float64x1_t) are not mangled correctly. The current 4386 typedef declarations also unintentionally allow implicit casting 4387 between singleton vector types and their base types. These issues 4388 will be resolved in a near future release. See [2]PR60825 for more 4389 information. 4390 4391 More information on porting to GCC 4.9 from previous versions of GCC 4392 can be found in the [3]porting guide for this release. 4393 4394 General Optimizer Improvements 4395 4396 * AddressSanitizer, a fast memory error detector, is now available on 4397 ARM. 4398 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 4399 detector, has been added and can be enabled via 4400 -fsanitize=undefined. Various computations will be instrumented to 4401 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 4402 currently available for the C and C++ languages. 4403 * Link-time optimization (LTO) improvements: 4404 + Type merging was rewritten. The new implementation is 4405 significantly faster and uses less memory. 4406 + Better partitioning algorithm resulting in less streaming 4407 during link time. 4408 + Early removal of virtual methods reduces the size of object 4409 files and improves link-time memory usage and compile time. 4410 + Function bodies are now loaded on-demand and released early 4411 improving overall memory usage at link time. 4412 + C++ hidden keyed methods can now be optimized out. 4413 + When using a linker plugin, compiling with the -flto option 4414 now generates slim object files (.o) which only contain 4415 intermediate language representation for LTO. Use 4416 -ffat-lto-objects to create files which contain additionally 4417 the object code. To generate static libraries suitable for LTO 4418 processing, use gcc-ar and gcc-ranlib; to list symbols from a 4419 slim object file use gcc-nm. (This requires that ar, ranlib 4420 and nm have been compiled with plugin support.) 4421 Memory usage building Firefox with debug enabled was reduced from 4422 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 4423 * Inter-procedural optimization improvements: 4424 + New type inheritance analysis module improving 4425 devirtualization. Devirtualization now takes into account 4426 anonymous name-spaces and the C++11 final keyword. 4427 + New speculative devirtualization pass (controlled by 4428 -fdevirtualize-speculatively. 4429 + Calls that were speculatively made direct are turned back to 4430 indirect where direct call is not cheaper. 4431 + Local aliases are introduced for symbols that are known to be 4432 semantically equivalent across shared libraries improving 4433 dynamic linking times. 4434 * Feedback directed optimization improvements: 4435 + Profiling of programs using C++ inline functions is now more 4436 reliable. 4437 + New time profiling determines typical order in which functions 4438 are executed. 4439 + A new function reordering pass (controlled by 4440 -freorder-functions) significantly reduces startup time of 4441 large applications. Until binutils support is completed, it is 4442 effective only with link-time optimization. 4443 + Feedback driven indirect call removal and devirtualization now 4444 handle cross-module calls when link-time optimization is 4445 enabled. 4446 4447 New Languages and Language specific improvements 4448 4449 * Version 4.0 of the [4]OpenMP specification is now supported in the 4450 C and C++ compilers and starting with the 4.9.1 release also in the 4451 Fortran compiler. The new -fopenmp-simd option can be used to 4452 enable OpenMP's SIMD directives while ignoring other OpenMP 4453 directives. The new [5]-fsimd-cost-model= option permits to tune 4454 the vectorization cost model for loops annotated with OpenMP and 4455 Cilk Plus simd directives. -Wopenmp-simd warns when the current 4456 cost model overrides simd directives set by the user. 4457 * The -Wdate-time option has been added for the C, C++ and Fortran 4458 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 4459 macros are used. Those macros might prevent bit-wise-identical 4460 reproducible compilations. 4461 4462 Ada 4463 4464 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 4465 4466 C family 4467 4468 * Support for colorizing diagnostics emitted by GCC has been added. 4469 The [6]-fdiagnostics-color=auto will enable it when outputting to 4470 terminals, -fdiagnostics-color=always unconditionally. The 4471 GCC_COLORS environment variable can be used to customize the colors 4472 or disable coloring. If GCC_COLORS variable is present in the 4473 environment, the default is -fdiagnostics-color=auto, otherwise 4474 -fdiagnostics-color=never. 4475 Sample diagnostics output: 4476 $ g++ -fdiagnostics-color=always -S -Wall test.C 4477 test.C: In function int foo(): 4478 test.C:1:14: warning: no return statement in function returning non-void [-W 4479 return-type] 4480 int foo () { } 4481 ^ 4482 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 4483 -ftemplate-depth= to increase the maximum) instantiating struct X<100> 4484 template <int N> struct X { static const int value = X<N-1>::value; }; temp 4485 late struct X<1000>; 4486 ^ 4487 test.C:2:46: recursively required from const int X<999>::value 4488 test.C:2:46: required from const int X<1000>::value 4489 test.C:2:88: required from here 4490 4491 test.C:2:46: error: incomplete type X<100> used in nested name specifier 4492 4493 * With the new [7]#pragma GCC ivdep, the user can assert that there 4494 are no loop-carried dependencies which would prevent concurrent 4495 execution of consecutive iterations using SIMD (single instruction 4496 multiple data) instructions. 4497 * Support for [8]Cilk Plus has been added and can be enabled with the 4498 -fcilkplus option. Cilk Plus is an extension to the C and C++ 4499 languages to support data and task parallelism. The present 4500 implementation follows ABI version 1.2; all features but _Cilk_for 4501 have been implemented. 4502 4503 C 4504 4505 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 4506 <stdatomic.h> header) are now supported. 4507 * ISO C11 generic selections (_Generic keyword) are now supported. 4508 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 4509 __thread) is now supported. 4510 * ISO C11 support is now at a similar level of completeness to ISO 4511 C99 support: substantially complete modulo bugs, extended 4512 identifiers (supported except for corner cases when 4513 -fextended-identifiers is used), floating-point issues (mainly but 4514 not entirely relating to optional C99 features from Annexes F and 4515 G) and the optional Annexes K (Bounds-checking interfaces) and L 4516 (Analyzability). 4517 * A new C extension __auto_type provides a subset of the 4518 functionality of C++11 auto in GNU C. 4519 4520 C++ 4521 4522 * The G++ implementation of [9]C++1y return type deduction for normal 4523 functions has been updated to conform to [10]N3638, the proposal 4524 accepted into the working paper. Most notably, it adds 4525 decltype(auto) for getting decltype semantics rather than the 4526 template argument deduction semantics of plain auto: 4527 4528 int& f(); 4529 auto i1 = f(); // int 4530 decltype(auto) i2 = f(); // int& 4531 4532 * G++ supports [11]C++1y lambda capture initializers: 4533 4534 [x = 42]{ ... }; 4535 4536 Actually, they have been accepted since GCC 4.5, but now the 4537 compiler doesn't warn about them with -std=c++1y, and supports 4538 parenthesized and brace-enclosed initializers as well. 4539 * G++ supports [12]C++1y variable length arrays. G++ has supported 4540 GNU/C99-style VLAs for a long time, but now additionally supports 4541 initializers and lambda capture by reference. In C++1y mode G++ 4542 will complain about VLA uses that are not permitted by the draft 4543 standard, such as forming a pointer to VLA type or applying sizeof 4544 to a VLA variable. Note that it now appears that VLAs will not be 4545 part of C++14, but will be part of a separate document and then 4546 perhaps C++17. 4547 4548 void f(int n) { 4549 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 4550 [&a]{ for (int i : a) { cout << i << endl; } }(); 4551 &a; // error, taking address of VLA 4552 } 4553 4554 * G++ supports the [13]C++1y [[deprecated]] attribute modulo bugs in 4555 the underlying [[gnu::deprecated]] attribute. Classes and functions 4556 can be marked deprecated and a diagnostic message added: 4557 4558 class A; 4559 int bar(int n); 4560 #if __cplusplus > 201103 4561 class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 4562 [[deprecated("bar is unsafe; use foo() instead")]] 4563 int bar(int n); 4564 4565 int foo(int n); 4566 class B; 4567 #endif 4568 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 4569 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 4570 o() instead 4571 4572 * G++ supports [14]C++1y digit separators. Long numeric literals can 4573 be subdivided with a single quote ' to enhance readability: 4574 4575 int i = 1048576; 4576 int j = 1'048'576; 4577 int k = 0x10'0000; 4578 int m = 0'004'000'000; 4579 int n = 0b0001'0000'0000'0000'0000'0000; 4580 4581 double x = 1.602'176'565e-19; 4582 double y = 1.602'176'565e-1'9; 4583 4584 * G++ supports [15]C++1y generic (polymorphic) lambdas. 4585 4586 // a functional object that will increment any type 4587 auto incr = [](auto x) { return x++; }; 4588 4589 * As a GNU extension, G++ supports explicit template parameter syntax 4590 for generic lambdas. This can be combined in the expected way with 4591 the standard auto syntax. 4592 4593 // a functional object that will add two like-type objects 4594 auto add = [] <typename T> (T a, T b) { return a + b; }; 4595 4596 * G++ supports unconstrained generic functions as specified by 4.1.2 4597 and 5.1.1 of [16]N3889: Concepts Lite Specification. Briefly, auto 4598 may be used as a type-specifier in a parameter declaration of any 4599 function declarator in order to introduce an implicit function 4600 template parameter, akin to generic lambdas. 4601 4602 // the following two function declarations are equivalent 4603 auto incr(auto x) { return x++; } 4604 template <typename T> 4605 auto incr(T x) { return x++; } 4606 4607 Runtime Library (libstdc++) 4608 4609 * [17]Improved support for C++11, including: 4610 + support for <regex>; 4611 + The associative containers in <map> and <set> and the 4612 unordered associative containers in <unordered_map> and 4613 <unordered_set> meet the allocator-aware container 4614 requirements; 4615 * [18]Improved experimental support for the upcoming ISO C++ 4616 standard, C++14, including: 4617 + fixing constexpr member functions without const; 4618 + implementation of the std::exchange() utility function; 4619 + addressing tuples by type; 4620 + implemention of std::make_unique; 4621 + implemention of std::shared_lock; 4622 + making std::result_of SFINAE-friendly; 4623 + adding operator() to std::integral_constant; 4624 + adding user-defined literals for standard library types 4625 std::basic_string, std::chrono::duration, and std::complex; 4626 + adding two range overloads to non-modifying sequence oprations 4627 std::equal and std::mismatch; 4628 + adding IO manipulators for quoted strings; 4629 + adding constexpr members to <utility>, <complex>, <chrono>, 4630 and some containers; 4631 + adding compile-time std::integer_sequence; 4632 + adding cleaner transformation traits; 4633 + making <functional>s operator functors easier to use and more 4634 generic; 4635 * An implementation of std::experimental::optional. 4636 * An implementation of std::experimental::string_view. 4637 * The non-standard function std::copy_exception has been deprecated 4638 and will be removed in a future version. std::make_exception_ptr 4639 should be used instead. 4640 4641 Fortran 4642 4643 * Compatibility notice: 4644 + Module files: The version of the module files (.mod) has been 4645 incremented; additionally, module files are now compressed. 4646 Fortran MODULEs compiled by earlier GCC versions have to be 4647 recompiled, when they are USEd by files compiled with GCC 4.9. 4648 GCC 4.9 is not able to read .mod files of earlier GCC 4649 versions; attempting to do so gives an error message. Note: 4650 The ABI of the produced assembler data itself has not changed: 4651 object files and libraries are fully compatible with older 4652 versions (except as stated below). 4653 + ABI changes: 4654 o The [19]argument passing ABI has changed for scalar dummy 4655 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 4656 which have both the VALUE and the OPTIONAL attributes. 4657 o To support finalization the virtual table associated with 4658 polymorphic variables has changed. Code containing CLASS 4659 should be recompiled, including all files which define 4660 derived types involved in the type definition used by 4661 polymorphic variables. (Note: Due to the incremented 4662 module version, trying to mix old code with new code will 4663 usually give an error message.) 4664 + GNU Fortran no longer deallocates allocatable variables or 4665 allocatable components of variables declared in the main 4666 program. Since Fortran 2008, the standard explicitly states 4667 that variables declared in the Fortran main program 4668 automatically have the SAVE attribute. 4669 + When opening files, the close-on-exec flag is set if the 4670 system supports such a feature. This is generally considered 4671 good practice these days, but if there is a need to pass file 4672 descriptors to child processes the parent process must now 4673 remember to clear the close-on-exec flag by calling fcntl(), 4674 e.g. via ISO_C_BINDING, before executing the child process. 4675 * The deprecated command-line option -fno-whole-file has been 4676 removed. (-fwhole-file is the default since GCC 4.6.) 4677 -fwhole-file/-fno-whole-file continue to be accepted but do not 4678 influence the code generation. 4679 * The compiler no longer unconditionally warns about DO loops with 4680 zero iterations. This warning is now controlled by the -Wzerotrip 4681 option, which is implied by -Wall. 4682 * The new NO_ARG_CHECK attribute of the [20]!GCC$ directive can be 4683 used to disable the type-kind-rank (TKR) argument check for a dummy 4684 argument. The feature is similar to ISO/IEC TS 29133:2012's 4685 TYPE(*), except that it additionally also disables the rank check. 4686 Variables with NO_ARG_CHECK have to be dummy arguments and may only 4687 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 4688 to another NO_ARG_CHECK dummy argument; also the other constraints 4689 of TYPE(*) apply. The dummy arguments should be declared as scalar 4690 or assumed-size variable of type type(*) (recommended) or of type 4691 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 4692 the data without further type or shape information is passed, 4693 similar to C's void*. Note that also TS 29113's 4694 type(*),dimension(..) accepts arguments of any type and rank; 4695 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 4696 descriptor which contains the array shape and stride of the 4697 argument. 4698 * [21]Fortran 2003: 4699 + Finalization is now supported. It is currently only done for a 4700 subset of those situations in which it should occur. 4701 + Experimental support for scalar character components with 4702 deferred length (i.e. allocatable string length) in derived 4703 types has been added. (Deferred-length character variables are 4704 supported since GCC 4.6.) 4705 * [22]Fortran 2008: 4706 + When STOP or ERROR STOP are used to terminate the execution 4707 and any exception (but inexact) is signaling, a warning is 4708 printed to ERROR_UNIT, indicating which exceptions are 4709 signaling. The [23]-ffpe-summary= command-line option can be 4710 used to fine-tune for which exceptions the warning should be 4711 shown. 4712 + Rounding on input (READ) is now handled on systems where 4713 strtod honours the rounding mode. (For output, rounding is 4714 supported since GCC 4.5.) Note that for input, the compatible 4715 rounding mode is handled as nearest (i.e., rounding to an even 4716 least significant [cf. IEC 60559:1989] for a tie, while 4717 compatible rounds away from zero in that case). 4718 4719 Go 4720 4721 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 4722 4723 New Targets and Target Specific Improvements 4724 4725 AArch64 4726 4727 * The ARMv8-A crypto and CRC instructions are now supported through 4728 intrinsics. These are enabled when the architecture supports these 4729 and are available through the -march=armv8-a+crc and 4730 -march=armv8-a+crypto options. 4731 * Initial support for ILP32 has now been added to the compiler. This 4732 is now available through the command-line option -mabi=ilp32. 4733 Support for ILP32 is considered experimental as the ABI 4734 specification is still beta. 4735 * Coverage of more of the ISA including the SIMD extensions has been 4736 added. The Advanced SIMD intrinsics have also been improved. 4737 * The new local register allocator (LRA) is now on by default for the 4738 AArch64 backend. 4739 * The REE (Redundant extension elimination) pass has now been enabled 4740 by default for the AArch64 backend. 4741 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 4742 * Initial big.LITTLE tuning support for the combination of Cortex-A57 4743 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 4744 option. 4745 * A number of structural changes have been made to both the ARM and 4746 AArch64 backends to facilitate improved code-generation. 4747 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 4748 has been added and can be enabled by giving the 4749 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 4750 default by configuring GCC with the --enable-fix-cortex-a53-835769 4751 option. 4752 4753 ARC 4754 4755 * A port for Synopsys Designware ARC has been contributed by Embecosm 4756 and Synopsys Inc. 4757 4758 ARM 4759 4760 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 4761 disabled by default. This was found to generate better code in only 4762 a small number of cases. It can be turned back on with the 4763 -mneon-for-64bits option. 4764 * Further support for the ARMv8-A architecture, notably implementing 4765 the restriction around IT blocks in the Thumb32 instruction set has 4766 been added. The -mrestrict-it option can be used with 4767 -march=armv7-a or the -march=armv7ve options to make code 4768 generation fully compatible with the deprecated instructions in 4769 ARMv8-A. 4770 * Support has now been added for the ARMv7ve variant of the 4771 architecture. This can be used by the -march=armv7ve option. 4772 * The ARMv8-A crypto and CRC instructions are now supported through 4773 intrinsics and are available through the -march=armv8-a+crc and 4774 mfpu=crypto-neon-fp-armv8 options. 4775 * LRA is now on by default for the ARM target. This can be turned off 4776 using the -mno-lra option. This option is a purely transitionary 4777 command-line option and will be removed in a future release. We are 4778 interested in any bug reports regarding functional and performance 4779 regressions with LRA. 4780 * A new option -mslow-flash-data to improve performance of programs 4781 fetching data on slow flash memory has now been introduced for the 4782 ARMv7-M profile cores. 4783 * A new option -mpic-data-is-text-relative for targets that allows 4784 data segments to be relative to text segments has been added. This 4785 is on by default for all targets except VxWorks RTP. 4786 * A number of infrastructural changes have been made to both the ARM 4787 and AArch64 backends to facilitate improved code-generation. 4788 * GCC now supports Cortex-A12 and the Cortex-R7 through the 4789 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 4790 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 4791 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 4792 * Initial big.LITTLE tuning support for the combination of Cortex-A57 4793 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 4794 option. Similar support was added for the combination of Cortex-A15 4795 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 4796 * Further performance optimizations for the Cortex-A15 and the 4797 Cortex-M4 have been added. 4798 * A number of code generation improvements for Thumb2 to reduce code 4799 size when compiling for the M-profile processors. 4800 4801 AVR 4802 4803 * A new command-line option -mfract-convert-truncate has been added. 4804 It allows compiler to use truncation instead of rounding towards 4805 zero for fractional fixed-point types. 4806 4807 IA-32/x86-64 4808 4809 * -mfpmath=sse is now implied by -ffast-math on all targets where 4810 SSE2 is supported. 4811 * Intel AVX-512 support was added to GCC. That includes inline 4812 assembly support, new registers and extending existing ones, new 4813 intrinsics (covered by corresponding testsuite), and basic 4814 autovectorization. AVX-512 instructions are available via the 4815 following GCC switches: AVX-512 foundation instructions: -mavx512f, 4816 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 4817 reciprocal instructions: -mavx512er, AVX-512 conflict detection 4818 instructions: -mavx512cd. 4819 * It is now possible to call x86 intrinsics from select functions in 4820 a file that are tagged with the corresponding target attribute 4821 without having to compile the entire file with the -mxxx option. 4822 This improves the usability of x86 intrinsics and is particularly 4823 useful when doing [24]Function Multiversioning. 4824 * GCC now supports the new Intel microarchitecture named Silvermont 4825 through -march=silvermont. 4826 * GCC now supports the new Intel microarchitecture named Broadwell 4827 through -march=broadwell. 4828 * Optimizing for other Intel microarchitectures have been renamed to 4829 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 4830 * -march=generic has been retuned for better support of Intel core 4831 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 4832 Pentium-M, and Pentium4 based CPUs is no longer considered 4833 important for generic. 4834 * -mtune=intel can now be used to generate code running well on the 4835 most current Intel processors, which are Haswell and Silvermont for 4836 GCC 4.9. 4837 * Support to encode 32-bit assembly instructions in 16-bit format is 4838 now available through the -m16 command-line option. 4839 * Better inlining of memcpy and memset that is aware of value ranges 4840 and produces shorter alignment prologues. 4841 * -mno-accumulate-outgoing-args is now honored when unwind 4842 information is output. Argument accumulation is also now turned off 4843 for portions of programs optimized for size. 4844 * Support for new AMD family 15h processors (Excavator core) is now 4845 available through the -march=bdver4 and -mtune=bdver4 options. 4846 4847 MSP430 4848 4849 * A new command-line option -mcpu= has been added to the MSP430 4850 backend. This option is used to specify the ISA to be used. 4851 Accepted values are msp430 (the default), msp430x and msp430xv2. 4852 The ISA is no longer deduced from the -mmcu= option as there are 4853 far too many different MCU names. The -mmcu= option is still 4854 supported, and this is still used to select linker scripts and 4855 generate a C preprocessor symbol that will be recognised by the 4856 msp430.h header file. 4857 4858 NDS32 4859 4860 * A new nds32 port supports the 32-bit architecture from Andes 4861 Technology Corporation. 4862 * The port provides initial support for the V2, V3, V3m instruction 4863 set architectures. 4864 4865 Nios II 4866 4867 * A port for the Altera Nios II has been contributed by Mentor 4868 Graphics. 4869 4870 PowerPC / PowerPC64 / RS6000 4871 4872 * GCC now supports Power ISA 2.07, which includes support for 4873 Hardware Transactional Memory (HTM), Quadword atomics and several 4874 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 4875 integer and decimal integer operations. 4876 * Support for the POWER8 processor is now available through the 4877 -mcpu=power8 and -mtune=power8 options. 4878 * The libitm library has been modified to add a HTM fastpath that 4879 automatically uses POWER's HTM hardware instructions when it is 4880 executing on a HTM enabled processor. 4881 * Support for the new powerpc64le-linux platform has been added. It 4882 defaults to generating code that conforms to the ELFV2 ABI. 4883 4884 S/390, System z 4885 4886 * Support for the Transactional Execution Facility included with the 4887 IBM zEnterprise zEC12 processor has been added. A set of GCC style 4888 builtins as well as XLC style builtins are provided. The builtins 4889 are enabled by default when using the -march=zEC12 option but can 4890 explicitly be disabled with -mno-htm. Using the GCC builtins also 4891 libitm supports hardware transactions on S/390. 4892 * The hotpatch features allows to prepare functions for hotpatching. 4893 A certain amount of bytes is reserved before the function entry 4894 label plus a NOP is inserted at its very beginning to implement a 4895 backward jump when applying a patch. The feature can either be 4896 enabled per compilation unit via the command-line option -mhotpatch 4897 or per function using the hotpatch attribute. 4898 * The shrink wrap optimization is now supported on S/390 and enabled 4899 by default. 4900 * A major rework of the routines to determine which registers need to 4901 be saved and restored in function prologue/epilogue now allow to 4902 use floating point registers as save slots. This will happen for 4903 certain leaf function with -march=z10 or higher. 4904 * The LRA rtl pass replaces reload by default on S/390. 4905 4906 RX 4907 4908 * The port now allows to specify the RX100, RX200, and RX600 4909 processors with the command-line options -mcpu=rx100, -mcpu=rx200 4910 and -mcpu=rx600. 4911 4912 SH 4913 4914 * Minor improvements to code generated for integer arithmetic and 4915 code that involves the T bit. 4916 * Added support for the SH2A clips and clipu instructions. The 4917 compiler will now try to utilize them for min/max expressions such 4918 as max (-128, min (127, x)). 4919 * Added support for the cmp/str instruction through built-in 4920 functions such as __builtin_strlen. When not optimizing for size, 4921 the compiler will now expand calls to e.g. strlen as an inlined 4922 sequences which utilize the cmp/str instruction. 4923 * Improved code generated around volatile memory loads and stores. 4924 * The option -mcbranchdi has been deprecated. Specifying it will 4925 result in a warning and will not influence code generation. 4926 * The option -mcmpeqdi has been deprecated. Specifying it will result 4927 in a warning and will not influence code generation. 4928 4929 GCC 4.9.1 4930 4931 This is the [25]list of problem reports (PRs) from GCC's bug tracking 4932 system that are known to be fixed in the 4.9.1 release. This list might 4933 not be complete (that is, it is possible that some PRs that have been 4934 fixed are not listed here). 4935 4936 Version 4.0 of the OpenMP specification is supported even in Fortran, 4937 not just C and C++. 4938 4939 GCC 4.9.2 4940 4941 This is the [26]list of problem reports (PRs) from GCC's bug tracking 4942 system that are known to be fixed in the 4.9.2 release. This list might 4943 not be complete (that is, it is possible that some PRs that have been 4944 fixed are not listed here). 4945 4946 GCC 4.9.3 4947 4948 This is the [27]list of problem reports (PRs) from GCC's bug tracking 4949 system that are known to be fixed in the 4.9.3 release. This list might 4950 not be complete (that is, it is possible that some PRs that have been 4951 fixed are not listed here). 4952 4953 GCC 4.9.4 4954 4955 This is the [28]list of problem reports (PRs) from GCC's bug tracking 4956 system that are known to be fixed in the 4.9.4 release. This list might 4957 not be complete (that is, it is possible that some PRs that have been 4958 fixed are not listed here). 4959 4960 4961 For questions related to the use of GCC, please consult these web 4962 pages and the [29]GCC manuals. If that fails, the 4963 [30]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4964 web pages and the development of GCC are welcome on our developer 4965 list at [31]gcc (a] gcc.gnu.org. All of [32]our lists have public 4966 archives. 4967 4968 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 4969 distribution of this entire article is permitted in any medium, 4970 provided this notice is preserved. 4971 4972 These pages are [34]maintained by the GCC team. Last modified 4973 2019-11-28[35]. 4974 4975 References 4976 4977 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 4978 2. https://gcc.gnu.org/PR60825 4979 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 4980 4. https://www.openmp.org/specifications/ 4981 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 4982 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 4983 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 4984 8. https://www.cilkplus.org/ 4985 9. http://gcc.gnu.org/projects/cxx1y.html 4986 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 4987 11. http://gcc.gnu.org/projects/cxx1y.html 4988 12. http://gcc.gnu.org/projects/cxx1y.html 4989 13. http://gcc.gnu.org/projects/cxx1y.html 4990 14. http://gcc.gnu.org/projects/cxx1y.html 4991 15. http://gcc.gnu.org/projects/cxx1y.html 4992 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 4993 17. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 4994 18. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 4995 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 4996 20. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 4997 21. https://gcc.gnu.org/wiki/Fortran2003Status 4998 22. https://gcc.gnu.org/wiki/Fortran2008Status 4999 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 5000 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 5001 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 5002 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 5003 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 5004 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4 5005 29. https://gcc.gnu.org/onlinedocs/ 5006 30. mailto:gcc-help (a] gcc.gnu.org 5007 31. mailto:gcc (a] gcc.gnu.org 5008 32. https://gcc.gnu.org/lists.html 5009 33. https://www.fsf.org/ 5010 34. https://gcc.gnu.org/about.html 5011 35. http://validator.w3.org/check/referer 5012 ====================================================================== 5013 http://gcc.gnu.org/gcc-4.8/index.html 5014 GCC 4.8 Release Series 5015 5016 (This release series is no longer supported.) 5017 5018 June 23, 2015 5019 5020 The [1]GNU project and the GCC developers are pleased to announce the 5021 release of GCC 4.8.5. 5022 5023 This release is a bug-fix release, containing fixes for regressions in 5024 GCC 4.8.4 relative to previous releases of GCC. 5025 5026 Release History 5027 5028 GCC 4.8.5 5029 June 23, 2015 ([2]changes, [3]documentation) 5030 5031 GCC 4.8.4 5032 December 19, 2014 ([4]changes, [5]documentation) 5033 5034 GCC 4.8.3 5035 May 22, 2014 ([6]changes, [7]documentation) 5036 5037 GCC 4.8.2 5038 October 16, 2013 ([8]changes, [9]documentation) 5039 5040 GCC 4.8.1 5041 May 31, 2013 ([10]changes, [11]documentation) 5042 5043 GCC 4.8.0 5044 March 22, 2013 ([12]changes, [13]documentation) 5045 5046 References and Acknowledgements 5047 5048 GCC used to stand for the GNU C Compiler, but since the compiler 5049 supports several other languages aside from C, it now stands for the 5050 GNU Compiler Collection. 5051 5052 A list of [14]successful builds is updated as new information becomes 5053 available. 5054 5055 The GCC developers would like to thank the numerous people that have 5056 contributed new features, improvements, bug fixes, and other changes as 5057 well as test results to GCC. This [15]amazing group of volunteers is 5058 what makes GCC successful. 5059 5060 For additional information about GCC please refer to the [16]GCC 5061 project web site or contact the [17]GCC development mailing list. 5062 5063 To obtain GCC please use [18]our mirror sites or [19]our version 5064 control system. 5065 5066 5067 For questions related to the use of GCC, please consult these web 5068 pages and the [20]GCC manuals. If that fails, the 5069 [21]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5070 web pages and the development of GCC are welcome on our developer 5071 list at [22]gcc (a] gcc.gnu.org. All of [23]our lists have public 5072 archives. 5073 5074 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 5075 distribution of this entire article is permitted in any medium, 5076 provided this notice is preserved. 5077 5078 These pages are [25]maintained by the GCC team. Last modified 5079 2020-01-14[26]. 5080 5081 References 5082 5083 1. http://www.gnu.org/ 5084 2. http://gcc.gnu.org/gcc-4.8/changes.html 5085 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 5086 4. http://gcc.gnu.org/gcc-4.8/changes.html 5087 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 5088 6. http://gcc.gnu.org/gcc-4.8/changes.html 5089 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 5090 8. http://gcc.gnu.org/gcc-4.8/changes.html 5091 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 5092 10. http://gcc.gnu.org/gcc-4.8/changes.html 5093 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 5094 12. http://gcc.gnu.org/gcc-4.8/changes.html 5095 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 5096 14. http://gcc.gnu.org/gcc-4.8/buildstat.html 5097 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5098 16. http://gcc.gnu.org/index.html 5099 17. mailto:gcc (a] gcc.gnu.org 5100 18. http://gcc.gnu.org/mirrors.html 5101 19. http://gcc.gnu.org/git.html 5102 20. https://gcc.gnu.org/onlinedocs/ 5103 21. mailto:gcc-help (a] gcc.gnu.org 5104 22. mailto:gcc (a] gcc.gnu.org 5105 23. https://gcc.gnu.org/lists.html 5106 24. https://www.fsf.org/ 5107 25. https://gcc.gnu.org/about.html 5108 26. http://validator.w3.org/check/referer 5109 ====================================================================== 5110 http://gcc.gnu.org/gcc-4.8/changes.html 5111 GCC 4.8 Release Series 5112 Changes, New Features, and Fixes 5113 5114 Caveats 5115 5116 GCC now uses C++ as its implementation language. This means that to 5117 build GCC from sources, you will need a C++ compiler that understands 5118 C++ 2003. For more details on the rationale and specific changes, 5119 please refer to the [1]C++ conversion page. 5120 5121 To enable the Graphite framework for loop optimizations you now need 5122 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 5123 the [2]GCC infrastructure directory. The installation manual contains 5124 more information about requirements to build GCC. 5125 5126 GCC now uses a more aggressive analysis to derive an upper bound for 5127 the number of iterations of loops using constraints imposed by language 5128 standards. This may cause non-conforming programs to no longer work as 5129 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 5130 option, -fno-aggressive-loop-optimizations, was added to disable this 5131 aggressive analysis. In some loops that have known constant number of 5132 iterations, but undefined behavior is known to occur in the loop before 5133 reaching or during the last iteration, GCC will warn about the 5134 undefined behavior in the loop instead of deriving lower upper bound of 5135 the number of iterations for the loop. The warning can be disabled with 5136 -Wno-aggressive-loop-optimizations. 5137 5138 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 5139 for the layout of vectors that could lead to wrong code being 5140 generated. Vectors larger than 8 bytes in size are now by default 5141 aligned to an 8-byte boundary. This is an ABI change: code that makes 5142 explicit use of vector types may be incompatible with binary objects 5143 built with older versions of GCC. Auto-vectorized code is not affected 5144 by this change. 5145 5146 On AVR, support has been removed for the command-line option 5147 -mshort-calls deprecated in GCC 4.7. 5148 5149 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2 5150 is turned on per default for all non-RTEMS configurations. This option 5151 arranges for a better integration of [3]AVR Libc with avr-gcc. For 5152 technical details, see [4]PR54461. To turn off the option in non-RTEMS 5153 configurations, use --with-avrlibc=no. If the compiler is configured 5154 for RTEMS, the option is always turned off. 5155 5156 More information on porting to GCC 4.8 from previous versions of GCC 5157 can be found in the [5]porting guide for this release. 5158 5159 General Optimizer Improvements (and Changes) 5160 5161 * DWARF4 is now the default when generating DWARF debug information. 5162 When -g is used on a platform that uses DWARF debugging 5163 information, GCC will now default to -gdwarf-4 5164 -fno-debug-types-section. 5165 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 5166 consumers support DWARF4 by default. Before GCC 4.8 the default 5167 version used was DWARF2. To make GCC 4.8 generate an older DWARF 5168 version use -g together with -gdwarf-2 or -gdwarf-3. The default 5169 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 5170 * A new general optimization level, -Og, has been introduced. It 5171 addresses the need for fast compilation and a superior debugging 5172 experience while providing a reasonable level of run-time 5173 performance. Overall experience for development should be better 5174 than the default optimization level -O0. 5175 * A new option -ftree-partial-pre was added to control the partial 5176 redundancy elimination (PRE) optimization. This option is enabled 5177 by default at the -O3 optimization level, and it makes PRE more 5178 aggressive. 5179 * The option -fconserve-space has been removed; it was no longer 5180 useful on most targets since GCC supports putting variables into 5181 BSS without making them common. 5182 * The struct reorg and matrix reorg optimizations (command-line 5183 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 5184 removed. They did not always work correctly, nor did they work with 5185 link-time optimization (LTO), hence were only applicable to 5186 programs consisting of a single translation unit. 5187 * Several scalability bottle-necks have been removed from GCC's 5188 optimization passes. Compilation of extremely large functions, e.g. 5189 due to the use of the flatten attribute in the "Eigen" C++ linear 5190 algebra templates library, is significantly faster than previous 5191 releases of GCC. 5192 * Link-time optimization (LTO) improvements: 5193 + LTO partitioning has been rewritten for better reliability and 5194 maintanibility. Several important bugs leading to link 5195 failures have been fixed. 5196 * Interprocedural optimization improvements: 5197 + A new symbol table has been implemented. It builds on existing 5198 callgraph and varpool modules and provide a new API. Unusual 5199 symbol visibilities and aliases are handled more consistently 5200 leading to, for example, more aggressive unreachable code 5201 removal with LTO. 5202 + The inline heuristic can now bypass limits on the size of of 5203 inlined functions when the inlining is particularly 5204 profitable. This happens, for example, when loop bounds or 5205 array strides get propagated. 5206 + Values passed through aggregates (either by value or 5207 reference) are now propagated at the inter-procedural level 5208 leading to better inlining decisions (for example in the case 5209 of Fortran array descriptors) and devirtualization. 5210 * [6]AddressSanitizer , a fast memory error detector, has been added 5211 and can be enabled via -fsanitize=address. Memory access 5212 instructions will be instrumented to detect heap-, stack-, and 5213 global-buffer overflow as well as use-after-free bugs. To get nicer 5214 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 5215 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 5216 x86-64 Darwin. 5217 * [7]ThreadSanitizer has been added and can be enabled via 5218 -fsanitize=thread. Instructions will be instrumented to detect data 5219 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 5220 * A new local register allocator (LRA) has been implemented, which 5221 replaces the 26 year old reload pass and improves generated code 5222 quality. For now it is active on the IA-32 and x86-64 targets. 5223 * Support for transactional memory has been implemented on the 5224 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 5225 Alpha. 5226 5227 New Languages and Language specific improvements 5228 5229 C family 5230 5231 * Each diagnostic emitted now includes the original source line and a 5232 caret '^' indicating the column. The option 5233 -fno-diagnostics-show-caret suppresses this information. 5234 * The option -ftrack-macro-expansion=2 is now enabled by default. 5235 This allows the compiler to display the macro expansion stack in 5236 diagnostics. Combined with the caret information, an example 5237 diagnostic showing these two features is: 5238 5239 t.c:1:94: error: invalid operands to binary < (have struct mystruct and float 5240 ) 5241 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 5242 _b = (B); __a < __b ? __b : __a; }) 5243 5244 ^ 5245 t.c:7:7: note: in expansion of macro 'MYMAX' 5246 X = MYMAX(P, F); 5247 ^ 5248 5249 * A new -Wsizeof-pointer-memaccess warning has been added (also 5250 enabled by -Wall) to warn about suspicious length parameters to 5251 certain string and memory built-in functions if the argument uses 5252 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 5253 (ptr)); if ptr is not an array, but a pointer, and suggests a 5254 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 5255 * The new option -Wpedantic is an alias for -pedantic, which is now 5256 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 5257 -Wno-error=pedantic work in the same way as for any other -W 5258 option. One caveat is that -Werror=pedantic is not equivalent to 5259 -pedantic-errors, since the latter makes into errors some warnings 5260 that are not controlled by -Wpedantic, and the former only affects 5261 diagnostics that are disabled when using -Wno-pedantic. 5262 * The option -Wshadow no longer warns if a declaration shadows a 5263 function declaration, unless the former declares a function or 5264 pointer to function, because this is [8]a common and valid case in 5265 real-world code. 5266 5267 C++ 5268 5269 * G++ now implements the [9]C++11 thread_local keyword; this differs 5270 from the GNU __thread keyword primarily in that it allows dynamic 5271 initialization and destruction semantics. Unfortunately, this 5272 support requires a run-time penalty for references to 5273 non-function-local thread_local variables defined in a different 5274 translation unit even if they don't need dynamic initialization, so 5275 users may want to continue to use __thread for TLS variables with 5276 static initialization semantics. 5277 If the programmer can be sure that no use of the variable in a 5278 non-defining TU needs to trigger dynamic initialization (either 5279 because the variable is statically initialized, or a use of the 5280 variable in the defining TU will be executed before any uses in 5281 another TU), they can avoid this overhead with the 5282 -fno-extern-tls-init option. 5283 OpenMP threadprivate variables now also support dynamic 5284 initialization and destruction by the same mechanism. 5285 * G++ now implements the [10]C++11 attribute syntax, e.g. 5286 5287 [[noreturn]] void f(); 5288 5289 and also the alignment specifier, e.g. 5290 5291 alignas(double) int i; 5292 5293 * G++ now implements [11]C++11 inheriting constructors, e.g. 5294 5295 struct A { A(int); }; 5296 struct B: A { using A::A; }; // defines B::B(int) 5297 B b(42); // OK 5298 5299 * As of GCC 4.8.1, G++ implements the change to decltype semantics 5300 from [12]N3276. 5301 5302 struct A f(); 5303 decltype(f()) g(); // OK, return type of f() is not required to be complete. 5304 5305 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g. 5306 5307 struct A { int f() &; }; 5308 int i = A().f(); // error, f() requires an lvalue object 5309 5310 * G++ now supports a -std=c++1y option for experimentation with 5311 features proposed for the next revision of the standard, expected 5312 around 2014. Currently the only difference from -std=c++11 is 5313 support for return type deduction in normal functions, as proposed 5314 in [14]N3386. Status of C++1y features in GCC 4.8 can be found 5315 [15]here. 5316 * The G++ namespace association extension, __attribute ((strong)), 5317 has been deprecated. Inline namespaces should be used instead. 5318 * G++ now supports a -fext-numeric-literal option to control whether 5319 GNU numeric literal suffixes are accepted as extensions or 5320 processed as C++11 user-defined numeric literal suffixes. The flag 5321 is on (use suffixes for GNU literals) by default for -std=gnu++*, 5322 and -std=c++98. The flag is off (use suffixes for user-defined 5323 literals) by default for -std=c++11 and later. 5324 5325 Runtime Library (libstdc++) 5326 5327 * [16]Improved experimental support for the new ISO C++ standard, 5328 C++11, including: 5329 + forward_list meets the allocator-aware container requirements; 5330 + this_thread::sleep_for(), this_thread::sleep_until() and 5331 this_thread::yield() are defined without requiring the 5332 configure option --enable-libstdcxx-time; 5333 * Improvements to <random>: 5334 + SSE optimized normal_distribution. 5335 + Use of hardware RNG instruction for random_device on new x86 5336 processors (requires the assembler to support the 5337 instruction.) 5338 and <ext/random>: 5339 + New random number engine simd_fast_mersenne_twister_engine 5340 with an optimized SSE implementation. 5341 + New random number distributions beta_distribution, 5342 normal_mv_distribution, rice_distribution, 5343 nakagami_distribution, pareto_distribution, k_distribution, 5344 arcsine_distribution, hoyt_distribution. 5345 * Added --disable-libstdcxx-verbose configure option to disable 5346 diagnostic messages issued when a process terminates abnormally. 5347 This may be useful for embedded systems to reduce the size of 5348 executables that link statically to the library. 5349 5350 Fortran 5351 5352 * Compatibility notice: 5353 + Module files: The version of module files (.mod) has been 5354 incremented. Fortran MODULEs compiled by earlier GCC versions 5355 have to be recompiled, when they are USEd by files compiled 5356 with GCC 4.8. GCC 4.8 is not able to read .mod files created 5357 by earlier versions; attempting to do so gives an error 5358 message. 5359 Note: The ABI of the produced assembler data itself has not 5360 changed; object files and libraries are fully compatible with 5361 older versions except as noted below. 5362 + ABI: Some internal names (used in the assembler/object file) 5363 have changed for symbols declared in the specification part of 5364 a module. If an affected module or a file using it via use 5365 association is recompiled, the module and all files which 5366 directly use such symbols have to be recompiled as well. This 5367 change only affects the following kind of module symbols: 5368 o Procedure pointers. Note: C-interoperable function 5369 pointers (type(c_funptr)) are not affected nor are 5370 procedure-pointer components. 5371 o Deferred-length character strings. 5372 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 5373 backtrace at an arbitrary place in user code; program execution 5374 continues normally afterwards. 5375 * The [18]-Wc-binding-type warning option has been added (disabled by 5376 default). It warns if the a variable might not be C interoperable; 5377 in particular, if the variable has been declared using an intrinsic 5378 type with default kind instead of using a kind parameter defined 5379 for C interoperability in the intrinsic ISO_C_Binding module. 5380 Before, this warning was always printed. The -Wc-binding-type 5381 option is enabled by -Wall. 5382 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 5383 options have been added, which diagnose when code is inserted for 5384 automatic (re)allocation of a variable during assignment. This 5385 option can be used to decide whether it is safe to use 5386 [20]-fno-realloc-lhs. Additionally, it can be used to find 5387 automatic (re)allocation in hot loops. (For arrays, replacing 5388 "var=" by "var(:)=" disables the automatic reallocation.) 5389 * The [21]-Wcompare-reals command-line option has been added. When 5390 this is set, warnings are issued when comparing REAL or COMPLEX 5391 types for equality and inequality; consider replacing a == b by 5392 abs(ab) < eps with a suitable eps. -Wcompare-reals is enabled by 5393 -Wextra. 5394 * The [22]-Wtarget-lifetime command-line option has been added 5395 (enabled with -Wall), which warns if the pointer in a pointer 5396 assignment might outlive its target. 5397 * Reading floating point numbers which use "q" for the exponential 5398 (such as 4.0q0) is now supported as vendor extension for better 5399 compatibility with old data files. It is strongly recommended to 5400 use for I/O the equivalent but standard conforming "e" (such as 5401 4.0e0). 5402 (For Fortran source code, consider replacing the "q" in 5403 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 5404 suitable qp). Note that in Fortran source code replacing "q" by 5405 a simple "e" is not equivalent.) 5406 * The GFORTRAN_TMPDIR environment variable for specifying a 5407 non-default directory for files opened with STATUS="SCRATCH", is 5408 not used anymore. Instead gfortran checks the POSIX/GNU standard 5409 TMPDIR environment variable. If TMPDIR is not defined, gfortran 5410 falls back to other methods to determine the directory for 5411 temporary files as documented in the [23]user manual. 5412 * [24]Fortran 2003: 5413 + Support for unlimited polymorphic variables (CLASS(*)) has 5414 been added. Nonconstant character lengths are not yet 5415 supported. 5416 * [25]TS 29113: 5417 + Assumed types (TYPE(*)) are now supported. 5418 + Experimental support for assumed-rank arrays (dimension(..)) 5419 has been added. Note that currently gfortran's own array 5420 descriptor is used, which is different from the one defined in 5421 TS29113, see [26]gfortran's header file or use the [27]Chasm 5422 Language Interoperability Tools. 5423 5424 Go 5425 5426 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 5427 release. 5428 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 5429 release. The library support is not quite complete. 5430 * Go has been tested on GNU/Linux and Solaris platforms for various 5431 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 5432 work on other platforms as well. 5433 5434 New Targets and Target Specific Improvements 5435 5436 AArch64 5437 5438 * A new port has been added to support AArch64, the new 64-bit 5439 architecture from ARM. Note that this is a separate port from the 5440 existing 32-bit ARM port. 5441 * The port provides initial support for the Cortex-A53 and the 5442 Cortex-A57 processors with the command line options 5443 -mcpu=cortex-a53 and -mcpu=cortex-a57. 5444 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 5445 has been added and can be enabled by giving the 5446 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 5447 default by configuring GCC with the --enable-fix-cortex-a53-835769 5448 option. 5449 5450 ARM 5451 5452 * Initial support has been added for the AArch32 extensions defined 5453 in the ARMv8 architecture. 5454 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 5455 * A new option, -mcpu=marvell-pj4, has been added to generate code 5456 for the Marvell PJ4 processor. 5457 * The compiler can now automatically generate the VFMA, VFMS, REVSH 5458 and REV16 instructions. 5459 * A new vectorizer cost model for Advanced SIMD configurations to 5460 improve the auto-vectorization strategies used. 5461 * The scheduler now takes into account the number of live registers 5462 to reduce the amount of spilling that can occur. This should 5463 improve code performance in large functions. The limit can be 5464 removed by using the option -fno-sched-pressure. 5465 * Improvements have been made to the Marvell iWMMX code generation 5466 and support for the iWMMX2 SIMD unit has been added. The option 5467 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 5468 * A number of code generation improvements for Thumb2 to reduce code 5469 size when compiling for the M-profile processors. 5470 * The RTEMS (arm-rtems) port has been updated to use the EABI. 5471 * Code generation support for the old FPA and Maverick floating-point 5472 architectures has been removed. Ports that previously relied on 5473 these features have also been removed. This includes the targets: 5474 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 5475 + arm*-*-elf (use arm*-*-eabi) 5476 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 5477 + arm*-*-ecos-elf (no alternative) 5478 + arm*-*-freebsd (no alternative) 5479 + arm*-wince-pe* (no alternative). 5480 5481 AVR 5482 5483 * Support for the "Embedded C" fixed-point has been added. For 5484 details, see the [28]GCC wiki and the [29]user manual. The support 5485 is not complete. 5486 * A new print modifier %r for register operands in inline assembler 5487 is supported. It will print the raw register number without the 5488 register prefix 'r': 5489 /* Return the most significant byte of 'val', a 64-bit value. */ 5490 5491 unsigned char msb (long long val) 5492 { 5493 unsigned char c; 5494 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 5495 return c; 5496 } 5497 The inline assembler in this example will generate code like 5498 mov r24, 8+7 5499 provided c is allocated to R24 and val is allocated to R8R15. This 5500 works because the GNU assembler accepts plain register numbers 5501 without register prefix. 5502 * Static initializers with 3-byte symbols are supported now: 5503 extern const __memx char foo; 5504 const __memx void *pfoo = &foo; 5505 This requires at least Binutils 2.23. 5506 5507 IA-32/x86-64 5508 5509 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 5510 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 5511 stack alignment, this is ABI incompatible and intended to be used 5512 in controlled environments where stack space is an important 5513 limitation. This option will lead to wrong code when functions 5514 compiled with 16 byte stack alignment (such as functions from a 5515 standard library) are called with misaligned stack. In this case, 5516 SSE instructions may lead to misaligned memory access traps. In 5517 addition, variable arguments will be handled incorrectly for 16 5518 byte aligned objects (including x87 long double and __int128), 5519 leading to wrong results. You must build all modules with 5520 -mpreferred-stack-boundary=3, including any libraries. This 5521 includes the system libraries and startup modules. 5522 * Support for the new Intel processor codename Broadwell with RDSEED, 5523 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 5524 -mrdseed command-line options. 5525 * Support for the Intel RTM and HLE intrinsics, built-in functions 5526 and code generation is available via -mrtm and -mhle. 5527 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 5528 Intrinsics and built-in functions are available via -mfxsr, -mxsave 5529 and -mxsaveopt respectively. 5530 * New -maddress-mode=[short|long] options for x32. 5531 -maddress-mode=short overrides default 64-bit addresses to 32-bit 5532 by emitting the 0x67 address-size override prefix. This is the 5533 default address mode for x32. 5534 * New built-in functions to detect run-time CPU type and ISA: 5535 + A built-in function __builtin_cpu_is has been added to detect 5536 if the run-time CPU is of a particular type. It returns a 5537 positive integer on a match and zero otherwise. It accepts one 5538 string literal argument, the CPU name. For example, 5539 __builtin_cpu_is("westmere") returns a positive integer if the 5540 run-time CPU is an Intel Core i7 Westmere processor. Please 5541 refer to the [30]user manual for the list of valid CPU names 5542 recognized. 5543 + A built-in function __builtin_cpu_supports has been added to 5544 detect if the run-time CPU supports a particular ISA feature. 5545 It returns a positive integer on a match and zero otherwise. 5546 It accepts one string literal argument, the ISA feature. For 5547 example, __builtin_cpu_supports("ssse3") returns a positive 5548 integer if the run-time CPU supports SSSE3 instructions. 5549 Please refer to the [31]user manual for the list of valid ISA 5550 names recognized. 5551 Caveat: If these built-in functions are called before any static 5552 constructors are invoked, like during IFUNC initialization, then 5553 the CPU detection initialization must be explicitly run using this 5554 newly provided built-in function, __builtin_cpu_init. The 5555 initialization needs to be done only once. For example, this is how 5556 the invocation would look like inside an IFUNC initializer: 5557 static void (*some_ifunc_resolver(void))(void) 5558 { 5559 __builtin_cpu_init(); 5560 if (__builtin_cpu_is("amdfam10h") ... 5561 if (__builtin_cpu_supports("popcnt") ... 5562 } 5563 5564 * Function Multiversioning Support with G++: 5565 It is now possible to create multiple function versions each 5566 targeting a specific processor and/or ISA. Function versions have 5567 the same signature but different target attributes. For example, 5568 here is a program with function versions: 5569 __attribute__ ((target ("default"))) 5570 int foo(void) 5571 { 5572 return 1; 5573 } 5574 5575 __attribute__ ((target ("sse4.2"))) 5576 int foo(void) 5577 { 5578 return 2; 5579 } 5580 5581 int main (void) 5582 { 5583 int (*p) = &foo; 5584 assert ((*p)() == foo()); 5585 return 0; 5586 } 5587 5588 Please refer to this [32]wiki for more information. 5589 * The x86 back end has been improved to allow option -fschedule-insns 5590 to work reliably. This option can be used to schedule instructions 5591 better and leads to improved performace in certain cases. 5592 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 5593 from the Mingw-w64 trunk. 5594 * Support for new AMD family 15h processors (Steamroller core) is now 5595 available through the -march=bdver3 and -mtune=bdver3 options. 5596 * Support for new AMD family 16h processors (Jaguar core) is now 5597 available through the -march=btver2 and -mtune=btver2 options. 5598 5599 FRV 5600 5601 * This target now supports the -fstack-usage command-line option. 5602 5603 MIPS 5604 5605 * GCC can now generate code specifically for the R4700, Broadcom XLP 5606 and MIPS 34kn processors. The associated -march options are 5607 -march=r4700, -march=xlp and -march=34kn respectively. 5608 * GCC now generates better DSP code for MIPS 74k cores thanks to 5609 further scheduling optimizations. 5610 * The MIPS port now supports the -fstack-check option. 5611 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 5612 * Previous versions of GCC would silently accept -fpic and -fPIC for 5613 -mno-abicalls targets like mips*-elf. This combination was not 5614 intended or supported, and did not generate position-independent 5615 code. GCC 4.8 now reports an error when this combination is used. 5616 5617 PowerPC / PowerPC64 / RS6000 5618 5619 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 5620 restore or update the VRSAVE register by default. The respective 5621 operating systems manage the VRSAVE register directly. 5622 * Large TOC support has been added for AIX through the command line 5623 option -mcmodel=large. 5624 * Native Thread-Local Storage support has been added for AIX. 5625 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 5626 when targetting processors that support those hardware features on 5627 AIX 6.1 and above. 5628 5629 RX 5630 5631 * This target will now issue a warning message whenever multiple fast 5632 interrupt handlers are found in the same compilation unit. This 5633 feature can be turned off by the new 5634 -mno-warn-multiple-fast-interrupts command-line option. 5635 5636 S/390, System z 5637 5638 * Support for the IBM zEnterprise zEC12 processor has been added. 5639 When using the -march=zEC12 option, the compiler will generate code 5640 making use of the following new instructions: 5641 + load and trap instructions 5642 + 2 new compare and trap instructions 5643 + rotate and insert selected bits - without CC clobber 5644 The -mtune=zEC12 option enables zEC12 specific instruction 5645 scheduling without making use of new instructions. 5646 * Register pressure sensitive instruction scheduling is enabled by 5647 default. 5648 * The ifunc function attribute is enabled by default. 5649 * memcpy and memcmp invokations on big memory chunks or with run time 5650 lengths are not generated inline anymore when tuning for z10 or 5651 higher. The purpose is to make use of the IFUNC optimized versions 5652 in Glibc. 5653 5654 SH 5655 5656 * The default alignment settings have been reduced to be less 5657 aggressive. This results in more compact code for optimization 5658 levels other than -Os. 5659 * Improved support for the __atomic built-in functions: 5660 + A new option -matomic-model=model selects the model for the 5661 generated atomic sequences. The following models are 5662 supported: 5663 5664 soft-gusa 5665 Software gUSA sequences (SH3* and SH4* only). On 5666 SH4A targets this will now also partially utilize 5667 the movco.l and movli.l instructions. This is the 5668 default when the target is sh3*-*-linux* or 5669 sh4*-*-linux*. 5670 5671 hard-llcs 5672 Hardware movco.l / movli.l sequences (SH4A only). 5673 5674 soft-tcb 5675 Software thread control block sequences. 5676 5677 soft-imask 5678 Software interrupt flipping sequences (privileged 5679 mode only). This is the default when the target is 5680 sh1*-*-linux* or sh2*-*-linux*. 5681 5682 none 5683 Generates function calls to the respective __atomic 5684 built-in functions. This is the default for SH64 5685 targets or when the target is not sh*-*-linux*. 5686 5687 + The option -msoft-atomic has been deprecated. It is now an 5688 alias for -matomic-model=soft-gusa. 5689 + A new option -mtas makes the compiler generate the tas.b 5690 instruction for the __atomic_test_and_set built-in function 5691 regardless of the selected atomic model. 5692 + The __sync functions in libgcc now reflect the selected atomic 5693 model when building the toolchain. 5694 * Added support for the mov.b and mov.w instructions with 5695 displacement addressing. 5696 * Added support for the SH2A instructions movu.b and movu.w. 5697 * Various improvements to code generated for integer arithmetic. 5698 * Improvements to conditional branches and code that involves the T 5699 bit. A new option -mzdcbranch tells the compiler to favor 5700 zero-displacement branches. This is enabled by default for SH4* 5701 targets. 5702 * The pref instruction will now be emitted by the __builtin_prefetch 5703 built-in function for SH3* targets. 5704 * The fmac instruction will now be emitted by the fmaf standard 5705 function and the __builtin_fmaf built-in function. 5706 * The -mfused-madd option has been deprecated in favor of the 5707 machine-independent -ffp-contract option. Notice that the fmac 5708 instruction will now be generated by default for expressions like a 5709 * b + c. This is due to the compiler default setting 5710 -ffp-contract=fast. 5711 * Added new options -mfsrra and -mfsca to allow the compiler using 5712 the fsrra and fsca instructions on targets other than SH4A (where 5713 they are already enabled by default). 5714 * Added support for the __builtin_bswap32 built-in function. It is 5715 now expanded as a sequence of swap.b and swap.w instructions 5716 instead of a library function call. 5717 * The behavior of the -mieee option has been fixed and the negative 5718 form -mno-ieee has been added to control the IEEE conformance of 5719 floating point comparisons. By default -mieee is now enabled and 5720 the option -ffinite-math-only implicitly sets -mno-ieee. 5721 * Added support for the built-in functions __builtin_thread_pointer 5722 and __builtin_set_thread_pointer. This assumes that GBR is used to 5723 hold the thread pointer of the current thread. Memory loads and 5724 stores relative to the address returned by __builtin_thread_pointer 5725 will now also utilize GBR based displacement address modes. 5726 * The -mdiv= option for targets other than SHmedia has been fixed and 5727 documented. 5728 5729 SPARC 5730 5731 * Added optimized instruction scheduling for Niagara4. 5732 5733 TILE-Gx 5734 5735 * Added support for the -mcmodel=MODEL command-line option. The 5736 models supported are small and large. 5737 5738 V850 5739 5740 * This target now supports the E3V5 architecture via the use of the 5741 new -mv850e3v5 command-line option. It also has experimental 5742 support for the e3v5 LOOP instruction which can be enabled via the 5743 new -mloop command-line option. 5744 5745 XStormy16 5746 5747 * This target now supports the -fstack-usage command-line option. 5748 5749 Operating Systems 5750 5751 OpenBSD 5752 5753 * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and 5754 support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated. 5755 5756 Windows (Cygwin) 5757 5758 * Executables are now linked against shared libgcc by default. The 5759 previous default was to link statically, which can still be done by 5760 explicitly specifying -static or static-libgcc on the command line. 5761 However it is strongly advised against, as it will cause problems 5762 for any application that makes use of DLLs compiled by GCC. It 5763 should be alright for a monolithic stand-alone application that 5764 only links against the Windows DLLs, but offers little or no 5765 benefit. 5766 5767 GCC 4.8.1 5768 5769 This is the [33]list of problem reports (PRs) from GCC's bug tracking 5770 system that are known to be fixed in the 4.8.1 release. This list might 5771 not be complete (that is, it is possible that some PRs that have been 5772 fixed are not listed here). 5773 5774 The C++11 <chrono> std::chrono::system_clock and 5775 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 5776 both are now separate (never typedefs of each other), both use 5777 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 5778 std::chrono::steady_clock is now finally monotonic, and both classes 5779 are mangled differently than in the previous GCC releases. 5780 std::chrono::system_clock::now() with std::chrono::microseconds resp. 5781 std::chrono::seconds resolution is still exported for backwards 5782 compatibility with default configured libstdc++. Note that libstdc++ 5783 configured with --enable-libstdcxx-time= used to be ABI incompatible 5784 with default configured libstdc++ for those two classes and no ABI 5785 compatibility can be offered for those configurations, so any C++11 5786 code that uses those classes and has been compiled and linked against 5787 libstdc++ configured with the non-default --enable-libstdcxx-time= 5788 configuration option needs to be recompiled. 5789 5790 GCC 4.8.2 5791 5792 This is the [34]list of problem reports (PRs) from GCC's bug tracking 5793 system that are known to be fixed in the 4.8.2 release. This list might 5794 not be complete (that is, it is possible that some PRs that have been 5795 fixed are not listed here). 5796 5797 GCC 4.8.3 5798 5799 This is the [35]list of problem reports (PRs) from GCC's bug tracking 5800 system that are known to be fixed in the 4.8.3 release. This list might 5801 not be complete (that is, it is possible that some PRs that have been 5802 fixed are not listed here). 5803 5804 Support for the new powerpc64le-linux platform has been added. It 5805 defaults to generating code that conforms to the ELFV2 ABI. 5806 5807 GCC 4.8.4 5808 5809 This is the [36]list of problem reports (PRs) from GCC's bug tracking 5810 system that are known to be fixed in the 4.8.4 release. This list might 5811 not be complete (that is, it is possible that some PRs that have been 5812 fixed are not listed here). 5813 5814 GCC 4.8.5 5815 5816 This is the [37]list of problem reports (PRs) from GCC's bug tracking 5817 system that are known to be fixed in the 4.8.5 release. This list might 5818 not be complete (that is, it is possible that some PRs that have been 5819 fixed are not listed here). 5820 5821 5822 For questions related to the use of GCC, please consult these web 5823 pages and the [38]GCC manuals. If that fails, the 5824 [39]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5825 web pages and the development of GCC are welcome on our developer 5826 list at [40]gcc (a] gcc.gnu.org. All of [41]our lists have public 5827 archives. 5828 5829 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and 5830 distribution of this entire article is permitted in any medium, 5831 provided this notice is preserved. 5832 5833 These pages are [43]maintained by the GCC team. Last modified 5834 2019-11-28[44]. 5835 5836 References 5837 5838 1. https://gcc.gnu.org/wiki/cxx-conversion 5839 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 5840 3. http://www.nongnu.org/avr-libc/ 5841 4. https://gcc.gnu.org/PR54461 5842 5. https://gcc.gnu.org/gcc-4.8/porting_to.html 5843 6. https://github.com/google/sanitizers 5844 7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 5845 8. https://lkml.org/lkml/2006/11/28/239 5846 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 5847 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 5848 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 5849 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 5850 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 5851 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 5852 15. http://gcc.gnu.org/projects/cxx1y.html 5853 16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 5854 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 5855 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 5856 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 5857 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 5858 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 5859 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 5860 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 5861 24. https://gcc.gnu.org/wiki/Fortran2003Status 5862 25. https://gcc.gnu.org/wiki/TS29113Status 5863 26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text/plain&view=co 5864 27. http://chasm-interop.sourceforge.net/ 5865 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 5866 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 5867 30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 5868 31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 5869 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning 5870 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 5871 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 5872 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 5873 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 5874 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 5875 38. https://gcc.gnu.org/onlinedocs/ 5876 39. mailto:gcc-help (a] gcc.gnu.org 5877 40. mailto:gcc (a] gcc.gnu.org 5878 41. https://gcc.gnu.org/lists.html 5879 42. https://www.fsf.org/ 5880 43. https://gcc.gnu.org/about.html 5881 44. http://validator.w3.org/check/referer 5882 ====================================================================== 5883 http://gcc.gnu.org/gcc-4.7/index.html 5884 GCC 4.7 Release Series 5885 5886 (This release series is no longer supported.) 5887 5888 June 12, 2014 5889 5890 The [1]GNU project and the GCC developers are pleased to announce the 5891 release of GCC 4.7.4. 5892 5893 This release is a bug-fix release, containing fixes for regressions in 5894 GCC 4.7.3 relative to previous releases of GCC. 5895 5896 Release History 5897 5898 GCC 4.7.4 5899 June 12, 2014 ([2]changes, [3]documentation) 5900 5901 GCC 4.7.3 5902 April 11, 2013 ([4]changes, [5]documentation) 5903 5904 GCC 4.7.2 5905 September 20, 2012 ([6]changes, [7]documentation) 5906 5907 GCC 4.7.1 5908 June 14, 2012 ([8]changes, [9]documentation) 5909 5910 GCC 4.7.0 5911 March 22, 2012 ([10]changes, [11]documentation) 5912 5913 References and Acknowledgements 5914 5915 GCC used to stand for the GNU C Compiler, but since the compiler 5916 supports several other languages aside from C, it now stands for the 5917 GNU Compiler Collection. 5918 5919 A list of [12]successful builds is updated as new information becomes 5920 available. 5921 5922 The GCC developers would like to thank the numerous people that have 5923 contributed new features, improvements, bug fixes, and other changes as 5924 well as test results to GCC. This [13]amazing group of volunteers is 5925 what makes GCC successful. 5926 5927 For additional information about GCC please refer to the [14]GCC 5928 project web site or contact the [15]GCC development mailing list. 5929 5930 To obtain GCC please use [16]our mirror sites or [17]our version 5931 control system. 5932 5933 5934 For questions related to the use of GCC, please consult these web 5935 pages and the [18]GCC manuals. If that fails, the 5936 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5937 web pages and the development of GCC are welcome on our developer 5938 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 5939 archives. 5940 5941 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 5942 distribution of this entire article is permitted in any medium, 5943 provided this notice is preserved. 5944 5945 These pages are [23]maintained by the GCC team. Last modified 5946 2020-01-14[24]. 5947 5948 References 5949 5950 1. http://www.gnu.org/ 5951 2. http://gcc.gnu.org/gcc-4.7/changes.html 5952 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 5953 4. http://gcc.gnu.org/gcc-4.7/changes.html 5954 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 5955 6. http://gcc.gnu.org/gcc-4.7/changes.html 5956 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 5957 8. http://gcc.gnu.org/gcc-4.7/changes.html 5958 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 5959 10. http://gcc.gnu.org/gcc-4.7/changes.html 5960 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 5961 12. http://gcc.gnu.org/gcc-4.7/buildstat.html 5962 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5963 14. http://gcc.gnu.org/index.html 5964 15. mailto:gcc (a] gcc.gnu.org 5965 16. http://gcc.gnu.org/mirrors.html 5966 17. http://gcc.gnu.org/git.html 5967 18. https://gcc.gnu.org/onlinedocs/ 5968 19. mailto:gcc-help (a] gcc.gnu.org 5969 20. mailto:gcc (a] gcc.gnu.org 5970 21. https://gcc.gnu.org/lists.html 5971 22. https://www.fsf.org/ 5972 23. https://gcc.gnu.org/about.html 5973 24. http://validator.w3.org/check/referer 5974 ====================================================================== 5975 http://gcc.gnu.org/gcc-4.7/changes.html 5976 GCC 4.7 Release Series 5977 Changes, New Features, and Fixes 5978 5979 Caveats 5980 5981 * The -fconserve-space flag has been deprecated. The flag had no 5982 effect for most targets: only targets without a global .bss section 5983 and without support for switchable sections. Furthermore, the flag 5984 only had an effect for G++, where it could result in wrong 5985 semantics (please refer to the GCC manual for further details). The 5986 flag will be removed in GCC 4.8 5987 * Support for a number of older systems and recently unmaintained or 5988 untested target ports of GCC has been declared obsolete in GCC 4.7. 5989 Unless there is activity to revive them, the next release of GCC 5990 will have their sources permanently removed. 5991 All GCC ports for the following processor architectures have been 5992 declared obsolete: 5993 + picoChip (picochip-*) 5994 The following ports for individual systems on particular 5995 architectures have been obsoleted: 5996 + IRIX 6.5 (mips-sgi-irix6.5) 5997 + MIPS OpenBSD (mips*-*-openbsd*) 5998 + Solaris 8 (*-*-solaris2.8). Details can be found in the 5999 [1]announcement. 6000 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 6001 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 6002 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 6003 default, which for some sources generates code that accesses memory 6004 on unaligned addresses. This requires the kernel of those systems 6005 to enable such accesses (controlled by CP15 register c1, refer to 6006 ARM documentation). Alternatively, or for compatibility with 6007 kernels where unaligned accesses are not supported, all code has to 6008 be compiled with -mno-unaligned-access. Upstream Linux kernel 6009 releases have automatically and unconditionally supported unaligned 6010 accesses as emitted by GCC due to this option being active since 6011 version 2.6.28. 6012 * Support on ARM for the legacy floating-point accelerator (FPA) and 6013 the mixed-endian floating-point format that it used has been 6014 obsoleted. The ports that still use this format have been obsoleted 6015 as well. Many legacy ARM ports already provide an alternative that 6016 uses the VFP floating-point format. The obsolete ports will be 6017 deleted in the next release. 6018 The obsolete ports with alternatives are: 6019 + arm*-*-rtems (use arm*-*-rtemseabi) 6020 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 6021 + arm*-*-elf (use arm*-*-eabi) 6022 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 6023 Note, however, that these alternatives are not binary compatible 6024 with their legacy counterparts (although some can support running 6025 legacy applications). 6026 The obsolete ports that currently lack a modern alternative are: 6027 + arm*-*-ecos-elf 6028 + arm*-*-freebsd 6029 + arm*-wince-pe* 6030 New ports that support more recent versions of the architecture are 6031 welcome. 6032 * Support for the Maverick co-processor on ARM has been obsoleted. 6033 Code to support it will be deleted in the next release. 6034 * Support has been removed for Unix International threads on Solaris 6035 2, so the --enable-threads=solaris configure option and the 6036 -threads compiler option don't work any longer. 6037 * Support has been removed for the Solaris BSD Compatibility Package, 6038 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 6039 from Solaris 11, and was only intended as a migration aid from 6040 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 6041 recognized any longer. 6042 * The AVR port's libgcc has been improved and its multilib structure 6043 has been enhanced. As a result, all objects contributing to an 6044 application must either be compiled with GCC versions up to 4.6.x 6045 or with GCC versions 4.7.1 or later. If the compiler is used with 6046 AVR Libc, you need a version that supports the new layout, i.e. 6047 implements [2]#35407. 6048 * The AVR port's -mshort-calls command-line option has been 6049 deprecated. It will be removed in the GCC 4.8 release. See -mrelax 6050 for a replacement. 6051 * The AVR port only references startup code that clears .bss and the 6052 common section resp. initializes the .data and .rodata section 6053 provided respective sections (or subsections thereof) are not 6054 empty, see [3]PR18145. Applications that put all static storage 6055 objects into non-standard sections and / or define all static 6056 storage objects in assembler modules, must reference __do_clear_bss 6057 resp. __do_copy_data by hand or undefine the symbol(s) by means of 6058 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 6059 * The ARM port's -mwords-little-endian option has been deprecated. It 6060 will be removed in a future release. 6061 * Support has been removed for the NetWare x86 configuration 6062 obsoleted in GCC 4.6. 6063 * It is no longer possible to use the "l" constraint in MIPS16 asm 6064 statements. 6065 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 6066 library which affected the ABI in C++11 mode: a data member was 6067 added to std::list changing its size and altering the definitions 6068 of some member functions, and std::pair's move constructor was 6069 non-trivial which altered the calling convention for functions with 6070 std::pair arguments or return types. The ABI incompatibilities have 6071 been fixed for GCC version 4.7.2 but as a result C++11 code 6072 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 6073 code compiled with different GCC versions and with C++98/C++03 code 6074 compiled with any version. 6075 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 6076 rules for the layout of vectors that could lead to wrong code being 6077 generated. Vectors larger than 8 bytes in size are now by default 6078 aligned to an 8-byte boundary. This is an ABI change: code that 6079 makes explicit use of vector types may be incompatible with binary 6080 objects built with older versions of GCC. Auto-vectorized code is 6081 not affected by this change. (This change affects GCC versions 6082 4.7.2 and later.) 6083 * More information on porting to GCC 4.7 from previous versions of 6084 GCC can be found in the [4]porting guide for this release. 6085 6086 General Optimizer Improvements 6087 6088 * Support for a new parameter --param case-values-threshold=n was 6089 added to allow users to control the cutoff between doing switch 6090 statements as a series of if statements and using a jump table. 6091 * Link-time optimization (LTO) improvements: 6092 + Improved scalability and reduced memory usage. Link time 6093 optimization of Firefox now requires 3GB of RAM on a 64-bit 6094 system, while over 8GB was needed previously. Linking time has 6095 been improved, too. The serial stage of linking Firefox has 6096 been sped up by about a factor of 10. 6097 + Reduced size of object files and temporary storage used during 6098 linking. 6099 + Streaming performance (both outbound and inbound) has been 6100 improved. 6101 + ld -r is now supported with LTO. 6102 + Several bug fixes, especially in symbol table handling and 6103 merging. 6104 * Interprocedural optimization improvements: 6105 + Heuristics now take into account that after inlining code will 6106 be optimized out because of known values (or properties) of 6107 function parameters. For example: 6108 void foo(int a) 6109 { 6110 if (a > 10) 6111 ... huge code ... 6112 } 6113 void bar (void) 6114 { 6115 foo (0); 6116 } 6117 6118 The call of foo will be inlined into bar even when optimizing 6119 for code size. Constructs based on __builtin_constant_p are 6120 now understood by the inliner and code size estimates are 6121 evaluated a lot more realistically. 6122 + The representation of C++ virtual thunks and aliases (both 6123 implicit and defined via the alias attribute) has been 6124 re-engineered. Aliases no longer pose optimization barriers 6125 and calls to an alias can be inlined and otherwise optimized. 6126 + The inter-procedural constant propagation pass has been 6127 rewritten. It now performs generic function specialization. 6128 For example when compiling the following: 6129 void foo(bool flag) 6130 { 6131 if (flag) 6132 ... do something ... 6133 else 6134 ... do something else ... 6135 } 6136 void bar (void) 6137 { 6138 foo (false); 6139 foo (true); 6140 foo (false); 6141 foo (true); 6142 foo (false); 6143 foo (true); 6144 } 6145 6146 GCC will now produce two copies of foo. One with flag being 6147 true, while other with flag being false. This leads to 6148 performance improvements previously possible only by inlining 6149 all calls. Cloning causes a lot less code size growth. 6150 * A string length optimization pass has been added. It attempts to 6151 track string lengths and optimize various standard C string 6152 functions like strlen, strchr, strcpy, strcat, stpcpy and their 6153 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 6154 enabled by default at -O2 or above, unless optimizing for size, and 6155 can be disabled by the -fno-optimize-strlen option. The pass can 6156 e.g. optimize 6157 char *bar (const char *a) 6158 { 6159 size_t l = strlen (a) + 2; 6160 char *p = malloc (l); if (p == NULL) return p; 6161 strcpy (p, a); strcat (p, "/"); return p; 6162 } 6163 6164 into: 6165 char *bar (const char *a) 6166 { 6167 size_t tmp = strlen (a); 6168 char *p = malloc (tmp + 2); if (p == NULL) return p; 6169 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 6170 } 6171 6172 or for hosted compilations where stpcpy is available in the runtime 6173 and headers provide its prototype, e.g. 6174 void foo (char *a, const char *b, const char *c, const char *d) 6175 { 6176 strcpy (a, b); strcat (a, c); strcat (a, d); 6177 } 6178 6179 can be optimized into: 6180 void foo (char *a, const char *b, const char *c, const char *d) 6181 { 6182 strcpy (stpcpy (stpcpy (a, b), c), d); 6183 } 6184 6185 New Languages and Language specific improvements 6186 6187 * Version 3.1 of the OpenMP specification is now supported for the C, 6188 C++, and Fortran compilers. 6189 6190 Ada 6191 6192 * The command-line option -feliminate-unused-debug-types has been 6193 re-enabled by default, as it is for the other languages, leading to 6194 a reduction in debug info size of 12.5% and more for relevant 6195 cases, as well as to a small compilation speedup. 6196 6197 C family 6198 6199 * A new built-in, __builtin_assume_aligned, has been added, through 6200 which the compiler can be hinted about pointer alignment and can 6201 use it to improve generated code. 6202 * A new warning option -Wunused-local-typedefs was added for C, C++, 6203 Objective-C and Objective-C++. This warning diagnoses typedefs 6204 locally defined in a function, and otherwise not used. 6205 * A new experimental command-line option -ftrack-macro-expansion was 6206 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 6207 the compiler to emit diagnostic about the current macro expansion 6208 stack when a compilation error occurs in a macro expansion. 6209 * Experimental support for transactional memory has been added. It 6210 includes support in the compiler, as well as a supporting runtime 6211 library called libitm. To compile code with transactional memory 6212 constructs, use the -fgnu-tm option. 6213 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 6214 and 32-bit/64-bit x86 platforms. 6215 For more details on transactional memory see [5]the GCC WiKi. 6216 * Support for atomic operations specifying the C++11/C11 memory model 6217 has been added. These new __atomic routines replace the existing 6218 __sync built-in routines. 6219 Atomic support is also available for memory blocks. Lock-free 6220 instructions will be used if a memory block is the same size and 6221 alignment as a supported integer type. Atomic operations which do 6222 not have lock-free support are left as function calls. A set of 6223 library functions is available on the GCC atomic wiki in the 6224 "External Atomics Library" section. 6225 For more details on the memory models and features, see the 6226 [6]atomic wiki. 6227 * When a binary operation is performed on vector types and one of the 6228 operands is a uniform vector, it is possible to replace the vector 6229 with the generating element. For example: 6230 typedef int v4si __attribute__ ((vector_size (16))); 6231 v4si res, a = {1,2,3,4}; 6232 int x; 6233 6234 res = 2 + a; /* means {2,2,2,2} + a */ 6235 res = a - x; /* means a - {x,x,x,x} */ 6236 6237 C 6238 6239 * There is support for some more features from the C11 revision of 6240 the ISO C standard. GCC now accepts the options -std=c11 and 6241 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 6242 + Unicode strings (previously supported only with options such 6243 as -std=gnu11, now supported with -std=c11), and the 6244 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 6245 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 6246 + Alignment support (_Alignas, _Alignof, max_align_t, 6247 <stdalign.h>). 6248 + A built-in function __builtin_complex is provided to support C 6249 library implementation of the CMPLX family of macros. 6250 6251 C++ 6252 6253 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 6254 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 6255 -Wc++0x-compat, respectively. 6256 * G++ now implements [7]C++11 extended friend syntax: 6257 6258 template<class W> 6259 class Q 6260 { 6261 static const int I = 2; 6262 public: 6263 friend W; 6264 }; 6265 6266 struct B 6267 { 6268 int ar[Q<B>::I]; 6269 }; 6270 6271 * Thanks to Ville Voutilainen, G++ now implements [8]C++11 explicit 6272 override control. 6273 6274 struct B { 6275 virtual void f() const final; 6276 virtual void f(int); 6277 }; 6278 6279 struct D : B { 6280 void f() const; // error: D::f attempts to override final B::f 6281 void f(long) override; // error: doesn't override anything 6282 void f(int) override; // ok 6283 }; 6284 6285 struct E final { }; 6286 struct F: E { }; // error: deriving from final class 6287 6288 * G++ now implements [9]C++11 non-static data member initializers. 6289 6290 struct A { 6291 int i = 42; 6292 } a; // initializes a.i to 42 6293 6294 * Thanks to Ed Smith-Rowland, G++ now implements [10]C++11 6295 user-defined literals. 6296 6297 // Not actually a good approximation. :) 6298 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 6299 long double pi = 180.0_degrees; 6300 6301 * G++ now implements [11]C++11 alias-declarations. 6302 6303 template <class T> using Ptr = T*; 6304 Ptr<int> ip; // decltype(ip) is int* 6305 6306 * Thanks to Ville Voutilainen and Pedro Lamaro, G++ now implements 6307 [12]C++11 delegating constructors. 6308 6309 struct A { 6310 A(int); 6311 A(): A(42) { } // delegate to the A(int) constructor 6312 }; 6313 6314 * G++ now fully implements C++11 atomic classes rather than just 6315 integer derived classes. 6316 6317 class POD { 6318 int a; 6319 int b; 6320 }; 6321 std::atomic<POD> my_atomic_POD; 6322 6323 * G++ now sets the predefined macro __cplusplus to the correct value, 6324 199711L for C++98/03, and 201103L for C++11. 6325 * G++ now correctly implements the two-phase lookup rules such that 6326 an unqualified name used in a template must have an appropriate 6327 declaration found either in scope at the point of definition of the 6328 template or by argument-dependent lookup at the point of 6329 instantiation. As a result, code that relies on a second 6330 unqualified lookup at the point of instantiation to find functions 6331 declared after the template or in dependent bases will be rejected. 6332 The compiler will suggest ways to fix affected code, and using the 6333 -fpermissive compiler flag will allow the code to compile with a 6334 warning. 6335 6336 template <class T> 6337 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 6338 void g(int) { } // fix by moving this declaration before the declaration of f 6339 6340 template <class T> 6341 struct A: T { 6342 // error, B::g(B) not found by argument-dependent lookup 6343 void f() { g(T()); } // fix by using this->g or A::g 6344 }; 6345 6346 struct B { void g(B); }; 6347 6348 int main() 6349 { 6350 f<int>(); 6351 A<B>().f(); 6352 } 6353 6354 * G++ now properly re-uses stack space allocated for temporary 6355 objects when their lifetime ends, which can significantly lower 6356 stack consumption for some C++ functions. As a result of this, some 6357 code with undefined behavior will now break: 6358 6359 const int &f(const int &i) { return i; } 6360 .... 6361 const int &x = f(1); 6362 const int &y = f(2); 6363 6364 Here, x refers to the temporary allocated to hold the 1 argument, 6365 which only lives until the end of the initialization; it 6366 immediately becomes a dangling reference. So the next statement 6367 re-uses the stack slot to hold the 2 argument, and users of x get 6368 that value instead. 6369 Note that this should not cause any change of behavior for 6370 temporaries of types with non-trivial destructors, as they are 6371 already destroyed at end of full-expression; the change is that now 6372 the storage is released as well. 6373 * A new command-line option -Wdelete-non-virtual-dtor has been added 6374 to warn when delete is used to destroy an instance of a class which 6375 has virtual functions and non-virtual destructor. It is unsafe to 6376 delete an instance of a derived class through a pointer to a base 6377 class if the base class does not have a virtual destructor. This 6378 warning is enabled by -Wall. 6379 * A new command-line option -Wzero-as-null-pointer-constant has been 6380 added to warn when a literal '0' is used as null pointer constant. 6381 It can be useful to facilitate the conversion to nullptr in C++11. 6382 * As per C++98, access-declarations are now deprecated by G++. 6383 Using-declarations are to be used instead. Furthermore, some 6384 efforts have been made to improve the support of class scope 6385 using-declarations. In particular, using-declarations referring to 6386 a dependent type now work as expected ([13]bug c++/14258). 6387 * The ELF symbol visibility of a template instantiation is now 6388 properly constrained by the visibility of its template arguments 6389 ([14]bug c++/35688). 6390 6391 Runtime Library (libstdc++) 6392 6393 * [15]Improved experimental support for the new ISO C++ standard, 6394 C++11, including: 6395 + using noexcept in most of the library; 6396 + implementations of pointer_traits, allocator_traits and 6397 scoped_allocator_adaptor; 6398 + uses-allocator construction for tuple; 6399 + vector meets the allocator-aware container requirements; 6400 + replacing monotonic_clock with steady_clock; 6401 + enabling the thread support library on most POSIX targets; 6402 + many small improvements to conform to the FDIS. 6403 * Added --enable-clocale=newlib configure option. 6404 * Debug Mode iterators for unordered associative containers. 6405 * Avoid polluting the global namespace and do not include <unistd.h>. 6406 6407 Fortran 6408 6409 * The compile flag [16]-fstack-arrays has been added, which causes 6410 all local arrays to be put on stack memory. For some programs this 6411 will improve the performance significantly. If your program uses 6412 very large local arrays, it is possible that you will have to 6413 extend your runtime limits for stack memory. 6414 * The [17]-Ofast flag now also implies [18]-fno-protect-parens and 6415 [19]-fstack-arrays. 6416 * Front-end optimizations can now be selected by the 6417 [20]-ffrontend-optimize option and deselected by the 6418 -fno-frontend-optimize option. 6419 * When front-end optimization removes a function call, 6420 [21]-Wfunction-elimination warns about that. 6421 * When performing front-end-optimization, the 6422 [22]-faggressive-function-elimination option allows the removal of 6423 duplicate function calls even for impure functions. 6424 * The flag [23]-Wreal-q-constant has been added, which warns if 6425 floating-point literals have been specified using q (such as 6426 1.0q0); the q marker is now supported as a vendor extension to 6427 denote quad precision (REAL(16) or, if not available, REAL(10)). 6428 Consider using a kind parameter (such as in 1.0_qp) instead, which 6429 can be obtained via [24]SELECTED_REAL_KIND. 6430 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 6431 Fortran now always prints error messages to standard error. If you 6432 wish to redirect standard error, please consult the manual for your 6433 OS, shell, batch environment etc. as appropriate. 6434 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 6435 variable have been removed. When encountering a serious error, 6436 gfortran will now always abort the program. Whether a core dump is 6437 generated depends on the user environment settings; see the ulimit 6438 -c setting for POSIX shells, limit coredumpsize for C shells, and 6439 the [25]WER user-mode dumps settings on Windows. 6440 * The [26]-fbacktrace option is now enabled by default. When 6441 encountering a fatal error, gfortran will attempt to print a 6442 backtrace to standard error before aborting. It can be disabled 6443 with -fno-backtrace. Note: On POSIX targets with the addr2line 6444 utility from GNU binutils, GNU Fortran can print a backtrace with 6445 function name, file name, line number information in addition to 6446 the addresses; otherwise only the addresses are printed. 6447 * [27]Fortran 2003: 6448 + Generic interface names which have the same name as derived 6449 types are now supported, which allows to write constructor 6450 functions. Note that Fortran does not support static 6451 constructor functions; only default initialization or an 6452 explicit structure-constructor initialization are available. 6453 + [28]Polymorphic (class) arrays are now supported. 6454 * [29]Fortran 2008: 6455 + Support for the DO CONCURRENT construct has been added, which 6456 allows the user to specify that individual loop iterations 6457 have no interdependencies. 6458 + [30]Coarrays: Full single-image support except for polymorphic 6459 coarrays. Additionally, preliminary support for multiple 6460 images via an MPI-based [31]coarray communication library has 6461 been added. Note: The library version is not yet usable as 6462 remote coarray access is not yet possible. 6463 * [32]TS 29113: 6464 + New flag [33]-std=f2008ts permits programs that are expected 6465 to conform to the Fortran 2008 standard and the draft 6466 Technical Specification (TS) 29113 on Further Interoperability 6467 of Fortran with C. 6468 + The OPTIONAL attribute is now allowed for dummy arguments of 6469 BIND(C) procedures. 6470 + The RANK intrinsic has been added. 6471 + The implementation of the ASYNCHRONOUS attribute in GCC is 6472 compatible with the candidate draft of TS 29113 (since GCC 6473 4.6). 6474 6475 Go 6476 6477 * GCC 4.7 implements the [34]Go 1 language standard. The library 6478 support in 4.7.0 is not quite complete, due to release timing. 6479 Release 4.7.1 includes complete support for Go 1. The Go library is 6480 from the Go 1.0.1 release. 6481 * Go has been tested on GNU/Linux and Solaris platforms. It may work 6482 on other platforms as well. 6483 6484 New Targets and Target Specific Improvements 6485 6486 ARM 6487 6488 * GCC now supports the Cortex-A7 processor implementing the v7-a 6489 version of the architecture using the option -mcpu=cortex-a7. 6490 * The default vector size in auto-vectorization for NEON is now 128 6491 bits. If vectorization fails thusly, the vectorizer tries again 6492 with 64-bit vectors. 6493 * A new option -mvectorize-with-neon-double was added to allow users 6494 to change the vector size to 64 bits. 6495 6496 AVR 6497 6498 * GCC now supports the XMEGA architecture. This requires GNU binutils 6499 2.22 or later. 6500 * Support for the [35]named address spaces __flash, __flash1, , 6501 __flash5 and __memx has been added. These address spaces locate 6502 read-only data in flash memory and allow reading from flash memory 6503 by means of ordinary C code, i.e. without the need of (inline) 6504 assembler code: 6505 6506 const __flash int values[] = { 42, 31 }; 6507 6508 int add_values (const __flash int *p, int i) 6509 { 6510 return values[i] + *p; 6511 } 6512 6513 * Support has been added for the AVR-specific configure option 6514 --with-avrlibc=yes in order to arrange for better integration of 6515 [36]AVR-Libc. This configure option is supported in avr-gcc 4.7.2 6516 and newer and will only take effect in non-RTEMS configurations. If 6517 avr-gcc is configured for RTEMS, the option will be ignored which 6518 is the same as specifying --with-avrlibc=no. See [37]PR54461 for 6519 more technical details. 6520 * Support for AVR-specific [38]built-in functions has been added. 6521 * Support has been added for the signed and unsigned 24-bit scalar 6522 integer types __int24 and __uint24. 6523 * New command-line options -maccumulate-args, -mbranch-cost=cost and 6524 -mstrict-X were added to allow better fine-tuning of code 6525 optimization. 6526 * The command-line option -fdata-sections now also takes affect on 6527 the section names of variables with the progmem attribute. 6528 * A new inline assembler print modifier %i to print a RAM address as 6529 I/O address has been added: 6530 6531 #include <avr/io.h> /* Port Definitions from AVR-LibC */ 6532 6533 void set_portb (uint8_t value) 6534 { 6535 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 6536 } 6537 6538 The offset between an I/O address and the RAM address for that I/O 6539 location is device-specific. This offset is taken into account when 6540 printing a RAM address with the %i modifier so that the address is 6541 suitable to be used as operand in an I/O command. The address must 6542 be a constant integer known at compile time. 6543 * The inline assembler constraint "R" to represent integers in the 6544 range 6 5 has been removed without replacement. 6545 * Many optimizations to: 6546 + 64-bit integer arithmetic 6547 + Widening multiplication 6548 + Integer division by a constant 6549 + Avoid constant reloading in multi-byte instructions. 6550 + Micro-optimizations for special instruction sequences. 6551 + Generic built-in functions like __builtin_ffs*, 6552 __builtin_clz*, etc. 6553 + If-else decision trees generated by switch instructions 6554 + Merging of data located in flash memory 6555 + New libgcc variants for devices with 8-bit wide stack pointer 6556 + 6557 * Better documentation: 6558 + Handling of EIND and indirect jumps on devices with more than 6559 128 KiB of program memory. 6560 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 6561 registers. 6562 + Function attributes OS_main and OS_task. 6563 + AVR-specific built-in macros. 6564 6565 C6X 6566 6567 * Support has been added for the Texas Instruments C6X family of 6568 processors. 6569 6570 CR16 6571 6572 * Support has been added for National Semiconductor's CR16 6573 architecture. 6574 6575 Epiphany 6576 6577 * Support has been added for Adapteva's Epiphany architecture. 6578 6579 IA-32/x86-64 6580 6581 * Support for Intel AVX2 intrinsics, built-in functions and code 6582 generation is available via -mavx2. 6583 * Support for Intel BMI2 intrinsics, built-in functions and code 6584 generation is available via -mbmi2. 6585 * Implementation and automatic generation of __builtin_clz* using the 6586 lzcnt instruction is available via -mlzcnt. 6587 * Support for Intel FMA3 intrinsics and code generation is available 6588 via -mfma. 6589 * A new -mfsgsbase command-line option is available that makes GCC 6590 generate new segment register read/write instructions through 6591 dedicated built-ins. 6592 * Support for the new Intel rdrnd instruction is available via 6593 -mrdrnd. 6594 * Two additional AVX vector conversion instructions are available via 6595 -mf16c. 6596 * Support for new Intel processor codename IvyBridge with RDRND, 6597 FSGSBASE and F16C is available through -march=core-avx-i. 6598 * Support for the new Intel processor codename Haswell with AVX2, 6599 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 6600 * Support for new AMD family 15h processors (Piledriver core) is now 6601 available through -march=bdver2 and -mtune=bdver2 options. 6602 * Support for [39]the x32 psABI is now available through the -mx32 6603 option. 6604 * Windows mingw targets are using the -mms-bitfields option by 6605 default. 6606 * Windows x86 targets are using the __thiscall calling convention for 6607 C++ class-member functions. 6608 * Support for the configure option --with-threads=posix for Windows 6609 mingw targets. 6610 6611 MIPS 6612 6613 * GCC now supports thread-local storage (TLS) for MIPS16. This 6614 requires GNU binutils 2.22 or later. 6615 * GCC can now generate code specifically for the Cavium Octeon+ and 6616 Octeon2 processors. The associated command-line options are 6617 -march=octeon+ and -march=octeon2 respectively. Both options 6618 require GNU binutils 2.22 or later. 6619 * GCC can now work around certain 24k errata, under the control of 6620 the command-line option -mfix-24k. These workarounds require GNU 6621 binutils 2.20 or later. 6622 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 6623 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 6624 toolchain that generates 32-bit code by default. Use the 6625 configure-time option --enable-targets=all to select these extra 6626 multilibs. 6627 * Passing -fno-delayed-branch now also stops the assembler from 6628 automatically filling delay slots. 6629 6630 PowerPC/PowerPC64 6631 6632 * Vectors of type vector long long or vector long are passed and 6633 returned using the same method as other vectors with the VSX 6634 instruction set. Previously GCC did not adhere to the ABI for 6635 128-bit vectors with 64-bit integer base types (PR 48857). This 6636 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 6637 * A new option -mno-pointers-to-nested-functions was added to allow 6638 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 6639 that the compiler should not load up the chain register (r11) 6640 before calling a function through a pointer. If you use this 6641 option, you cannot call nested functions through a pointer, or call 6642 other languages that might use the static chain. 6643 * A new option msave-toc-indirect was added to allow AIX 6644 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 6645 save the TOC in the prologue for indirect calls or generate the 6646 save inline. This can speed up some programs that call through a 6647 function pointer a lot, but it can slow down other functions that 6648 only call through a function pointer in exceptional cases. 6649 * The PowerPC port will now enable machine-specific built-in 6650 functions when the user switches the target machine using the 6651 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 6652 sequences. In addition, the target macros are updated. However, due 6653 to the way the -save-temps switch is implemented, you won't see the 6654 effect of these additional macros being defined in preprocessor 6655 output. 6656 6657 SH 6658 6659 * A new option -msoft-atomic has been added. When it is specified, 6660 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 6661 the new __atomic routines. 6662 * Since it is neither supported by GAS nor officially documented, 6663 code generation for little endian SH2A has been disabled. 6664 Specifying -ml with -m2a* will now result in a compiler error. 6665 * The defunct -mbranch-cost option has been fixed. 6666 * Some improvements to the generated code of: 6667 + Utilization of the tst #imm,R0 instruction. 6668 + Dynamic shift instructions on SH2A. 6669 + Integer absolute value calculations. 6670 * The -mdiv= option for targets other than SHmedia has been fixed and 6671 documented. 6672 6673 SPARC 6674 6675 * The option -mflat has been reinstated. When it is specified, the 6676 compiler will generate code for a single register window model. 6677 This is essentially a new implementation and the corresponding 6678 debugger support has been added to GDB 7.4. 6679 * Support for the options -mtune=native and -mcpu=native has been 6680 added on selected native platforms (GNU/Linux and Solaris). 6681 * Support for the SPARC T3 (Niagara 3) processor has been added. 6682 * VIS: 6683 + An intrinsics header visintrin.h has been added. 6684 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 6685 compare instructions have been added. 6686 + The little-endian version of alignaddr is now supported. 6687 + When possible, VIS builtins are marked const, which should 6688 increase the compiler's ability to optimize VIS operations. 6689 + The compiler now properly tracks the %gsr register and how it 6690 behaves as an input for various VIS instructions. 6691 + Akin to fzero, the compiler can now generate fone instructions 6692 in order to set all of the bits of a floating-point register 6693 to 1. 6694 + The documentation for the VIS intrinsics in the GCC manual has 6695 been brought up to date and many inaccuracies were fixed. 6696 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 6697 non-condition-code setting edge instructions have been added. 6698 Their availability is controlled by the new -mvis2 and 6699 -mno-vis2 options. They are enabled by default on 6700 UltraSPARC-III and later CPUs. 6701 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 6702 has been added. These instructions are enabled by default on SPARC 6703 T3 (Niagara 3) and later CPUs. 6704 6705 TILE-Gx/TILEPro 6706 6707 * Support has been added for the Tilera TILE-Gx and TILEPro families 6708 of processors. 6709 6710 Other significant improvements 6711 6712 * A new option (-grecord-gcc-switches) was added that appends 6713 compiler command-line options that might affect code generation to 6714 the DW_AT_producer attribute string in the DWARF debugging 6715 information. 6716 * GCC now supports various new GNU extensions to the DWARF debugging 6717 information format, like [40]entry value and [41]call site 6718 information, [42]typed DWARF stack or [43]a more compact macro 6719 representation. Support for these extensions has been added to GDB 6720 7.4. They can be disabled through the -gstrict-dwarf command-line 6721 option. 6722 6723 GCC 4.7.1 6724 6725 This is the [44]list of problem reports (PRs) from GCC's bug tracking 6726 system that are known to be fixed in the 4.7.1 release. This list might 6727 not be complete (that is, it is possible that some PRs that have been 6728 fixed are not listed here). 6729 6730 The Go front end in the 4.7.1 release fully supports the [45]Go 1 6731 language standard. 6732 6733 GCC 4.7.2 6734 6735 This is the [46]list of problem reports (PRs) from GCC's bug tracking 6736 system that are known to be fixed in the 4.7.2 release. This list might 6737 not be complete (that is, it is possible that some PRs that have been 6738 fixed are not listed here). 6739 6740 GCC 4.7.3 6741 6742 This is the [47]list of problem reports (PRs) from GCC's bug tracking 6743 system that are known to be fixed in the 4.7.3 release. This list might 6744 not be complete (that is, it is possible that some PRs that have been 6745 fixed are not listed here). 6746 6747 GCC 4.7.4 6748 6749 This is the [48]list of problem reports (PRs) from GCC's bug tracking 6750 system that are known to be fixed in the 4.7.4 release. This list might 6751 not be complete (that is, it is possible that some PRs that have been 6752 fixed are not listed here). 6753 6754 6755 For questions related to the use of GCC, please consult these web 6756 pages and the [49]GCC manuals. If that fails, the 6757 [50]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6758 web pages and the development of GCC are welcome on our developer 6759 list at [51]gcc (a] gcc.gnu.org. All of [52]our lists have public 6760 archives. 6761 6762 Copyright (C) [53]Free Software Foundation, Inc. Verbatim copying and 6763 distribution of this entire article is permitted in any medium, 6764 provided this notice is preserved. 6765 6766 These pages are [54]maintained by the GCC team. Last modified 6767 2019-11-28[55]. 6768 6769 References 6770 6771 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 6772 2. http://savannah.nongnu.org/bugs/?35407 6773 3. https://gcc.gnu.org/PR18145 6774 4. https://gcc.gnu.org/gcc-4.7/porting_to.html 6775 5. https://gcc.gnu.org/wiki/TransactionalMemory 6776 6. https://gcc.gnu.org/wiki/Atomic/GCCMM 6777 7. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6778 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6779 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6780 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6781 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6782 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6783 13. https://gcc.gnu.org/PR14258 6784 14. https://gcc.gnu.org/PR35688 6785 15. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 6786 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 6787 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 6788 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 6789 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 6790 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 6791 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 6792 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 6793 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149 6794 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 6795 25. https://docs.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps 6796 26. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 6797 27. https://gcc.gnu.org/wiki/Fortran2003Status 6798 28. https://gcc.gnu.org/wiki/OOP 6799 29. https://gcc.gnu.org/wiki/Fortran2008Status 6800 30. https://gcc.gnu.org/wiki/Coarray 6801 31. https://gcc.gnu.org/wiki/CoarrayLib 6802 32. https://gcc.gnu.org/wiki/TS29113Status 6803 33. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bstd_003d_007d_0040var_007bstd_007d-option-53 6804 34. https://golang.org/doc/go1 6805 35. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 6806 36. http://nongnu.org/avr-libc/ 6807 37. https://gcc.gnu.org/PR54461 6808 38. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built_002din-Functions.html 6809 39. https://sites.google.com/site/x32abi/ 6810 40. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1 6811 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2 6812 42. http://www.dwarfstd.org/ShowIssue.php?issue=140425.1 6813 43. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1 6814 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 6815 45. https://golang.org/doc/go1 6816 46. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 6817 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 6818 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 6819 49. https://gcc.gnu.org/onlinedocs/ 6820 50. mailto:gcc-help (a] gcc.gnu.org 6821 51. mailto:gcc (a] gcc.gnu.org 6822 52. https://gcc.gnu.org/lists.html 6823 53. https://www.fsf.org/ 6824 54. https://gcc.gnu.org/about.html 6825 55. http://validator.w3.org/check/referer 6826 ====================================================================== 6827 http://gcc.gnu.org/gcc-4.6/index.html 6828 GCC 4.6 Release Series 6829 6830 (This release series is no longer supported.) 6831 6832 April 12, 2013 6833 6834 The [1]GNU project and the GCC developers are pleased to announce the 6835 release of GCC 4.6.4. 6836 6837 This release is a bug-fix release, containing fixes for regressions in 6838 GCC 4.6.3 relative to previous releases of GCC. 6839 6840 Release History 6841 6842 GCC 4.6.4 6843 April 12, 2013 ([2]changes, [3]documentation) 6844 6845 GCC 4.6.3 6846 March 1, 2012 ([4]changes, [5]documentation) 6847 6848 GCC 4.6.2 6849 October 26, 2011 ([6]changes, [7]documentation) 6850 6851 GCC 4.6.1 6852 June 27, 2011 ([8]changes, [9]documentation) 6853 6854 GCC 4.6.0 6855 March 25, 2011 ([10]changes, [11]documentation) 6856 6857 References and Acknowledgements 6858 6859 GCC used to stand for the GNU C Compiler, but since the compiler 6860 supports several other languages aside from C, it now stands for the 6861 GNU Compiler Collection. 6862 6863 A list of [12]successful builds is updated as new information becomes 6864 available. 6865 6866 The GCC developers would like to thank the numerous people that have 6867 contributed new features, improvements, bug fixes, and other changes as 6868 well as test results to GCC. This [13]amazing group of volunteers is 6869 what makes GCC successful. 6870 6871 For additional information about GCC please refer to the [14]GCC 6872 project web site or contact the [15]GCC development mailing list. 6873 6874 To obtain GCC please use [16]our mirror sites or [17]our version 6875 control system. 6876 6877 6878 For questions related to the use of GCC, please consult these web 6879 pages and the [18]GCC manuals. If that fails, the 6880 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6881 web pages and the development of GCC are welcome on our developer 6882 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 6883 archives. 6884 6885 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 6886 distribution of this entire article is permitted in any medium, 6887 provided this notice is preserved. 6888 6889 These pages are [23]maintained by the GCC team. Last modified 6890 2020-01-14[24]. 6891 6892 References 6893 6894 1. http://www.gnu.org/ 6895 2. http://gcc.gnu.org/gcc-4.6/changes.html 6896 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 6897 4. http://gcc.gnu.org/gcc-4.6/changes.html 6898 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 6899 6. http://gcc.gnu.org/gcc-4.6/changes.html 6900 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 6901 8. http://gcc.gnu.org/gcc-4.6/changes.html 6902 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 6903 10. http://gcc.gnu.org/gcc-4.6/changes.html 6904 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 6905 12. http://gcc.gnu.org/gcc-4.6/buildstat.html 6906 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 6907 14. http://gcc.gnu.org/index.html 6908 15. mailto:gcc (a] gcc.gnu.org 6909 16. http://gcc.gnu.org/mirrors.html 6910 17. http://gcc.gnu.org/git.html 6911 18. https://gcc.gnu.org/onlinedocs/ 6912 19. mailto:gcc-help (a] gcc.gnu.org 6913 20. mailto:gcc (a] gcc.gnu.org 6914 21. https://gcc.gnu.org/lists.html 6915 22. https://www.fsf.org/ 6916 23. https://gcc.gnu.org/about.html 6917 24. http://validator.w3.org/check/referer 6918 ====================================================================== 6919 http://gcc.gnu.org/gcc-4.6/changes.html 6920 GCC 4.6 Release Series 6921 Changes, New Features, and Fixes 6922 6923 Caveats 6924 6925 * The options -b <machine> and -V <version> have been removed because 6926 they were unreliable. Instead, users should directly run 6927 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 6928 run a different version of gcc. 6929 * GCC now has stricter checks for invalid command-line options. In 6930 particular, when gcc was called to link object files rather than 6931 compile source code, it would previously accept and ignore all 6932 options starting with --, including linker options such as 6933 --as-needed and --export-dynamic, although such options would 6934 result in errors if any source code was compiled. Such options, if 6935 unknown to the compiler, are now rejected in all cases; if the 6936 intent was to pass them to the linker, options such as 6937 -Wl,--as-needed should be used. 6938 * Versions of the GNU C library up to and including 2.11.1 included 6939 an [1]incorrect implementation of the cproj function. GCC optimizes 6940 its builtin cproj according to the behavior specified and allowed 6941 by the ISO C99 standard. If you want to avoid discrepancies between 6942 the C library and GCC's builtin transformations when using cproj in 6943 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 6944 and actually rely on the incorrect behavior of cproj, then you can 6945 disable GCC's transformations using -fno-builtin-cproj. 6946 * The C-only intermodule optimization framework (IMA, enabled by 6947 -combine) has been removed in favor of the new generic link-time 6948 optimization framework (LTO) introduced in [2]GCC 4.5.0. 6949 * GCC now ships with the LGPL-licensed libquadmath library, which 6950 provides quad-precision mathematical functions for targets with a 6951 __float128 datatype. __float128 is available for targets on 32-bit 6952 x86, x86-64 and Itanium architectures. The libquadmath library is 6953 automatically built on such targets when building the Fortran 6954 compiler. 6955 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 6956 warnings were added for C, C++, Objective-C and Objective-C++. 6957 These warnings diagnose variables respective parameters which are 6958 only set in the code and never otherwise used. Usually such 6959 variables are useless and often even the value assigned to them is 6960 computed needlessly, sometimes expensively. The 6961 -Wunused-but-set-variable warning is enabled by default by -Wall 6962 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 6963 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 6964 rules for the layout of vectors that could lead to wrong code being 6965 generated. Vectors larger than 8 bytes in size are now by default 6966 aligned to an 8-byte boundary. This is an ABI change: code that 6967 makes explicit use of vector types may be incompatible with binary 6968 objects built with older versions of GCC. Auto-vectorized code is 6969 not affected by this change. (This change affects GCC versions 6970 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 6971 * On AVR, variables with the progmem attribute to locate data in 6972 flash memory must be qualified as const. 6973 * Support for a number of older systems and recently unmaintained or 6974 untested target ports of GCC has been declared obsolete in GCC 4.6. 6975 Unless there is activity to revive them, the next release of GCC 6976 will have their sources permanently removed. 6977 All GCC ports for the following processor architectures have been 6978 declared obsolete: 6979 + Argonaut ARC (arc-*) 6980 + National Semiconductor CRX (crx-*) 6981 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 6982 m68hc12-*-*, m6812-*-*) 6983 + Sunplus S+core (score-*) 6984 The following ports for individual systems on particular 6985 architectures have been obsoleted: 6986 + Interix (i[34567]86-*-interix3*) 6987 + NetWare x86 (i[3456x]86-*-netware*) 6988 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 6989 + MCore PE (mcore-*-pe*) 6990 + SH SymbianOS (sh*-*-symbianelf*) 6991 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 6992 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 6993 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 6994 vax-*-netbsd*, but not *-*-netbsdelf*) 6995 The i[34567]86-*-pe alias for Cygwin targets has also been 6996 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 6997 Certain configure options to control the set of libraries built 6998 with GCC on some targets have been obsoleted. On ARM targets, the 6999 options --disable-fpu, --disable-26bit, --disable-underscore, 7000 --disable-interwork, --disable-biendian and --disable-nofmult have 7001 been obsoleted. On MIPS targets, the options 7002 --disable-single-float, --disable-biendian and --disable-softfloat 7003 have been obsoleted. 7004 * Support has been removed for all the [3]configurations obsoleted in 7005 GCC 4.5. 7006 * More information on porting to GCC 4.6 from previous versions of 7007 GCC can be found in the [4]porting guide for this release. 7008 7009 General Optimizer Improvements 7010 7011 * A new general optimization level, -Ofast, has been introduced. It 7012 combines the existing optimization level -O3 with options that can 7013 affect standards compliance but result in better optimized code. 7014 For example, -Ofast enables -ffast-math. 7015 * Link-time optimization improvements: 7016 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 7017 stabilized to the point of being usable. It has become the 7018 default mode when using the LTO optimization model. Link time 7019 optimization can now split itself into multiple parallel 7020 compilations. Parallelism is controlled with -flto=n (where n 7021 specifies the number of compilations to execute in parallel). 7022 GCC can also cooperate with a GNU make job server by 7023 specifying the -flto=jobserver option and adding + to the 7024 beginning of the Makefile rule executing the linker. 7025 Classical LTO mode can be enforced by -flto-partition=none. 7026 This may result in small code quality improvements. 7027 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 7028 and other large applications can be built with LTO enabled. 7029 + The linker plugin support improvements 7030 o Linker plugin is now enabled by default when the linker 7031 is detected to have plugin support. This is the case for 7032 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 7033 the Gold linker on ELF targets. Plugin support of the 7034 Apple linker on Darwin is not compatible with GCC. The 7035 linker plugin can also be controlled by the 7036 -fuse-linker-plugin command-line option. 7037 o Resolution information from the linker plugin is used to 7038 drive whole program assumptions. Use of the linker plugin 7039 results in more aggressive optimization on binaries and 7040 on shared libraries that use the hidden visibility 7041 attribute. Consequently the use of -fwhole-program is not 7042 necessary in addition to LTO. 7043 + Hidden symbols used from non-LTO objects now have to be 7044 explicitly annotated with externally_visible when the linker 7045 plugin is not used. 7046 + C++ inline functions and virtual tables are now privatized 7047 more aggressively, leading to better inter-procedural 7048 optimization and faster dynamic linking. 7049 + Memory usage and intermediate language streaming performance 7050 have been improved. 7051 + Static constructors and destructors from individual units are 7052 inlined into a single function. This can significantly improve 7053 startup times of large C++ applications where static 7054 constructors are very common. For example, static constructors 7055 are used when including the iostream header. 7056 + Support for the Ada language has been added. 7057 * Interprocedural optimization improvements 7058 + The interprocedural framework was re-tuned for link time 7059 optimization. Several scalability issues were resolved. 7060 + Improved auto-detection of const and pure functions. Newly, 7061 noreturn functions are auto-detected. 7062 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 7063 available that informs users when adding attributes to headers 7064 might improve code generation. 7065 + A number of inlining heuristic improvements. In particular: 7066 o Partial inlining is now supported and enabled by default 7067 at -O2 and greater. The feature can be controlled via 7068 -fpartial-inlining. 7069 Partial inlining splits functions with short hot path to 7070 return. This allows more aggressive inlining of the hot 7071 path leading to better performance and often to code size 7072 reductions (because cold parts of functions are not 7073 duplicated). 7074 o Scalability for large compilation units was improved 7075 significantly. 7076 o Inlining of callbacks is now more aggressive. 7077 o Virtual methods are considered for inlining when the 7078 caller is inlined and devirtualization is then possible. 7079 o Inlining when optimizing for size (either in cold regions 7080 of a program or when compiling with -Os) was improved to 7081 better handle C++ programs with larger abstraction 7082 penalty, leading to smaller and faster code. 7083 + The IPA reference optimization pass detecting global variables 7084 used or modified by functions was strengthened and sped up. 7085 + Functions whose address was taken are now optimized out when 7086 all references to them are dead. 7087 + A new inter-procedural static profile estimation pass detects 7088 functions that are executed once or unlikely to be executed. 7089 Unlikely executed functions are optimized for size. Functions 7090 executed once are optimized for size except for the inner 7091 loops. 7092 + On most targets with named section support, functions used 7093 only at startup (static constructors and main), functions used 7094 only at exit and functions detected to be cold are placed into 7095 separate text segment subsections. This extends the 7096 -freorder-functions feature and is controlled by the same 7097 switch. The goal is to improve the startup time of large C++ 7098 programs. 7099 Proper function placement requires linker support. GNU ld 7100 2.21.51 on ELF targets was updated to place those functions 7101 together within the text section leading to better code 7102 locality and faster startup times of large C++ programs. The 7103 feature is also supported in the Apple linker. Support in the 7104 gold linker is planned. 7105 * A new switch -fstack-usage has been added. It makes the compiler 7106 output stack usage information for the program, on a per-function 7107 basis, in an auxiliary file. 7108 * A new switch -fcombine-stack-adjustments has been added. It can be 7109 used to enable or disable the compiler's stack-slot combining pass 7110 which before was enabled automatically at -O1 and above, but could 7111 not be controlled on its own. 7112 * A new switch -fstrict-volatile-bitfields has been added. Using it 7113 indicates that accesses to volatile bitfields should use a single 7114 access of the width of the field's type. This option can be useful 7115 for precisely defining and accessing memory-mapped peripheral 7116 registers from C or C++. 7117 7118 Compile time and memory usage improvements 7119 7120 * Datastructures used by the dataflow framework in GCC were 7121 reorganized for better memory usage and more cache locality. 7122 Compile time is improved especially on units with large functions 7123 (possibly resulting from a lot of inlining) not fitting into the 7124 processor cache. The compile time of the GCC C compiler binary with 7125 link-time optimization went down by over 10% (benchmarked on x86-64 7126 target). 7127 7128 New Languages and Language specific improvements 7129 7130 Ada 7131 7132 * Stack checking has been improved on selected architectures (Alpha, 7133 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 7134 overflows in all cases on these architectures. 7135 * Initial support for Ada 2012 has been added. 7136 7137 C family 7138 7139 * A new warning, enabled by -Wdouble-promotion, has been added that 7140 warns about cases where a value of type float is implicitly 7141 promoted to double. This is especially helpful for CPUs that handle 7142 the former in hardware, but emulate the latter in software. 7143 * A new function attribute leaf was introduced. This attribute allows 7144 better inter-procedural optimization across calls to functions that 7145 return to the current unit only via returning or exception 7146 handling. This is the case for most library functions that have no 7147 callbacks. 7148 * Support for a new data type __int128 for targets having wide enough 7149 machine-mode support. 7150 * The new function attribute callee_pop_aggregate allows to specify 7151 if the caller or callee is responsible for popping the aggregate 7152 return pointer value from the stack. 7153 * Support for selectively enabling and disabling warnings via #pragma 7154 GCC diagnostic has been added. For instance: 7155 #pragma GCC diagnostic error "-Wuninitialized" 7156 foo(a); /* error is given for this one */ 7157 #pragma GCC diagnostic push 7158 #pragma GCC diagnostic ignored "-Wuninitialized" 7159 foo(b); /* no diagnostic for this one */ 7160 #pragma GCC diagnostic pop 7161 foo(c); /* error is given for this one */ 7162 #pragma GCC diagnostic pop 7163 foo(d); /* depends on command-line options */ 7164 7165 * The -fmax-errors=N option is now supported. Using this option 7166 causes the compiler to exit after N errors have been issued. 7167 7168 C 7169 7170 * There is now experimental support for some features from the 7171 upcoming C1X revision of the ISO C standard. This support may be 7172 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 7173 Note that this support is experimental and may change incompatibly 7174 in future releases for consistency with changes to the C1X standard 7175 draft. The following features are newly supported as described in 7176 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 7177 meeting); some other features were already supported with no 7178 compiler changes being needed, or have some support but not in full 7179 accord with N1539 (as amended). 7180 + Static assertions (_Static_assert keyword) 7181 + Typedef redefinition 7182 + New macros in <float.h> 7183 + Anonymous structures and unions 7184 * The new -fplan9-extensions option directs the compiler to support 7185 some extensions for anonymous struct fields which are implemented 7186 by the Plan 9 compiler. A pointer to a struct may be automatically 7187 converted to a pointer to an anonymous field when calling a 7188 function, in order to make the types match. An anonymous struct 7189 field whose type is a typedef name may be referred to using the 7190 typedef name. 7191 7192 C++ 7193 7194 * Improved [7]experimental support for the upcoming C++0x ISO C++ 7195 standard, including support for constexpr (thanks to Gabriel Dos 7196 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 7197 noexcept, unrestricted unions, range-based for loops (thanks to 7198 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 7199 Rodrigo), implicitly deleted functions and implicit move 7200 constructors. 7201 * When an extern declaration within a function does not match a 7202 declaration in the enclosing context, G++ now properly declares the 7203 name within the namespace of the function rather than the namespace 7204 which was open just before the function definition ([8]c++/43145). 7205 * GCC now warns by default when casting integers to larger pointer 7206 types. These warnings can be disabled with the option 7207 -Wno-int-to-pointer-cast, which is now also available in C++. 7208 * G++ no longer optimizes using the assumption that a value of 7209 enumeration type will fall within the range specified by the 7210 standard, since that assumption is easily violated with a 7211 conversion from integer type ([9]c++/43680). The old behavior can 7212 be restored with -fstrict-enums. 7213 * The new -fnothrow-opt flag changes the semantics of a throw() 7214 exception specification to match the proposed semantics of the 7215 noexcept specification: just call terminate if an exception tries 7216 to propagate out of a function with such an exception 7217 specification. This dramatically reduces or eliminates the code 7218 size overhead from adding the exception specification. 7219 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 7220 a function that the compiler can tell doesn't throw if it would 7221 change the value of a noexcept expression. 7222 * The -Wshadow option now warns if a local variable or type 7223 declaration shadows another type in C++. Note that the compiler 7224 will not warn if a local variable shadows a struct/class/enum, but 7225 will warn if it shadows an explicit typedef. 7226 * When an identifier is not found in the current scope, G++ now 7227 offers suggestions about which identifier might have been intended. 7228 * G++ now issues clearer diagnostics for missing semicolons after 7229 class, struct, and union definitions. 7230 * G++ now issues clearer diagnostics for missing semicolons after 7231 class member declarations. 7232 * G++ now issues clearer diagnostics when a colon is used in a place 7233 where a double-colon was intended. 7234 * G++ no longer accepts mutable on reference members ([10]c++/33558). 7235 Use -fpermissive to allow the old, non-conforming behaviour. 7236 * A few mangling fixes have been made, to attribute const/volatile on 7237 function pointer types, decltype of a plain decl, and use of a 7238 function parameter in the declaration of another parameter. By 7239 default the compiler still uses the old mangling, but emits aliases 7240 with the new mangling on targets that support strong aliases. Users 7241 can switch over entirely to the new mangling with -fabi-version=5 7242 or -fabi-version=0. -Wabi will now warn about code that uses the 7243 old mangling. 7244 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 7245 type to be default initialized unless the type has a user-declared 7246 default constructor. In 4.6.2 G++ implements the proposed 7247 resolution of [11]DR 253, so default initialization is allowed if 7248 it initializes all subobjects. Code that fails to compile can be 7249 fixed by providing an initializer e.g. 7250 struct A { A(); }; 7251 struct B : A { int i; }; 7252 const B b = B(); 7253 Use -fpermissive to allow the old, non-conforming behaviour. 7254 7255 Runtime Library (libstdc++) 7256 7257 * [12]Improved experimental support for the upcoming ISO C++ 7258 standard, C++0x, including using constexpr and nullptr. 7259 * Performance improvements to the [13]Debug Mode, thanks to Franois 7260 Dumont. 7261 * Atomic operations used for reference-counting are annotated so that 7262 they can be understood by race detectors such as Helgrind, see 7263 [14]Data Race Hunting. 7264 * Most libstdc++ standard headers have been changed to no longer 7265 include the cstddef header as an implementation detail. Code that 7266 relied on that header being included as side-effect of including 7267 other standard headers will need to include cstddef explicitly. 7268 7269 Fortran 7270 7271 * On systems supporting the libquadmath library, GNU Fortran now also 7272 supports a quad-precision, kind=16 floating-point data type 7273 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 7274 hardware, calculations might be one to two orders of magnitude 7275 slower than with the 4, 8 or 10 bytes floating-point data types. 7276 This change does not affect systems which support REAL(16) in 7277 hardware nor those which do not support libquadmath. 7278 * Much improved compile time for large array constructors. 7279 * In order to reduce execution time and memory consumption, use of 7280 temporary arrays in assignment expressions is avoided for many 7281 cases. The compiler now reverses loops in order to avoid generating 7282 a temporary array where possible. 7283 * Improved diagnostics, especially with -fwhole-file. 7284 * The -fwhole-file flag is now enabled by default. This improves code 7285 generation and diagnostics. It can be disabled using the deprecated 7286 -fno-whole-file flag. 7287 * Support the generation of Makefile dependencies via the [15]-M... 7288 flags of GCC; you may need to specify the -cpp option in addition. 7289 The dependencies take modules, Fortran's include, and CPP's 7290 #include into account. Note: Using -M for the module path is no 7291 longer supported, use -J instead. 7292 * The flag -Wconversion has been modified to only issue warnings 7293 where a conversion leads to information loss. This drastically 7294 reduces the number of warnings; -Wconversion is thus now enabled 7295 with -Wall. The flag -Wconversion-extra has been added and also 7296 warns about other conversions; -Wconversion-extra typically issues 7297 a huge number of warnings, most of which can be ignored. 7298 * A new command-line option -Wunused-dummy-argument warns about 7299 unused dummy arguments and is included in -Wall. Before, 7300 -Wunused-variable also warned about unused dummy arguments. 7301 * Fortran 2003 support has been extended: 7302 + Improved support for polymorphism between libraries and 7303 programs and for complicated inheritance patterns (cf. 7304 [16]object-oriented programming). 7305 + Experimental support of the ASSOCIATE construct. 7306 + In pointer assignments it is now possible to specify the lower 7307 bounds of the pointer and, for a rank-1 or a simply contiguous 7308 data-target, to remap the bounds. 7309 + Automatic (re)allocation: In intrinsic assignments to 7310 allocatable variables the left-hand side will be automatically 7311 allocated (if unallocated) or reallocated (if the shape or 7312 type parameter is different). To avoid the small performance 7313 penalty, you can use a(:) = ... instead of a = ... for arrays 7314 and character strings or disable the feature using -std=f95 7315 or -fno-realloc-lhs. 7316 + Deferred type parameter: For scalar allocatable and pointer 7317 variables the character length can be deferred. 7318 + Namelist variables with allocatable and pointer attribute and 7319 nonconstant length type parameter are supported. 7320 * Fortran 2008 support has been extended: 7321 + Experimental [17]coarray support (for one image only, i.e. 7322 num_images() == 1); use the [18]-fcoarray=single flag to 7323 enable it. 7324 + The STOP and the new ERROR STOP statements now support all 7325 constant expressions. 7326 + Support for the CONTIGUOUS attribute. 7327 + Support for ALLOCATE with MOLD. 7328 + Support for the STORAGE_SIZE intrinsic inquiry function. 7329 + Support of the NORM2 and PARITY intrinsic functions. 7330 + The following bit intrinsics were added: POPCNT and POPPAR for 7331 counting the number of 1 bits and returning the parity; BGE, 7332 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 7333 for combined left and right shifts, MASKL and MASKR for simple 7334 left and right justified masks, MERGE_BITS for a bitwise merge 7335 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 7336 and the transformational bit intrinsics IALL, IANY and 7337 IPARITY. 7338 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 7339 + Support for the IMPURE attribute for procedures, which allows 7340 for ELEMENTAL procedures without the restrictions of PURE. 7341 + Null pointers (including NULL()) and not allocated variables 7342 can be used as actual argument to optional non-pointer, 7343 non-allocatable dummy arguments, denoting an absent argument. 7344 + Non-pointer variables with TARGET attribute can be used as 7345 actual argument to POINTER dummies with INTENT(IN) 7346 + Pointers including procedure pointers and those in a derived 7347 type (pointer components) can now be initialized by a target 7348 instead of only by NULL. 7349 + The EXIT statement (with construct-name) can now be used to 7350 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 7351 SELECT CASE and SELECT TYPE constructs. 7352 + Internal procedures can now be used as actual argument. 7353 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 7354 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 7355 have been added; these arrays contain the supported kind 7356 values for the respective types. 7357 + The module procedures C_SIZEOF of the intrinsic module 7358 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 7359 ISO_FORTRAN_ENV have been implemented. 7360 + Minor changes: obsolescence diagnostics for ENTRY was added 7361 for -std=f2008; a line may start with a semicolon; for 7362 internal and module procedures END can be used instead of END 7363 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 7364 a RADIX argument; intrinsic types are supported for 7365 TYPE(intrinsic-type-spec); multiple type-bound procedures can 7366 be declared in a single PROCEDURE statement; implied-shape 7367 arrays are supported for named constants (PARAMETER). The 7368 transformational, three argument versions of BESSEL_JN and 7369 BESSEL_YN were added the elemental, two-argument version had 7370 been added in GCC 4.4; note that the transformational 7371 functions use a recurrence algorithm. 7372 7373 Go 7374 7375 Support for the [19]Go programming language has been added to GCC. It 7376 is not enabled by default when you build GCC; use the 7377 --enable-languages configure option to build it. The driver program for 7378 compiling Go code is gccgo. 7379 7380 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 7381 is in progress. It may or may not work on other platforms. 7382 7383 Objective-C and Objective-C++ 7384 7385 * The -fobjc-exceptions flag is now required to enable Objective-C 7386 exception and synchronization syntax (introduced by the keywords 7387 @try, @catch, @finally and @synchronized). 7388 * A number of Objective-C 2.0 features and extensions are now 7389 supported by GCC. These features are enabled by default; you can 7390 disable them by using the new -fobjc-std=objc1 command-line option. 7391 * The Objective-C 2.0 dot-syntax is now supported. It is an 7392 alternative syntax for using getters and setters; object.count is 7393 automatically converted into [object count] or [object setCount: 7394 ...] depending on context; for example if (object.count > 0) is 7395 automatically compiled into the equivalent of if ([object count] > 7396 0) while object.count = 0; is automatically compiled into the 7397 equivalent ot [object setCount: 0];. The dot-syntax can be used 7398 with instance and class objects and with any setters or getters, no 7399 matter if they are part of a declared property or not. 7400 * Objective-C 2.0 declared properties are now supported. They are 7401 declared using the new @property keyword, and are most commonly 7402 used in conjunction with the new Objective-C 2.0 dot-syntax. The 7403 nonatomic, readonly, readwrite, assign, retain, copy, setter and 7404 getter attributes are all supported. Marking declared properties 7405 with __attribute__ ((deprecated)) is supported too. 7406 * The Objective-C 2.0 @synthesize and @dynamic keywords are 7407 supported. @synthesize causes the compiler to automatically 7408 synthesize a declared property, while @dynamic is used to disable 7409 all warnings for a declared property for which no implementation is 7410 provided at compile time. Synthesizing declared properties requires 7411 runtime support in most useful cases; to be able to use it with the 7412 GNU runtime, appropriate helper functions have been added to the 7413 GNU Objective-C runtime ABI, and are implemented by the GNU 7414 Objective-C runtime library shipped with GCC. 7415 * The Objective-C 2.0 fast enumeration syntax is supported in 7416 Objective-C. This is currently not yet available in Objective-C++. 7417 Fast enumeration requires support in the runtime, and such support 7418 has been added to the GNU Objective-C runtime library (shipped with 7419 GCC). 7420 * The Objective-C 2.0 @optional keyword is supported. It allows you 7421 to mark methods or properties in a protocol as optional as opposed 7422 to required. 7423 * The Objective-C 2.0 @package keyword is supported. It has currently 7424 the same effect as the @public keyword. 7425 * Objective-C 2.0 method attributes are supported. Currently the 7426 supported attributes are deprecated, sentinel, noreturn and format. 7427 * Objective-C 2.0 method argument attributes are supported. The most 7428 widely used attribute is unused, to mark an argument as unused in 7429 the implementation. 7430 * Objective-C 2.0 class and protocol attributes are supported. 7431 Currently the only supported attribute is deprecated. 7432 * Objective-C 2.0 class extensions are supported. A class extension 7433 has the same syntax as a category declaration with no category 7434 name, and the methods and properties declared in it are added 7435 directly to the main class. It is mostly used as an alternative to 7436 a category to add methods to a class without advertising them in 7437 the public headers, with the advantage that for class extensions 7438 the compiler checks that all the privately declared methods are 7439 actually implemented. 7440 * As a result of these enhancements, GCC can now be used to build 7441 Objective-C and Objective-C++ software that uses Foundation and 7442 other important system frameworks with the NeXT runtime on Darwin 9 7443 and Darwin 10 (OSX 10.5 and 10.6). 7444 * Many bugs in the compiler have been fixed in this release; in 7445 particular, LTO can now be used when compiling Objective-C and 7446 Objective-C++ and the parser is much more robust in dealing with 7447 invalid code. 7448 7449 Runtime Library (libobjc) 7450 7451 * The GNU Objective-C runtime library now defines the macro 7452 __GNU_LIBOBJC__ (with a value that is increased at every release 7453 where there is any change to the API) in objc/objc.h, making it 7454 easy to determine if the GNU Objective-C runtime library is being 7455 used, and if so, which version. Previous versions of the GNU 7456 Objective-C runtime library (and other Objective-C runtime 7457 libraries such as the Apple one) do not define this macro. 7458 * A new Objective-C 2.0 API, almost identical to the one implemented 7459 by the Apple Objective-C runtime, has been implemented in the GNU 7460 Objective-C runtime library. The new API hides the internals of 7461 most runtime structures but provides a more extensive set of 7462 functions to operate on them. It is much easier, for example, to 7463 create or modify classes at runtime. The new API also makes it 7464 easier to port software from Apple to GNU as almost no changes 7465 should be required. The old API is still supported for backwards 7466 compatibility; including the old objc/objc-api.h header file 7467 automatically selects the old API, while including the new 7468 objc/runtime.h header file automatically selects the new API. 7469 Support for the old API is being phased out and upgrading the 7470 software to use the new API is strongly recommended. To check for 7471 the availability of the new API, the __GNU_LIBOBJC__ macro can be 7472 used as older versions of the GNU Objective-C runtime library, 7473 which do not support the new API, do not define such a macro. 7474 * Runtime support for @synchronized has been added. 7475 * Runtime support for Objective-C 2.0 synthesized property accessors 7476 has been added. 7477 * Runtime support for Objective-C 2.0 fast enumeration has been 7478 added. 7479 7480 New Targets and Target Specific Improvements 7481 7482 ARM 7483 7484 * GCC now supports the Cortex-M4 processor implementing the v7-em 7485 version of the architecture using the option -mcpu=cortex-m4. 7486 * Scheduling descriptions for the Cortex-M4, the Neon and the 7487 floating point units of the Cortex-A9 and a pipeline description 7488 for the Cortex-A5 have been added. 7489 * Synchronization primitives such as __sync_fetch_and_add and friends 7490 are now inlined for supported architectures rather than calling 7491 into a kernel helper function. 7492 * SSA loop prefetching is enabled by default for the Cortex-A9 at 7493 -O3. 7494 * Several improvements were committed to improve code generation for 7495 the ARM architecture including a rewritten implementation for load 7496 and store multiples. 7497 * Several enhancements were committed to improve SIMD code generation 7498 for NEON by adding support for widening instructions, misaligned 7499 loads and stores, vector conditionals and support for 64 bit 7500 arithmetic. 7501 * Support was added for the Faraday cores fa526, fa606te, fa626te, 7502 fmp626te, fmp626 and fa726te and can be used with the respective 7503 names as parameters to the -mcpu= option. 7504 * Basic support was added for Cortex-A15 and is available through 7505 -mcpu=cortex-a15. 7506 * GCC for AAPCS configurations now more closely adheres to the AAPCS 7507 specification by enabling -fstrict-volatile-bitfields by default. 7508 7509 IA-32/x86-64 7510 7511 * The new -fsplit-stack option permits programs to use a 7512 discontiguous stack. This is useful for threaded programs, in that 7513 it is no longer necessary to specify the maximum stack size when 7514 creating a thread. This feature is currently only implemented for 7515 32-bit and 64-bit x86 GNU/Linux targets. 7516 * Support for emitting profiler counter calls before function 7517 prologues. This is enabled via a new command-line option -mfentry. 7518 * Optimization for the Intel Core 2 processors is now available 7519 through the -march=core2 and -mtune=core2 options. 7520 * Support for Intel Core i3/i5/i7 processors is now available through 7521 the -march=corei7 and -mtune=corei7 options. 7522 * Support for Intel Core i3/i5/i7 processors with AVX is now 7523 available through the -march=corei7-avx and -mtune=corei7-avx 7524 options. 7525 * Support for AMD Bobcat (family 14) processors is now available 7526 through the -march=btver1 and -mtune=btver1 options. 7527 * Support for AMD Bulldozer (family 15) processors is now available 7528 through the -march=bdver1 and -mtune=bdver1 options. 7529 * The default setting (when not optimizing for size) for 32-bit 7530 GNU/Linux and Darwin x86 targets has been changed to 7531 -fomit-frame-pointer. The default can be reverted to 7532 -fno-omit-frame-pointer by configuring GCC with the 7533 --enable-frame-pointer configure option. 7534 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 7535 __float128 on 32-bit and 64-bit x86 targets. 7536 * AVX floating-point arithmetic can now be enabled by default at 7537 configure time with the new --with-fpmath=avx option. 7538 * The SSA loop prefetching pass is enabled when using -O3 when 7539 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 7540 than K6). 7541 * Support for TBM (Trailing Bit Manipulation) built-in functions and 7542 code generation is available via -mtbm. 7543 * Support for AMD's BMI (Bit Manipulation) built-in functions and 7544 code generation is available via -mbmi. 7545 7546 MicroBlaze 7547 7548 * Support has been added for the Xilinx MicroBlaze softcore processor 7549 (microblaze-elf) embedded target. This configurable processor is 7550 supported on several Xilinx Spartan and Virtex FPGAs. 7551 7552 MIPS 7553 7554 * GCC now supports the Loongson 3A processor. Its canonical -march= 7555 and -mtune= name is loongson3a. 7556 7557 MN10300 / AM33 7558 7559 * The inline assembly register constraint "A" has been renamed "c". 7560 This constraint is used to select a floating-point register that 7561 can be used as the destination of a multiply-accumulate 7562 instruction. 7563 * New inline assembly register constraints "A" and "D" have been 7564 added. These constraint letters resolve to all general registers 7565 when compiling for AM33, and resolve to address registers only or 7566 data registers only when compiling for MN10300. 7567 * The MDR register is represented in the compiler. One can access the 7568 register via the "z" constraint in inline assembly. It can be 7569 marked as clobbered or used as a local register variable via the 7570 "mdr" name. The compiler uses the RETF instruction if the function 7571 does not modify the MDR register, so it is important that inline 7572 assembly properly annotate any usage of the register. 7573 7574 PowerPC/PowerPC64 7575 7576 * GCC now supports the Applied Micro Titan processor with 7577 -mcpu=titan. 7578 * The -mrecip option has been added, which indicates whether the 7579 reciprocal and reciprocal square root instructions should be used. 7580 * The -mveclibabi=mass option can be used to enable the compiler to 7581 autovectorize mathematical functions using the Mathematical 7582 Acceleration Subsystem library. 7583 * The -msingle-pic-base option has been added, which instructs the 7584 compiler to avoid loading the PIC base register in function 7585 prologues. The PIC base register must be initialized by the runtime 7586 system. 7587 * The -mblock-move-inline-limit option has been added, which enables 7588 the user to control the maximum size of inlined memcpy calls and 7589 similar. 7590 * PowerPC64 GNU/Linux support for applications requiring a large TOC 7591 section has been improved. A new command-line option, 7592 -mcmodel=MODEL, controls this feature; valid values for MODEL are 7593 small, medium, or large. 7594 * The Altivec builtin functions vec_ld and vec_st have been modified 7595 to generate the Altivec memory instructions LVX and STVX, even if 7596 the -mvsx option is used. In the initial GCC 4.5 release, these 7597 builtin functions were changed to generate VSX memory reference 7598 instructions instead of Altivec memory instructions, but there are 7599 differences between the two instructions. If the VSX instruction 7600 set is available, you can now use the new builtin functions 7601 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 7602 instructions. 7603 * The GCC compiler on AIX now defaults to a process layout with a 7604 larger data space allowing larger programs to be compiled. 7605 * The GCC long double type on AIX 6.1 and above has reverted to 64 7606 bit double precision, matching the AIX XL compiler default, because 7607 of missing C99 symbols required by the GCC runtime. 7608 * The default processor scheduling model and tuning for PowerPC64 7609 GNU/Linux and for AIX 6.1 and above now is POWER7. 7610 * Starting with GCC 4.6.1, vectors of type vector long long or vector 7611 long are passed and returned in the same method as other vectors 7612 with the VSX instruction set. Previously the GCC compiler did not 7613 adhere to the ABI for 128-bit vectors with 64-bit integer base 7614 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 7615 7616 S/390, zSeries and System z9/z10, IBM zEnterprise z196 7617 7618 * Support for the zEnterprise z196 processor has been added. When 7619 using the -march=z196 option, the compiler will generate code 7620 making use of the following instruction facilities: 7621 + Conditional load/store 7622 + Distinct-operands 7623 + Floating-point-extension 7624 + Interlocked-access 7625 + Population-count 7626 The -mtune=z196 option avoids the compare and branch instructions 7627 as well as the load address instruction with an index register as 7628 much as possible and performs instruction scheduling appropriate 7629 for the new out-of-order pipeline architecture. 7630 * When using the -m31 -mzarch options the generated code still 7631 conforms to the 32-bit ABI but uses the general purpose registers 7632 as 64-bit registers internally. This requires a Linux kernel saving 7633 the whole 64-bit registers when doing a context switch. Kernels 7634 providing that feature indicate that by the 'highgprs' string in 7635 /proc/cpuinfo. 7636 * The SSA loop prefetching pass is enabled when using -O3. 7637 7638 SPARC 7639 7640 * GCC now supports the LEON series of SPARC V8 processors. The code 7641 generated by the compiler can either be tuned to it by means of the 7642 --with-tune=leon configure option and -mtune=leon compilation 7643 option, or the compiler can be built for the sparc-leon-{elf,linux} 7644 and sparc-leon3-{elf,linux} targets directly. 7645 * GCC has stopped sign/zero-extending parameter registers in the 7646 callee for functions taking parameters with sub-word size in 32-bit 7647 mode, since this is redundant with the specification of the ABI. 7648 GCC has never done so in 64-bit mode since this is also redundant. 7649 * The command-line option -mfix-at697f has been added to enable the 7650 documented workaround for the single erratum of the Atmel AT697F 7651 processor. 7652 7653 Operating Systems 7654 7655 Android 7656 7657 * GCC now supports the Bionic C library and provides a convenient way 7658 of building native libraries and applications for the Android 7659 platform. Refer to the documentation of the -mandroid and -mbionic 7660 options for details on building native code. At the moment, Android 7661 support is enabled only for ARM. 7662 7663 Darwin/Mac OS X 7664 7665 * General 7666 + Initial support for CFString types has been added. 7667 This allows GCC to build projects including the system Core 7668 Foundation frameworks. The GCC Objective-C family supports 7669 CFString "toll-free bridged" as per the Mac OS X system tools. 7670 CFString is also recognized in the context of format 7671 attributes and arguments (see the documentation for format 7672 attributes for limitations). At present, 8-bit character types 7673 are supported. 7674 + Object file size reduction. 7675 The Darwin zeroed memory allocators have been re-written to 7676 make more use of .zerofill sections. For non-debug code, this 7677 can reduce object file size significantly. 7678 + Objective-C family 64-bit support (NeXT ABI 2). 7679 Initial support has been added to support 64-bit Objective-C 7680 code using the Darwin/OS X native (NeXT) runtime. ABI version 7681 2 will be selected automatically when 64-bit code is built. 7682 + Objective-C family 32-bit ABI 1. 7683 For 32-bit code ABI 1 is also now also allowed. At present it 7684 must be selected manually using -fobjc-abi-version=1 where 7685 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 7686 * x86 Architecture 7687 + The -mdynamic-no-pic option has been enabled. 7688 Code supporting -mdynamic-no-pic optimization has been added 7689 and is applicable to -m32 builds. The compiler bootstrap uses 7690 the option where appropriate. 7691 + The default value for -mtune= has been changed. 7692 Since Darwin systems are primarily Xeon, Core-2 or similar the 7693 default tuning has been changed to -mtune=core2. 7694 + Enable 128-bit long double (__float128) support on Darwin. 7695 * PPC Architecture 7696 + Darwin64 ABI. 7697 Several significant bugs have been fixed, such that GCC now 7698 produces code compatible with the Darwin64 PowerPC ABI. 7699 + libffi and boehm-gc. 7700 The Darwin ports of the libffi and boehm-gc libraries have 7701 been upgraded to include a Darwin64 implementation. This means 7702 that powerpc*-*-darwin9 platforms may now, for example, build 7703 Java applications with -m64 enabled. 7704 + Plug-in support has been enabled. 7705 + The -fsection-anchors option is now available although, 7706 presently, not heavily tested. 7707 7708 Solaris 2 7709 7710 New Features 7711 7712 * Support symbol versioning with the Sun linker. 7713 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 7714 * Support thread-local storage (TLS) with the Sun assembler on 7715 Solaris 2/x86. 7716 * Support TLS on Solaris 8/9 if prerequisites are met. 7717 * Support COMDAT group with the GNU assembler and recent Sun linker. 7718 * Support the Sun assembler visibility syntax. 7719 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 7720 -march=pentiumpro (Solaris 8/9). 7721 * Don't use SSE on Solaris 8/9 x86 by default. 7722 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 7723 7724 ABI Change 7725 7726 * Change the ABI for returning 8-byte vectors like __m64 in MMX 7727 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 7728 compilers. This is an incompatible change. If you use such types, 7729 you must either recompile all your code with the new compiler or 7730 use the new -mvect8-ret-in-mem option to remain compatible with 7731 previous versions of GCC and Sun Studio. 7732 7733 Windows x86/x86_64 7734 7735 * Initial support for decimal floating point. 7736 * Support for the __thiscall calling-convention. 7737 * Support for hot-patchable function prologues via the 7738 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 7739 * Improvements of stack-probing and stack-allocation mechanisms. 7740 * Support of push/pop-macro pragma as preprocessor command. 7741 With #pragma push_macro("macro-name") the current definition of 7742 macro-name is saved and can be restored with #pragma 7743 pop_macro("macro-name") to its saved definition. 7744 * Enable 128-bit long double (__float128) support on MinGW and 7745 Cygwin. 7746 7747 Other significant improvements 7748 7749 Installation changes 7750 7751 * An install-strip make target is provided that installs stripped 7752 executables, and may install libraries with unneeded or debugging 7753 sections stripped. 7754 * On Power7 systems, there is a potential problem if you build the 7755 GCC compiler with a host compiler using options that enable the VSX 7756 instruction set generation. If the host compiler has been patched 7757 so that the vec_ld and vec_st builtin functions generate Altivec 7758 memory instructions instead of VSX memory instructions, then you 7759 should be able to build the compiler with VSX instruction 7760 generation. 7761 7762 Changes for GCC Developers 7763 7764 Note: these changes concern developers that develop GCC itself or 7765 software that integrates with GCC, such as plugins, and not the general 7766 GCC users. 7767 * The gengtype utility, which previously was internal to the GCC 7768 build process, has been enchanced to provide GC root information 7769 for plugins as necessary. 7770 * The old GC allocation interface of ggc_alloc and friends was 7771 replaced with a type-safe alternative. 7772 7773 GCC 4.6.1 7774 7775 This is the [20]list of problem reports (PRs) from GCC's bug tracking 7776 system that are known to be fixed in the 4.6.1 release. This list might 7777 not be complete (that is, it is possible that some PRs that have been 7778 fixed are not listed here). 7779 7780 GCC 4.6.2 7781 7782 This is the [21]list of problem reports (PRs) from GCC's bug tracking 7783 system that are known to be fixed in the 4.6.2 release. This list might 7784 not be complete (that is, it is possible that some PRs that have been 7785 fixed are not listed here). 7786 7787 GCC 4.6.3 7788 7789 This is the [22]list of problem reports (PRs) from GCC's bug tracking 7790 system that are known to be fixed in the 4.6.3 release. This list might 7791 not be complete (that is, it is possible that some PRs that have been 7792 fixed are not listed here). 7793 7794 GCC 4.6.4 7795 7796 This is the [23]list of problem reports (PRs) from GCC's bug tracking 7797 system that are known to be fixed in the 4.6.4 release. This list might 7798 not be complete (that is, it is possible that some PRs that have been 7799 fixed are not listed here). 7800 7801 7802 For questions related to the use of GCC, please consult these web 7803 pages and the [24]GCC manuals. If that fails, the 7804 [25]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7805 web pages and the development of GCC are welcome on our developer 7806 list at [26]gcc (a] gcc.gnu.org. All of [27]our lists have public 7807 archives. 7808 7809 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and 7810 distribution of this entire article is permitted in any medium, 7811 provided this notice is preserved. 7812 7813 These pages are [29]maintained by the GCC team. Last modified 7814 2019-11-28[30]. 7815 7816 References 7817 7818 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 7819 2. http://gcc.gnu.org/gcc-4.5/changes.html 7820 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 7821 4. http://gcc.gnu.org/gcc-4.6/porting_to.html 7822 5. http://gcc.gnu.org/projects/lto/whopr.pdf 7823 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 7824 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html 7825 8. https://gcc.gnu.org/PR43145 7826 9. https://gcc.gnu.org/PR43680 7827 10. https://gcc.gnu.org/PR33558 7828 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 7829 12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 7830 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 7831 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 7832 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 7833 16. https://gcc.gnu.org/wiki/OOP 7834 17. https://gcc.gnu.org/wiki/Coarray 7835 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 7836 19. https://golang.org/ 7837 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 7838 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 7839 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 7840 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 7841 24. https://gcc.gnu.org/onlinedocs/ 7842 25. mailto:gcc-help (a] gcc.gnu.org 7843 26. mailto:gcc (a] gcc.gnu.org 7844 27. https://gcc.gnu.org/lists.html 7845 28. https://www.fsf.org/ 7846 29. https://gcc.gnu.org/about.html 7847 30. http://validator.w3.org/check/referer 7848 ====================================================================== 7849 http://gcc.gnu.org/gcc-4.5/index.html 7850 GCC 4.5 Release Series 7851 7852 (This release series is no longer supported.) 7853 7854 Jul 2, 2012 7855 7856 The [1]GNU project and the GCC developers are pleased to announce the 7857 release of GCC 4.5.4. 7858 7859 This release is a bug-fix release, containing fixes for regressions in 7860 GCC 4.5.3 relative to previous releases of GCC. 7861 7862 Release History 7863 7864 GCC 4.5.4 7865 Jul 2, 2012 ([2]changes) 7866 7867 GCC 4.5.3 7868 Apr 28, 2011 ([3]changes) 7869 7870 GCC 4.5.2 7871 Dec 16, 2010 ([4]changes) 7872 7873 GCC 4.5.1 7874 Jul 31, 2010 ([5]changes) 7875 7876 GCC 4.5.0 7877 April 14, 2010 ([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 version 7897 control system. 7898 7899 7900 For questions related to the use of GCC, please consult these web 7901 pages and the [13]GCC manuals. If that fails, the 7902 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7903 web pages and the development of GCC are welcome on our developer 7904 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 7905 archives. 7906 7907 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 7908 distribution of this entire article is permitted in any medium, 7909 provided this notice is preserved. 7910 7911 These pages are [18]maintained by the GCC team. Last modified 7912 2020-01-14[19]. 7913 7914 References 7915 7916 1. http://www.gnu.org/ 7917 2. http://gcc.gnu.org/gcc-4.5/changes.html 7918 3. http://gcc.gnu.org/gcc-4.5/changes.html 7919 4. http://gcc.gnu.org/gcc-4.5/changes.html 7920 5. http://gcc.gnu.org/gcc-4.5/changes.html 7921 6. http://gcc.gnu.org/gcc-4.5/changes.html 7922 7. http://gcc.gnu.org/gcc-4.5/buildstat.html 7923 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7924 9. http://gcc.gnu.org/index.html 7925 10. mailto:gcc (a] gcc.gnu.org 7926 11. http://gcc.gnu.org/mirrors.html 7927 12. http://gcc.gnu.org/git.html 7928 13. https://gcc.gnu.org/onlinedocs/ 7929 14. mailto:gcc-help (a] gcc.gnu.org 7930 15. mailto:gcc (a] gcc.gnu.org 7931 16. https://gcc.gnu.org/lists.html 7932 17. https://www.fsf.org/ 7933 18. https://gcc.gnu.org/about.html 7934 19. http://validator.w3.org/check/referer 7935 ====================================================================== 7936 http://gcc.gnu.org/gcc-4.5/changes.html 7937 GCC 4.5 Release Series 7938 Changes, New Features, and Fixes 7939 7940 Caveats 7941 7942 * GCC now requires the [1]MPC library in order to build. See the 7943 [2]prerequisites page for version requirements. 7944 * Support for a number of older systems and recently unmaintained or 7945 untested target ports of GCC has been declared obsolete in GCC 4.5. 7946 Unless there is activity to revive them, the next release of GCC 7947 will have their sources permanently removed. 7948 The following ports for individual systems on particular 7949 architectures have been obsoleted: 7950 + IRIX releases before 6.5 (mips-sgi-irix5*, 7951 mips-sgi-irix6.[0-4]) 7952 + Solaris 7 (*-*-solaris2.7) 7953 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 7954 alpha-dec-osf5.0*) 7955 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 7956 can be found in the [3]announcement. 7957 Support for the classic POWER architecture implemented in the 7958 original RIOS and RIOS2 processors of the old IBM RS/6000 product 7959 line has been obsoleted in the rs6000 port. This does not affect 7960 the new generation Power and PowerPC architectures. 7961 * Support has been removed for all the [4]configurations obsoleted in 7962 GCC 4.4. 7963 * Support has been removed for the protoize and unprotoize utilities, 7964 obsoleted in GCC 4.4. 7965 * Support has been removed for tuning for Itanium1 (Merced) variants. 7966 Note that code tuned for Itanium2 should also run correctly on 7967 Itanium1. 7968 * GCC now generates unwind info also for epilogues. DWARF debuginfo 7969 generated by GCC now uses more features of DWARF3 than before, and 7970 also some DWARF4 features. GDB older than 7.0 is not able to handle 7971 either of these, so to debug GCC 4.5 generated binaries or 7972 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 7973 features with the -gdwarf-3 -gstrict-dwarf options, or use 7974 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 7975 epilogue unwind info is emitted unconditionally whenever unwind 7976 info is emitted. 7977 * On x86 targets, code containing floating-point calculations may run 7978 significantly more slowly when compiled with GCC 4.5 in strict C99 7979 conformance mode than they did with earlier GCC versions. This is 7980 due to stricter standard conformance of the compiler and can be 7981 avoided by using the option -fexcess-precision=fast; also see 7982 [5]below. 7983 * The function attribute noinline no longer prevents GCC from cloning 7984 the function. A new attribute noclone has been introduced for this 7985 purpose. Cloning a function means that it is duplicated and the new 7986 copy is specialized for certain contexts (for example when a 7987 parameter is a known constant). 7988 7989 General Optimizer Improvements 7990 7991 * The -save-temps now takes an optional argument. The -save-temps and 7992 -save-temps=cwd switches write the temporary files in the current 7993 working directory based on the original source file. The 7994 -save-temps=obj switch will write files into the directory 7995 specified with the -o option, and the intermediate filenames are 7996 based on the output file. This will allow the user to get the 7997 compiler intermediate files when doing parallel builds without two 7998 builds of the same filename located in different directories from 7999 interfering with each other. 8000 * Debugging dumps are now created in the same directory as the object 8001 file rather than in the current working directory. This allows the 8002 user to get debugging dumps when doing parallel builds without two 8003 builds of the same filename interfering with each other. 8004 * GCC has been integrated with the MPC library. This allows GCC to 8005 evaluate complex arithmetic at compile time [6]more accurately. It 8006 also allows GCC to evaluate calls to complex built-in math 8007 functions having constant arguments and replace them at compile 8008 time with their mathematically equivalent results. In doing so, GCC 8009 can generate correct results regardless of the math library 8010 implementation or floating point precision of the host platform. 8011 This also allows GCC to generate identical results regardless of 8012 whether one compiles in native or cross-compile configurations to a 8013 particular target. The following built-in functions take advantage 8014 of this new capability: cacos, cacosh, casin, casinh, catan, 8015 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 8016 and ctanh. The float and long double variants of these functions 8017 (e.g. csinf and csinl) are also handled. 8018 * A new link-time optimizer has been added ([7]-flto). When this 8019 option is used, GCC generates a bytecode representation of each 8020 input file and writes it to specially-named sections in each object 8021 file. When the object files are linked together, all the function 8022 bodies are read from these named sections and instantiated as if 8023 they had been part of the same translation unit. This enables 8024 interprocedural optimizations to work across different files (and 8025 even different languages), potentially improving the performance of 8026 the generated code. To use the link-timer optimizer, -flto needs to 8027 be specified at compile time and during the final link. If the 8028 program does not require any symbols to be exported, it is possible 8029 to combine -flto and the experimental [8]-fwhopr with 8030 [9]-fwhole-program to allow the interprocedural optimizers to use 8031 more aggressive assumptions. 8032 * The automatic parallelization pass was enhanced to support 8033 parallelization of outer loops. 8034 * Automatic parallelization can be enabled as part of Graphite. In 8035 addition to -ftree-parallelize-loops=, specify 8036 -floop-parallelize-all to enable the Graphite-based optimization. 8037 * The infrastructure for optimizing based on [10]restrict qualified 8038 pointers has been rewritten and should result in code generation 8039 improvements. Optimizations based on restrict qualified pointers 8040 are now also available when using -fno-strict-aliasing. 8041 * There is a new optimization pass that attempts to change prototype 8042 of functions to avoid unused parameters, pass only relevant parts 8043 of structures and turn arguments passed by reference to arguments 8044 passed by value when possible. It is enabled by -O2 and above as 8045 well as -Os and can be manually invoked using the new command-line 8046 switch -fipa-sra. 8047 * GCC now optimize exception handling code. In particular cleanup 8048 regions that are proved to not have any effect are optimized out. 8049 8050 New Languages and Language specific improvements 8051 8052 All languages 8053 8054 * The -fshow-column option is now on by default. This means error 8055 messages now have a column associated with them. 8056 8057 Ada 8058 8059 * Compilation of programs heavily using discriminated record types 8060 with variant parts has been sped up and generates more compact 8061 code. 8062 * Stack checking now works reasonably well on most plaforms. In some 8063 specific cases, stack overflows may still fail to be detected, but 8064 a compile-time warning will be issued for these cases. 8065 8066 C family 8067 8068 * If a header named in a #include directive is not found, the 8069 compiler exits immediately. This avoids a cascade of errors arising 8070 from declarations expected to be found in that header being 8071 missing. 8072 * A new built-in function __builtin_unreachable() has been added that 8073 tells the compiler that control will never reach that point. It may 8074 be used after asm statements that terminate by transferring control 8075 elsewhere, and in other places that are known to be unreachable. 8076 * The -Wlogical-op option now warns for logical expressions such as 8077 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 8078 mistakes. This option is disabled by default. 8079 * An asm goto feature has been added to allow asm statements that 8080 jump to C labels. 8081 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 8082 * The deprecated attribute now takes an optional string argument, for 8083 example, __attribute__((deprecated("text string"))), that will be 8084 printed together with the deprecation warning. 8085 8086 C 8087 8088 * The -Wenum-compare option, which warns when comparing values of 8089 different enum types, now works for C. It formerly only worked for 8090 C++. This warning is enabled by -Wall. It may be avoided by using a 8091 type cast. 8092 * The -Wcast-qual option now warns about casts which are unsafe in 8093 that they permit const-correctness to be violated without further 8094 warnings. Specifically, it warns about cases where a qualifier is 8095 added when all the lower types are not const. For example, it warns 8096 about a cast from char ** to const char **. 8097 * The -Wc++-compat option is significantly improved. It issues new 8098 warnings for: 8099 + Using C++ reserved operator names as identifiers. 8100 + Conversions to enum types without explicit casts. 8101 + Using va_arg with an enum type. 8102 + Using different enum types in the two branches of ?:. 8103 + Using ++ or -- on a variable of enum type. 8104 + Using the same name as both a struct, union or enum tag and a 8105 typedef, unless the typedef refers to the tagged type itself. 8106 + Using a struct, union, or enum which is defined within another 8107 struct or union. 8108 + A struct field defined using a typedef if there is a field in 8109 the struct, or an enclosing struct, whose name is the typedef 8110 name. 8111 + Duplicate definitions at file scope. 8112 + Uninitialized const variables. 8113 + A global variable with an anonymous struct, union, or enum 8114 type. 8115 + Using a string constant to initialize a char array whose size 8116 is the length of the string. 8117 * The new -Wjump-misses-init option warns about cases where a goto or 8118 switch skips the initialization of a variable. This sort of branch 8119 is an error in C++ but not in C. This warning is enabled by 8120 -Wc++-compat. 8121 * GCC now ensures that a C99-conforming <stdint.h> is present on most 8122 targets, and uses information about the types in this header to 8123 implement the Fortran bindings to those types. GCC does not ensure 8124 the presence of such a header, and does not implement the Fortran 8125 bindings, on the following targets: NetBSD, VxWorks, VMS, 8126 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 8127 * GCC now implements C90- and C99-conforming rules for constant 8128 expressions. This may cause warnings or errors for some code using 8129 expressions that can be folded to a constant but are not constant 8130 expressions as defined by ISO C. 8131 * All known target-independent C90 and C90 Amendment 1 conformance 8132 bugs, and all known target-independent C99 conformance bugs not 8133 related to floating point or extended identifiers, have been fixed. 8134 * The C decimal floating point support now includes support for the 8135 FLOAT_CONST_DECIMAL64 pragma. 8136 * The named address space feature from ISO/IEC TR 18037 is now 8137 supported. This is currently only implemented for the SPU 8138 processor. 8139 8140 C++ 8141 8142 * Improved [11]experimental support for the upcoming C++0x ISO C++ 8143 standard, including support for raw strings, lambda expressions and 8144 explicit type conversion operators. 8145 * When printing the name of a class template specialization, G++ will 8146 now omit any template arguments which come from default template 8147 arguments. This behavior (and the pretty-printing of function 8148 template specializations as template signature and arguments) can 8149 be disabled with the -fno-pretty-templates option. 8150 * Access control is now applied to typedef names used in a template, 8151 which may cause G++ to reject some ill-formed code that was 8152 accepted by earlier releases. The -fno-access-control option can be 8153 used as a temporary workaround until the code is corrected. 8154 * Compilation time for code that uses templates should now scale 8155 linearly with the number of instantiations rather than 8156 quadratically, as template instantiations are now looked up using 8157 hash tables. 8158 * Declarations of functions that look like builtin declarations of 8159 library functions are only considered to be redeclarations if they 8160 are declared with extern "C". This may cause problems with code 8161 that omits extern "C" on hand-written declarations of C library 8162 functions such as abort or memcpy. Such code is ill-formed, but was 8163 accepted by earlier releases. 8164 * Diagnostics that used to complain about passing non-POD types to 8165 ... or jumping past the declaration of a non-POD variable now check 8166 for triviality rather than PODness, as per C++0x. 8167 * In C++0x mode local and anonymous classes are now allowed as 8168 template arguments, and in declarations of variables and functions 8169 with linkage, so long as any such declaration that is used is also 8170 defined ([12]DR 757). 8171 * Labels may now have attributes, as has been permitted for a while 8172 in C. This is only permitted when the label definition and the 8173 attribute specifier is followed by a semicoloni.e., the label 8174 applies to an empty statement. The only useful attribute for a 8175 label is unused. 8176 * G++ now implements [13]DR 176. Previously G++ did not support using 8177 the injected-class-name of a template base class as a type name, 8178 and lookup of the name found the declaration of the template in the 8179 enclosing scope. Now lookup of the name finds the 8180 injected-class-name, which can be used either as a type or as a 8181 template, depending on whether or not the name is followed by a 8182 template argument list. As a result of this change, some code that 8183 was previously accepted may be ill-formed because 8184 1. The injected-class-name is not accessible because it's from a 8185 private base, or 8186 2. The injected-class-name cannot be used as an argument for a 8187 template template parameter. 8188 In either of these cases, the code can be fixed by adding a 8189 nested-name-specifier to explicitly name the template. The first 8190 can be worked around with -fno-access-control; the second is only 8191 rejected with -pedantic. 8192 * A new standard mangling for SIMD vector types has been added, to 8193 avoid name clashes on systems with vectors of varying length. By 8194 default the compiler still uses the old mangling, but emits aliases 8195 with the new mangling on targets that support strong aliases. Users 8196 can switch over entirely to the new mangling with -fabi-version=4 8197 or -fabi-version=0. -Wabi will now warn about code that uses the 8198 old mangling. 8199 * The command-line option -ftemplate-depth-N is now written as 8200 -ftemplate-depth=N and the old form is deprecated. 8201 * Conversions between NULL and non-pointer types are now warned by 8202 default. The new option -Wno-conversion-null disables these 8203 warnings. Previously these warnings were only available when using 8204 -Wconversion explicitly. 8205 8206 Runtime Library (libstdc++) 8207 8208 * Improved experimental support for the upcoming ISO C++ standard, 8209 C++0x, including: 8210 + Support for <future>, <functional>, and <random>. 8211 + Existing facilities now exploit explicit operators and the 8212 newly implemented core C++0x features. 8213 + The header <cstdatomic> has been renamed to <atomic>. 8214 * An experimental [14]profile mode has been added. This is an 8215 implementation of many C++ standard library constructs with an 8216 additional analysis layer that gives performance improvement advice 8217 based on recognition of suboptimal usage patterns. For example, 8218 #include <vector> 8219 int main() 8220 { 8221 std::vector<int> v; 8222 for (int k = 0; k < 1024; ++k) 8223 v.insert(v.begin(), k); 8224 } 8225 8226 When instrumented via the profile mode, can return suggestions 8227 about the initial size and choice of the container used as follows: 8228 vector-to-list: improvement = 5: call stack = 0x804842c ... 8229 : advice = change std::vector to std::list 8230 vector-size: improvement = 3: call stack = 0x804842c ... 8231 : advice = change initial container size from 0 to 1024 8232 8233 These constructs can be substituted for the normal libstdc++ 8234 constructs on a piecemeal basis, or all existing components can be 8235 transformed via the -D_GLIBCXX_PROFILE macro. 8236 * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR 8237 24733) has been added. This support is in header file 8238 <decimal/decimal>, uses namespace std::decimal, and includes 8239 classes decimal32, decimal64, and decimal128. 8240 * Sources have been audited for application of function attributes 8241 nothrow, const, pure, and noreturn. 8242 * Python pretty-printers have been added for many standard library 8243 components that simplify the internal representation and present a 8244 more intuitive view of components when used with 8245 appropriately-advanced versions of GDB. For more information, 8246 please consult the more [16]detailed description. 8247 * The default behavior for comparing typeinfo names has changed, so 8248 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 8249 * The new -static-libstdc++ option directs g++ to link the C++ 8250 library statically, even if the default would normally be to link 8251 it dynamically. 8252 8253 Fortran 8254 8255 * The COMMON default padding has been changed instead of adding the 8256 padding before a variable it is now added afterwards, which 8257 increases the compatibility with other vendors and helps to obtain 8258 the correct output in some cases. Cf. also the -falign-commons 8259 option ([17]added in 4.4). 8260 * The -finit-real= option now also supports the value snan for 8261 signaling not-a-number; to be effective, one additionally needs to 8262 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 8263 optimizations can turn a signaling NaN into a quiet one. 8264 * The new option -fcheck= has been added with the options bounds, 8265 array-temps, do, pointer, and recursive. The bounds and array-temps 8266 options are equivalent to -fbounds-check and 8267 -fcheck-array-temporaries. The do option checks for invalid 8268 modification of loop iteration variables, and the recursive option 8269 tests for recursive calls to subroutines/functions which are not 8270 marked as recursive. With pointer pointer association checks in 8271 calls are performed; however, neither undefined pointers nor 8272 pointers in expressions are handled. Using -fcheck=all enables all 8273 these run-time checks. 8274 * The run-time checking -fcheck=bounds now warns about invalid string 8275 lengths of character dummy arguments. Additionally, more 8276 compile-time checks have been added. 8277 * The new option [18]-fno-protect-parens has been added; if set, the 8278 compiler may reorder REAL and COMPLEX expressions without regard to 8279 parentheses. 8280 * GNU Fortran no longer links against libgfortranbegin. As before, 8281 MAIN__ (assembler symbol name) is the actual Fortran main program, 8282 which is invoked by the main function. However, main is now 8283 generated and put in the same object file as MAIN__. For the time 8284 being, libgfortranbegin still exists for backward compatibility. 8285 For details see the new [19]Mixed-Language Programming chapter in 8286 the manual. 8287 * The I/O library was restructured for performance and cleaner code. 8288 * Array assignments and WHERE are now run in parallel when OpenMP's 8289 WORKSHARE is used. 8290 * The experimental option -fwhole-file was added. The option allows 8291 whole-file checking of procedure arguments and allows for better 8292 optimizations. It can also be used with -fwhole-program, which is 8293 now also supported in gfortran. 8294 * More Fortran 2003 and Fortran 2008 mathematical functions can now 8295 be used as initialization expressions. 8296 * Some extended attributes such as STDCALL are now supported via the 8297 [20]GCC$ compiler directive. 8298 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 8299 intrinsic behaves now as if zero were always positive. 8300 * For legacy compatibiliy: On Cygwin and MinGW, the special files 8301 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 8302 supported. 8303 * Fortran 2003 support has been extended: 8304 + Procedure-pointer function results and procedure-pointer 8305 components (including PASS), 8306 + allocatable scalars (experimental), 8307 + DEFERRED type-bound procedures, 8308 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 8309 have been implemented. 8310 + The ALLOCATE statement supports type-specs and the SOURCE= 8311 argument. 8312 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 8313 type-bound procedure (i.e. as type-bound operators). 8314 + Rounding (ROUND=, RZ, ...) for output is now supported. 8315 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 8316 intrinsic module ISO_C_BINDING are now supported, except for 8317 the targets listed above as ones where GCC does not have 8318 <stdint.h> type information. 8319 + Extensible derived types with type-bound procedure or 8320 procedure pointer with PASS attribute now have to use CLASS in 8321 line with the Fortran 2003 standard; the workaround to use 8322 TYPE is no longer supported. 8323 + [21]Experimental, incomplete support for polymorphism, 8324 including CLASS, SELECT TYPE and dynamic dispatch of 8325 type-bound procedure calls. Some features do not work yet such 8326 as unlimited polymorphism (CLASS(*)). 8327 * Fortran 2008 support has been extended: 8328 + The OPEN statement now supports the NEWUNIT= option, which 8329 returns a unique file unit, thus preventing inadvertent use of 8330 the same unit in different parts of the program. 8331 + Support for unlimited format items has been added. 8332 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 8333 the intrinsic module ISO_FORTRAN_ENV are now supported. 8334 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 8335 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 8336 and ATANH have been added (for real and complex arguments) and 8337 ATAN(Y,X) is now an alias for ATAN2(Y,X). 8338 + The BLOCK construct has been implemented. 8339 8340 New Targets and Target Specific Improvements 8341 8342 AIX 8343 8344 * Full cross-toolchain support now available with GNU Binutils 8345 8346 ARM 8347 8348 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 8349 * GCC now supports the ARM v7E-M architecture. 8350 * GCC now supports VFPv4-based FPUs and FPUs with 8351 single-precision-only VFP. 8352 * GCC has many improvements to optimization for other ARM processors, 8353 including scheduling support for the integer pipeline on Cortex-A9. 8354 * GCC now supports the IEEE 754-2008 half-precision floating-point 8355 type, and a variant ARM-specific half-precision type. This type is 8356 specified using __fp16, with the layout determined by 8357 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 8358 VFPv4 half-precision instructions will be used. 8359 * GCC now supports the variant of AAPCS that uses VFP registers for 8360 parameter passing and return values. 8361 8362 AVR 8363 8364 * The -mno-tablejump option has been removed because it has the same 8365 effect as the -fno-jump-tables option. 8366 * Added support for these new AVR devices: 8367 + ATmega8U2 8368 + ATmega16U2 8369 + ATmega32U2 8370 8371 IA-32/x86-64 8372 8373 * GCC now will set the default for -march= based on the configure 8374 target. 8375 * GCC now supports handling floating-point excess precision arising 8376 from use of the x87 floating-point unit in a way that conforms to 8377 ISO C99. This is enabled with -fexcess-precision=standard and with 8378 standards conformance options such as -std=c99, and may be disabled 8379 using -fexcess-precision=fast. 8380 * Support for the Intel Atom processor is now available through the 8381 -march=atom and -mtune=atom options. 8382 * A new -mcrc32 option is now available to enable crc32 intrinsics. 8383 * A new -mmovbe option is now available to enable GCC to use the 8384 movbe instruction to implement __builtin_bswap32 and 8385 __builtin_bswap64. 8386 * SSE math now can be enabled by default at configure time with the 8387 new --with-fpmath=sse option. 8388 * There is a new intrinsic header file, <x86intrin.h>. It should be 8389 included before using any IA-32/x86-64 intrinsics. 8390 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 8391 Orochi processors are now available with the -mxop, -mfma4, and 8392 -mlwp options. 8393 * The -mabm option enables GCC to use the popcnt and lzcnt 8394 instructions on AMD processors. 8395 * The -mpopcnt option enables GCC to use the popcnt instructions on 8396 both AMD and Intel processors. 8397 8398 M68K/ColdFire 8399 8400 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 8401 and 5441x devices. 8402 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 8403 processors. 8404 8405 MeP 8406 8407 Support has been added for the Toshiba Media embedded Processor (MeP, 8408 or mep-elf) embedded target. 8409 8410 MIPS 8411 8412 * GCC now supports MIPS 1004K processors. 8413 * GCC can now be configured with options --with-arch-32, 8414 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 8415 default optimization separately for 32-bit and 64-bit modes. 8416 * MIPS targets now support an alternative _mcount interface, in which 8417 register $12 points to the function's save slot for register $31. 8418 This interface is selected by the -mcount-ra-address option; see 8419 the documentation for more details. 8420 * GNU/Linux targets can now generate read-only .eh_frame sections. 8421 This optimization requires GNU binutils 2.20 or above, and is only 8422 available if GCC is configured with a suitable version of binutils. 8423 * GNU/Linux targets can now attach special relocations to indirect 8424 calls, so that the linker can turn them into direct jumps or 8425 branches. This optimization requires GNU binutils 2.20 or later, 8426 and is automatically selected if GCC is configured with an 8427 appropriate version of binutils. It can be explicitly enabled or 8428 disabled using the -mrelax-pic-calls command-line option. 8429 * GCC now generates more heavily-optimized atomic operations on 8430 Octeon processors. 8431 * MIPS targets now support the -fstack-protector option. 8432 * GCC now supports an -msynci option, which specifies that synci is 8433 enough to flush the instruction cache, without help from the 8434 operating system. GCC uses this information to optimize 8435 automatically-generated cache flush operations, such as those used 8436 for nested functions in C. There is also a --with-synci 8437 configure-time option, which makes -msynci the default. 8438 * GCC supports four new function attributes for interrupt handlers: 8439 interrupt, use_shadow_register_set, keep_interrupts_masked and 8440 use_debug_exception_return. See the documentation for more details 8441 about these attributes. 8442 8443 RS/6000 (POWER/PowerPC) 8444 8445 * GCC now supports the Power ISA 2.06, which includes the VSX 8446 instructions that add vector 64-bit floating point support, new 8447 population count instructions, and conversions between floating 8448 point and unsigned types. 8449 * Support for the power7 processor is now available through the 8450 -mcpu=power7 and -mtune=power7. 8451 * GCC will now vectorize loops that contain simple math functions 8452 like copysign when generating code for altivec or VSX targets. 8453 * Support for the A2 processor is now available through the -mcpu=a2 8454 and -mtune=a2 options. 8455 * Support for the 476 processor is now available through the 8456 -mcpu={476,476fp} and -mtune={476,476fp} options. 8457 * Support for the e500mc64 processor is now available through the 8458 -mcpu=e500mc64 and -mtune=e500mc64 options. 8459 * GCC can now be configured with options --with-cpu-32, 8460 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 8461 default optimization separately for 32-bit and 64-bit modes. 8462 * Starting with GCC 4.5.4, vectors of type vector long long or vector 8463 long are passed and returned in the same method as other vectors 8464 with the VSX instruction set. Previously the GCC compiler did not 8465 adhere to the ABI for 128-bit vectors with 64-bit integer base 8466 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 8467 8468 RX 8469 8470 Support has been added for the Renesas RX Processor (rx-elf) target. 8471 8472 Operating Systems 8473 8474 Windows (Cygwin and MinGW) 8475 8476 * GCC now installs all the major language runtime libraries as DLLs 8477 when configured with the --enable-shared option. 8478 * GCC now makes use of the new support for aligned common variables 8479 in versions of binutils >= 2.20 to fix bugs in the support for SSE 8480 data types. 8481 * Improvements to the libffi support library increase the reliability 8482 of code generated by GCJ on all Windows platforms. Libgcj is 8483 enabled by default for the first time. 8484 * Libtool improvements simplify installation by placing the generated 8485 DLLs in the correct binaries directory. 8486 * Numerous other minor bugfixes and improvements, and substantial 8487 enhancements to the Fortran language support library. 8488 8489 > 8490 8491 Other significant improvements 8492 8493 Plugins 8494 8495 * It is now possible to extend the compiler without having to modify 8496 its source code. A new option -fplugin=file.so tells GCC to load 8497 the shared object file.so and execute it as part of the compiler. 8498 The internal documentation describes the details on how plugins can 8499 interact with the compiler. 8500 8501 Installation changes 8502 8503 * The move to newer autotools changed default installation 8504 directories and switches to control them: The --with-datarootdir, 8505 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 8506 used any more. Instead, you can now use --datarootdir, --docdir, 8507 --htmldir, and --pdfdir. The default installation directories have 8508 changed as follows according to the GNU Coding Standards: 8509 8510 datarootdir read-only architecture-independent data root [PREFIX/share] 8511 localedir locale-specific message catalogs [DATAROOTDIR/locale] 8512 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 8513 htmldir html documentation [DOCDIR] 8514 dvidir dvi documentation [DOCDIR] 8515 pdfdir pdf documentation [DOCDIR] 8516 psdir ps documentation [DOCDIR] 8517 The following variables have new default values: 8518 8519 datadir read-only architecture-independent data [DATAROOTDIR] 8520 infodir info documentation [DATAROOTDIR/info] 8521 mandir man documentation [DATAROOTDIR/man] 8522 8523 GCC 4.5.1 8524 8525 This is the [22]list of problem reports (PRs) from GCC's bug tracking 8526 system that are known to be fixed in the 4.5.1 release. This list might 8527 not be complete (that is, it is possible that some PRs that have been 8528 fixed are not listed here). 8529 8530 All languages 8531 8532 * GCC's new link-time optimizer ([23]-flto) now also works on a few 8533 non-ELF targets: 8534 + Cygwin (*-cygwin*) 8535 + MinGW (*-mingw*) 8536 + Darwin on x86-64 (x86_64-apple-darwin*) 8537 LTO is not enabled by default for these targets. To enable LTO, you 8538 should configure with the --enable-lto option. 8539 8540 GCC 4.5.2 8541 8542 This is the [24]list of problem reports (PRs) from GCC's bug tracking 8543 system that are known to be fixed in the 4.5.2 release. This list might 8544 not be complete (that is, it is possible that some PRs that have been 8545 fixed are not listed here). 8546 8547 GCC 4.5.3 8548 8549 This is the [25]list of problem reports (PRs) from GCC's bug tracking 8550 system that are known to be fixed in the 4.5.3 release. This list might 8551 not be complete (that is, it is possible that some PRs that have been 8552 fixed are not listed here). 8553 8554 On the PowerPC compiler, the Altivec builtin functions vec_ld and 8555 vec_st have been modified to generate the Altivec memory instructions 8556 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 8557 release, these builtin functions were changed to generate VSX memory 8558 reference instructions instead of Altivec memory instructions, but 8559 there are differences between the two instructions. If the VSX 8560 instruction set is available, you can now use the new builtin functions 8561 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 8562 instructions. 8563 8564 GCC 4.5.4 8565 8566 This is the [26]list of problem reports (PRs) from GCC's bug tracking 8567 system that are known to be fixed in the 4.5.4 release. This list might 8568 not be complete (that is, it is possible that some PRs that have been 8569 fixed are not listed here). 8570 8571 8572 For questions related to the use of GCC, please consult these web 8573 pages and the [27]GCC manuals. If that fails, the 8574 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8575 web pages and the development of GCC are welcome on our developer 8576 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 8577 archives. 8578 8579 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 8580 distribution of this entire article is permitted in any medium, 8581 provided this notice is preserved. 8582 8583 These pages are [32]maintained by the GCC team. Last modified 8584 2019-11-28[33]. 8585 8586 References 8587 8588 1. http://www.multiprecision.org/mpc/ 8589 2. https://gcc.gnu.org/install/prerequisites.html 8590 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 8591 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 8592 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86 8593 6. https://gcc.gnu.org/PR30789 8594 7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 8595 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 8596 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 8597 10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 8598 11. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html 8599 12. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 8600 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 8601 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 8602 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 8603 16. https://sourceware.org/gdb/wiki/STLSupport 8604 17. http://gcc.gnu.org/gcc-4.4/changes.html 8605 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 8606 19. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 8607 20. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 8608 21. https://gcc.gnu.org/wiki/OOP 8609 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 8610 23. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 8611 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 8612 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 8613 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 8614 27. https://gcc.gnu.org/onlinedocs/ 8615 28. mailto:gcc-help (a] gcc.gnu.org 8616 29. mailto:gcc (a] gcc.gnu.org 8617 30. https://gcc.gnu.org/lists.html 8618 31. https://www.fsf.org/ 8619 32. https://gcc.gnu.org/about.html 8620 33. http://validator.w3.org/check/referer 8621 ====================================================================== 8622 http://gcc.gnu.org/gcc-4.4/index.html 8623 GCC 4.4 Release Series 8624 8625 This release series is no longer maintained. 8626 8627 March 13, 2012 8628 8629 The [1]GNU project and the GCC developers are pleased to announce the 8630 release of GCC 4.4.7. 8631 8632 This release is a bug-fix release, containing fixes for regressions in 8633 GCC 4.4.6 relative to previous releases of GCC. 8634 8635 Release History 8636 8637 GCC 4.4.7 8638 March 13, 2012 ([2]changes) 8639 8640 GCC 4.4.6 8641 April 16, 2011 ([3]changes) 8642 8643 GCC 4.4.5 8644 October 1, 2010 ([4]changes) 8645 8646 GCC 4.4.4 8647 April 29, 2010 ([5]changes) 8648 8649 GCC 4.4.3 8650 January 21, 2010 ([6]changes) 8651 8652 GCC 4.4.2 8653 October 15, 2009 ([7]changes) 8654 8655 GCC 4.4.1 8656 July 22, 2009 ([8]changes) 8657 8658 GCC 4.4.0 8659 April 21, 2009 ([9]changes) 8660 8661 References and Acknowledgements 8662 8663 GCC used to stand for the GNU C Compiler, but since the compiler 8664 supports several other languages aside from C, it now stands for the 8665 GNU Compiler Collection. 8666 8667 A list of [10]successful builds is updated as new information becomes 8668 available. 8669 8670 The GCC developers would like to thank the numerous people that have 8671 contributed new features, improvements, bug fixes, and other changes as 8672 well as test results to GCC. This [11]amazing group of volunteers is 8673 what makes GCC successful. 8674 8675 For additional information about GCC please refer to the [12]GCC 8676 project web site or contact the [13]GCC development mailing list. 8677 8678 To obtain GCC please use [14]our mirror sites or [15]our version 8679 control system. 8680 8681 8682 For questions related to the use of GCC, please consult these web 8683 pages and the [16]GCC manuals. If that fails, the 8684 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8685 web pages and the development of GCC are welcome on our developer 8686 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 8687 archives. 8688 8689 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 8690 distribution of this entire article is permitted in any medium, 8691 provided this notice is preserved. 8692 8693 These pages are [21]maintained by the GCC team. Last modified 8694 2020-01-14[22]. 8695 8696 References 8697 8698 1. http://www.gnu.org/ 8699 2. http://gcc.gnu.org/gcc-4.4/changes.html 8700 3. http://gcc.gnu.org/gcc-4.4/changes.html 8701 4. http://gcc.gnu.org/gcc-4.4/changes.html 8702 5. http://gcc.gnu.org/gcc-4.4/changes.html 8703 6. http://gcc.gnu.org/gcc-4.4/changes.html 8704 7. http://gcc.gnu.org/gcc-4.4/changes.html 8705 8. http://gcc.gnu.org/gcc-4.4/changes.html 8706 9. http://gcc.gnu.org/gcc-4.4/changes.html 8707 10. http://gcc.gnu.org/gcc-4.4/buildstat.html 8708 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8709 12. http://gcc.gnu.org/index.html 8710 13. mailto:gcc (a] gcc.gnu.org 8711 14. http://gcc.gnu.org/mirrors.html 8712 15. http://gcc.gnu.org/git.html 8713 16. https://gcc.gnu.org/onlinedocs/ 8714 17. mailto:gcc-help (a] gcc.gnu.org 8715 18. mailto:gcc (a] gcc.gnu.org 8716 19. https://gcc.gnu.org/lists.html 8717 20. https://www.fsf.org/ 8718 21. https://gcc.gnu.org/about.html 8719 22. http://validator.w3.org/check/referer 8720 ====================================================================== 8721 http://gcc.gnu.org/gcc-4.4/changes.html 8722 GCC 4.4 Release Series 8723 Changes, New Features, and Fixes 8724 8725 The latest release in the 4.4 release series is [1]GCC 4.4.7. 8726 8727 Caveats 8728 8729 * __builtin_stdarg_start has been completely removed from GCC. 8730 Support for <varargs.h> had been deprecated since GCC 4.0. Use 8731 __builtin_va_start as a replacement. 8732 * Some of the errors issued by the C++ front end that could be 8733 downgraded to warnings in previous releases by using -fpermissive 8734 are now warnings by default. They can be converted into errors by 8735 using -pedantic-errors. 8736 * Use of the cpp assertion extension will now emit a warning when 8737 -Wdeprecated or -pedantic is used. This extension has been 8738 deprecated for many years, but never warned about. 8739 * Packed bit-fields of type char were not properly bit-packed on many 8740 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 8741 causes an ABI change. For example there is no longer a 4-bit 8742 padding between field a and b in this structure: 8743 struct foo 8744 { 8745 char a:4; 8746 char b:8; 8747 } __attribute__ ((packed)); 8748 There is a new warning to help identify fields that are affected: 8749 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 8750 The warning can be disabled with -Wno-packed-bitfield-compat. 8751 * On ARM EABI targets, the C++ mangling of the va_list type has been 8752 changed to conform to the current revision of the EABI. This does 8753 not affect the libstdc++ library included with GCC. 8754 * The SCOUNT and POS bits of the MIPS DSP control register are now 8755 treated as global. Previous versions of GCC treated these fields as 8756 call-clobbered instead. 8757 * The MIPS port no longer recognizes the h asm constraint. It was 8758 necessary to remove this constraint in order to avoid generating 8759 unpredictable code sequences. 8760 One of the main uses of the h constraint was to extract the high 8761 part of a multiplication on 64-bit targets. For example: 8762 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 8763 You can now achieve the same effect using 128-bit types: 8764 typedef unsigned int uint128_t __attribute__((mode(TI))); 8765 result = ((uint128_t) x * y) >> 64; 8766 The second sequence is better in many ways. For example, if x and y 8767 are constants, the compiler can perform the multiplication at 8768 compile time. If x and y are not constants, the compiler can 8769 schedule the runtime multiplication better than it can schedule an 8770 asm statement. 8771 * Support for a number of older systems and recently unmaintained or 8772 untested target ports of GCC has been declared obsolete in GCC 4.4. 8773 Unless there is activity to revive them, the next release of GCC 8774 will have their sources permanently removed. 8775 The following ports for individual systems on particular 8776 architectures have been obsoleted: 8777 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 8778 m68k-*-aout*) 8779 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 8780 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 8781 sh-*-*). This does not affect other more specific targets 8782 using the COFF object format on those architectures, or the 8783 more specific H8300 and SH targets (h8300-*-rtems*, 8784 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 8785 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 8786 + 2BSD on PDP-11 (pdp11-*-bsd) 8787 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 8788 powerpc-ibm-aix4.[12]*) 8789 + Tuning support for Itanium1 (Merced) variants. Note that code 8790 tuned for Itanium2 should also run correctly on Itanium1. 8791 * The protoize and unprotoize utilities have been obsoleted and will 8792 be removed in GCC 4.5. These utilities have not been installed by 8793 default since GCC 3.0. 8794 * Support has been removed for all the [2]configurations obsoleted in 8795 GCC 4.3. 8796 * Unknown -Wno-* options are now silently ignored by GCC if no other 8797 diagnostics are issued. If other diagnostics are issued, then GCC 8798 warns about the unknown options. 8799 * More information on porting to GCC 4.4 from previous versions of 8800 GCC can be found in the [3]porting guide for this release. 8801 8802 General Optimizer Improvements 8803 8804 * A new command-line switch -findirect-inlining has been added. When 8805 turned on it allows the inliner to also inline indirect calls that 8806 are discovered to have known targets at compile time thanks to 8807 previous inlining. 8808 * A new command-line switch -ftree-switch-conversion has been added. 8809 This new pass turns simple initializations of scalar variables in 8810 switch statements into initializations from a static array, given 8811 that all the values are known at compile time and the ratio between 8812 the new array size and the original switch branches does not exceed 8813 the parameter --param switch-conversion-max-branch-ratio (default 8814 is eight). 8815 * A new command-line switch -ftree-builtin-call-dce has been added. 8816 This optimization eliminates unnecessary calls to certain builtin 8817 functions when the return value is not used, in cases where the 8818 calls can not be eliminated entirely because the function may set 8819 errno. This optimization is on by default at -O2 and above. 8820 * A new command-line switch -fconserve-stack directs the compiler to 8821 minimize stack usage even if it makes the generated code slower. 8822 This affects inlining decisions. 8823 * When the assembler supports it, the compiler will now emit unwind 8824 information using assembler .cfi directives. This makes it possible 8825 to use such directives in inline assembler code. The new option 8826 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 8827 directives. 8828 * The [4]Graphite branch has been merged. This merge has brought in a 8829 new framework for loop optimizations based on a polyhedral 8830 intermediate representation. These optimizations apply to all the 8831 languages supported by GCC. The following new code transformations 8832 are available in GCC 4.4: 8833 + -floop-interchange performs loop interchange transformations 8834 on loops. Interchanging two nested loops switches the inner 8835 and outer loops. For example, given a loop like: 8836 DO J = 1, M 8837 DO I = 1, N 8838 A(J, I) = A(J, I) * C 8839 ENDDO 8840 ENDDO 8841 8842 loop interchange will transform the loop as if the user had 8843 written: 8844 DO I = 1, N 8845 DO J = 1, M 8846 A(J, I) = A(J, I) * C 8847 ENDDO 8848 ENDDO 8849 8850 which can be beneficial when N is larger than the caches, 8851 because in Fortran, the elements of an array are stored in 8852 memory contiguously by column, and the original loop iterates 8853 over rows, potentially creating at each access a cache miss. 8854 + -floop-strip-mine performs loop strip mining transformations 8855 on loops. Strip mining splits a loop into two nested loops. 8856 The outer loop has strides equal to the strip size and the 8857 inner loop has strides of the original loop within a strip. 8858 For example, given a loop like: 8859 DO I = 1, N 8860 A(I) = A(I) + C 8861 ENDDO 8862 8863 loop strip mining will transform the loop as if the user had 8864 written: 8865 DO II = 1, N, 4 8866 DO I = II, min (II + 3, N) 8867 A(I) = A(I) + C 8868 ENDDO 8869 ENDDO 8870 8871 + -floop-block performs loop blocking transformations on loops. 8872 Blocking strip mines each loop in the loop nest such that the 8873 memory accesses of the element loops fit inside caches. For 8874 example, given a loop like: 8875 DO I = 1, N 8876 DO J = 1, M 8877 A(J, I) = B(I) + C(J) 8878 ENDDO 8879 ENDDO 8880 8881 loop blocking will transform the loop as if the user had 8882 written: 8883 DO II = 1, N, 64 8884 DO JJ = 1, M, 64 8885 DO I = II, min (II + 63, N) 8886 DO J = JJ, min (JJ + 63, M) 8887 A(J, I) = B(I) + C(J) 8888 ENDDO 8889 ENDDO 8890 ENDDO 8891 ENDDO 8892 8893 which can be beneficial when M is larger than the caches, 8894 because the innermost loop will iterate over a smaller amount 8895 of data that can be kept in the caches. 8896 * A new register allocator has replaced the old one. It is called 8897 integrated register allocator (IRA) because coalescing, register 8898 live range splitting, and hard register preferencing are done 8899 on-the-fly during coloring. It also has better integration with the 8900 reload pass. IRA is a regional register allocator which uses modern 8901 Chaitin-Briggs coloring instead of Chow's priority coloring used in 8902 the old register allocator. More info about IRA internals and 8903 options can be found in the GCC manuals. 8904 * A new instruction scheduler and software pipeliner, based on the 8905 selective scheduling approach, has been added. The new pass 8906 performs instruction unification, register renaming, substitution 8907 through register copies, and speculation during scheduling. The 8908 software pipeliner is able to pipeline non-countable loops. The new 8909 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 8910 it is available for the Intel Itanium platform working by default 8911 as the second scheduling pass (after register allocation) at the 8912 -O3 optimization level. 8913 * When using -fprofile-generate with a multi-threaded program, the 8914 profile counts may be slightly wrong due to race conditions. The 8915 new -fprofile-correction option directs the compiler to apply 8916 heuristics to smooth out the inconsistencies. By default the 8917 compiler will give an error message when it finds an inconsistent 8918 profile. 8919 * The new -fprofile-dir=PATH option permits setting the directory 8920 where profile data files are stored when using -fprofile-generate 8921 and friends, and the directory used when reading profile data files 8922 using -fprofile-use and friends. 8923 8924 New warning options 8925 8926 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 8927 warning if any stack frame is larger than NUMBER bytes. This may be 8928 used to help ensure that code fits within a limited amount of stack 8929 space. 8930 * The command-line option -Wlarger-than-N is now written as 8931 -Wlarger-than=N and the old form is deprecated. 8932 * The new -Wno-mudflap option disables warnings about constructs 8933 which can not be instrumented when using -fmudflap. 8934 8935 New Languages and Language specific improvements 8936 8937 * Version 3.0 of the OpenMP specification is now supported for the C, 8938 C++, and Fortran compilers. 8939 * New character data types, per [5]TR 19769: New character types in 8940 C, are now supported for the C compiler in -std=gnu99 mode, as 8941 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 8942 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 8943 8944 C family 8945 8946 * A new optimize attribute was added to allow programmers to change 8947 the optimization level and particular optimization options for an 8948 individual function. You can also change the optimization options 8949 via the GCC optimize pragma for functions defined after the pragma. 8950 The GCC push_options pragma and the GCC pop_options pragma allow 8951 you temporarily save and restore the options used. The GCC 8952 reset_options pragma restores the options to what was specified on 8953 the command line. 8954 * Uninitialized warnings do not require enabling optimization 8955 anymore, that is, -Wuninitialized can be used together with -O0. 8956 Nonetheless, the warnings given by -Wuninitialized will probably be 8957 more accurate if optimization is enabled. 8958 * -Wparentheses now warns about expressions such as (!x | y) and (!x 8959 & y). Using explicit parentheses, such as in ((!x) | y), silences 8960 this warning. 8961 * -Wsequence-point now warns within if, while,do while and for 8962 conditions, and within for begin/end expressions. 8963 * A new option -dU is available to dump definitions of preprocessor 8964 macros that are tested or expanded. 8965 8966 C++ 8967 8968 * [6]Improved experimental support for the upcoming ISO C++ standard, 8969 C++0x. Including support for auto, inline namespaces, generalized 8970 initializer lists, defaulted and deleted functions, new character 8971 types, and scoped enums. 8972 * Those errors that may be downgraded to warnings to build legacy 8973 code now mention -fpermissive when -fdiagnostics-show-option is 8974 enabled. 8975 * -Wconversion now warns if the result of a static_cast to enumeral 8976 type is unspecified because the value is outside the range of the 8977 enumeral type. 8978 * -Wuninitialized now warns if a non-static reference or non-static 8979 const member appears in a class without constructors. 8980 * G++ now properly implements value-initialization, so objects with 8981 an initializer of () and an implicitly defined default constructor 8982 will be zero-initialized before the default constructor is called. 8983 8984 Runtime Library (libstdc++) 8985 8986 * Improved experimental support for the upcoming ISO C++ standard, 8987 C++0x, including: 8988 + Support for <chrono>, <condition_variable>, <cstdatomic>, 8989 <forward_list>, <initializer_list>, <mutex>, <ratio>, 8990 <system_error>, and <thread>. 8991 + unique_ptr, <algorithm> additions, exception propagation, and 8992 support for the new character types in <string> and <limits>. 8993 + Existing facilities now exploit initializer lists, defaulted 8994 and deleted functions, and the newly implemented core C++0x 8995 features. 8996 + Some standard containers are more efficient together with 8997 stateful allocators, i.e., no allocator is constructed on the 8998 fly at element construction time. 8999 * Experimental support for non-standard pointer types in containers. 9000 * The long standing libstdc++/30928 has been fixed for targets 9001 running glibc 2.10 or later. 9002 * As usual, many small and larger bug fixes, in particular quite a 9003 few corner cases in <locale>. 9004 9005 Fortran 9006 9007 * GNU Fortran now employs libcpp directly instead of using cc1 as an 9008 external preprocessor. The [7]-cpp option was added to allow manual 9009 invocation of the preprocessor without relying on filename 9010 extensions. 9011 * The [8]-Warray-temporaries option warns about array temporaries 9012 generated by the compiler, as an aid to optimization. 9013 * The [9]-fcheck-array-temporaries option has been added, printing a 9014 notification at run time, when an array temporary had to be created 9015 for an function argument. Contrary to -Warray-temporaries the 9016 warning is only printed if the array is noncontiguous. 9017 * Improved generation of DWARF debugging symbols 9018 * If using an intrinsic not part of the selected standard (via -std= 9019 and -fall-intrinsics) gfortran will now treat it as if this 9020 procedure were declared EXTERNAL and try to link to a user-supplied 9021 procedure. -Wintrinsics-std will warn whenever this happens. The 9022 now-useless option -Wnonstd-intrinsic was removed. 9023 * The flag -falign-commons has been added to control the alignment of 9024 variables in COMMON blocks, which is enabled by default in line 9025 with previous GCC version. Using -fno-align-commons one can force 9026 commons to be contiguous in memory as required by the Fortran 9027 standard, however, this slows down the memory access. The option 9028 -Walign-commons, which is enabled by default, warns when padding 9029 bytes were added for alignment. The proper solution is to sort the 9030 common objects by decreasing storage size, which avoids the 9031 alignment problems. 9032 * Fortran 2003 support has been extended: 9033 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 9034 now supported (except internal reads from/writes to wide 9035 strings). [10]-fbackslash now supports also \unnnn and 9036 \Unnnnnnnn to enter Unicode characters. 9037 + Asynchronous I/O (implemented as synchronous I/O) and the 9038 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 9039 are now supported in I/O statements. 9040 + Support for Fortran 2003 structure constructors and for array 9041 constructor with typespec has been added. 9042 + Procedure Pointers (but not yet as component in derived types 9043 and as function results) are now supported. 9044 + Abstract types, type extension, and type-bound procedures 9045 (both PROCEDURE and GENERIC but not as operators). Note: As 9046 CLASS/polymorphyic types are not implemented, type-bound 9047 procedures with PASS accept as non-standard extension TYPE 9048 arguments. 9049 * Fortran 2008 support has been added: 9050 + The -std=f2008 option and support for the file extensions 9051 .f2008 and .F2008 has been added. 9052 + The g0 format descriptor is now supported. 9053 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 9054 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 9055 are now available (some of them existed as GNU extension 9056 before). Note: The hyperbolic functions are not yet supporting 9057 complex arguments and the three- argument version of BESSEL_*N 9058 is not available. 9059 + The bit intrinsics LEADZ and TRAILZ have been added. 9060 9061 Java (GCJ) 9062 9063 Ada 9064 9065 * The Ada runtime now supports multilibs on many platforms including 9066 x86_64, SPARC and PowerPC. Their build is enabled by default. 9067 9068 New Targets and Target Specific Improvements 9069 9070 ARM 9071 9072 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 9073 Cortex-R4F processors and has many other improvements to 9074 optimization for ARM processors. 9075 * GCC now supports the VFPv3 variant with 16 double-precision 9076 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 9077 renamed to -mfpu=vfpv3. 9078 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 9079 erratum on Cortex-M3 processors. 9080 * GCC now supports the __sync_* atomic operations for ARM EABI 9081 GNU/Linux. 9082 * The section anchors optimization is now enabled by default when 9083 optimizing for ARM. 9084 * GCC now uses a new EABI-compatible profiling interface for EABI 9085 targets. This requires a function __gnu_mcount_nc, which is 9086 provided by GNU libc versions 2.8 and later. 9087 9088 AVR 9089 9090 * The -mno-tablejump option has been deprecated because it has the 9091 same effect as the -fno-jump-tables option. 9092 * Added support for these new AVR devices: 9093 + ATA6289 9094 + ATtiny13A 9095 + ATtiny87 9096 + ATtiny167 9097 + ATtiny327 9098 + ATmega8C1 9099 + ATmega16C1 9100 + ATmega32C1 9101 + ATmega8M1 9102 + ATmega16M1 9103 + ATmega32M1 9104 + ATmega32U4 9105 + ATmega16HVB 9106 + ATmega4HVD 9107 + ATmega8HVD 9108 + ATmega64C1 9109 + ATmega64M1 9110 + ATmega16U4 9111 + ATmega32U6 9112 + ATmega128RFA1 9113 + AT90PWM81 9114 + AT90SCR100 9115 + M3000F 9116 + M3000S 9117 + M3001B 9118 9119 IA-32/x86-64 9120 9121 * Support for Intel AES built-in functions and code generation is 9122 available via -maes. 9123 * Support for Intel PCLMUL built-in function and code generation is 9124 available via -mpclmul. 9125 * Support for Intel AVX built-in functions and code generation is 9126 available via -mavx. 9127 * Automatically align the stack for local variables with alignment 9128 requirement. 9129 * GCC can now utilize the SVML library for vectorizing calls to a set 9130 of C99 functions if -mveclibabi=svml is specified and you link to 9131 an SVML ABI compatible library. 9132 * On x86-64, the ABI has been changed in the following cases to 9133 conform to the x86-64 ABI: 9134 + Passing/returning structures with flexible array member: 9135 struct foo 9136 { 9137 int i; 9138 int flex[]; 9139 }; 9140 + Passing/returning structures with complex float member: 9141 struct foo 9142 { 9143 int i; 9144 __complex__ float f; 9145 }; 9146 + Passing/returning unions with long double member: 9147 union foo 9148 { 9149 int x; 9150 long double ld; 9151 }; 9152 Code built with previous versions of GCC that uses any of these is 9153 not compatible with code built with GCC 4.4.0 or later. 9154 * A new target attribute was added to allow programmers to change the 9155 target options like -msse2 or -march=k8 for an individual function. 9156 You can also change the target options via the GCC target pragma 9157 for functions defined after the pragma. 9158 * GCC can now be configured with options --with-arch-32, 9159 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 9160 --with-tune-64 to control the default optimization separately for 9161 32-bit and 64-bit modes. 9162 9163 IA-32/IA64 9164 9165 * Support for __float128 (TFmode) IEEE quad type and corresponding 9166 TCmode IEEE complex quad type is available via the soft-fp library 9167 on IA-32/IA64 targets. This includes basic arithmetic operations 9168 (addition, subtraction, negation, multiplication and division) on 9169 __float128 real and TCmode complex values, the full set of IEEE 9170 comparisons between __float128 values, conversions to and from 9171 float, double and long double floating point types, as well as 9172 conversions to and from signed or unsigned integer, signed or 9173 unsigned long integer and signed or unsigned quad (TImode, IA64 9174 only) integer types. Additionally, all operations generate the full 9175 set of IEEE exceptions and support the full set of IEEE rounding 9176 modes. 9177 9178 M68K/ColdFire 9179 9180 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 9181 processors. (Scheduling support for ColdFire V2 processors was 9182 added in GCC 4.3.) 9183 * GCC now supports the -mxgot option to support programs requiring 9184 many GOT entries on ColdFire. 9185 * The m68k-*-linux-gnu target now builds multilibs by default. 9186 9187 MIPS 9188 9189 * MIPS Technologies have extended the original MIPS SVR4 ABI to 9190 include support for procedure linkage tables (PLTs) and copy 9191 relocations. These extensions allow GNU/Linux executables to use a 9192 significantly more efficient code model than the one defined by the 9193 original ABI. 9194 GCC support for this code model is available via a new command-line 9195 option, -mplt. There is also a new configure-time option, 9196 --with-mips-plt, to make -mplt the default. 9197 The new code model requires support from the assembler, the linker, 9198 and the runtime C library. This support is available in binutils 9199 2.19 and GLIBC 2.9. 9200 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 9201 and 32-bit GNU/Linux shared libraries. This feature requires GNU 9202 binutils 2.19 or above. 9203 * Support for RMI's XLR processor is now available through the 9204 -march=xlr and -mtune=xlr options. 9205 * 64-bit targets can now perform 128-bit multiplications inline, 9206 instead of relying on a libgcc function. 9207 * Native GNU/Linux toolchains now support -march=native and 9208 -mtune=native, which select the host processor. 9209 * GCC now supports the R10K, R12K, R14K and R16K processors. The 9210 canonical -march= and -mtune= names for these processors are 9211 r10000, r12000, r14000 and r16000 respectively. 9212 * GCC can now work around the side effects of speculative execution 9213 on R10K processors. Please see the documentation of the 9214 -mr10k-cache-barrier option for details. 9215 * Support for the MIPS64 Release 2 instruction set has been added. 9216 The option -march=mips64r2 enables generation of these 9217 instructions. 9218 * GCC now supports Cavium Networks' Octeon processor. This support is 9219 available through the -march=octeon and -mtune=octeon options. 9220 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 9221 canonical -march= and -mtune= names for these processors are 9222 loongson2e and loongson2f. 9223 9224 picochip 9225 9226 Picochip is a 16-bit processor. A typical picoChip contains over 250 9227 small cores, each with small amounts of memory. There are three 9228 processor variants (STAN, MEM and CTRL) with different instruction sets 9229 and memory configurations and they can be chosen using the -mae option. 9230 9231 This port is intended to be a "C" only port. 9232 9233 Power Architecture and PowerPC 9234 9235 * GCC now supports the e300c2, e300c3 and e500mc processors. 9236 * GCC now supports Xilinx processors with a single-precision FPU. 9237 * Decimal floating point is now supported for e500 processors. 9238 9239 S/390, zSeries and System z9/z10 9240 9241 * Support for the IBM System z10 EC/BC processor has been added. When 9242 using the -march=z10 option, the compiler will generate code making 9243 use of instructions provided by the General-Instruction-Extension 9244 Facility and the Execute-Extension Facility. 9245 9246 VxWorks 9247 9248 * GCC now supports the thread-local storage mechanism used on 9249 VxWorks. 9250 9251 Xtensa 9252 9253 * GCC now supports thread-local storage (TLS) for Xtensa processor 9254 configurations that include the Thread Pointer option. TLS also 9255 requires support from the assembler and linker; this support is 9256 provided in the GNU binutils beginning with version 2.19. 9257 9258 Documentation improvements 9259 9260 Other significant improvements 9261 9262 GCC 4.4.1 9263 9264 This is the [11]list of problem reports (PRs) from GCC's bug tracking 9265 system that are known to be fixed in the 4.4.1 release. This list might 9266 not be complete (that is, it is possible that some PRs that have been 9267 fixed are not listed here). 9268 9269 GCC 4.4.2 9270 9271 This is the [12]list of problem reports (PRs) from GCC's bug tracking 9272 system that are known to be fixed in the 4.4.2 release. This list might 9273 not be complete (that is, it is possible that some PRs that have been 9274 fixed are not listed here). 9275 9276 GCC 4.4.3 9277 9278 This is the [13]list of problem reports (PRs) from GCC's bug tracking 9279 system that are known to be fixed in the 4.4.3 release. This list might 9280 not be complete (that is, it is possible that some PRs that have been 9281 fixed are not listed here). 9282 9283 GCC 4.4.4 9284 9285 This is the [14]list of problem reports (PRs) from GCC's bug tracking 9286 system that are known to be fixed in the 4.4.4 release. This list might 9287 not be complete (that is, it is possible that some PRs that have been 9288 fixed are not listed here). 9289 9290 GCC 4.4.5 9291 9292 This is the [15]list of problem reports (PRs) from GCC's bug tracking 9293 system that are known to be fixed in the 4.4.5 release. This list might 9294 not be complete (that is, it is possible that some PRs that have been 9295 fixed are not listed here). 9296 9297 GCC 4.4.6 9298 9299 This is the [16]list of problem reports (PRs) from GCC's bug tracking 9300 system that are known to be fixed in the 4.4.6 release. This list might 9301 not be complete (that is, it is possible that some PRs that have been 9302 fixed are not listed here). 9303 9304 GCC 4.4.7 9305 9306 This is the [17]list of problem reports (PRs) from GCC's bug tracking 9307 system that are known to be fixed in the 4.4.7 release. This list might 9308 not be complete (that is, it is possible that some PRs that have been 9309 fixed are not listed here). 9310 9311 9312 For questions related to the use of GCC, please consult these web 9313 pages and the [18]GCC manuals. If that fails, the 9314 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9315 web pages and the development of GCC are welcome on our developer 9316 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 9317 archives. 9318 9319 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 9320 distribution of this entire article is permitted in any medium, 9321 provided this notice is preserved. 9322 9323 These pages are [23]maintained by the GCC team. Last modified 9324 2019-11-28[24]. 9325 9326 References 9327 9328 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 9329 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 9330 3. http://gcc.gnu.org/gcc-4.4/porting_to.html 9331 4. https://gcc.gnu.org/wiki/Graphite 9332 5. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 9333 6. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html 9334 7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 9335 8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 9336 9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 9337 10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 9338 11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 9339 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 9340 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 9341 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 9342 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 9343 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 9344 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 9345 18. https://gcc.gnu.org/onlinedocs/ 9346 19. mailto:gcc-help (a] gcc.gnu.org 9347 20. mailto:gcc (a] gcc.gnu.org 9348 21. https://gcc.gnu.org/lists.html 9349 22. https://www.fsf.org/ 9350 23. https://gcc.gnu.org/about.html 9351 24. http://validator.w3.org/check/referer 9352 ====================================================================== 9353 http://gcc.gnu.org/gcc-4.3/index.html 9354 GCC 4.3 Release Series 9355 9356 (This release series is no longer supported.) 9357 9358 Jun 27, 2011 9359 9360 The [1]GNU project and the GCC developers are pleased to announce the 9361 release of GCC 4.3.6. 9362 9363 This release is a bug-fix release, containing fixes for regressions in 9364 GCC 4.3.5 relative to previous releases of GCC. 9365 9366 Release History 9367 9368 GCC 4.3.6 9369 Jun 27, 2011 ([2]changes) 9370 9371 GCC 4.3.5 9372 May 22, 2010 ([3]changes) 9373 9374 GCC 4.3.4 9375 August 4, 2009 ([4]changes) 9376 9377 GCC 4.3.3 9378 January 24, 2009 ([5]changes) 9379 9380 GCC 4.3.2 9381 August 27, 2008 ([6]changes) 9382 9383 GCC 4.3.1 9384 June 6, 2008 ([7]changes) 9385 9386 GCC 4.3.0 9387 March 5, 2008 ([8]changes) 9388 9389 References and Acknowledgements 9390 9391 GCC used to stand for the GNU C Compiler, but since the compiler 9392 supports several other languages aside from C, it now stands for the 9393 GNU Compiler Collection. 9394 9395 A list of [9]successful builds is updated as new information becomes 9396 available. 9397 9398 The GCC developers would like to thank the numerous people that have 9399 contributed new features, improvements, bug fixes, and other changes as 9400 well as test results to GCC. This [10]amazing group of volunteers is 9401 what makes GCC successful. 9402 9403 For additional information about GCC please refer to the [11]GCC 9404 project web site or contact the [12]GCC development mailing list. 9405 9406 To obtain GCC please use [13]our mirror sites or [14]our version 9407 control system. 9408 9409 9410 For questions related to the use of GCC, please consult these web 9411 pages and the [15]GCC manuals. If that fails, the 9412 [16]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9413 web pages and the development of GCC are welcome on our developer 9414 list at [17]gcc (a] gcc.gnu.org. All of [18]our lists have public 9415 archives. 9416 9417 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 9418 distribution of this entire article is permitted in any medium, 9419 provided this notice is preserved. 9420 9421 These pages are [20]maintained by the GCC team. Last modified 9422 2020-01-14[21]. 9423 9424 References 9425 9426 1. http://www.gnu.org/ 9427 2. http://gcc.gnu.org/gcc-4.3/changes.html 9428 3. http://gcc.gnu.org/gcc-4.3/changes.html 9429 4. http://gcc.gnu.org/gcc-4.3/changes.html 9430 5. http://gcc.gnu.org/gcc-4.3/changes.html 9431 6. http://gcc.gnu.org/gcc-4.3/changes.html 9432 7. http://gcc.gnu.org/gcc-4.3/changes.html 9433 8. http://gcc.gnu.org/gcc-4.3/changes.html 9434 9. http://gcc.gnu.org/gcc-4.3/buildstat.html 9435 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9436 11. http://gcc.gnu.org/index.html 9437 12. mailto:gcc (a] gcc.gnu.org 9438 13. http://gcc.gnu.org/mirrors.html 9439 14. http://gcc.gnu.org/git.html 9440 15. https://gcc.gnu.org/onlinedocs/ 9441 16. mailto:gcc-help (a] gcc.gnu.org 9442 17. mailto:gcc (a] gcc.gnu.org 9443 18. https://gcc.gnu.org/lists.html 9444 19. https://www.fsf.org/ 9445 20. https://gcc.gnu.org/about.html 9446 21. http://validator.w3.org/check/referer 9447 ====================================================================== 9448 http://gcc.gnu.org/gcc-4.3/changes.html 9449 GCC 4.3 Release Series 9450 Changes, New Features, and Fixes 9451 9452 The latest release in the 4.3 release series is [1]GCC 4.3.5. 9453 9454 Caveats 9455 9456 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 9457 various front-end languages it supports. See the [4]prerequisites 9458 page for version requirements. 9459 * ColdFire targets now treat long double as having the same format as 9460 double. In earlier versions of GCC, they used the 68881 long double 9461 format instead. 9462 * The m68k-uclinux target now uses the same calling conventions as 9463 m68k-linux-gnu. You can select the original calling conventions by 9464 configuring for m68k-uclinuxoldabi instead. Note that 9465 m68k-uclinuxoldabi also retains the original 80-bit long double on 9466 ColdFire targets. 9467 * The -fforce-mem option has been removed because it has had no 9468 effect in the last few GCC releases. 9469 * The i386 -msvr3-shlib option has been removed since it is no longer 9470 used. 9471 * Fastcall for i386 has been changed not to pass aggregate arguments 9472 in registers, following Microsoft compilers. 9473 * Support for the AOF assembler has been removed from the ARM back 9474 end; this affects only the targets arm-semi-aof and armel-semi-aof, 9475 which are no longer recognized. We removed these targets without a 9476 deprecation period because we discovered that they have been 9477 unusable since GCC 4.0.0. 9478 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 9479 has been removed. This support had been deprecated since GCC 4.0.0. 9480 * Support for a number of older systems and recently unmaintained or 9481 untested target ports of GCC has been declared obsolete in GCC 4.3. 9482 Unless there is activity to revive them, the next release of GCC 9483 will have their sources permanently removed. 9484 All GCC ports for the following processor architectures have been 9485 declared obsolete: 9486 + Morpho MT (mt-*) 9487 The following aliases for processor architectures have been 9488 declared obsolete. Users should use the indicated generic target 9489 names instead, with compile-time options such as -mcpu or 9490 configure-time options such as --with-cpu to control the 9491 configuration more precisely. 9492 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 9493 instead). 9494 + parisc*-*-* (use hppa*-*-* instead). 9495 + m680[012]0-*-* (use m68k-*-* instead). 9496 All GCC ports for the following operating systems have been 9497 declared obsolete: 9498 + BeOS (*-*-beos*) 9499 + kaOS (*-*-kaos*) 9500 + GNU/Linux using the a.out object format (*-*-linux*aout*) 9501 + GNU/Linux using version 1 of the GNU C Library 9502 (*-*-linux*libc1*) 9503 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 9504 *-*-solaris2.[0-6].*) 9505 + Miscellaneous System V (*-*-sysv*) 9506 + WindISS (*-*-windiss*) 9507 Also, those for some individual systems on particular architectures 9508 have been obsoleted: 9509 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 9510 + CRIS with a.out object format (cris-*-aout) 9511 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 9512 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 9513 + PRO on PA-RISC (hppa1.1-*-pro*) 9514 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 9515 i[34567]86-sequent-sysv4*) 9516 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 9517 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 9518 was previously [5]removed in 2001, leaving only the support 9519 for UWIN as a target now being deprecated) 9520 + ChorusOS on PowerPC (powerpc-*-chorusos*) 9521 + All VAX configurations apart from NetBSD and OpenBSD 9522 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 9523 * The [6]-Wconversion option has been modified. Its purpose now is to 9524 warn for implicit conversions that may alter a value. This new 9525 behavior is available for both C and C++. Warnings about 9526 conversions between signed and unsigned integers can be disabled by 9527 using -Wno-sign-conversion. In C++, they are disabled by default 9528 unless -Wsign-conversion is explicitly requested. The old behavior 9529 of -Wconversion, that is, warn for prototypes causing a type 9530 conversion that is different from what would happen to the same 9531 argument in the absence of a prototype, has been moved to a new 9532 option -Wtraditional-conversion, which is only available for C. 9533 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 9534 been removed because they were deprecated for more than 3 GCC major 9535 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 9536 -mtune=pentiumpro as a replacement. 9537 * The -funsafe-math-optimizations option now automatically turns on 9538 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 9539 reassociation and thus may introduce or remove traps. 9540 * The -ftree-vectorize option is now on by default under -O3. In 9541 order to generate code for a SIMD extension, it has to be enabled 9542 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 9543 i?86 and x86_64. 9544 * More information on porting to GCC 4.3 from previous versions of 9545 GCC can be found in the [7]porting guide for this release. 9546 9547 General Optimizer Improvements 9548 9549 * The GCC middle-end has been integrated with the MPFR library. This 9550 allows GCC to evaluate and replace at compile-time calls to 9551 built-in math functions having constant arguments with their 9552 mathematically equivalent results. In making use of MPFR, GCC can 9553 generate correct results regardless of the math library 9554 implementation or floating point precision of the host platform. 9555 This also allows GCC to generate identical results regardless of 9556 whether one compiles in native or cross-compile configurations to a 9557 particular target. The following built-in functions take advantage 9558 of this new capability: acos, acosh, asin, asinh, atan2, atan, 9559 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 9560 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 9561 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 9562 tan, tanh, tgamma, y0, y1 and yn. The float and long double 9563 variants of these functions (e.g. sinf and sinl) are also handled. 9564 The sqrt and cabs functions with constant arguments were already 9565 optimized in prior GCC releases. Now they also use MPFR. 9566 * A new forward propagation pass on RTL was added. The new pass 9567 replaces several slower transformations, resulting in compile-time 9568 improvements as well as better code generation in some cases. 9569 * A new command-line switch -frecord-gcc-switches has been added to 9570 GCC, although it is only enabled for some targets. The switch 9571 causes the command line that was used to invoke the compiler to be 9572 recorded into the object file that is being created. The exact 9573 format of this recording is target and binary file format 9574 dependent, but it usually takes the form of a note section 9575 containing ASCII text. The switch is related to the -fverbose-asm 9576 switch, but that one only records the information in the assembler 9577 output file as comments, so the information never reaches the 9578 object file. 9579 * The inliner heuristic is now aware of stack frame consumption. New 9580 command-line parameters --param large-stack-frame and --param 9581 large-stack-frame-growth can be used to limit stack frame size 9582 growth caused by inlining. 9583 * During feedback directed optimizations, the expected block size the 9584 memcpy, memset and bzero functions operate on is discovered and for 9585 cases of commonly used small sizes, specialized inline code is 9586 generated. 9587 * __builtin_expect no longer requires its argument to be a compile 9588 time constant. 9589 * Interprocedural optimization was reorganized to work on functions 9590 in SSA form. This enables more precise and cheaper dataflow 9591 analysis and makes writing interprocedural optimizations easier. 9592 The following improvements have been implemented on top of this 9593 framework: 9594 + Pre-inline optimization: Selected local optimization passes 9595 are run before the inliner (and other interprocedural passes) 9596 are executed. This significantly improves the accuracy of code 9597 growth estimates used by the inliner and reduces the overall 9598 memory footprint for large compilation units. 9599 + Early inlining (a simple bottom-up inliner pass inlining only 9600 functions whose body is smaller than the expected call 9601 overhead) is now executed with the early optimization passes, 9602 thus inlining already optimized function bodies into an 9603 unoptimized function that is subsequently optimized by early 9604 optimizers. This enables the compiler to quickly eliminate 9605 abstraction penalty in C++ programs. 9606 + Interprocedural constant propagation now operate on SSA form 9607 increasing accuracy of the analysis. 9608 * A new internal representation for GIMPLE statements has been 9609 contributed, resulting in compile-time memory savings. 9610 * The vectorizer was enhanced to support vectorization of outer 9611 loops, intra-iteration parallelism (loop-aware SLP), vectorization 9612 of strided accesses and loops with multiple data-types. Run-time 9613 dependency testing using loop versioning was added. The cost model, 9614 turned on by -fvect-cost-model, was developed. 9615 9616 New Languages and Language specific improvements 9617 9618 * We have added new command-line options 9619 -finstrument-functions-exclude-function-list and 9620 -finstrument-functions-exclude-file-list. They provide more control 9621 over which functions are annotated by the -finstrument-functions 9622 option. 9623 9624 C family 9625 9626 * Implicit conversions between generic vector types are now only 9627 permitted when the two vectors in question have the same number of 9628 elements and compatible element types. (Note that the restriction 9629 involves compatible element types, not implicitly-convertible 9630 element types: thus, a vector type with element type int may not be 9631 implicitly converted to a vector type with element type unsigned 9632 int.) This restriction, which is in line with specifications for 9633 SIMD architectures such as AltiVec, may be relaxed using the flag 9634 -flax-vector-conversions. This flag is intended only as a 9635 compatibility measure and should not be used for new code. 9636 * -Warray-bounds has been added and is now enabled by default for 9637 -Wall . It produces warnings for array subscripts that can be 9638 determined at compile time to be always out of bounds. 9639 -Wno-array-bounds will disable the warning. 9640 * The constructor and destructor function attributes now accept 9641 optional priority arguments which control the order in which the 9642 constructor and destructor functions are run. 9643 * New [8]command-line options -Wtype-limits, -Wold-style-declaration, 9644 -Wmissing-parameter-type, -Wempty-body, -Wclobbered and 9645 -Wignored-qualifiers have been added for finer control of the 9646 diverse warnings enabled by -Wextra. 9647 * A new function attribute alloc_size has been added to mark up 9648 malloc style functions. For constant sized allocations this can be 9649 used to find out the size of the returned pointer using the 9650 __builtin_object_size() function for buffer overflow checking and 9651 similar. This supplements the already built-in malloc and calloc 9652 constant size handling. 9653 * Integer constants written in binary are now supported as a GCC 9654 extension. They consist of a prefix 0b or 0B, followed by a 9655 sequence of 0 and 1 digits. 9656 * A new predefined macro __COUNTER__ has been added. It expands to 9657 sequential integral values starting from 0. In conjunction with the 9658 ## operator, this provides a convenient means to generate unique 9659 identifiers. 9660 * A new command-line option -fdirectives-only has been added. It 9661 enables a special preprocessing mode which improves the performance 9662 of applications like distcc and ccache. 9663 * Fixed-point data types and operators have been added. They are 9664 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 9665 Currently, only MIPS targets are supported. 9666 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 9667 N1241, is now supported as a GCC extension to C for targets 9668 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 9669 and x86_64-*-linux-gnu. The feature introduces new data types 9670 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 9671 DD, and DL. 9672 9673 C++ 9674 9675 * [9]Experimental support for the upcoming ISO C++ standard, C++0x. 9676 * -Wc++0x-compat has been added and is now enabled by default for 9677 -Wall. It produces warnings for constructs whose meaning differs 9678 between ISO C++ 1998 and C++0x. 9679 * The -Wparentheses option now works for C++ as it does for C. It 9680 warns if parentheses are omitted when operators with confusing 9681 precedence are nested. It also warns about ambiguous else 9682 statements. Since -Wparentheses is enabled by -Wall, this may cause 9683 additional warnings with existing C++ code which uses -Wall. These 9684 new warnings may be disabled by using -Wall -Wno-parentheses. 9685 * The -Wmissing-declarations now works for C++ as it does for C. 9686 * The -fvisibility-ms-compat flag was added, to make it easier to 9687 port larger projects using shared libraries from Microsoft's Visual 9688 Studio to ELF and Mach-O systems. 9689 * C++ attribute handling has been overhauled for template arguments 9690 (ie dependent types). In particular, __attribute__((aligned(T))); 9691 works for C++ types. 9692 9693 Runtime Library (libstdc++) 9694 9695 * [10]Experimental support for the upcoming ISO C++ standard, C++0x. 9696 * Support for TR1 mathematical special functions and regular 9697 expressions. 9698 * Default what implementations give more elaborate exception strings 9699 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 9700 * Header dependencies have been streamlined, reducing unnecessary 9701 includes and pre-processed bloat. 9702 * Variadic template implementations of items in <tuple> and 9703 <functional>. 9704 * An experimental [11]parallel mode has been added. This is a 9705 parallel implementation of many C++ Standard library algorithms, 9706 like std::accumulate, std::for_each, std::transform, or std::sort, 9707 to give but four examples. These algorithms can be substituted for 9708 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 9709 or all existing algorithms can be transformed via the 9710 -D_GLIBCXX_PARALLEL macro. 9711 * Debug mode versions of classes in <unordered_set> and 9712 <unordered_map>. 9713 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 9714 now <backward/hash_set> and <backward/hash_map>. This code: 9715 #include <ext/hash_set> 9716 __gnu_cxx::hash_set<int> s; 9717 9718 Can be transformed (in order of preference) to: 9719 #include <tr1/unordered_set> 9720 std::tr1::unordered_set<int> s; 9721 9722 or 9723 #include <backward/hash_set> 9724 __gnu_cxx::hash_set<int> s; 9725 9726 Similar transformations apply to __gnu_cxx::hash_map, 9727 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 9728 __gnu_cxx::hash_multiset. 9729 9730 Fortran 9731 9732 * Due to the fact that the GMP and MPFR libraries are required for 9733 all languages, Fortran is no longer special in this regard and is 9734 available by default. 9735 * The [12]-fexternal-blas option has been added, which generates 9736 calls to BLAS routines for intrinsic matrix operations such as 9737 matmul rather than using the built-in algorithms. 9738 * Support to give a backtrace (compiler flag -fbacktrace or 9739 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 9740 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 9741 run-time error occured. 9742 * GNU Fortran now defines __GFORTRAN__ when it runs the C 9743 preprocessor (CPP). 9744 * The [13]-finit-local-zero, -finit-real, -finit-integer, 9745 -finit-character, and -finit-logical options have been added, which 9746 can be used to initialize local variables. 9747 * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added, 9748 which calculate the Gamma function and its logarithm. Use EXTERNAL 9749 gamma if you want to use your own gamma function. 9750 * GNU Fortran now regards the backslash character as literal (as 9751 required by the Fortran 2003 standard); using [16]-fbackslash GNU 9752 Fortran interprets backslashes as C-style escape characters. 9753 * The [17]interpretation of binary, octal and hexadecimal (BOZ) 9754 literal constants has been changed. Before they were always 9755 interpreted as integer; now they are bit-wise transferred as 9756 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 9757 2003 standard, and for real and complex variables in DATA 9758 statements or when directly assigned to real and complex variables. 9759 Everywhere else and especially in expressions they are still 9760 regarded as integer constants. 9761 * Fortran 2003 support has been extended: 9762 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 9763 + Pointer intent 9764 + Intrinsic module ISO_ENV_FORTRAN 9765 + Interoperability with C (ISO C Bindings) 9766 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 9767 attribute) 9768 + Fortran 2003 BOZ 9769 9770 Java (GCJ) 9771 9772 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 9773 This enables the use of all 1.5 language features, and fixes most 9774 existing front end bugs. 9775 * libgcj now supports all 1.5 language features which require runtime 9776 support: foreach, enum, annotations, generics, and auto-boxing. 9777 * We've made many changes to the tools shipped with gcj. 9778 + The old jv-scan tool has been removed. This tool never really 9779 worked properly. There is no replacement. 9780 + gcjh has been rewritten. Some of its more obscure options no 9781 longer work, but are still recognized in an attempt at 9782 compatibility. gjavah is a new program with similar 9783 functionality but different command-line options. 9784 + grmic and grmiregistry have been rewritten. grmid has been 9785 added. 9786 + gjar replaces the old fastjar. 9787 + gjarsigner (used for signing jars), gkeytool (used for key 9788 management), gorbd (for CORBA), gserialver (computes 9789 serialization UIDs), and gtnameserv (also for CORBA) are now 9790 installed. 9791 * The ability to dump the contents of the java run time heap to a 9792 file for off-line analysis has been added. The heap dumps may be 9793 analyzed with the new gc-analyze tool. They may be generated on 9794 out-of-memory conditions or on demand and are controlled by the new 9795 run time class gnu.gcj.util.GCInfo. 9796 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 9797 provide correct, updated, timezone information. This means that 9798 packagers no longer have to update libgcj when a time zone change 9799 is published. 9800 9801 New Targets and Target Specific Improvements 9802 9803 IA-32/x86-64 9804 9805 * Tuning for Intel Core 2 processors is available via -mtune=core2 9806 and -march=core2. 9807 * Tuning for AMD Geode processors is available via -mtune=geode and 9808 -march=geode. 9809 * Code generation of block move (memcpy) and block set (memset) was 9810 rewritten. GCC can now pick the best algorithm (loop, unrolled 9811 loop, instruction with rep prefix or a library call) based on the 9812 size of the block being copied and the CPU being optimized for. A 9813 new option -minline-stringops-dynamically has been added. With this 9814 option string operations of unknown size are expanded such that 9815 small blocks are copied by in-line code, while for large blocks a 9816 library call is used. This results in faster code than 9817 -minline-all-stringops when the library implementation is capable 9818 of using cache hierarchy hints. The heuristic choosing the 9819 particular algorithm can be overwritten via -mstringop-strategy. 9820 Newly also memset of values different from 0 is inlined. 9821 * GCC no longer places the cld instruction before string operations. 9822 Both i386 and x86-64 ABI documents mandate the direction flag to be 9823 clear at the entry of a function. It is now invalid to set the flag 9824 in asm statement without reseting it afterward. 9825 * Support for SSSE3 built-in functions and code generation are 9826 available via -mssse3. 9827 * Support for SSE4.1 built-in functions and code generation are 9828 available via -msse4.1. 9829 * Support for SSE4.2 built-in functions and code generation are 9830 available via -msse4.2. 9831 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 9832 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 9833 allow explicit control of x87 floating point precision. 9834 * Support for __float128 (TFmode) IEEE quad type and corresponding 9835 TCmode IEEE complex quad type is available via the soft-fp library 9836 on x86_64 targets. This includes basic arithmetic operations 9837 (addition, subtraction, negation, multiplication and division) on 9838 __float128 real and TCmode complex values, the full set of IEEE 9839 comparisons between __float128 values, conversions to and from 9840 float, double and long double floating point types, as well as 9841 conversions to and from signed or unsigned integer, signed or 9842 unsigned long integer and signed or unsigned quad (TImode) integer 9843 types. Additionally, all operations generate the full set of IEEE 9844 exceptions and support the full set of IEEE rounding modes. 9845 * GCC can now utilize the ACML library for vectorizing calls to a set 9846 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 9847 link to an ACML ABI compatible library. 9848 9849 ARM 9850 9851 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 9852 has been added. 9853 9854 CRIS 9855 9856 New features 9857 9858 * Compiler and Library support for the CRIS v32 architecture, as 9859 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 9860 added. 9861 9862 Configuration changes 9863 9864 * The cris-*-elf target now includes support for CRIS v32, including 9865 libraries, through the -march=v32 option. 9866 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 9867 * A new crisv32-*-linux* target defaults to generate code for CRIS 9868 v32. 9869 * The cris-*-aout target has been obsoleted. 9870 9871 Improved support for built-in functions 9872 9873 * GCC can now use the lz and swapwbr instructions to implement the 9874 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 9875 * __builtin_bswap32 is now implemented using the swapwb instruction, 9876 when available. 9877 9878 m68k and ColdFire 9879 9880 New features 9881 9882 * Support for several new ColdFire processors has been added. You can 9883 generate code for them using the new -mcpu option. 9884 * All targets now support ColdFire processors. 9885 * m68k-uclinux targets have improved support for C++ constructors and 9886 destructors, and for shared libraries. 9887 * It is now possible to set breakpoints on the first or last line of 9888 a function, even if there are no statements on that line. 9889 9890 Optimizations 9891 9892 * Support for sibling calls has been added. 9893 * More use is now made of the ColdFire mov3q instruction. 9894 * __builtin_clz is now implemented using the ff1 ColdFire 9895 instruction, when available. 9896 * GCC now honors the -m68010 option. 68010 code now uses clr rather 9897 than move to zero volatile memory. 9898 * 68020 targets and above can now use symbol(index.size*scale) 9899 addresses for indexed array accesses. Earlier compilers would 9900 always load the symbol into a base register first. 9901 9902 Configuration changes 9903 9904 * All m68k and ColdFire targets now allow the default processor to be 9905 set at configure time using --with-cpu. 9906 * A --with-arch configuration option has been added. This option 9907 allows you to restrict a target to ColdFire or non-ColdFire 9908 processors. 9909 9910 Preprocessor macros 9911 9912 * An __mcfv*__ macro is now defined for all ColdFire targets. 9913 (Earlier versions of GCC only defined __mcfv4e__.) 9914 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 9915 * All targets now define __mc68010 and __mc68010__ when generating 9916 68010 code. 9917 9918 Command-line changes 9919 9920 * New command-line options -march, -mcpu, -mtune and -mhard-float 9921 have been added. These options apply to both m68k and ColdFire 9922 targets. 9923 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 9924 versions of -mshort, etc. 9925 * -fforce-addr has been removed. It is now ignored by the compiler. 9926 9927 Other improvements 9928 9929 * ColdFire targets now try to maintain a 4-byte-aligned stack where 9930 possible. 9931 * m68k-uclinux targets now try to avoid situations that lead to the 9932 load-time error: BINFMT_FLAT: reloc outside program. 9933 9934 MIPS 9935 9936 Changes to existing configurations 9937 9938 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 9939 and n64. Every GNU/Linux configuration now builds these libraries 9940 by default. 9941 * GNU/Linux configurations now generate -mno-shared code unless 9942 overridden by -fpic, -fPIC, -fpie or -fPIE. 9943 * mipsisa32*-linux-gnu configurations now generate hard-float code by 9944 default, just like other mipsisa32* and mips*-linux-gnu 9945 configurations. You can build a soft-float version of any 9946 mips*-linux-gnu configuration by passing --with-float=soft to 9947 configure. 9948 * mips-wrs-vxworks now supports run-time processes (RTPs). 9949 9950 Changes to existing command-line options 9951 9952 * The -march and -mtune options no longer accept 24k as a processor 9953 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 9954 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 9955 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 9956 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 9957 24kx, 24kex and 34kx. 9958 9959 New configurations 9960 9961 GCC now supports the following configurations: 9962 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 9963 default. Earlier releases also recognized this configuration, but 9964 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 9965 you can customize any mips*-linux-gnu* configuration to a 9966 particular ISA or processor by passing an appropriate --with-arch 9967 option to configure. 9968 * mipsisa*-sde-elf*, which provides compatibility with MIPS 9969 Technologies' SDE toolchains. The configuration uses the SDE 9970 libraries by default, but you can use it like other newlib-based 9971 ELF configurations by passing --with-newlib to configure. It is the 9972 only configuration besides mips64vr*-elf* to build MIPS16 as well 9973 as non-MIPS16 libraries. 9974 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 9975 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 9976 and 64-bit forms of the EABI. 9977 9978 New processors and application-specific extensions 9979 9980 * Support for the SmartMIPS ASE is available through the new 9981 -msmartmips option. 9982 * Support for revision 2 of the DSP ASE is available through the new 9983 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 9984 indicates the revision of the ASE in use. 9985 * Support for the 4KS and 74K families of processors is available 9986 through the -march and -mtune options. 9987 9988 Improved support for built-in functions 9989 9990 * GCC can now use load-linked, store-conditional and sync 9991 instructions to implement atomic built-in functions such as 9992 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 9993 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 9994 * GCC can now use the clz and dclz instructions to implement the 9995 __builtin_ctz and __builtin_ffs families of functions. 9996 * There is a new __builtin___clear_cache function for flushing the 9997 instruction cache. GCC expands this function inline on MIPS32 9998 revision 2 targets, otherwise it calls the function specified by 9999 -mcache-flush-func. 10000 10001 MIPS16 improvements 10002 10003 * GCC can now compile objects that contain a mixture of MIPS16 and 10004 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 10005 for specifying which mode a function should use. 10006 * A new option called -minterlink-mips16 makes non-MIPS16 code 10007 link-compatible with MIPS16 code. 10008 * After many bug fixes, the long-standing MIPS16 -mhard-float support 10009 should now work fairly reliably. 10010 * GCC can now use the MIPS16e save and restore instructions. 10011 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 10012 with -G0 -fsection-anchors is often smaller than code compiled with 10013 -G8. However, please note that you must usually compile all objects 10014 in your application with the same -G option; see the documentation 10015 of -G for details. 10016 * A new option called-mcode-readable specifies which instructions are 10017 allowed to load from the code segment. -mcode-readable=yes is the 10018 default and says that any instruction may load from the code 10019 segment. The other alternatives are -mcode-readable=pcrel, which 10020 says that only PC-relative MIPS16 instructions may load from the 10021 code segment, and -mcode-readable=no, which says that no 10022 instruction may do so. Please see the documentation for more 10023 details, including example uses. 10024 10025 Small-data improvements 10026 10027 There are three new options for controlling small data: 10028 * -mno-extern-sdata, which disables small-data accesses for 10029 externally-defined variables. Code compiled with -Gn 10030 -mno-extern-sdata will be link-compatible with any -G setting 10031 between -G0 and -Gn inclusive. 10032 * -mno-local-sdata, which disables the use of small-data sections for 10033 data that is not externally visible. This option can be a useful 10034 way of reducing small-data usage in less performance-critical parts 10035 of an application. 10036 * -mno-gpopt, which disables the use of the $gp register while still 10037 honoring the -G limit when placing externally-visible data. This 10038 option implies -mno-extern-sdata and -mno-local-sdata and it can be 10039 useful in situations where $gp does not necessarily hold the 10040 expected value. 10041 10042 Miscellaneous improvements 10043 10044 * There is a new option called -mbranch-cost for tweaking the 10045 perceived cost of branches. 10046 * If GCC is configured to use a version of GAS that supports the 10047 .gnu_attribute directive, it will use that directive to record 10048 certain properties of the output code. .gnu_attribute is new to GAS 10049 2.18. 10050 * There are two new function attributes, near and far, for overriding 10051 the command-line setting of -mlong-calls on a function-by-function 10052 basis. 10053 * -mfp64, which previously required a 64-bit target, now works with 10054 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 10055 mipsisa*-sde-elf* configurations provide suitable library support. 10056 * GCC now recognizes the -mdmx and -mmt options and passes them down 10057 to the assembler. It does nothing else with the options at present. 10058 10059 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 10060 (BEA) 10061 10062 * Support has been added for this new architecture. 10063 10064 RS6000 (POWER/PowerPC) 10065 10066 * Support for the PowerPC 750CL paired-single instructions has been 10067 added with a new powerpc-*-linux*paired* target configuration. It 10068 is enabled by an associated -mpaired option and can be accessed 10069 using new built-in functions. 10070 * Support for auto-detecting architecture and system configuration to 10071 auto-select processor optimization tuning. 10072 * Support for VMX on AIX 5.3 has been added. 10073 * Support for AIX Version 6.1 has been added. 10074 10075 S/390, zSeries and System z9 10076 10077 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 10078 added. When using the -march=z9-ec option, the compiler will 10079 generate code making use of instructions provided by the decimal 10080 floating point facility and the floating point conversion facility 10081 (pfpo). Besides the instructions used to implement decimal floating 10082 point operations these facilities also contain instructions to move 10083 between general purpose and floating point registers and to modify 10084 and copy the sign-bit of floating point values. 10085 * When the -march=z9-ec option is used the new 10086 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 10087 decimal floating point hardware instructions will be used or not. 10088 If none of them is given the hardware support is enabled by 10089 default. 10090 * The -mstack-guard option can now be omitted when using stack 10091 checking via -mstack-size in order to let GCC choose a sensible 10092 stack guard value according to the frame size of each function. 10093 * Various changes to improve performance of generated code have been 10094 implemented, including: 10095 + The condition code set by an add logical with carry 10096 instruction is now available for overflow checks like: a + b + 10097 carry < b. 10098 + The test data class instruction is now used to implement 10099 sign-bit and infinity checks of binary and decimal floating 10100 point numbers. 10101 10102 SPARC 10103 10104 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 10105 added. 10106 10107 Xtensa 10108 10109 * Stack unwinding for exception handling now uses by default a 10110 specialized version of DWARF unwinding. This is not 10111 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 10112 Xtensa with previous versions of GCC. 10113 * For Xtensa processors that include the Conditional Store option, 10114 the built-in functions for atomic memory access are now implemented 10115 using S32C1I instructions. 10116 * If the Xtensa NSA option is available, GCC will use it to implement 10117 the __builtin_ctz and __builtin_clz functions. 10118 10119 Documentation improvements 10120 10121 * Existing libstdc++ documentation has been edited and restructured 10122 into a single DocBook XML manual. The results can be viewed online 10123 [18]here. 10124 10125 Other significant improvements 10126 10127 * The compiler's --help command-line option has been extended so that 10128 it now takes an optional set of arguments. These arguments restrict 10129 the information displayed to specific classes of command-line 10130 options, and possibly only a subset of those options. It is also 10131 now possible to replace the descriptive text associated with each 10132 displayed option with an indication of its current value, or for 10133 binary options, whether it has been enabled or disabled. 10134 Here are some examples. The following will display all the options 10135 controlling warning messages: 10136 --help=warnings 10137 10138 Whereas this will display all the undocumented, target specific 10139 options: 10140 --help=target,undocumented 10141 10142 This sequence of commands will display the binary optimizations 10143 that are enabled by -O3: 10144 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 10145 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 10146 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 10147 10148 * The configure options --with-pkgversion and --with-bugurl have been 10149 added. These allow distributors of GCC to include a 10150 distributor-specific string in manuals and --version output and to 10151 specify the URL for reporting bugs in their versions of GCC. 10152 10153 GCC 4.3.1 10154 10155 This is the [19]list of problem reports (PRs) from GCC's bug tracking 10156 system that are known to be fixed in the 4.3.1 release. This list might 10157 not be complete (that is, it is possible that some PRs that have been 10158 fixed are not listed here). 10159 10160 Target Specific Changes 10161 10162 IA-32/x86-64 10163 10164 ABI changes 10165 10166 * Starting with GCC 4.3.1, decimal floating point variables are 10167 aligned to their natural boundaries when they are passed on the 10168 stack for i386. 10169 10170 Command-line changes 10171 10172 * Starting with GCC 4.3.1, the -mcld option has been added to 10173 automatically generate a cld instruction in the prologue of 10174 functions that use string instructions. This option is used for 10175 backward compatibility on some operating systems and can be enabled 10176 by default for 32-bit x86 targets by configuring GCC with the 10177 --enable-cld configure option. 10178 10179 GCC 4.3.2 10180 10181 This is the [20]list of problem reports (PRs) from GCC's bug tracking 10182 system that are known to be fixed in the 4.3.2 release. This list might 10183 not be complete (that is, it is possible that some PRs that have been 10184 fixed are not listed here). 10185 10186 GCC 4.3.3 10187 10188 This is the [21]list of problem reports (PRs) from GCC's bug tracking 10189 system that are known to be fixed in the 4.3.3 release. This list might 10190 not be complete (that is, it is possible that some PRs that have been 10191 fixed are not listed here). 10192 10193 GCC 4.3.4 10194 10195 This is the [22]list of problem reports (PRs) from GCC's bug tracking 10196 system that are known to be fixed in the 4.3.4 release. This list might 10197 not be complete (that is, it is possible that some PRs that have been 10198 fixed are not listed here). 10199 10200 GCC 4.3.5 10201 10202 This is the [23]list of problem reports (PRs) from GCC's bug tracking 10203 system that are known to be fixed in the 4.3.5 release. This list might 10204 not be complete (that is, it is possible that some PRs that have been 10205 fixed are not listed here). 10206 10207 GCC 4.3.6 10208 10209 This is the [24]list of problem reports (PRs) from GCC's bug tracking 10210 system that are known to be fixed in the 4.3.6 release. This list might 10211 not be complete (that is, it is possible that some PRs that have been 10212 fixed are not listed here). 10213 10214 10215 For questions related to the use of GCC, please consult these web 10216 pages and the [25]GCC manuals. If that fails, the 10217 [26]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10218 web pages and the development of GCC are welcome on our developer 10219 list at [27]gcc (a] gcc.gnu.org. All of [28]our lists have public 10220 archives. 10221 10222 Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and 10223 distribution of this entire article is permitted in any medium, 10224 provided this notice is preserved. 10225 10226 These pages are [30]maintained by the GCC team. Last modified 10227 2019-11-28[31]. 10228 10229 References 10230 10231 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 10232 2. https://gmplib.org/ 10233 3. https://www.mpfr.org/ 10234 4. https://gcc.gnu.org/install/prerequisites.html 10235 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 10236 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 10237 7. http://gcc.gnu.org/gcc-4.3/porting_to.html 10238 8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 10239 9. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 10240 10. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 10241 11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 10242 12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 10243 13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 10244 14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 10245 15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 10246 16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 10247 17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 10248 18. https://gcc.gnu.org/onlinedocs/libstdc++/ 10249 19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 10250 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 10251 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 10252 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 10253 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 10254 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 10255 25. https://gcc.gnu.org/onlinedocs/ 10256 26. mailto:gcc-help (a] gcc.gnu.org 10257 27. mailto:gcc (a] gcc.gnu.org 10258 28. https://gcc.gnu.org/lists.html 10259 29. https://www.fsf.org/ 10260 30. https://gcc.gnu.org/about.html 10261 31. http://validator.w3.org/check/referer 10262 ====================================================================== 10263 http://gcc.gnu.org/gcc-4.2/index.html 10264 GCC 4.2 Release Series 10265 10266 (This release series is no longer supported.) 10267 10268 May 19, 2008 10269 10270 The [1]GNU project and the GCC developers are pleased to announce the 10271 release of GCC 4.2.4. 10272 10273 This release is a bug-fix release, containing fixes for regressions in 10274 GCC 4.2.3 relative to previous releases of GCC. 10275 10276 Release History 10277 10278 GCC 4.2.4 10279 May 19, 2008 ([2]changes) 10280 10281 GCC 4.2.3 10282 February 1, 2008 ([3]changes) 10283 10284 GCC 4.2.2 10285 October 7, 2007 ([4]changes) 10286 10287 GCC 4.2.1 10288 July 18, 2007 ([5]changes) 10289 10290 GCC 4.2.0 10291 May 13, 2007 ([6]changes) 10292 10293 References and Acknowledgements 10294 10295 GCC used to stand for the GNU C Compiler, but since the compiler 10296 supports several other languages aside from C, it now stands for the 10297 GNU Compiler Collection. 10298 10299 A list of [7]successful builds is updated as new information becomes 10300 available. 10301 10302 The GCC developers would like to thank the numerous people that have 10303 contributed new features, improvements, bug fixes, and other changes as 10304 well as test results to GCC. This [8]amazing group of volunteers is 10305 what makes GCC successful. 10306 10307 For additional information about GCC please refer to the [9]GCC project 10308 web site or contact the [10]GCC development mailing list. 10309 10310 To obtain GCC please use [11]our mirror sites or [12]our version 10311 control system. 10312 10313 10314 For questions related to the use of GCC, please consult these web 10315 pages and the [13]GCC manuals. If that fails, the 10316 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10317 web pages and the development of GCC are welcome on our developer 10318 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 10319 archives. 10320 10321 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 10322 distribution of this entire article is permitted in any medium, 10323 provided this notice is preserved. 10324 10325 These pages are [18]maintained by the GCC team. Last modified 10326 2020-01-14[19]. 10327 10328 References 10329 10330 1. http://www.gnu.org/ 10331 2. http://gcc.gnu.org/gcc-4.2/changes.html 10332 3. http://gcc.gnu.org/gcc-4.2/changes.html 10333 4. http://gcc.gnu.org/gcc-4.2/changes.html 10334 5. http://gcc.gnu.org/gcc-4.2/changes.html 10335 6. http://gcc.gnu.org/gcc-4.2/changes.html 10336 7. http://gcc.gnu.org/gcc-4.2/buildstat.html 10337 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10338 9. http://gcc.gnu.org/index.html 10339 10. mailto:gcc (a] gcc.gnu.org 10340 11. http://gcc.gnu.org/mirrors.html 10341 12. http://gcc.gnu.org/git.html 10342 13. https://gcc.gnu.org/onlinedocs/ 10343 14. mailto:gcc-help (a] gcc.gnu.org 10344 15. mailto:gcc (a] gcc.gnu.org 10345 16. https://gcc.gnu.org/lists.html 10346 17. https://www.fsf.org/ 10347 18. https://gcc.gnu.org/about.html 10348 19. http://validator.w3.org/check/referer 10349 ====================================================================== 10350 http://gcc.gnu.org/gcc-4.2/changes.html 10351 GCC 4.2 Release Series 10352 Changes, New Features, and Fixes 10353 10354 Caveats 10355 10356 * GCC no longer accepts the -fshared-data option. This option has had 10357 no effect in any GCC 4 release; the targets to which the option 10358 used to apply had been removed before GCC 4.0. 10359 10360 General Optimizer Improvements 10361 10362 * New command-line options specify the possible relationships among 10363 parameters and between parameters and global data. For example, 10364 -fargument-noalias-anything specifies that arguments do not alias 10365 any other storage. 10366 Each language will automatically use whatever option is required by 10367 the language standard. You should not need to use these options 10368 yourself. 10369 10370 New Languages and Language specific improvements 10371 10372 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 10373 * New command-line options -fstrict-overflow and -Wstrict-overflow 10374 have been added. -fstrict-overflow tells the compiler that it may 10375 assume that the program follows the strict signed overflow 10376 semantics permitted for the language: for C and C++ this means that 10377 the compiler may assume that signed overflow does not occur. For 10378 example, a loop like 10379 for (i = 1; i > 0; i *= 2) 10380 10381 is presumably intended to continue looping until i overflows. With 10382 -fstrict-overflow, the compiler may assume that signed overflow 10383 will not occur, and transform this into an infinite loop. 10384 -fstrict-overflow is turned on by default at -O2, and may be 10385 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 10386 be used to warn about cases where the compiler assumes that signed 10387 overflow will not occur. It takes five different levels: 10388 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 10389 -Wstrict-overflow=1 is enabled by -Wall. 10390 * The new command-line option -fno-toplevel-reorder directs GCC to 10391 emit top-level functions, variables, and asm statements in the same 10392 order that they appear in the input file. This is intended to 10393 support existing code which relies on a particular ordering (for 10394 example, code which uses top-level asm statements to switch 10395 sections). For new code, it is generally better to use function and 10396 variable attributes. The -fno-toplevel-reorder option may be used 10397 for most cases which currently use -fno-unit-at-a-time. The 10398 -fno-unit-at-a-time option will be removed in some future version 10399 of GCC. If you know of a case which requires -fno-unit-at-a-time 10400 which is not fixed by -fno-toplevel-reorder, please open a bug 10401 report. 10402 10403 C family 10404 10405 * The pragma redefine_extname will now macro expand its tokens for 10406 compatibility with SunPRO. 10407 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 10408 GCC to handle inline functions as specified in the C99 standard. In 10409 preparation for this, GCC 4.2 will warn about any use of non-static 10410 inline functions in gnu99 or c99 mode. This new warning may be 10411 disabled with the new gnu_inline function attribute or the new 10412 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 10413 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 10414 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 10415 in the current compilation. 10416 * A new command-line option -Waddress has been added to warn about 10417 suspicious uses of memory addresses as, for example, using the 10418 address of a function in a conditional expression, and comparisons 10419 against the memory address of a string literal. This warning is 10420 enabled by -Wall. 10421 10422 C++ 10423 10424 * C++ visibility handling has been overhauled. 10425 Restricted visiblity is propagated from classes to members, from 10426 functions to local statics, and from templates and template 10427 arguments to instantiations, unless the latter has explicitly 10428 declared visibility. 10429 The visibility attribute for a class must come between the 10430 class-key and the name, not after the closing brace. 10431 Attributes are now allowed for enums and elaborated-type-specifiers 10432 that only declare a type. 10433 Members of the anonymous namespace are now local to a particular 10434 translation unit, along with any other declarations which use them, 10435 though they are still treated as having external linkage for 10436 language semantics. 10437 * The (undocumented) extension which permitted templates with default 10438 arguments to be bound to template template parameters with fewer 10439 parameters has been removed. For example: 10440 template <template <typename> class C> 10441 void f(C<double>) {} 10442 10443 template <typename T, typename U = int> 10444 struct S {}; 10445 10446 template void f(S<double>); 10447 10448 is no longer accepted by G++. The reason this code is not accepted 10449 is that S is a template with two parameters; therefore, it cannot 10450 be bound to C which has only one parameter. 10451 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 10452 releases, have been removed. 10453 * The command-line option -fconst-strings, deprecated in previous GCC 10454 releases, has been removed. 10455 * The configure variable enable-__cxa_atexit is now enabled by 10456 default for more targets. Enabling this variable is necessary in 10457 order for static destructors to be executed in the correct order, 10458 but it depends upon the presence of a non-standard C library in the 10459 target library in order to work. The variable is now enabled for 10460 more targets which are known to have suitable C libraries. 10461 * -Wextra will produce warnings for if statements with a semicolon as 10462 the only body, to catch code like: 10463 if (a); 10464 return 1; 10465 return 0; 10466 10467 To suppress the warning in valid cases, use { } instead. 10468 * The C++ front end now also produces strict aliasing warnings when 10469 -fstrict-aliasing -Wstrict-aliasing is in effect. 10470 10471 Runtime Library (libstdc++) 10472 10473 * Added support for TR1 <random>, <complex>, and C compatibility 10474 headers. In addition, a lock-free version of shared_ptr was 10475 contributed as part of Phillip Jordan's Google Summer of Code 10476 project on lock-free containers. 10477 * In association with the Summer of Code work on lock-free 10478 containers, the interface for atomic builtins was adjusted, 10479 creating simpler alternatives for non-threaded code paths. Also, 10480 usage was consolidated and all elements were moved from namespace 10481 std to namespace__gnu_cxx. Affected interfaces are the functions 10482 __exchange_and_add, __atomic_add, and the objects __mutex, 10483 __recursive_mutex, and __scoped_lock. 10484 * Support for versioning weak symbol names via namespace association 10485 was added. However, as this changes the names of exported symbols, 10486 this is turned off by default in the current ABI. Intrepid users 10487 can enable this feature by using 10488 --enable-symvers=gnu-versioned-namespace during configuration. 10489 * Revised, simplified, and expanded policy-based associative 10490 containers, including data types for tree and trie forms 10491 (basic_tree, tree, trie), lists (list_update), and both 10492 collision-chaining and probing hash-based containers 10493 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 10494 the [3]documentation. 10495 * The implementation of the debug mode was modified, whereby the 10496 debug namespaces were nested inside of namespace std and namespace 10497 __gnu_cxx in order to resolve some long standing corner cases 10498 involving name lookup. Debug functionality from the policy-based 10499 data structures was consolidated and enabled with the single macro, 10500 _GLIBCXX_DEBUG. See PR 26142 for more information. 10501 * Added extensions for type traits: __conditional_type, 10502 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 10503 * Added a typelist implementation for compile-time meta-programming. 10504 Elements for typelist construction and operation can be found 10505 within namespace __gnu_cxx::typelist. 10506 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 10507 exception-safety. 10508 * Enabled library-wide visibility control, allowing -fvisibility to 10509 be used. 10510 * Consolidated all nested namespaces and the conversion of 10511 __gnu_internal implementation-private details to anonymous 10512 namespaces whenever possible. 10513 * Implemented LWG resolutions DR 431 and DR 538. 10514 10515 Fortran 10516 10517 * Support for allocatable components has been added (TR 15581 and 10518 Fortran 2003). 10519 * Support for the Fortran 2003 streaming IO extension has been added. 10520 * The GNU Fortran compiler now uses 4-byte record markers by default 10521 for unformatted files to be compatible with g77 and most other 10522 compilers. The implementation allows for records greater than 2 GB 10523 and is compatible with several other compilers. Older versions of 10524 gfortran used 8-byte record markers by default (on most systems). 10525 In order to change the length of the record markers, e.g. to read 10526 unformatted files created by older gfortran versions, the 10527 [4]-frecord-marker=8 option can be used. 10528 10529 Java (GCJ) 10530 10531 * A new command-line option -static-libgcj has been added for targets 10532 that use a linker compatible with GNU Binutils. As its name 10533 implies, this causes libgcj to be linked statically. In some cases 10534 this causes the resulting executable to start faster and use less 10535 memory than if the shared version of libgcj were used. However 10536 caution should be used as it can also cause essential parts of the 10537 library to be omitted. Some of these issues are discussed in: 10538 [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj 10539 * fastjar is no longer bundled with GCC. To build libgcj, you will 10540 need either InfoZIP (both zip and unzip) or an external jar 10541 program. In the former case, the GCC build will install a jar shell 10542 script that is based on InfoZIP and provides the same functionality 10543 as fastjar. 10544 10545 New Targets and Target Specific Improvements 10546 10547 IA-32/x86-64 10548 10549 * -mtune=generic can now be used to generate code running well on 10550 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 10551 Pentium-M, Intel Pentium 4 and Intel Core 2. 10552 * -mtune=native and -march=native will produce code optimized for the 10553 host architecture as detected using the cpuid instruction. 10554 * Added a new command-line option -fstackrealign and and 10555 __attribute__ ((force_align_arg_pointer)) to realign the stack at 10556 runtime. This allows functions compiled with a vector-aligned stack 10557 to be invoked from legacy objects that keep only word-alignment. 10558 10559 SPARC 10560 10561 * The default CPU setting has been changed from V7 to V9 in 32-bit 10562 mode on Solaris 7 and above. This is already the case in 64-bit 10563 mode. It can be overridden by specifying --with-cpu at configure 10564 time. 10565 * Back-end support of built-in functions for atomic memory access has 10566 been implemented. 10567 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 10568 added. 10569 10570 M32C 10571 10572 * Various bug fixes have made some functions (notably, functions 10573 returning structures) incompatible with previous releases. 10574 Recompiling all libraries is recommended. Note that code quality 10575 has considerably improved since 4.1, making a recompile even more 10576 beneficial. 10577 10578 MIPS 10579 10580 * Added support for the Broadcom SB-1A core. 10581 10582 IA-64 10583 10584 * Added support for IA-64 data and control speculation. By default 10585 speculation is enabled only during second scheduler pass. A number 10586 of machine flags was introduced to control the usage of speculation 10587 for both scheduler passes. 10588 10589 HPPA 10590 10591 * Added Java language support (libffi and libjava) for 32-bit HP-UX 10592 11 target. 10593 10594 Obsolete Systems 10595 10596 Documentation improvements 10597 10598 PDF Documentation 10599 10600 * A make pdf target has been added to the top-level makefile, 10601 enabling automated production of PDF documentation files. 10602 (Front-ends external to GCC should modify their Make-lang.in file 10603 to add a lang.pdf: target.) 10604 10605 Other significant improvements 10606 10607 Build system improvements 10608 10609 * All the components of the compiler are now bootstrapped by default. 10610 This improves the resilience to bugs in the system compiler or 10611 binary compatibility problems, as well as providing better testing 10612 of GCC 4.2 itself. In addition, if you build the compiler from a 10613 combined tree, the assembler, linker, etc. will also be 10614 bootstrapped (i.e. built with themselves). 10615 You can disable this behavior, and go back to the pre-GCC 4.2 set 10616 up, by configuring GCC with --disable-bootstrap. 10617 * The rules that configure follows to find target tools resemble more 10618 closely the locations that the built compiler will search. In 10619 addition, you can use the new configure option --with-target-tools 10620 to specify where to find the target tools used during the build, 10621 without affecting what the built compiler will use. 10622 This can be especially useful when building packages of GCC. For 10623 example, you may want to build GCC with GNU as or ld, even if the 10624 resulting compiler to work with the native assembler and linker. To 10625 do so, you can use --with-target-tools to point to the native 10626 tools. 10627 10628 Incompatible changes to the build system 10629 10630 * Front-ends external to GCC should modify their Make-lang.in file to 10631 replace double-colon rules (e.g. dvi::) with normal rules (like 10632 lang.dvi:). Front-end makefile hooks do not use double-colon rules 10633 anymore. 10634 * Up to GCC 4.1, a popular way to specify the target tools used 10635 during the build was to create directories named gas, binutils, 10636 etc. in the build tree, and create links to the tools from there. 10637 This does not work any more when the compiler is bootstrapped. The 10638 new configure option --with-target-tools provides a better way to 10639 achieve the same effect, and works for all native and cross 10640 settings. 10641 10642 10643 For questions related to the use of GCC, please consult these web 10644 pages and the [6]GCC manuals. If that fails, the 10645 [7]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10646 web pages and the development of GCC are welcome on our developer 10647 list at [8]gcc (a] gcc.gnu.org. All of [9]our lists have public archives. 10648 10649 Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and 10650 distribution of this entire article is permitted in any medium, 10651 provided this notice is preserved. 10652 10653 These pages are [11]maintained by the GCC team. Last modified 10654 2019-11-28[12]. 10655 10656 References 10657 10658 1. http://gcc.gnu.org/projects/gomp/ 10659 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 10660 3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 10661 4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 10662 5. https://gcc.gnu.org/wiki/Statically_linking_libgcj 10663 6. https://gcc.gnu.org/onlinedocs/ 10664 7. mailto:gcc-help (a] gcc.gnu.org 10665 8. mailto:gcc (a] gcc.gnu.org 10666 9. https://gcc.gnu.org/lists.html 10667 10. https://www.fsf.org/ 10668 11. https://gcc.gnu.org/about.html 10669 12. http://validator.w3.org/check/referer 10670 ====================================================================== 10671 http://gcc.gnu.org/gcc-4.1/index.html 10672 GCC 4.1 Release Series 10673 10674 (This release series is no longer supported.) 10675 10676 February 13, 2007 10677 10678 The [1]GNU project and the GCC developers are pleased to announce the 10679 release of GCC 4.1.2. 10680 10681 This release is a bug-fix release, containing fixes for regressions in 10682 GCC 4.1.1 relative to previous releases of GCC. 10683 10684 Release History 10685 10686 GCC 4.1.2 10687 February 13, 2007 ([2]changes) 10688 10689 GCC 4.1.1 10690 May 24, 2006 ([3]changes) 10691 10692 GCC 4.1.0 10693 February 28, 2006 ([4]changes) 10694 10695 References and Acknowledgements 10696 10697 GCC used to stand for the GNU C Compiler, but since the compiler 10698 supports several other languages aside from C, it now stands for the 10699 GNU Compiler Collection. 10700 10701 A list of [5]successful builds is updated as new information becomes 10702 available. 10703 10704 The GCC developers would like to thank the numerous people that have 10705 contributed new features, improvements, bug fixes, and other changes as 10706 well as test results to GCC. This [6]amazing group of volunteers is 10707 what makes GCC successful. 10708 10709 For additional information about GCC please refer to the [7]GCC project 10710 web site or contact the [8]GCC development mailing list. 10711 10712 To obtain GCC please use [9]our mirror sites or [10]our version control 10713 system. 10714 10715 10716 For questions related to the use of GCC, please consult these web 10717 pages and the [11]GCC manuals. If that fails, the 10718 [12]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10719 web pages and the development of GCC are welcome on our developer 10720 list at [13]gcc (a] gcc.gnu.org. All of [14]our lists have public 10721 archives. 10722 10723 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 10724 distribution of this entire article is permitted in any medium, 10725 provided this notice is preserved. 10726 10727 These pages are [16]maintained by the GCC team. Last modified 10728 2020-01-14[17]. 10729 10730 References 10731 10732 1. http://www.gnu.org/ 10733 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 10734 3. http://gcc.gnu.org/gcc-4.1/changes.html 10735 4. http://gcc.gnu.org/gcc-4.1/changes.html 10736 5. http://gcc.gnu.org/gcc-4.1/buildstat.html 10737 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10738 7. http://gcc.gnu.org/index.html 10739 8. mailto:gcc (a] gcc.gnu.org 10740 9. http://gcc.gnu.org/mirrors.html 10741 10. http://gcc.gnu.org/git.html 10742 11. https://gcc.gnu.org/onlinedocs/ 10743 12. mailto:gcc-help (a] gcc.gnu.org 10744 13. mailto:gcc (a] gcc.gnu.org 10745 14. https://gcc.gnu.org/lists.html 10746 15. https://www.fsf.org/ 10747 16. https://gcc.gnu.org/about.html 10748 17. http://validator.w3.org/check/referer 10749 ====================================================================== 10750 http://gcc.gnu.org/gcc-4.1/changes.html 10751 GCC 4.1 Release Series 10752 Changes, New Features, and Fixes 10753 10754 The latest release in the 4.1 release series is [1]GCC 4.1.2. 10755 10756 Caveats 10757 10758 General Optimizer Improvements 10759 10760 * GCC now has infrastructure for inter-procedural optimizations and 10761 the following inter-procedural optimizations are implemented: 10762 + Profile guided inlining. When doing profile feedback guided 10763 optimization, GCC can now use the profile to make better 10764 informed decisions on whether inlining of a function is 10765 profitable or not. This means that GCC will no longer inline 10766 functions at call sites that are not executed very often, and 10767 that functions at hot call sites are more likely to be 10768 inlined. 10769 A new parameter min-inline-recursive-probability is also now 10770 available to throttle recursive inlining of functions with 10771 small average recursive depths. 10772 + Discovery of pure and const functions, a form of side-effects 10773 analysis. While older GCC releases could also discover such 10774 special functions, the new IPA-based pass runs earlier so that 10775 the results are available to more optimizers. The pass is also 10776 simply more powerful than the old one. 10777 + Analysis of references to static variables and type escape 10778 analysis, also forms of side-effects analysis. The results of 10779 these passes allow the compiler to be less conservative about 10780 call-clobbered variables and references. This results in more 10781 redundant loads being eliminated and in making static 10782 variables candidates for register promotion. 10783 + Improvement of RTL-based alias analysis. The results of type 10784 escape analysis are fed to the RTL type-based alias analyzer, 10785 allowing it to disambiguate more memory references. 10786 + Interprocedural constant propagation and function versioning. 10787 This pass looks for functions that are always called with the 10788 same constant value for one or more of the function arguments, 10789 and propagates those constants into those functions. 10790 + GCC will now eliminate static variables whose usage was 10791 optimized out. 10792 + -fwhole-program --combine can now be used to make all 10793 functions in program static allowing whole program 10794 optimization. As an exception, the main function and all 10795 functions marked with the new externally_visible attribute are 10796 kept global so that programs can link with runtime libraries. 10797 * GCC can now do a form of partial dead code elimination (PDCE) that 10798 allows code motion of expressions to the paths where the result of 10799 the expression is actually needed. This is not always a win, so the 10800 pass has been limited to only consider profitable cases. Here is an 10801 example: 10802 int foo (int *, int *); 10803 int 10804 bar (int d) 10805 { 10806 int a, b, c; 10807 b = d + 1; 10808 c = d + 2; 10809 a = b + c; 10810 if (d) 10811 { 10812 foo (&b, &c); 10813 a = b + c; 10814 } 10815 printf ("%d\n", a); 10816 } 10817 10818 The a = b + c can be sunk to right before the printf. Normal code 10819 sinking will not do this, it will sink the first one above into the 10820 else-branch of the conditional jump, which still gives you two 10821 copies of the code. 10822 * GCC now has a value range propagation pass. This allows the 10823 compiler to eliminate bounds checks and branches. The results of 10824 the pass can also be used to accurately compute branch 10825 probabilities. 10826 * The pass to convert PHI nodes to straight-line code (a form of 10827 if-conversion for GIMPLE) has been improved significantly. The two 10828 most significant improvements are an improved algorithm to 10829 determine the order in which the PHI nodes are considered, and an 10830 improvement that allow the pass to consider if-conversions of basic 10831 blocks with more than two predecessors. 10832 * Alias analysis improvements. GCC can now differentiate between 10833 different fields of structures in Tree-SSA's virtual operands form. 10834 This lets stores/loads from non-overlapping structure fields not 10835 conflict. A new algorithm to compute points-to sets was contributed 10836 that can allows GCC to see now that p->a and p->b, where p is a 10837 pointer to a structure, can never point to the same field. 10838 * Various enhancements to auto-vectorization: 10839 + Incrementally preserve SSA form when vectorizing. 10840 + Incrementally preserve loop-closed form when vectorizing. 10841 + Improvements to peeling for alignment: generate better code 10842 when the misalignment of an access is known at compile time, 10843 or when different accesses are known to have the same 10844 misalignment, even if the misalignment amount itself is 10845 unknown. 10846 + Consider dependence distance in the vectorizer. 10847 + Externalize generic parts of data reference analysis to make 10848 this analysis available to other passes. 10849 + Vectorization of conditional code. 10850 + Reduction support. 10851 * GCC can now partition functions in sections of hot and cold code. 10852 This can significantly improve performance due to better 10853 instruction cache locality. This feature works best together with 10854 profile feedback driven optimization. 10855 * A new pass to avoid saving of unneeded arguments to the stack in 10856 vararg functions if the compiler can prove that they will not be 10857 needed. 10858 * Transition of basic block profiling to tree level implementation 10859 has been completed. The new implementation should be considerably 10860 more reliable (hopefully avoiding profile mismatch errors when 10861 using -fprofile-use or -fbranch-probabilities) and can be used to 10862 drive higher level optimizations, such as inlining. 10863 The -ftree-based-profiling command-line option was removed and 10864 -fprofile-use now implies disabling old RTL level loop optimizer 10865 (-fno-loop-optimize). Speculative prefetching optimization 10866 (originally enabled by -fspeculative-prefetching) was removed. 10867 10868 New Languages and Language specific improvements 10869 10870 C and Objective-C 10871 10872 * The old Bison-based C and Objective-C parser has been replaced by a 10873 new, faster hand-written recursive-descent parser. 10874 10875 Ada 10876 10877 * The build infrastructure for the Ada runtime library and tools has 10878 been changed to be better integrated with the rest of the build 10879 infrastructure of GCC. This should make doing cross builds of Ada a 10880 bit easier. 10881 10882 C++ 10883 10884 * ARM-style name-injection of friend declarations is no longer the 10885 default. For example: 10886 struct S { 10887 friend void f(); 10888 }; 10889 10890 void g() { f(); } 10891 will not be accepted; instead a declaration of f will need to be 10892 present outside of the scope of S. The new -ffriend-injection 10893 option will enable the old behavior. 10894 * The (undocumented) extension which permitted templates with default 10895 arguments to be bound to template template parameters with fewer 10896 parameters has been deprecated, and will be removed in the next 10897 major release of G++. For example: 10898 template <template <typename> class C> 10899 void f(C<double>) {} 10900 10901 template <typename T, typename U = int> 10902 struct S {}; 10903 10904 template void f(S<double>); 10905 10906 makes use of the deprecated extension. The reason this code is not 10907 valid ISO C++ is that S is a template with two parameters; 10908 therefore, it cannot be bound to C which has only one parameter. 10909 10910 Runtime Library (libstdc++) 10911 10912 * Optimization work: 10913 + A new implementation of std::search_n is provided, better 10914 performing in case of random access iterators. 10915 + Added further efficient specializations of istream functions, 10916 i.e., character array and string extractors. 10917 + Other smaller improvements throughout. 10918 * Policy-based associative containers, designed for high-performance, 10919 flexibility and semantic safety are delivered in ext/pb_assoc. 10920 * A versatile string class, __gnu_cxx::__versa_string, providing 10921 facilities conforming to the standard requirements for 10922 basic_string, is delivered in <ext/vstring.h>. In particular: 10923 + Two base classes are provided: the default one avoids 10924 reference counting and is optimized for short strings; the 10925 alternate one, still uses it while improving in a few low 10926 level areas (e.g., alignment). See vstring_fwd.h for some 10927 useful typedefs. 10928 + Various algorithms have been rewritten (e.g., replace), the 10929 code streamlined and simple optimizations added. 10930 + Option 3 of DR 431 is implemented for both available bases, 10931 thus improving the support for stateful allocators. 10932 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 10933 libstdc++/23953) and LWG resolutions put into effect for the first 10934 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 10935 Issue 6.19). The implementation status of TR1 is now tracked in the 10936 docs in tr1.html. 10937 10938 Objective-C++ 10939 10940 * A new language front end for Objective-C++ has been added. This 10941 language allows users to mix the object oriented features of 10942 Objective-C with those of C++. 10943 10944 Java (GCJ) 10945 10946 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 10947 features (plus some 0.20 bug-fixes) 10948 + Networking 10949 o The java.net.HttpURLConnection implementation no longer 10950 buffers the entire response body in memory. This means 10951 that response bodies larger than available memory can now 10952 be handled. 10953 + (N)IO 10954 o NIO FileChannel.map implementation, fast bulk put 10955 implementation for DirectByteBuffer (speeds up this 10956 method 10x). 10957 o FileChannel.lock() and FileChannel.force() implemented. 10958 + XML 10959 o gnu.xml fix for nodes created outside a namespace 10960 context. 10961 o Add support for output indenting and 10962 cdata-section-elements output instruction in 10963 xml.transform. 10964 o xml.xpath corrections for cases where elements/attributes 10965 might have been created in non-namespace-aware mode. 10966 Corrections to handling of XSL variables and minor 10967 conformance updates. 10968 + AWT 10969 o GNU JAWT implementation, the AWT Native Interface, which 10970 allows direct access to native screen resources from 10971 within a Canvas's paint method. GNU Classpath Examples 10972 comes with a Demo, see libjava/classpath/examples/README. 10973 o awt.datatransfer updated to 1.5 with support for 10974 FlavorEvents. The gtk+ awt peers now allow copy/paste of 10975 text, images, URIs/files and serialized objects with 10976 other applications and tracking clipboard change events 10977 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 10978 objects are supported). A GNU Classpath Examples 10979 datatransfer Demo was added to show the new 10980 functionality. 10981 o Split gtk+ awt peers event handling in two threads and 10982 improve gdk lock handling (solves several awt lock ups). 10983 o Speed up awt Image loading. 10984 o Better gtk+ scrollbar peer implementation when using gtk+ 10985 >= 2.6. 10986 o Handle image loading errors correctly for gdkpixbuf and 10987 MediaTracker. 10988 o Better handle GDK lock. Properly prefix gtkpeer native 10989 functions (cp_gtk). 10990 o GdkGraphics2D has been updated to use Cairo 0.5.x or 10991 higher. 10992 o BufferedImage and GtkImage rewrites. All image drawing 10993 operations should now work correctly (flipping requires 10994 gtk+ >= 2.6) 10995 o Future Graphics2D, image and text work is documented at: 10996 [2]http://developer.classpath.org/mediation/ClasspathGrap 10997 hicsImagesText 10998 o When gtk+ 2.6 or higher is installed the default log 10999 handler will produce stack traces whenever a WARNING, 11000 CRITICAL or ERROR message is produced. 11001 + Free Swing 11002 o The RepaintManager has been reworked for more efficient 11003 painting, especially for large GUIs. 11004 o The layout manager OverlayLayout has been implemented, 11005 the BoxLayout has been rewritten to make use of the 11006 SizeRequirements utility class and caching for more 11007 efficient layout. 11008 o Improved accessibility support. 11009 o Significant progress has been made in the implementation 11010 of the javax.swing.plaf.metal package, with most UI 11011 delegates in a working state now. Please test this with 11012 your own applications and provide feedback that will help 11013 us to improve this package. 11014 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 11015 extended to highlight various features in our Free Swing 11016 implementation. And it includes a look and feel switcher 11017 for Metal (default), Ocean and GNU themes. 11018 o The javax.swing.plaf.multi package is now implemented. 11019 o Editing and several key actions for JTree and JTable were 11020 implemented. 11021 o Lots of icons and look and feel improvements for Free 11022 Swing basic and metal themes were added. Try running the 11023 GNU Classpath Swing Demo in examples 11024 (gnu.classpath.examples.swing.Demo) with: 11025 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 11026 l or 11027 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 11028 l 11029 o Start of styled text capabilites for java.swing.text. 11030 o DefaultMutableTreeNode pre-order, post-order, depth-first 11031 and breadth-first traversal enumerations implemented. 11032 o JInternalFrame colors and titlebar draw properly. 11033 o JTree is working up to par (icons, selection and keyboard 11034 traversal). 11035 o JMenus were made more compatible in visual and 11036 programmatic behavior. 11037 o JTable changeSelection and multiple selections 11038 implemented. 11039 o JButton and JToggleButton change states work properly 11040 now. 11041 o JFileChooser fixes. 11042 o revalidate() and repaint() fixes which make Free Swing 11043 much more responsive. 11044 o MetalIconFactory implemented. 11045 o Free Swing Top-Level Compatibility. JFrame, JDialog, 11046 JApplet, JInternalFrame, and JWindow are now 1.5 11047 compatible in the sense that you can call add() and 11048 setLayout() directly on them, which will have the same 11049 effect as calling getContentPane().add() and 11050 getContentPane().setLayout(). 11051 o The JTree interface has been completed. JTrees now 11052 recognizes mouse clicks and selections work. 11053 o BoxLayout works properly now. 11054 o Fixed GrayFilter to actually work. 11055 o Metal SplitPane implemented. 11056 o Lots of Free Swing text and editor stuff work now. 11057 + Free RMI and Corba 11058 o Andrew Watson, Vice President and Technical Director of 11059 the Object Management Group, has officially assigned us 11060 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 11061 mark remote classpath-specific system exceptions. 11062 Obtaining the VMCID means that GNU Classpath now is a 11063 recogniseable type of node in a highly interoperable 11064 CORBA world. 11065 o GNU Classpath now includes the first working draft to 11066 support the RMI over IIOP protocol. The current 11067 implementation is capable of remote invocations, 11068 transferring various Serializables and Externalizables 11069 via RMI-IIOP protocol. It can flatten graphs and, at 11070 least for the simple cases, is interoperable with 1.5 11071 JDKs. 11072 o org.omg.PortableInterceptor and related functionality in 11073 other packages is now implemented: 11074 # The sever and client interceptors work as required 11075 since 1.4. 11076 # The IOR interceptor works as needed for 1.5. 11077 o The org.omg.DynamicAny package is completed and passes 11078 the prepared tests. 11079 o The Portable Object Adapter should now support the output 11080 of the recent IDL to java compilers. These compilers now 11081 generate servants and not CORBA objects as before, making 11082 the output depend on the existing POA implementation. 11083 Completing POA means that such code can already be tried 11084 to run on Classpath. Our POA is tested for the following 11085 usager scenarios: 11086 # POA converts servant to the CORBA object. 11087 # Servant provides to the CORBA object. 11088 # POA activates new CORBA object with the given Object 11089 Id (byte array) that is later accessible for the 11090 servant. 11091 # During the first call, the ServantActivator provides 11092 servant for this and all subsequent calls on the 11093 current object. 11094 # During each call, the ServantLocator provides 11095 servant for this call only. 11096 # ServantLocator or ServantActivator forwards call to 11097 another server. 11098 # POA has a single servant, responsible for all 11099 objects. 11100 # POA has a default servant, but some objects are 11101 explicitly connected to they specific servants. 11102 The POA is verified using tests from the former 11103 cost.omg.org. 11104 o The CORBA implementation is now a working prototype that 11105 should support features up to 1.3 inclusive. We invite 11106 groups writing CORBA dependent applications to try 11107 Classpath implementation, reporting any possible bugs. 11108 The CORBA prototype is interoperable with Sun's 11109 implementation v 1.4, transferring object references, 11110 primitive types, narrow and wide strings, arrays, 11111 structures, trees, abstract interfaces and value types 11112 (feature of CORBA 2.3) between these two platforms. 11113 Remote exceptions are transferred and handled correctly. 11114 The stringified object references (IORs) from various 11115 sources are parsed as required. The transient (for 11116 current session) and permanent (till jre restart) 11117 redirections work. Both Little and Big Endian encoded 11118 messages are accepted. The implementation is verified 11119 using tests from the former cost.omg.org. The current 11120 release includes working examples (see the examples 11121 directory), demonstrating the client-server 11122 communication, using either CORBA Request or IDL-based 11123 stub (usually generated by a IDL to java compiler). These 11124 examples also show how to use the Classpath CORBA naming 11125 service. The IDL to java compiler is not yet written, but 11126 as our library must be compatible, it naturally accepts 11127 the output of other idlj implementations. 11128 + Misc 11129 o Updated TimeZone data against Olson tzdata2005l. 11130 o Make zip and jar packages UTF-8 clean. 11131 o "native" code builds and compiles (warning free) on 11132 Darwin and Solaris. 11133 o java.util.logging.FileHandler now rotates files. 11134 o Start of a generic JDWP framework in gnu/classpath/jdwp. 11135 This is unfinished, but feedback (at classpath (a] gnu.org) 11136 from runtime hackers is greatly appreciated. Although 11137 most of the work is currently being done around gcj/gij 11138 we want this framework to be as VM neutral as possible. 11139 Early design is described in: 11140 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 11141 o QT4 AWT peers, enable by giving configure 11142 --enable-qt-peer. Included, but not ready for production 11143 yet. They are explicitly disabled and not supported. But 11144 if you want to help with the development of these new 11145 features we are interested in feedback. You will have to 11146 explicitly enable them to try them out (and they will 11147 most likely contain bugs). 11148 o Documentation fixes all over the place. See 11149 [4]http://developer.classpath.org/doc/ 11150 11151 New Targets and Target Specific Improvements 11152 11153 IA-32/x86-64 11154 11155 * The x86-64 medium model (that allows building applications whose 11156 data segment exceeds 4GB) was redesigned to match latest ABI draft. 11157 New implementation split large datastructures into separate segment 11158 improving performance of accesses to small datastructures and also 11159 allows linking of small model libraries into medium model programs 11160 as long as the libraries are not accessing the large datastructures 11161 directly. Medium model is also supported in position independent 11162 code now. 11163 The ABI change results in partial incompatibility among medium 11164 model objects. Linking medium model libraries (or objects) compiled 11165 with new compiler into medium model program compiled with older 11166 will likely result in exceeding ranges of relocations. 11167 Binutils 2.16.91 or newer are required for compiling medium model 11168 now. 11169 11170 RS6000 (POWER/PowerPC) 11171 11172 * The AltiVec vector primitives in <altivec.h> are now implemented in 11173 a way that puts a smaller burden on the preprocessor, instead 11174 processing the "overloading" in the front ends. This should benefit 11175 compilation speed on AltiVec vector code. 11176 * AltiVec initializers now are generated more efficiently. 11177 * The popcountb instruction available on POWER5 now is generated. 11178 * The floating point round to integer instructions available on 11179 POWER5+ now is generated. 11180 * Floating point divides can be synthesized using the floating point 11181 reciprocal estimate instructions. 11182 * Double precision floating point constants are initialized as single 11183 precision values if they can be represented exactly. 11184 11185 S/390, zSeries and System z9 11186 11187 * Support for the IBM System z9 109 processor has been added. When 11188 using the -march=z9-109 option, the compiler will generate code 11189 making use of instructions provided by the extended immediate 11190 facility. 11191 * Support for 128-bit IEEE floating point has been added. When using 11192 the -mlong-double-128 option, the compiler will map the long double 11193 data type to 128-bit IEEE floating point. Using this option 11194 constitutes an ABI change, and requires glibc support. 11195 * Various changes to improve performance of generated code have been 11196 implemented, including: 11197 + In functions that do not require a literal pool, register %r13 11198 (which is traditionally reserved as literal pool pointer), can 11199 now be freely used for other purposes by the compiler. 11200 + More precise tracking of register use allows the compiler to 11201 generate more efficient function prolog and epilog code in 11202 certain cases. 11203 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 11204 instructions are now used to implement C string functions. 11205 + The MOVE CHARACTER instruction with single byte overlap is now 11206 used to implement the memset function with non-zero fill byte. 11207 + The LOAD ZERO instructions are now used where appropriate. 11208 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 11209 and INSERT IMMEDIATE instructions are now used more frequently 11210 to optimize bitfield operations. 11211 + The BRANCH ON COUNT instruction is now used more frequently. 11212 In particular, the fact that a loop contains a subroutine call 11213 no longer prevents the compiler from using this instruction. 11214 + The compiler is now aware that all shift and rotate 11215 instructions implicitly truncate the shift count to six bits. 11216 * Back-end support for the following generic features has been 11217 implemented: 11218 + The full set of [5]built-in functions for atomic memory 11219 access. 11220 + The -fstack-protector feature. 11221 + The optimization pass avoiding unnecessary stores of incoming 11222 argument registers in functions with variable argument list. 11223 11224 SPARC 11225 11226 * The default code model in 64-bit mode has been changed from 11227 Medium/Anywhere to Medium/Middle on Solaris. 11228 * TLS support is disabled by default on Solaris prior to release 10. 11229 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 11230 and later) by specifying --enable-tls at configure time. 11231 11232 MorphoSys 11233 11234 * Support has been added for this new architecture. 11235 11236 Obsolete Systems 11237 11238 Documentation improvements 11239 11240 Other significant improvements 11241 11242 * GCC can now emit code for protecting applications from 11243 stack-smashing attacks. The protection is realized by buffer 11244 overflow detection and reordering of stack variables to avoid 11245 pointer corruption. 11246 * Some built-in functions have been fortified to protect them against 11247 various buffer overflow (and format string) vulnerabilities. 11248 Compared to the mudflap bounds checking feature, the safe builtins 11249 have far smaller overhead. This means that programs built using 11250 safe builtins should not experience any measurable slowdown. 11251 11252 GCC 4.1.2 11253 11254 This is the [6]list of problem reports (PRs) from GCC's bug tracking 11255 system that are known to be fixed in the 4.1.2 release. This list might 11256 not be complete (that is, it is possible that some PRs that have been 11257 fixed are not listed here). 11258 11259 When generating code for a shared library, GCC now recognizes that 11260 global functions may be replaced when the program runs. Therefore, it 11261 is now more conservative in deducing information from the bodies of 11262 functions. For example, in this example: 11263 void f() {} 11264 void g() { 11265 try { f(); } 11266 catch (...) { 11267 cout << "Exception"; 11268 } 11269 } 11270 11271 G++ would previously have optimized away the catch clause, since it 11272 would have concluded that f cannot throw exceptions. Because users may 11273 replace f with another function in the main body of the program, this 11274 optimization is unsafe, and is no longer performed. If you wish G++ to 11275 continue to optimize as before, you must add a throw() clause to the 11276 declaration of f to make clear that it does not throw exceptions. 11277 11278 11279 For questions related to the use of GCC, please consult these web 11280 pages and the [7]GCC manuals. If that fails, the 11281 [8]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11282 web pages and the development of GCC are welcome on our developer 11283 list at [9]gcc (a] gcc.gnu.org. All of [10]our lists have public 11284 archives. 11285 11286 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and 11287 distribution of this entire article is permitted in any medium, 11288 provided this notice is preserved. 11289 11290 These pages are [12]maintained by the GCC team. Last modified 11291 2019-11-28[13]. 11292 11293 References 11294 11295 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 11296 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText 11297 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 11298 4. http://developer.classpath.org/doc/ 11299 5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 11300 6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 11301 7. https://gcc.gnu.org/onlinedocs/ 11302 8. mailto:gcc-help (a] gcc.gnu.org 11303 9. mailto:gcc (a] gcc.gnu.org 11304 10. https://gcc.gnu.org/lists.html 11305 11. https://www.fsf.org/ 11306 12. https://gcc.gnu.org/about.html 11307 13. http://validator.w3.org/check/referer 11308 ====================================================================== 11309 http://gcc.gnu.org/gcc-4.0/index.html 11310 GCC 4.0 Release Series 11311 11312 (This release series is no longer supported.) 11313 11314 January 31, 2007 11315 11316 The [1]GNU project and the GCC developers are pleased to announce the 11317 release of GCC 4.0.4. 11318 11319 This release is a bug-fix release, containing fixes for regressions in 11320 GCC 4.0.3 relative to previous releases of GCC. 11321 11322 Release History 11323 11324 GCC 4.0.4 11325 January 31, 2007 ([2]changes) 11326 11327 GCC 4.0.3 11328 March 10, 2006 ([3]changes) 11329 11330 GCC 4.0.2 11331 September 28, 2005 ([4]changes) 11332 11333 GCC 4.0.1 11334 July 7, 2005 ([5]changes) 11335 11336 GCC 4.0.0 11337 April 20, 2005 ([6]changes) 11338 11339 References and Acknowledgements 11340 11341 GCC used to stand for the GNU C Compiler, but since the compiler 11342 supports several other languages aside from C, it now stands for the 11343 GNU Compiler Collection. 11344 11345 A list of [7]successful builds is updated as new information becomes 11346 available. 11347 11348 The GCC developers would like to thank the numerous people that have 11349 contributed new features, improvements, bug fixes, and other changes as 11350 well as test results to GCC. This [8]amazing group of volunteers is 11351 what makes GCC successful. 11352 11353 For additional information about GCC please refer to the [9]GCC project 11354 web site or contact the [10]GCC development mailing list. 11355 11356 To obtain GCC please use [11]our mirror sites, or [12]our version 11357 control system. 11358 11359 11360 For questions related to the use of GCC, please consult these web 11361 pages and the [13]GCC manuals. If that fails, the 11362 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11363 web pages and the development of GCC are welcome on our developer 11364 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 11365 archives. 11366 11367 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 11368 distribution of this entire article is permitted in any medium, 11369 provided this notice is preserved. 11370 11371 These pages are [18]maintained by the GCC team. Last modified 11372 2020-01-14[19]. 11373 11374 References 11375 11376 1. http://www.gnu.org/ 11377 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 11378 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 11379 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 11380 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 11381 6. http://gcc.gnu.org/gcc-4.0/changes.html 11382 7. http://gcc.gnu.org/gcc-4.0/buildstat.html 11383 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11384 9. http://gcc.gnu.org/index.html 11385 10. mailto:gcc (a] gcc.gnu.org 11386 11. http://gcc.gnu.org/mirrors.html 11387 12. http://gcc.gnu.org/git.html 11388 13. https://gcc.gnu.org/onlinedocs/ 11389 14. mailto:gcc-help (a] gcc.gnu.org 11390 15. mailto:gcc (a] gcc.gnu.org 11391 16. https://gcc.gnu.org/lists.html 11392 17. https://www.fsf.org/ 11393 18. https://gcc.gnu.org/about.html 11394 19. http://validator.w3.org/check/referer 11395 ====================================================================== 11396 http://gcc.gnu.org/gcc-4.0/changes.html 11397 GCC 4.0 Release Series 11398 Changes, New Features, and Fixes 11399 11400 The latest release in the 4.0 release series is [1]GCC 4.0.4. 11401 11402 Caveats 11403 11404 * GCC now generates location lists by default when compiling with 11405 debug info and optimization. 11406 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 11407 or later is needed to debug binaries containing location 11408 lists. 11409 + When you are trying to view a value of a variable in a part of 11410 a function where it has no location (for example when the 11411 variable is no longer used and thus its location was used for 11412 something else) GDB will say that it is not available. 11413 You can disable generating location lists by -fno-var-tracking. 11414 * GCC no longer accepts the -fwritable-strings option. Use named 11415 character arrays when you need a writable string. 11416 * The options -freduce-all-givs and -fmove-all-movables have been 11417 discontinued. They were used to circumvent a shortcoming in the 11418 heuristics of the old loop optimization code with respect to common 11419 Fortran constructs. The new (tree) loop optimizer works differently 11420 and doesn't need those work-arounds. 11421 * The graph-coloring register allocator, formerly enabled by the 11422 option -fnew-ra, has been discontinued. 11423 * -I- has been deprecated. -iquote is meant to replace the need for 11424 this option. 11425 * The MIPS -membedded-pic and -mrnames options have been removed. 11426 * All MIPS targets now require the GNU assembler. In particular, IRIX 11427 configurations can no longer use the MIPSpro assemblers, although 11428 they do still support the MIPSpro linkers. 11429 * The SPARC option -mflat has been removed. 11430 * English-language diagnostic messages will now use Unicode quotation 11431 marks in UTF-8 locales. (Non-English messages already used the 11432 quotes appropriate for the language in previous releases.) If your 11433 terminal does not support UTF-8 but you are using a UTF-8 locale 11434 (such locales are the default on many GNU/Linux systems) then you 11435 should set LC_CTYPE=C in the environment to disable that locale. 11436 Programs that parse diagnostics and expect plain ASCII 11437 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 11438 explanation of Unicode quotation marks for more information. 11439 * The specs file is no longer installed on most platforms. Most users 11440 will be totally unaffected. However, if you are accustomed to 11441 editing the specs file yourself, you will now have to use the 11442 -dumpspecs option to generate the specs file, and then edit the 11443 resulting file. 11444 11445 General Optimizer Improvements 11446 11447 * The [3]tree ssa branch has been merged. This merge has brought in a 11448 completely new optimization framework based on a higher level 11449 intermediate representation than the existing RTL representation. 11450 Numerous new code transformations based on the new framework are 11451 available in GCC 4.0, including: 11452 + Scalar replacement of aggregates 11453 + Constant propagation 11454 + Value range propagation 11455 + Partial redundancy elimination 11456 + Load and store motion 11457 + Strength reduction 11458 + Dead store elimination 11459 + Dead and unreachable code elimination 11460 + [4]Autovectorization 11461 + Loop interchange 11462 + Tail recursion by accumulation 11463 Many of these passes outperform their counterparts from previous 11464 GCC releases. 11465 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 11466 scheduling optimization intended for loops that perform heavy 11467 computations. 11468 11469 New Languages and Language specific improvements 11470 11471 C family 11472 11473 * The sentinel attribute has been added to GCC. This function 11474 attribute allows GCC to warn when variadic functions such as execl 11475 are not NULL terminated. See the GCC manual for a complete 11476 description of its behavior. 11477 * Given __attribute__((alias("target"))) it is now an error if target 11478 is not a symbol, defined in the same translation unit. This also 11479 applies to aliases created by #pragma weak alias=target. This is 11480 because it's meaningless to define an alias to an undefined symbol. 11481 On Solaris, the native assembler would have caught this error, but 11482 GNU as does not. 11483 11484 C and Objective-C 11485 11486 * The -Wstrict-aliasing=2 option has been added. This warning catches 11487 all unsafe cases, but it may also give a warning for some cases 11488 that are safe. 11489 * The cast-as-lvalue, conditional-expression-as-lvalue and 11490 compound-expression-as-lvalue extensions, which were deprecated in 11491 3.3.4 and 3.4, have been removed. 11492 * The -fwritable-strings option, which was deprecated in 3.4, has 11493 been removed. 11494 * #pragma pack() semantics have been brought closer to those used by 11495 other compilers. This also applies to C++. 11496 * Taking the address of a variable with register storage is invalid 11497 in C. GCC now issues an error instead of a warning. 11498 * Arrays of incomplete element type are invalid in C. GCC now issues 11499 an error for such arrays. Declarations such as extern struct s x[]; 11500 (where struct s has not been defined) can be moved after the 11501 definition of struct s. Function parameters declared as arrays of 11502 incomplete type can instead be declared as pointers. 11503 11504 C++ 11505 11506 * When compiling without optimizations (-O0), the C++ front end is 11507 much faster than in any previous versions of GCC. Independent 11508 testers have measured speed-ups up to 25% in real-world production 11509 code, compared to the 3.4 family (which was already the fastest 11510 version to date). Upgrading from older versions might show even 11511 bigger improvements. 11512 * ELF visibility attributes can now be applied to a class type, so 11513 that it affects every member function of a class at once, without 11514 having to specify each individually: 11515 class __attribute__ ((visibility("hidden"))) Foo 11516 { 11517 int foo1(); 11518 void foo2(); 11519 }; 11520 The syntax is deliberately similar to the __declspec() system used 11521 by Microsoft Windows based compilers, allowing cross-platform 11522 projects to easily reuse their existing macro system for denoting 11523 exports and imports. By explicitly marking internal classes never 11524 used outside a binary as hidden, one can completely avoid PLT 11525 indirection overheads during their usage by the compiler. You can 11526 find out more about the advantages of this at 11527 [6]https://www.akkadia.org/drepper/dsohowto.pdf 11528 * The -fvisibility-inlines-hidden option has been added which marks 11529 all inlineable functions as having hidden ELF visibility, thus 11530 removing their symbol and typeinfo from the exported symbol table 11531 of the output ELF binary. Using this option can reduce the exported 11532 symbol count of template-heavy code by up to 40% with no code 11533 change at all, thus notably improving link and load times for the 11534 binary as well as a reduction in size of up to 10%. Also, check the 11535 new [7]-fvisibility option. 11536 * The compiler now uses the library interface specified by the [8]C++ 11537 ABI for thread-safe initialization of function-scope static 11538 variables. Most users should leave this alone, but embedded 11539 programmers may want to disable this by specifying 11540 -fno-threadsafe-statics for a small savings in code size. 11541 * Taking the address of an explicit register variable is no longer 11542 supported. Note that C++ allows taking the address of variables 11543 with register storage so this will continue to compile with a 11544 warning. For example, assuming that r0 is a machine register: 11545 register int foo asm ("r0"); 11546 register int bar; 11547 &foo; // error, no longer accepted 11548 &bar; // OK, with a warning 11549 * G++ has an undocumented extension to virtual function covariancy 11550 rules that allowed the overrider to return a type that was 11551 implicitly convertable to the overridden function's return type. 11552 For instance a function returning void * could be overridden by a 11553 function returning T *. This is now deprecated and will be removed 11554 in a future release. 11555 * The G++ minimum and maximum operators (<? and >?) and their 11556 compound forms (<?=) and >?=) have been deprecated and will be 11557 removed in a future version. Code using these operators should be 11558 modified to use std::min and std::max instead. 11559 * Declaration of nested classes of class templates as friends are 11560 supported: 11561 template <typename T> struct A { 11562 class B {}; 11563 }; 11564 class C { 11565 template <typename T> friend class A<T>::B; 11566 }; 11567 This complements the feature member functions of class templates as 11568 friends introduced in GCC 3.4.0. 11569 * When declaring a friend class using an unqualified name, classes 11570 outside the innermost non-class scope are not searched: 11571 class A; 11572 namespace N { 11573 class B { 11574 friend class A; // Refer to N::A which has not been declared yet 11575 // because name outside namespace N are not searched 11576 friend class ::A; // Refer to ::A 11577 }; 11578 } 11579 Hiding the friend name until declaration is still not implemented. 11580 * Friends of classes defined outside their namespace are correctly 11581 handled: 11582 namespace N { 11583 class A; 11584 } 11585 class N::A { 11586 friend class B; // Refer to N::B in GCC 4.0.0 11587 // but ::B in earlier versions of GCC 11588 }; 11589 11590 Runtime Library (libstdc++) 11591 11592 * Optimization work: 11593 + Added efficient specializations of istream functions for char 11594 and wchar_t. 11595 + Further performance tuning of strings, in particular wrt 11596 single-char append and getline. 11597 + iter_swap - and therefore most of the mutating algorithms - 11598 now makes an unqualified call to swap when the value_type of 11599 the two iterators is the same. 11600 * A large subset of the features in Technical Report 1 (TR1 for 11601 short) is experimentally delivered (i.e., no guarantees about the 11602 implementation are provided. In particular it is not promised that 11603 the library will remain link-compatible when code using TR1 is 11604 used): 11605 + General utilities such as reference_wrapper and shared_ptr. 11606 + Function objects, i.e., result_of, mem_fn, bind, function. 11607 + Support for metaprogramming. 11608 + New containers such as tuple, array, unordered_set, 11609 unordered_map, unordered_multiset, unordered_multimap. 11610 * As usual, many bugs have been fixed and LWG resolutions implemented 11611 for the first time (e.g., DR 409). 11612 11613 Java 11614 11615 * In order to prevent naming conflicts with other implementations of 11616 these tools, some GCJ binaries have been renamed: 11617 + rmic is now grmic, 11618 + rmiregistry is now grmiregistry, and 11619 + jar is now fastjar. 11620 In particular, these names were problematic for the jpackage.org 11621 packaging conventions which install symlinks in /usr/bin that point 11622 to the preferred versions of these tools. 11623 * The -findirect-dispatch argument to the compiler now works and 11624 generates code following a new "binary compatibility" ABI. Code 11625 compiled this way follows the binary compatibility rules of the 11626 Java Language Specification. 11627 * libgcj now has support for using GCJ as a JIT, using the 11628 gnu.gcj.jit family of system properties. 11629 * libgcj can now find a shared library corresponding to the bytecode 11630 representation of a class. See the documentation for the new 11631 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 11632 property. 11633 * There have been many improvements to the class library. Here are 11634 some highlights: 11635 + Much more of AWT and Swing exist. 11636 + Many new packages and classes were added, including 11637 java.util.regex, java.net.URI, javax.crypto, 11638 javax.crypto.interfaces, javax.crypto.spec, javax.net, 11639 javax.net.ssl, javax.security.auth, 11640 javax.security.auth.callback, javax.security.auth.login, 11641 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 11642 javax.imageio, javax.imageio.event, javax.imageio.spi, 11643 javax.print, javax.print.attribute, 11644 javax.print.attribute.standard, javax.print.event, and 11645 javax.xml 11646 + Updated SAX and DOM, and imported GNU JAXP 11647 11648 Fortran 11649 11650 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 11651 front end. The new front end supports Fortran 90 and Fortran 95. It 11652 may not yet be as stable as the old Fortran front end. 11653 11654 Ada 11655 11656 * Ada (with tasking and Zero Cost Exceptions) is now available on 11657 many more targets, including but not limited to: alpha-linux, 11658 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 11659 s390x-linux, sparc-linux. 11660 * Some of the new Ada 2005 features are now implemented like 11661 Wide_Wide_Character and Ada.Containers. 11662 * Many bugs have been fixed, tools and documentation improved. 11663 * To compile Ada from the sources, install an older working Ada 11664 compiler and then use --enable-languages=ada at configuration time, 11665 since the Ada front end is not currently activated by default. See 11666 the [10]Installing GCC for details. 11667 11668 New Targets and Target Specific Improvements 11669 11670 H8/300 11671 11672 * The frame layout has changed. In the new layout, the prologue of a 11673 function first saves registers and then allocate space for locals, 11674 resulting in an 1% improvement on code size. 11675 11676 IA-32/x86-64 (AMD64) 11677 11678 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 11679 log1p, log2, logb and tan mathematical builtins (and their float 11680 and long double variants) are now implemented as inline x87 11681 intrinsics when using -ffast-math. 11682 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 11683 (and their float and long double variants) are now implemented as 11684 inline x87 intrinsics when using -ffast-math. 11685 * The x87's fsincos instruction is now used automatically with 11686 -ffast-math when calculating both the sin and cos of the same 11687 argument. 11688 * Instruction selection for multiplication and division by constants 11689 has been improved. 11690 11691 IA-64 11692 11693 * Floating point division, integer division and sqrt are now inlined, 11694 resulting in significant performance improvements on some codes. 11695 11696 MIPS 11697 11698 * Division by zero checks now use conditional traps if the target 11699 processor supports them. This decreases code size by one word per 11700 division operation. The old behavior (branch and break) can be 11701 obtained either at configure time by passing --with-divide=breaks 11702 to configure or at runtime by passing -mdivide-breaks to GCC. 11703 * Support for MIPS64 paired-single instructions has been added. It is 11704 enabled by -mpaired-single and can be accessed using both the 11705 target-independent vector extensions and new MIPS-specific built-in 11706 functions. 11707 * Support for the MIPS-3D ASE has been added. It is enabled by 11708 -mips3d and provides new MIPS-3D-specific built-in functions. 11709 * The -mexplicit-relocs option now supports static n64 code (as is 11710 used, for example, in 64-bit linux kernels). -mexplicit-relocs 11711 should now be feature-complete and is enabled by default when GCC 11712 is configured to use a compatible assembler. 11713 * Support for the NEC VR4130 series has been added. This support 11714 includes the use of VR-specific instructions and a new VR4130 11715 scheduler. Full VR4130 support can be selected with -march=vr4130 11716 while code for any ISA can be tuned for the VR4130 using 11717 -mtune=vr4130. There is also a new -mvr4130-align option that 11718 produces better schedules at the cost of increased code size. 11719 * Support for the Broadcom SB-1 has been extended. There is now an 11720 SB-1 scheduler as well as support for the SB-1-specific 11721 paired-single instructions. Full SB-1 support can be selected with 11722 -march=sb1 while code for any ISA can be optimized for the SB-1 11723 using -mtune=sb1. 11724 * The compiler can now work around errata in R4000, R4400, VR4120 and 11725 VR4130 processors. These workarounds are enabled by -mfix-r4000, 11726 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 11727 and VR4130 workarounds need binutils 2.16 or above. 11728 * IRIX shared libraries are now installed into the standard library 11729 directories: o32 libraries go into lib/, n32 libraries go into 11730 lib32/ and n64 libraries go into lib64/. 11731 * The compiler supports a new -msym32 option. It can be used to 11732 optimize n64 code in which all symbols are known to have 32-bit 11733 values. 11734 11735 S/390 and zSeries 11736 11737 * New command-line options help to generate code intended to run in 11738 an environment where stack space is restricted, e.g. Linux kernel 11739 code: 11740 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 11741 warnings for single functions that require large or dynamic 11742 stack frames. 11743 + -mstack-size and -mstack-guard generate code that checks for 11744 stack overflow at run time. 11745 + -mpacked-stack generates code that reduces the stack frame 11746 size of many functions by reusing unneeded parts of the stack 11747 bias area. 11748 * The -msoft-float option now ensures that generated code never 11749 accesses floating point registers. 11750 * The s390x-ibm-tpf target now fully supports C++, including 11751 exceptions and threads. 11752 * Various changes to improve performance of the generated code have 11753 been implemented, including: 11754 + GCC now uses sibling calls where possible. 11755 + Condition code handling has been optimized, allowing GCC to 11756 omit redundant comparisons in certain cases. 11757 + The cost function guiding many optimizations has been refined 11758 to more accurately represent the z900 and z990 processors. 11759 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 11760 instructions are now used to avoid conditional branches in 11761 certain cases. 11762 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 11763 optimize address arithmetic required to access large stack 11764 frames. 11765 + GCC now makes more efficient use of memory-to-memory type 11766 instructions (MVC, CLC, ...). 11767 + More precise tracking of special register use allows better 11768 instruction scheduling, in particular of the function prologue 11769 and epilogue sequences. 11770 + The Java front end now generates inline code to implement 11771 integer division, instead of calling library routines. 11772 11773 SPARC 11774 11775 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 11776 -mf934 have been removed. They have been replaced with -mcpu=xxx. 11777 * The internal model used to estimate the relative cost of each 11778 instruction has been updated. It is expected to give better results 11779 on recent UltraSPARC processors. 11780 * Code generation for function prologues and epilogues has been 11781 improved, resulting in better scheduling and allowing multiple exit 11782 points in functions. 11783 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 11784 It is enabled by -mvis and provides new built-in functions for VIS 11785 instructions on UltraSPARC processors. 11786 * The option -mapp-regs has been turned on by default on Solaris too. 11787 11788 NetWare 11789 11790 * Novell NetWare (on ix86, no other hardware platform was ever really 11791 supported by this OS) has been re-enabled and the ABI supported by 11792 GCC has been brought into sync with that of MetroWerks CodeWarrior 11793 (the ABI previously supported was that of some Unix systems, which 11794 NetWare never tried to support). 11795 11796 Obsolete Systems 11797 11798 Support for a number of older systems has been declared obsolete in GCC 11799 4.0. Unless there is activity to revive them, the next release of GCC 11800 will have their sources permanently removed. 11801 11802 All GCC ports for the following processor architectures have been 11803 declared obsolete: 11804 * Intel i860 11805 * Ubicom IP2022 11806 * National Semiconductor NS32K (ns32k) 11807 * Texas Instruments TMS320C[34]x 11808 11809 Also, those for some individual systems have been obsoleted: 11810 * SPARC family 11811 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 11812 sparc86x-*-elf) 11813 + OpenBSD 32-bit (sparc-*-openbsd*) 11814 11815 Documentation improvements 11816 11817 Other significant improvements 11818 11819 * Location lists are now generated by default when compiling with 11820 debug info and optimization. Location lists provide more accurate 11821 debug info about locations of variables and they allow debugging 11822 code compiled with -fomit-frame-pointer. 11823 * The -fvisibility option has been added which allows the default ELF 11824 visibility of all symbols to be set per compilation and the new 11825 #pragma GCC visibility preprocessor command allows the setting of 11826 default ELF visibility for a region of code. Using 11827 -fvisibility=hidden especially in combination with the new 11828 -fvisibility-inlines-hidden can yield substantial improvements in 11829 output binary quality including avoiding PLT indirection overheads, 11830 reduction of the exported symbol count by up to 60% (with resultant 11831 improvements to link and load times), better scope for the 11832 optimizer to improve code and up to a 20% reduction in binary size. 11833 Using these options correctly yields a binary with a similar symbol 11834 count to a Windows DLL. 11835 Perhaps more importantly, this new feature finally allows (with 11836 careful planning) complete avoidance of symbol clashes when 11837 manually loading shared objects with RTLD_GLOBAL, thus finally 11838 solving problems many projects such as python were forced to use 11839 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 11840 can find more information about using these options at 11841 [11]https://gcc.gnu.org/wiki/Visibility. 11842 __________________________________________________________________ 11843 11844 GCC 4.0.1 11845 11846 This is the [12]list of problem reports (PRs) from GCC's bug tracking 11847 system that are known to be fixed in the 4.0.1 release. This list might 11848 not be complete (that is, it is possible that some PRs that have been 11849 fixed are not listed here). 11850 11851 GCC 4.0.2 11852 11853 This is the [13]list of problem reports (PRs) from GCC's bug tracking 11854 system that are known to be fixed in the 4.0.2 release. This list might 11855 not be complete (that is, it is possible that some PRs that have been 11856 fixed are not listed here). 11857 11858 Unfortunately, due to a release engineering failure, this release has a 11859 regression on Solaris that will affect some C++ programs. We suggest 11860 that Solaris users apply a [14]patch that corrects the problem. Users 11861 who do not wish to apply the patch should explicitly link C++ programs 11862 with the -pthreads option, even if they do not use threads. This 11863 problem has been corrected in the current 4.0 branch sources and will 11864 not be present in GCC 4.0.3. 11865 11866 GCC 4.0.3 11867 11868 Starting with this release, the function getcontext is recognized by 11869 the compiler as having the same semantics as the setjmp function. In 11870 particular, the compiler will ensure that all registers are dead before 11871 calling such a function and will emit a warning about the variables 11872 that may be clobbered after the second return from the function. 11873 11874 GCC 4.0.4 11875 11876 This is the [15]list of problem reports (PRs) from GCC's bug tracking 11877 system that are known to be fixed in the 4.0.4 release. This list might 11878 not be complete (that is, it is possible that some PRs that have been 11879 fixed are not listed here). 11880 11881 The 4.0.4 release is provided for those that require a high degree of 11882 binary compatibility with previous 4.0.x releases. For most users, the 11883 GCC team recommends that version 4.1.1 or later be used instead." 11884 11885 11886 For questions related to the use of GCC, please consult these web 11887 pages and the [16]GCC manuals. If that fails, the 11888 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11889 web pages and the development of GCC are welcome on our developer 11890 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 11891 archives. 11892 11893 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 11894 distribution of this entire article is permitted in any medium, 11895 provided this notice is preserved. 11896 11897 These pages are [21]maintained by the GCC team. Last modified 11898 2019-11-28[22]. 11899 11900 References 11901 11902 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 11903 2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 11904 3. http://gcc.gnu.org/projects/tree-ssa/ 11905 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html 11906 5. http://gcc.gnu.org/news/sms.html 11907 6. https://www.akkadia.org/drepper/dsohowto.pdf 11908 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility 11909 8. https://itanium-cxx-abi.github.io/cxx-abi/ 11910 9. http://gcc.gnu.org/fortran/ 11911 10. https://gcc.gnu.org/install/ 11912 11. https://gcc.gnu.org/wiki/Visibility 11913 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 11914 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 11915 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 11916 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 11917 16. https://gcc.gnu.org/onlinedocs/ 11918 17. mailto:gcc-help (a] gcc.gnu.org 11919 18. mailto:gcc (a] gcc.gnu.org 11920 19. https://gcc.gnu.org/lists.html 11921 20. https://www.fsf.org/ 11922 21. https://gcc.gnu.org/about.html 11923 22. http://validator.w3.org/check/referer 11924 ====================================================================== 11925 http://gcc.gnu.org/gcc-3.4/index.html 11926 GCC 3.4 Release Series 11927 11928 (This release series is no longer supported.) 11929 11930 May 26, 2006 11931 11932 The [1]GNU project and the GCC developers are pleased to announce the 11933 release of GCC 3.4.6. 11934 11935 This release is a bug-fix release, containing fixes for regressions in 11936 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 11937 3.4.x series. 11938 11939 The GCC 3.4 release series includes numerous [2]new features, 11940 improvements, bug fixes, and other changes, thanks to an [3]amazing 11941 group of volunteers. 11942 11943 Release History 11944 11945 GCC 3.4.6 11946 March 6, 2006 ([4]changes) 11947 11948 GCC 3.4.5 11949 November 30, 2005 ([5]changes) 11950 11951 GCC 3.4.4 11952 May 18, 2005 ([6]changes) 11953 11954 GCC 3.4.3 11955 November 4, 2004 ([7]changes) 11956 11957 GCC 3.4.2 11958 September 6, 2004 ([8]changes) 11959 11960 GCC 3.4.1 11961 July 1, 2004 ([9]changes) 11962 11963 GCC 3.4.0 11964 April 18, 2004 ([10]changes) 11965 11966 References and Acknowledgements 11967 11968 GCC used to stand for the GNU C Compiler, but since the compiler 11969 supports several other languages aside from C, it now stands for the 11970 GNU Compiler Collection. 11971 11972 A list of [11]successful builds is updated as new information becomes 11973 available. 11974 11975 The GCC developers would like to thank the numerous people that have 11976 contributed new features, improvements, bug fixes, and other changes as 11977 well as test results to GCC. This [12]amazing group of volunteers is 11978 what makes GCC successful. 11979 11980 For additional information about GCC please refer to the [13]GCC 11981 project web site or contact the [14]GCC development mailing list. 11982 11983 To obtain GCC please use [15]our mirror sites, or [16]our version 11984 control system. 11985 11986 11987 For questions related to the use of GCC, please consult these web 11988 pages and the [17]GCC manuals. If that fails, the 11989 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11990 web pages and the development of GCC are welcome on our developer 11991 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 11992 archives. 11993 11994 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 11995 distribution of this entire article is permitted in any medium, 11996 provided this notice is preserved. 11997 11998 These pages are [22]maintained by the GCC team. Last modified 11999 2020-01-14[23]. 12000 12001 References 12002 12003 1. http://www.gnu.org/ 12004 2. http://gcc.gnu.org/gcc-3.4/changes.html 12005 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 12006 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 12007 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 12008 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 12009 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 12010 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 12011 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 12012 10. http://gcc.gnu.org/gcc-3.4/changes.html 12013 11. http://gcc.gnu.org/gcc-3.4/buildstat.html 12014 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 12015 13. http://gcc.gnu.org/index.html 12016 14. mailto:gcc (a] gcc.gnu.org 12017 15. http://gcc.gnu.org/mirrors.html 12018 16. http://gcc.gnu.org/git.html 12019 17. https://gcc.gnu.org/onlinedocs/ 12020 18. mailto:gcc-help (a] gcc.gnu.org 12021 19. mailto:gcc (a] gcc.gnu.org 12022 20. https://gcc.gnu.org/lists.html 12023 21. https://www.fsf.org/ 12024 22. https://gcc.gnu.org/about.html 12025 23. http://validator.w3.org/check/referer 12026 ====================================================================== 12027 http://gcc.gnu.org/gcc-3.4/changes.html 12028 GCC 3.4 Release Series 12029 Changes, New Features, and Fixes 12030 12031 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 12032 is now closed. 12033 12034 GCC 3.4 has [2]many improvements in the C++ front end. Before reporting 12035 a bug, please make sure it's really GCC, and not your code, that is 12036 broken. 12037 12038 Caveats 12039 12040 * GNU Make is now required to build GCC. 12041 * With -nostdinc the preprocessor used to ignore both standard 12042 include paths and include paths contained in environment variables. 12043 It was neither documented nor intended that environment variable 12044 paths be ignored, so this has been corrected. 12045 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 12046 -fvolatile-static. It is unlikely that they worked correctly in any 12047 3.x release. 12048 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 12049 * Support for all the systems [3]obsoleted in GCC 3.3 has been 12050 removed from GCC 3.4. See below for a [4]list of systems which are 12051 obsoleted in this release. 12052 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 12053 compilers will not work. 12054 * The implementation of the [5]MIPS ABIs has changed. As a result, 12055 the code generated for certain MIPS targets will not be binary 12056 compatible with earlier releases. 12057 * In previous releases, the MIPS port had a fake "hilo" register with 12058 the user-visible name accum. This register has been removed. 12059 * The implementation of the [6]SPARC ABIs has changed. As a result, 12060 the code generated will not be binary compatible with earlier 12061 releases in certain cases. 12062 * The configure option --enable-threads=pthreads has been removed; 12063 use --enable-threads=posix instead, which should have the same 12064 effect. 12065 * Code size estimates used by inlining heuristics for C, Objective-C, 12066 C++ and Java have been redesigned significantly. As a result the 12067 parameters of -finline-insns, --param max-inline-insns-single and 12068 --param max-inline-insns-auto need to be reconsidered. 12069 * --param max-inline-slope and --param min-inline-insns have been 12070 removed; they are not needed for the new bottom-up inlining 12071 heuristics. 12072 * The new unit-at-a-time compilation scheme has several compatibility 12073 issues: 12074 + The order in which functions, variables, and top-level asm 12075 statements are emitted may have changed. Code relying on some 12076 particular ordering needs to be updated. The majority of such 12077 top-level asm statements can be replaced by section 12078 attributes. 12079 + Unreferenced static variables and functions are removed. This 12080 may result in undefined references when an asm statement 12081 refers to the variable/function directly. In that case either 12082 the variable/function shall be listed in asm statement operand 12083 or in the case of top-level asm statements the attribute used 12084 shall be used to force function/variable to be always output 12085 and considered as a possibly used by unknown code. 12086 For variables the attribute is accepted only by GCC 3.4 and 12087 newer, while for earlier versions it is sufficient to use 12088 unused to silence warnings about the variables not being 12089 referenced. To keep code portable across different GCC 12090 versions, you can use appropriate preprocessor conditionals. 12091 + Static functions now can use non-standard passing conventions 12092 that may break asm statements calling functions directly. 12093 Again the attribute used shall be used to prevent this 12094 behavior. 12095 As a temporary workaround, -fno-unit-at-a-time can be used, but 12096 this scheme may not be supported by future releases of GCC. 12097 * GCC 3.4 automatically places zero-initialized variables in the .bss 12098 section on some operating systems. Versions of GNU Emacs up to (and 12099 including) 21.3 will not work correctly when using this 12100 optimization; you can use -fno-zero-initialized-in-bss to disable 12101 it. 12102 * If GCC 3.4 is configured with --enable-threads=posix (the default 12103 on most targets that support pthreads) then _REENTRANT will be 12104 defined unconditionally by some libstdc++ headers. C++ code which 12105 relies on that macro to detect whether multi-threaded code is being 12106 compiled might change in meaning, possibly resulting in linker 12107 errors for single-threaded programs. Affected users of [7]Boost 12108 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 12109 See Bugzilla for [8]more information. 12110 12111 General Optimizer Improvements 12112 12113 * Usability of the profile feedback and coverage testing has been 12114 improved. 12115 + Performance of profiled programs has been improved by faster 12116 profile merging code. 12117 + Better use of the profile feedback for optimization (loop 12118 unrolling and loop peeling). 12119 + File locking support allowing fork() calls and parallel runs 12120 of profiled programs. 12121 + Coverage file format has been redesigned. 12122 + gcov coverage tool has been improved. 12123 + make profiledbootstrap available to build a faster compiler. 12124 Experiments made on i386 hardware showed an 11% speedup on -O0 12125 and a 7.5% speedup on -O2 compilation of a [9]large C++ 12126 testcase. 12127 + New value profiling pass enabled via -fprofile-values 12128 + New value profile transformations pass enabled via -fvpt aims 12129 to optimize some code sequences by exploiting knowledge about 12130 value ranges or other properties of the operands. At the 12131 moment a conversion of expensive divisions into cheaper 12132 operations has been implemented. 12133 + New -fprofile-generate and -fprofile-use command-line options 12134 to simplify the use of profile feedback. 12135 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 12136 Java which is enabled via -funit-at-a-time (and implied by -O2). In 12137 this scheme a whole file is parsed first and optimized later. The 12138 following basic inter-procedural optimizations are implemented: 12139 + Removal of unreachable functions and variables 12140 + Discovery of local functions (functions with static linkage 12141 whose address is never taken) 12142 + On i386, these local functions use register parameter passing 12143 conventions. 12144 + Reordering of functions in topological order of the call graph 12145 to enable better propagation of optimizing hints (such as the 12146 stack alignments needed by functions) in the back end. 12147 + Call graph based out-of-order inlining heuristics which allows 12148 to limit overall compilation unit growth (--param 12149 inline-unit-growth). 12150 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 12151 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 12152 CPU). 12153 * More realistic code size estimates used by inlining for C, 12154 Objective-C, C++ and Java. The growth of large functions can now be 12155 limited via --param large-function-insns and --param 12156 large-function-growth. 12157 * A new cfg-level loop optimizer pass replaces the old loop unrolling 12158 pass and adds two other loop transformations -- loop peeling and 12159 loop unswitching -- and also uses the profile feedback to limit 12160 code growth. (The three optimizations are enabled by 12161 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 12162 respectively). 12163 The old loop unroller still can be enabled by -fold-unroll-loops 12164 and may produce better code in some cases, especially when the 12165 webizer optimization pass is not run. 12166 * A new web construction pass enabled via -fweb (and implied by -O3) 12167 improves the quality of register allocation, CSE, first scheduling 12168 pass and some other optimization passes by avoiding re-use of 12169 pseudo registers with non-overlapping live ranges. The pass almost 12170 always improves code quality but does make debugging difficult and 12171 thus is not enabled by default by -O2 12172 The pass is especially effective as cleanup after code duplication 12173 passes, such as the loop unroller or the tracer. 12174 * Experimental implementations of superblock or trace scheduling in 12175 the second scheduling pass can be enabled via 12176 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 12177 12178 New Languages and Language specific improvements 12179 12180 Ada 12181 12182 * The Ada front end has been updated to include numerous bug fixes 12183 and enhancements. These include: 12184 + Improved project file support 12185 + Additional set of warnings about potential wrong code 12186 + Improved error messages 12187 + Improved code generation 12188 + Improved cross reference information 12189 + Improved inlining 12190 + Better run-time check elimination 12191 + Better error recovery 12192 + More efficient implementation of unbounded strings 12193 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 12194 ... 12195 + New GNAT.xxxx packages (e.g. GNAT.Strings, 12196 GNAT.Exception_Action) 12197 + New pragmas 12198 + New -gnatS switch replacing gnatpsta 12199 + Implementation of new Ada features (in particular limited 12200 with, limited aggregates) 12201 12202 C/Objective-C/C++ 12203 12204 * Precompiled headers are now supported. Precompiled headers can 12205 dramatically speed up compilation of some projects. There are some 12206 known defects in the current precompiled header implementation that 12207 will result in compiler crashes in relatively rare situations. 12208 Therefore, precompiled headers should be considered a "technology 12209 preview" in this release. Read the manual for details about how to 12210 use precompiled headers. 12211 * File handling in the preprocessor has been rewritten. GCC no longer 12212 gets confused by symlinks and hardlinks, and now has a correct 12213 implementation of #import and #pragma once. These two directives 12214 have therefore been un-deprecated. 12215 * The undocumented extension that allowed C programs to have a label 12216 at the end of a compound statement, which has been deprecated since 12217 GCC 3.0, has been removed. 12218 * The cast-as-lvalue extension has been removed for C++ and 12219 deprecated for C and Objective-C. In particular, code like this: 12220 int i; 12221 (char) i = 5; 12222 12223 or this: 12224 char *p; 12225 ((int *) p)++; 12226 12227 is no longer accepted for C++ and will not be accepted for C and 12228 Objective-C in a future version. 12229 * The conditional-expression-as-lvalue extension has been deprecated 12230 for C and Objective-C. In particular, code like this: 12231 int a, b, c; 12232 (a ? b : c) = 2; 12233 12234 will not be accepted for C and Objective-C in a future version. 12235 * The compound-expression-as-lvalue extension has been deprecated for 12236 C and Objective-C. In particular, code like this: 12237 int a, b; 12238 (a, b) = 2; 12239 12240 will not be accepted for C and Objective-C in a future version. A 12241 possible non-intrusive workaround is the following: 12242 (*(a, &b)) = 2; 12243 12244 * Several [10]built-in functions such as __builtin_popcount for 12245 counting bits, finding the highest and lowest bit in a word, and 12246 parity have been added. 12247 * The -fwritable-strings option has been deprecated and will be 12248 removed. 12249 * Many C math library functions are now recognized as built-ins and 12250 optimized. 12251 * The C, C++, and Objective-C compilers can now handle source files 12252 written in any character encoding supported by the host C library. 12253 The default input character set is taken from the current locale, 12254 and may be overridden with the -finput-charset command line option. 12255 In the future we will add support for inline encoding markers. 12256 12257 C++ 12258 12259 * G++ is now much closer to full conformance to the ISO/ANSI C++ 12260 standard. This means, among other things, that a lot of invalid 12261 constructs which used to be accepted in previous versions will now 12262 be rejected. It is very likely that existing C++ code will need to 12263 be fixed. This document lists some of the most common issues. 12264 * A hand-written recursive-descent C++ parser has replaced the 12265 YACC-derived C++ parser from previous GCC releases. The new parser 12266 contains much improved infrastructure needed for better parsing of 12267 C++ source codes, handling of extensions, and clean separation 12268 (where possible) between proper semantics analysis and parsing. The 12269 new parser fixes many bugs that were found in the old parser. 12270 * You must now use the typename and template keywords to disambiguate 12271 dependent names, as required by the C++ standard. 12272 struct K { 12273 typedef int mytype_t; 12274 }; 12275 12276 template <class T1> struct A { 12277 template <class T2> struct B { 12278 void callme(void); 12279 }; 12280 12281 template <int N> void bar(void) 12282 { 12283 // Use 'typename' to tell the parser that T1::mytype_t names 12284 // a type. This is needed because the name is dependent (in 12285 // this case, on template parameter T1). 12286 typename T1::mytype_t x; 12287 x = 0; 12288 } 12289 }; 12290 12291 template <class T> void template_func(void) 12292 { 12293 // Use 'template' to prefix member templates within 12294 // dependent types (a has type A<T>, which depends on 12295 // the template parameter T). 12296 A<T> a; 12297 a.template bar<0>(); 12298 12299 // Use 'template' to tell the parser that B is a nested 12300 // template class (dependent on template parameter T), and 12301 // 'typename' because the whole A<T>::B<int> is 12302 // the name of a type (again, dependent). 12303 typename A<T>::template B<int> b; 12304 b.callme(); 12305 } 12306 12307 void non_template_func(void) 12308 { 12309 // Outside of any template class or function, no names can be 12310 // dependent, so the use of the keyword 'typename' and 'template' 12311 // is not needed (and actually forbidden). 12312 A<K> a; 12313 a.bar<0>(); 12314 A<K>::B<float> b; 12315 b.callme(); 12316 } 12317 * In a template definition, unqualified names will no longer find 12318 members of a dependent base (as specified by [temp.dep]/3 in the 12319 C++ standard). For example, 12320 template <typename T> struct B { 12321 int m; 12322 int n; 12323 int f (); 12324 int g (); 12325 }; 12326 int n; 12327 int g (); 12328 template <typename T> struct C : B<T> { 12329 void h () 12330 { 12331 m = 0; // error 12332 f (); // error 12333 n = 0; // ::n is modified 12334 g (); // ::g is called 12335 } 12336 }; 12337 You must make the names dependent, e.g. by prefixing them with 12338 this->. Here is the corrected definition of C<T>::h, 12339 template <typename T> void C<T>::h () 12340 { 12341 this->m = 0; 12342 this->f (); 12343 this->n = 0 12344 this->g (); 12345 } 12346 As an alternative solution (unfortunately not backwards compatible 12347 with GCC 3.3), you may use using declarations instead of this->: 12348 template <typename T> struct C : B<T> { 12349 using B<T>::m; 12350 using B<T>::f; 12351 using B<T>::n; 12352 using B<T>::g; 12353 void h () 12354 { 12355 m = 0; 12356 f (); 12357 n = 0; 12358 g (); 12359 } 12360 }; 12361 * In templates, all non-dependent names are now looked up and bound 12362 at definition time (while parsing the code), instead of later when 12363 the template is instantiated. For instance: 12364 void foo(int); 12365 12366 template <int> struct A { 12367 static void bar(void){ 12368 foo('a'); 12369 } 12370 }; 12371 12372 void foo(char); 12373 12374 int main() 12375 { 12376 A<0>::bar(); // Calls foo(int), used to call foo(char). 12377 } 12378 12379 * In an explicit instantiation of a class template, you must use 12380 class or struct before the template-id: 12381 template <int N> 12382 class A {}; 12383 12384 template A<0>; // error, not accepted anymore 12385 template class A<0>; // OK 12386 * The "named return value" and "implicit typename" extensions have 12387 been removed. 12388 * Default arguments in function types have been deprecated and will 12389 be removed. 12390 * ARM-style name-injection of friend declarations has been deprecated 12391 and will be removed. For example: struct S { friend void f(); }; 12392 void g() { f(); } will not be accepted by future versions of G++; 12393 instead a declaration of "f" will need to be present outside of the 12394 scope of "S". 12395 * Covariant returns are implemented for all but varadic functions 12396 that require an adjustment. 12397 * When -pedantic is used, G++ now issues errors about spurious 12398 semicolons. For example, 12399 namespace N {}; // Invalid semicolon. 12400 void f() {}; // Invalid semicolon. 12401 * G++ no longer accepts attributes for a declarator after the 12402 initializer associated with that declarator. For example, 12403 X x(1) __attribute__((...)); 12404 is no longer accepted. Instead, use: 12405 X x __attribute__((...)) (1); 12406 * Inside the scope of a template class, the name of the class itself 12407 can be treated as either a class or a template. So GCC used to 12408 accept the class name as argument of type template, and template 12409 template parameter. However this is not C++ standard compliant. Now 12410 the name is not treated as a valid template template argument 12411 unless you qualify the name by its scope. For example, the code 12412 below no longer compiles. 12413 template <template <class> class TT> class X {}; 12414 template <class T> class Y { 12415 X<Y> x; // Invalid, Y is always a type template parameter. 12416 }; 12417 The valid code for the above example is 12418 X< ::Y> x; // Valid. 12419 (Notice the space between < and : to prevent GCC to interpret this 12420 as a digraph for [.) 12421 * Friend declarations that refer to template specializations are 12422 rejected if the template has not already been declared. For 12423 example, 12424 template <typename T> 12425 class C { 12426 friend void f<> (C&); 12427 }; 12428 is rejected. You must first declare f as a template, 12429 template <typename T> 12430 void f(T); 12431 * In case of friend declarations, every name used in the friend 12432 declaration must be accessible at the point of that declaration. 12433 Previous versions of G++ used to be less strict about this and 12434 allowed friend declarations for private class members, for example. 12435 See the ISO C++ Standard Committee's [11]defect report #209 for 12436 details. 12437 * Declaration of member functions of class templates as friends are 12438 supported. For example, 12439 template <typename T> struct A { 12440 void f(); 12441 }; 12442 class C { 12443 template <typename T> friend void A<T>::f(); 12444 }; 12445 * You must use template <> to introduce template specializations, as 12446 required by the standard. For example, 12447 template <typename T> 12448 struct S; 12449 12450 struct S<int> { }; 12451 is rejected. You must write, 12452 template <> struct S<int> {}; 12453 * G++ used to accept code like this, 12454 struct S { 12455 int h(); 12456 void f(int i = g()); 12457 int g(int i = h()); 12458 }; 12459 This behavior is not mandated by the standard. Now G++ issues an 12460 error about this code. To avoid the error, you must move the 12461 declaration of g before the declaration of f. The default arguments 12462 for g must be visible at the point where it is called. 12463 * The C++ ABI Section 3.3.3 specifications for the array construction 12464 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 12465 NULL when the allocator argument returns NULL. These changes are 12466 incorporated into the libstdc++ runtime library. 12467 * Using a name introduced by a typedef in a friend declaration or in 12468 an explicit instantiation is now rejected, as specified by the ISO 12469 C++ standard. 12470 class A; 12471 typedef A B; 12472 class C { 12473 friend class B; // error, no typedef name here 12474 friend B; // error, friend always needs class/struct/enum 12475 friend class A; // OK 12476 }; 12477 12478 template <int> class Q {}; 12479 typedef Q<0> R; 12480 template class R; // error, no typedef name here 12481 template class Q<0>; // OK 12482 * When allocating an array with a new expression, GCC used to allow 12483 parentheses around the type name. This is actually ill-formed and 12484 it is now rejected: 12485 int* a = new (int)[10]; // error, not accepted anymore 12486 int* a = new int[10]; // OK 12487 * When binding an rvalue of class type to a reference, the copy 12488 constructor of the class must be accessible. For instance, consider 12489 the following code: 12490 class A 12491 { 12492 public: 12493 A(); 12494 12495 private: 12496 A(const A&); // private copy ctor 12497 }; 12498 12499 A makeA(void); 12500 void foo(const A&); 12501 12502 void bar(void) 12503 { 12504 foo(A()); // error, copy ctor is not accessible 12505 foo(makeA()); // error, copy ctor is not accessible 12506 12507 A a1; 12508 foo(a1); // OK, a1 is a lvalue 12509 } 12510 This might be surprising at first sight, especially since most 12511 popular compilers do not correctly implement this rule ([12]further 12512 details). 12513 * When forming a pointer to member or a pointer to member function, 12514 access checks for class visibility (public, protected, private) are 12515 now performed using the qualifying scope of the name itself. This 12516 is better explained with an example: 12517 class A 12518 { 12519 public: 12520 void pub_func(); 12521 protected: 12522 void prot_func(); 12523 private: 12524 void priv_func(); 12525 }; 12526 12527 class B : public A 12528 { 12529 public: 12530 void foo() 12531 { 12532 &A::pub_func; // OK, pub_func is accessible through A 12533 &A::prot_func; // error, cannot access prot_func through A 12534 &A::priv_func; // error, cannot access priv_func through A 12535 12536 &B::pub_func; // OK, pub_func is accessible through B 12537 &B::prot_func; // OK, can access prot_func through B (within B) 12538 &B::priv_func; // error, cannot access priv_func through B 12539 } 12540 }; 12541 12542 Runtime Library (libstdc++) 12543 12544 * Optimization work: 12545 + Streamlined streambuf, filebuf, separate synched with C 12546 Standard I/O streambuf. 12547 + All formatted I/O now uses cached locale information. 12548 + STL optimizations (memory/speed for list, red-black trees as 12549 used by sets and maps). 12550 + More use of GCC builtins. 12551 + String optimizations (avoid contention on 12552 increment/decrement-and-test of the reference count in the 12553 empty-string object, constructor from input_iterators 12554 speedup). 12555 * Static linkage size reductions. 12556 * Large File Support (files larger than 2 GB on 32-bit systems). 12557 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 12558 * Generic character traits. 12559 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 12560 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 12561 * The allocator class is now standard-conformant, and two additional 12562 extension allocators have been added, mt_alloc and 12563 bitmap_allocator. 12564 * PCH support: -include bits/stdc++.h (2x compile speedup). 12565 * Rewrote __cxa_demangle with support for C++ style allocators. 12566 * New debug modes for STL containers and iterators. 12567 * Testsuite rewrite: five times as many tests, plus increasingly 12568 sophisticated tests, including I/O, MT, multi-locale, wide and 12569 narrow characters. 12570 * Use current versions of GNU "autotools" for build/configuration. 12571 12572 Objective-C 12573 12574 * The Objective-C front end has been updated to include the numerous 12575 bug fixes and enhancements previously available only in Apple's 12576 version of GCC. These include: 12577 + Structured exception (@try... @catch... @finally, @throw) and 12578 synchronization (@synchronized) support. These are accessible 12579 via the -fobjc-exceptions switch; as of this writing, they may 12580 only be used in conjunction with -fnext-runtime on Mac OS X 12581 10.3 and later. See [13]Options Controlling Objective-C 12582 Dialect for more information. 12583 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 12584 may now be encoded as 'B'. In addition, the back-end/codegen 12585 dependencies have been removed. 12586 + An overhaul of message dispatch construction, ensuring that 12587 the various receiver types (and casts thereof) are handled 12588 properly, and that correct diagnostics are issued. 12589 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 12590 (-freplace-objc-classes) debugging modes, currently available 12591 on Mac OS X 10.3 and later. See [14]Options Controlling 12592 Objective-C Dialect for more information. 12593 + Access to optimized runtime entry points (-fno-nil-receivers ) 12594 on the assumption that message receivers are never nil. This 12595 is currently available on Mac OS X 10.3 and later. See 12596 [15]Options Controlling Objective-C Dialect for more 12597 information. 12598 12599 Java 12600 12601 * Compiling a .jar file will now cause non-.class entries to be 12602 automatically compiled as resources. 12603 * libgcj has been ported to Darwin. 12604 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 12605 to gcj. 12606 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 12607 code from shared libraries. 12608 * libgcj has been much more completely merged with [16]GNU Classpath. 12609 * Class loading is now much more correct; in particular the caller's 12610 class loader is now used when that is required. 12611 * [17]Eclipse 2.x will run out of the box using gij. 12612 * Parts of java.nio have been implemented. Direct and indirect 12613 buffers work, as do fundamental file and socket operations. 12614 * java.awt has been improved, though it is still not ready for 12615 general use. 12616 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 12617 method. 12618 * The MinGW port has matured. Enhancements include socket timeout 12619 support, thread interruption, improved Runtime.exec() handling and 12620 support for accented characters in filenames. 12621 12622 Fortran 12623 12624 * Fortran improvements are listed in the [18]Fortran documentation. 12625 12626 New Targets and Target Specific Improvements 12627 12628 Alpha 12629 12630 * Several [19]built-in functions have been added such as 12631 __builtin_alpha_zap to allow utilizing the more obscure 12632 instructions of the CPU. 12633 * Parameter passing of complex arguments has changed to match the 12634 ABI. This change is incompatible with previous GCC versions, but 12635 does fix compatibility with the Tru64 compiler and several corner 12636 cases where GCC was incompatible with itself. 12637 12638 ARM 12639 12640 * Nicolas Pitre has contributed his hand-coded floating-point support 12641 code for ARM. It is both significantly smaller and faster than the 12642 existing C-based implementation, even when building applications 12643 for Thumb. The arm-elf configuration has been converted to use the 12644 new code. 12645 * Support for the Intel's iWMMXt architecture, a second generation 12646 XScale processor, has been added. Enabled at run time with the 12647 -mcpu=iwmmxt command line switch. 12648 * A new ARM target has been added: arm-wince-pe. This is similar to 12649 the arm-pe target, but it defaults to using the APCS32 ABI. 12650 * The existing ARM pipeline description has been converted to the use 12651 the [20]DFA processor pipeline model. There is not much change in 12652 code performance, but the description is now [21]easier to 12653 understand. 12654 * Support for the Cirrus EP9312 Maverick floating point co-processor 12655 added. Enabled at run time with the -mcpu=ep9312 command line 12656 switch. Note however that the multilibs to support this chip are 12657 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 12658 enable their production you will have to uncomment the entries in 12659 that file. 12660 12661 H8/300 12662 12663 * Support for long long has been added. 12664 * Support for saveall attribute has been added. 12665 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 12666 for H8/300H and H8S, which is much faster than the previous 12667 implementation. 12668 * A lot of small performance improvements. 12669 12670 IA-32/AMD64 (x86-64) 12671 12672 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 12673 -march=k8 and -mcpu=k8. 12674 * Scalar SSE code generation carefully avoids reformatting penalties, 12675 hidden dependencies and minimizes the number of uops generated on 12676 both Intel and AMD CPUs. 12677 * Vector MMX and SSE operands are now passed in registers to improve 12678 performance and match the argument passing convention used by the 12679 Intel C++ Compiler. As a result it is not possible to call 12680 functions accepting vector arguments compiled by older GCC version. 12681 * Conditional jump elimination is now more aggressive on modern CPUs. 12682 * The Athlon ports has been converted to use the DFA processor 12683 pipeline description. 12684 * Optimization of indirect tail calls is now possible in a similar 12685 fashion as direct sibcall optimization. 12686 * Further small performance improvements. 12687 * -m128bit-long-double is now less buggy. 12688 * __float128 support in 64-bit compilation. 12689 * Support for data structures exceeding 2GB in 64-bit mode. 12690 * -mcpu has been renamed to -mtune. 12691 12692 IA-64 12693 12694 * Tuning code for the Itanium 2 processor has been added. The 12695 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 12696 enabled by default now. To generate code tuned for Itanium 1 the 12697 option -mtune=itanium1 should be used. 12698 * [22]DFA processor pipeline descriptions for the IA-64 processors 12699 have been added. This resulted in about 3% improvement on the 12700 SPECInt2000 benchmark for Itanium 2. 12701 * Instruction bundling for the IA-64 processors has been rewritten 12702 using the DFA pipeline hazard recognizer. It resulted in about 60% 12703 compiler speedup on the SPECInt2000 C programs. 12704 12705 M32R 12706 12707 * Support for the M32R/2 processor has been added by Renesas. 12708 * Support for an M32R GNU/Linux target and PIC code generation has 12709 been added by Renesas. 12710 12711 M68000 12712 12713 * Bernardo Innocenti (Develer S.r.l.) has contributed the 12714 m68k-uclinux target, based on former work done by Paul Dale 12715 (SnapGear Inc.). Code generation for the ColdFire processors family 12716 has been enhanced and extended to support the MCF 53xx and MCF 54xx 12717 cores, integrating former work done by Peter Barada (Motorola). 12718 12719 MIPS 12720 12721 Processor-specific changes 12722 12723 * Support for the RM7000 and RM9000 processors has been added. It can 12724 be selected using the -march compiler option and should work with 12725 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 12726 * Support for revision 2 of the MIPS32 ISA has been added. It can be 12727 selected with the command-line option -march=mips32r2. 12728 * There is a new option, -mfix-sb1, to work around certain SB-1 12729 errata. 12730 12731 Configuration 12732 12733 * It is possible to customize GCC using the following configure-time 12734 options: 12735 + --with-arch, which specifies the default value of the -march 12736 option. 12737 + --with-tune, which specifies the default value of the -mtune 12738 option. 12739 + --with-abi, which specifies the default ABI. 12740 + --with-float=soft, which tells GCC to use software floating 12741 point by default. 12742 + --with-float=hard, which tells GCC to use hardware floating 12743 point by default. 12744 * A 64-bit GNU/Linux port has been added. The associated 12745 configurations are mips64-linux-gnu and mips64el-linux-gnu. 12746 * The 32-bit GNU/Linux port now supports Java. 12747 * The IRIX 6 configuration now supports the o32 ABI and will build 12748 o32 multilibs by default. This support is compatible with both 12749 binutils and the SGI tools, but note that several features, 12750 including debugging information and DWARF2 exception handling, are 12751 only available when using the GNU assembler. Use of the GNU 12752 assembler and linker (version 2.15 or above) is strongly 12753 recommended. 12754 * The IRIX 6 configuration now supports 128-bit long doubles. 12755 * There are two new RTEMS-specific configurations, mips-rtems and 12756 mipsel-rtems. 12757 * There are two new *-elf configurations, mipsisa32r2-elf and 12758 mipsisa32r2el-elf. 12759 12760 General 12761 12762 * Several [23]ABI bugs have been fixed. Unfortunately, these changes 12763 will break binary compatibility with earlier releases. 12764 * GCC can now use explicit relocation operators when generating 12765 -mabicalls code. This behavior is controlled by -mexplicit-relocs 12766 and can have several performance benefits. For example: 12767 + It allows for more optimization of GOT accesses, including 12768 better scheduling and redundancy elimination. 12769 + It allows sibling calls to be implemented as jumps. 12770 + n32 and n64 leaf functions can use a call-clobbered global 12771 pointer instead of $28. 12772 + The code to set up $gp can be removed from functions that 12773 don't need it. 12774 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 12775 option is equivalent to the assembler's -xgot option and should be 12776 used instead of -Wa,-xgot. 12777 * Frame pointer elimination is now supported when generating 64-bit 12778 MIPS16 code. 12779 * Inline block moves have been optimized to take more account of 12780 alignment information. 12781 * Many internal changes have been made to the MIPS port, mostly aimed 12782 at reducing the reliance on assembler macros. 12783 12784 PowerPC 12785 12786 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 12787 [24]ABI incompatibilities regarding the way parameters are passed 12788 during functions calls. These changes may result in incompatibility 12789 between code compiled with GCC 3.3 and GCC 3.4. 12790 12791 PowerPC Darwin 12792 12793 * Support for shared/dylib gcc libraries has been added. It is 12794 enabled by default on powerpc-apple-darwin7.0.0 and up. 12795 * Libgcj is enabled by default. On systems older than 12796 powerpc-apple-darwin7.0.0 you need to install dlcompat. 12797 * 128-bit IBM extended precision format support added for long 12798 double. 12799 12800 PowerPC64 GNU/Linux 12801 12802 * By default, PowerPC64 GNU/Linux now uses natural alignment of 12803 structure elements. The old four byte alignment for double, with 12804 special rules for a struct starting with a double, can be chosen 12805 with -malign-power. This change may result in incompatibility 12806 between code compiled with GCC 3.3 and GCC 3.4. 12807 * -mabi=altivec is now the default rather than -mabi=no-altivec. 12808 * 128-bit IBM extended precision format support added for long 12809 double. 12810 12811 S/390 and zSeries 12812 12813 * New command-line options allow to specify the intended execution 12814 environment for generated code: 12815 + -mesa/-mzarch allows to specify whether to generate code 12816 running in ESA/390 mode or in z/Architecture mode (this is 12817 applicable to 31-bit code only). 12818 + -march allows to specify a minimum processor architecture 12819 level (g5, g6, z900, or z990). 12820 + -mtune allows to specify which processor to tune for. 12821 * It is possible to customize GCC using the following configure-time 12822 options: 12823 + --with-mode, which specifies whether to default to assuming 12824 ESA/390 or z/Architecture mode. 12825 + --with-arch, which specifies the default value of the -march 12826 option. 12827 + --with-tune, which specifies the default value of the -mtune 12828 option. 12829 * Support for the z990 processor has been added, and can be selected 12830 using -march=z990 or -mtune=z990. This includes instruction 12831 scheduling tuned for the superscalar instruction pipeline of the 12832 z990 processor as well as support for all new instructions provided 12833 by the long-displacement facility. 12834 * Support to generate 31-bit code optimized for zSeries processors 12835 (running in ESA/390 or in z/Architecture mode) has been added. This 12836 can be selected using -march=z900 and -mzarch respectively. 12837 * Instruction scheduling for the z900 and z990 processors now uses 12838 the DFA pipeline hazard recognizer. 12839 * GCC no longer generates code to maintain a stack backchain, 12840 previously used to generate stack backtraces for debugging 12841 purposes. As replacement that does not incur runtime overhead, 12842 DWARF-2 call frame information is provided by GCC; this is 12843 supported by GDB 6.1. The old behavior can be restored using the 12844 -mbackchain option. 12845 * The stack frame size of functions may now exceed 2 GB in 64-bit 12846 code. 12847 * A port for the 64-bit IBM TPF operating system has been added; the 12848 configuration is s390x-ibm-tpf. This configuration is supported as 12849 cross-compilation target only. 12850 * Various changes to improve the generated code have been 12851 implemented, including: 12852 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 12853 instructions to significantly speed up many floating-point 12854 applications. 12855 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 12856 WITH BORROW instructions to speed up long long arithmetic. 12857 + GCC now uses the SEARCH STRING instruction to implement 12858 strlen(). 12859 + In many cases, function call overhead for 31-bit code has been 12860 reduced by placing the literal pool after the function code 12861 instead of after the function prolog. 12862 + Register 14 is no longer reserved in 64-bit code. 12863 + Handling of global register variables has been improved. 12864 12865 SPARC 12866 12867 * The option -mflat is deprecated. 12868 * Support for large (> 2GB) frames has been added to the 64-bit port. 12869 * Several [25]ABI bugs have been fixed. Unfortunately, these changes 12870 will break binary compatibility with earlier releases. 12871 * The default debugging format has been switched from STABS to 12872 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 12873 the default debugging format for 64-bit code on Solaris. 12874 12875 SuperH 12876 12877 * Support for the SH2E processor has been added. Enabled at run time 12878 with the -m2e command line switch, or at configure time by 12879 specifying sh2e as the machine part of the target triple. 12880 12881 V850 12882 12883 * Support for the Mitsubishi V850E1 processor has been added. This is 12884 a variant of the V850E processor with some additional debugging 12885 instructions. 12886 12887 Xtensa 12888 12889 * Several ABI bugs have been fixed. Unfortunately, these changes 12890 break binary compatibility with earlier releases. 12891 + For big-endian processors, the padding of aggregate return 12892 values larger than a word has changed. If the size of an 12893 aggregate return value is not a multiple of 32 bits, previous 12894 versions of GCC inserted padding in the most-significant bytes 12895 of the first return value register. Aggregates larger than a 12896 word are now padded in the least-significant bytes of the last 12897 return value register used. Aggregates smaller than a word are 12898 still padded in the most-significant bytes. The return value 12899 padding has not changed for little-endian processors. 12900 + Function arguments with 16-byte alignment are now properly 12901 aligned. 12902 + The implementation of the va_list type has changed. A va_list 12903 value created by va_start from a previous release cannot be 12904 used with va_arg from this release, or vice versa. 12905 * More processor configuration options for Xtensa processors are 12906 supported: 12907 + the ABS instruction is now optional; 12908 + the ADDX* and SUBX* instructions are now optional; 12909 + an experimental CONST16 instruction can be used to synthesize 12910 constants instead of loading them from constant pools. 12911 These and other Xtensa processor configuration options can no 12912 longer be enabled or disabled by command-line options; the 12913 processor configuration must be specified by the xtensa-config.h 12914 header file when building GCC. Additionally, the 12915 -mno-serialize-volatile option is no longer supported. 12916 12917 Obsolete Systems 12918 12919 Support for a number of older systems has been declared obsolete in GCC 12920 3.4. Unless there is activity to revive them, the next release of GCC 12921 will have their sources permanently removed. 12922 12923 All configurations of the following processor architectures have been 12924 declared obsolete: 12925 * Mitsubishi D30V, d30v-* 12926 * AT&T DSP1600 and DSP1610, dsp16xx-* 12927 * Intel 80960, i960 12928 12929 Also, some individual systems have been obsoleted: 12930 * ARM Family 12931 + Support for generating code for operation in APCS/26 mode 12932 (-mapcs-26). 12933 * IBM ESA/390 12934 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 12935 maintained and supported.) 12936 * Intel 386 family 12937 + MOSS, i?86-moss-msdos and i?86-*-moss* 12938 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 12939 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 12940 i?86-*-freebsd2* 12941 + GNU/Linux with a.out object format, i?86-linux*aout* 12942 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 12943 + Interix versions before Interix 3, i?86-*-interix 12944 + Mach microkernel, i?86-mach* 12945 + SCO UnixWare with UDK, i?86-*-udk* 12946 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 12947 + VSTa microkernel, i386-*-vsta 12948 * Motorola M68000 family 12949 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 12950 + NetBSD with a.out object format (before NetBSD 1.4), 12951 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 12952 + Generic System V r.4, m68k-*-sysv4* 12953 * VAX 12954 + Generic VAX, vax-*-* (This is generic VAX only; we have not 12955 obsoleted any VAX triples for specific operating systems.) 12956 12957 Documentation improvements 12958 12959 Other significant improvements 12960 12961 * The build system has undergone several significant cleanups. 12962 Subdirectories will only be configured if they are being built, and 12963 all subdirectory configures are run from the make command. The top 12964 level has been autoconfiscated. 12965 * Building GCC no longer writes to its source directory. This should 12966 help those wishing to share a read-only source directory over NFS 12967 or build from a CD. The exceptions to this feature are if you 12968 configure with either --enable-maintainer-mode or 12969 --enable-generated-files-in-srcdir. 12970 * The -W warning option has been renamed to -Wextra, which is more 12971 easily understood. The older spelling will be retained for 12972 backwards compatibility. 12973 * Substantial improvements in compile time have been made, 12974 particularly for non-optimizing compilations. 12975 __________________________________________________________________ 12976 12977 GCC 3.4.0 12978 12979 Bug Fixes 12980 12981 A vast number of bugs have been fixed in 3.4.0, too many to publish a 12982 complete list here. [26]Follow this link to query the Bugzilla database 12983 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 12984 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 12985 regressions. 12986 __________________________________________________________________ 12987 12988 GCC 3.4.1 12989 12990 Bug Fixes 12991 12992 This section lists the problem reports (PRs) from GCC's bug tracking 12993 system that are known to be fixed in the 3.4.1 release. This list might 12994 not be complete (that is, it is possible that some PRs that have been 12995 fixed are not listed here). 12996 12997 Bootstrap failures 12998 12999 * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 13000 emitted - PIC related 13001 * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 13002 * [29]14760 A bug in configure.in prevents using both 13003 --program-suffix and --program-prefix 13004 * [30]14671 [hppa64] bootstrap fails: ICE in 13005 save_call_clobbered_regs, in caller_save.c 13006 * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on 13007 Alpha 13008 * [32]15178 Solaris 9/x86 fails linking after stage 3 13009 13010 Multi-platform internal compiler errors (ICEs) 13011 13012 * [33]12753 (preprocessor) Memory corruption in preprocessor on bad 13013 input 13014 * [34]13985 ICE in gcc.c-torture/compile/930621-1.c 13015 * [35]14810 (c++) tree check failures with invalid code involving 13016 templates 13017 * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 13018 cp/parser.c 13019 * [37]15044 (c++) ICE on syntax error, template header 13020 * [38]15057 (c++) Compiling of conditional value throw constructs 13021 cause a segmentation violation 13022 * [39]15064 (c++) typeid of template parameter gives ICE 13023 * [40]15142 (c++) ICE when passing a string where a char* is expected 13024 in a throw statement 13025 * [41]15159 ICE in rtl_verify_flow_info_1 13026 * [42]15165 (c++) ICE in instantiate_template 13027 * [43]15193 Unary minus using pointer to V4SF vector causes 13028 -fforce-mem to exhaust all memory 13029 * [44]15209 (c++) Runs out of memory with packed structs 13030 * [45]15227 (c++) Trouble with invalid function definition 13031 * [46]15285 (c++) instantiate_type ICE when forming pointer to 13032 template function 13033 * [47]15299 (c++) ICE in resolve_overloaded_unification 13034 * [48]15329 (c++) ICE on constructor of member template 13035 * [49]15550 ICE in extract_insn, in recog.c 13036 * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c 13037 * [51]15640 (c++) ICE on invalid code in arg_assoc, in 13038 cp/name-lookup.c 13039 * [52]15666 [unit-at-a-time] Gcc abort on valid code 13040 * [53]15696 (c++) ICE with bad pointer-to-member code 13041 * [54]15701 (c++) ICE with friends and template template parameter 13042 * [55]15761 ICE in do_SUBST, in combine.c 13043 * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 13044 13045 Ada 13046 13047 * [57]14538 All RTEMS targets broken for gnat 13048 13049 C front end 13050 13051 * [58]12391 missing warning about assigning to an incomplete type 13052 * [59]14649 atan(1.0) should not be a constant expression 13053 * [60]15004 [unit-at-a-time] no warning for unused paramater in 13054 static function 13055 * [61]15749 --pedantic-errors behaves differently from --pedantic 13056 with C-compiler on GNU/Linux 13057 13058 C++ compiler and library 13059 13060 * [62]10646 non-const reference is incorrectly matched in a "const T" 13061 partial specialization 13062 * [63]12077 wcin.rdbuf()->in_avail() return value too high 13063 * [64]13598 enc_filebuf doesn't work 13064 * [65]14211 const_cast returns lvalue but should be rvalue 13065 * [66]14220 num_put::do_put() undesired float/double behavior 13066 * [67]14245 problem with user-defined allocators in std::basic_string 13067 * [68]14340 libstdc++ Debug mode: failure to convert iterator to 13068 const_iterator 13069 * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 13070 FILE* 13071 * [70]14668 no warning anymore for reevaluation of declaration 13072 * [71]14775 LFS (large file support) tests missing 13073 * [72]14821 Duplicate namespace alias declaration should not conflict 13074 * [73]14930 Friend declaration ignored 13075 * [74]14932 cannot use offsetof to get offsets of array elements in 13076 g++ 3.4.0 13077 * [75]14950 [non unit-at-a-time] always_inline does not mix with 13078 templates and -O0 13079 * [76]14962 g++ ignores #pragma redefine_extname 13080 * [77]14975 Segfault on low-level write error during imbue 13081 * [78]15002 Linewise stream input is unusably slow (std::string slow) 13082 * [79]15025 compiler accepts redeclaration of template as 13083 non-template 13084 * [80]15046 [arm] Math functions misdetected by cross configuration 13085 * [81]15069 a bit test on a variable of enum type is miscompiled 13086 * [82]15074 g++ -lsupc++ still links against libstdc++ 13087 * [83]15083 spurious "statement has no effect" warning 13088 * [84]15096 parse error with templates and pointer to const member 13089 * [85]15287 combination of operator[] and operator .* fails in 13090 templates 13091 * [86]15317 __attribute__ unused in first parameter of constructor 13092 gives error 13093 * [87]15337 sizeof on incomplete type diagnostic 13094 * [88]15361 bitset<>::_Find_next fails 13095 * [89]15412 _GLIBCXX_ symbols symbols defined and used in different 13096 namespaces 13097 * [90]15427 valid code results in incomplete type error 13098 * [91]15471 Incorrect member pointer offsets in anonymous 13099 structs/unions 13100 * [92]15503 nested template problem 13101 * [93]15507 compiler hangs while laying out union 13102 * [94]15542 operator & and template definitions 13103 * [95]15565 SLES9: leading + sign for unsigned int with showpos 13104 * [96]15625 friend defined inside a template fails to find static 13105 function 13106 * [97]15629 Function templates, overloads, and friend name injection 13107 * [98]15742 'noreturn' attribute ignored in method of template 13108 functions. 13109 * [99]15775 Allocator::pointer consistently ignored 13110 * [100]15821 Duplicate namespace alias within namespace rejected 13111 * [101]15862 'enum yn' fails (confict with undeclared builtin) 13112 * [102]15875 rejects pointer to member in template 13113 * [103]15877 valid code using templates and anonymous enums is 13114 rejected 13115 * [104]15947 Puzzling error message for wrong destructor declaration 13116 in template class 13117 * [105]16020 cannot copy __gnu_debug::bitset 13118 * [106]16154 input iterator concept too restrictive 13119 * [107]16174 deducing top-level consts 13120 13121 Java 13122 13123 * [108]14315 Java compiler is not parallel make safe 13124 13125 Fortran 13126 13127 * [109]15151 [g77] incorrect logical i/o in 64-bit mode 13128 13129 Objective-C 13130 13131 * [110]7993 private variables cannot be shadowed in subclasses 13132 13133 Optimization bugs 13134 13135 * [111]15228 useless copies of floating point operands 13136 * [112]15345 [non-unit-at-a-time] unreferenced nested inline 13137 functions not optimized away 13138 * [113]15945 Incorrect floating point optimization 13139 * [114]15526 ftrapv aborts on 0 * (-1) 13140 * [115]14690 Miscompiled POOMA tests 13141 * [116]15112 GCC generates code to write to unchanging memory 13142 13143 Preprocessor 13144 13145 * [117]15067 Minor glitch in the source of cpp 13146 13147 Main driver program bugs 13148 13149 * [118]1963 collect2 interprets -oldstyle_liblookup as -o 13150 ldstyle_liblookup 13151 13152 x86-specific (Intel/AMD) 13153 13154 * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 13155 section} 13156 13157 HPPA-specific 13158 13159 * [120]14782 GCC produces an unaligned data access at -O2 13160 * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 13161 * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c 13162 13163 IA64-specific 13164 13165 * [123]14610 __float80 constants incorrectly emitted 13166 * [124]14813 init_array sections are initialized in the wrong order 13167 * [125]14857 GCC segfault on duplicated asm statement 13168 * [126]15598 Gcc 3.4 ICE on valid code 13169 * [127]15653 Gcc 3.4 ICE on valid code 13170 13171 MIPS-specific 13172 13173 * [128]15189 wrong filling of delay slot with -march=mips1 -G0 13174 -mno-split-addresses -mno-explicit-relocs 13175 * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 13176 2.14.91 13177 * [130]16144 Bogus reference to __divdf3 when -O1 13178 * [131]16176 Miscompilation of unaligned data in MIPS backend 13179 13180 PowerPC-specific 13181 13182 * [132]11591 ICE in gcc.dg/altivec-5.c 13183 * [133]12028 powerpc-eabispe produces bad sCOND operation 13184 * [134]14478 rs6000 geu/ltu patterns generate incorrect code 13185 * [135]14567 long double and va_arg complex args 13186 * [136]14715 Altivec stack layout may overlap gpr save with stack 13187 temps 13188 * [137]14902 (libstdc++) Stream checking functions fail when -pthread 13189 option is used. 13190 * [138]14924 Compiler ICE on valid code 13191 * [139]14960 -maltivec affects vector return with -mabi=no-altivec 13192 * [140]15106 vector varargs failure passing from altivec to 13193 non-altivec code for -m32 13194 * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 13195 half-word operation 13196 * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 13197 and stvx 13198 * [143]15662 Segmentation fault when an exception is thrown - even if 13199 try and catch are specified 13200 13201 s390-specific 13202 13203 * [144]15054 Bad code due to overlapping stack temporaries 13204 13205 SPARC-specific 13206 13207 * [145]15783 ICE with union assignment in 64-bit mode 13208 * [146]15626 GCC 3.4 emits "ld: warning: relocation error: 13209 R_SPARC_UA32" 13210 13211 x86-64-specific 13212 13213 * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 13214 * [148]14723 Backported -march=nocona from mainline 13215 * [149]15290 __float128 failed to pass to function properly 13216 13217 Cygwin/Mingw32-specific 13218 13219 * [150]15250 Option -mms-bitfields support on GCC 3.4 is not 13220 conformant to MS layout 13221 * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 13222 worker on windows32 targets 13223 13224 Bugs specific to embedded processors 13225 13226 * [152]8309 [m68k] -m5200 produces erroneous SImode set of short 13227 varaible on stack 13228 * [153]13250 [SH] Gcc code for rotation clobbers the register, but 13229 gcc continues to use the register as if it was not clobbered 13230 * [154]13803 [coldfire] movqi operand constraints too restrictivefor 13231 TARGET_COLDFIRE 13232 * [155]14093 [SH] ICE for code when using -mhitachi option in SH 13233 * [156]14457 [m6811hc] ICE with simple c++ source 13234 * [157]14542 [m6811hc] ICE on simple source 13235 * [158]15100 [SH] cc1plus got hang-up on 13236 libstdc++-v3/testsuite/abi_check.cc 13237 * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on 13238 cris-* 13239 * [160]15396 [SH] ICE with -O2 -fPIC 13240 * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 13241 ColdFire 13242 13243 Testsuite problems (compiler not affected) 13244 13245 * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely 13246 * [163]15488 (libstdc++) possibly insufficient file permissions for 13247 executing test suite 13248 * [164]15489 (libstdc++) testsuite_files determined incorrectly 13249 13250 Documentation bugs 13251 13252 * [165]13928 (libstdc++) no whatis info in some man pages generated 13253 by doxygen 13254 * [166]14150 Ada documentation out of date 13255 * [167]14949 (c++) Need to document method visibility changes 13256 * [168]15123 libstdc++-doc: Allocators.3 manpage is empty 13257 __________________________________________________________________ 13258 13259 GCC 3.4.2 13260 13261 Bug Fixes 13262 13263 This section lists the problem reports (PRs) from GCC's bug tracking 13264 system that are known to be fixed in the 3.4.2 release. This list might 13265 not be complete (that is, it is possible that some PRs that have been 13266 fixed are not listed here). 13267 13268 Bootstrap failures and issues 13269 13270 * [169]16469 [mips-sgi-irix5.3] bootstrap fails in 13271 libstdc++-v3/testsuite 13272 * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by 13273 profiledbootstrap does not work with the built compiler 13274 * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 13275 13276 Multi-platform internal compiler errors (ICEs) 13277 13278 * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 13279 cp_parser_class_specifier, in cp/parser.c 13280 * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 13281 * [174]15461 (c++) ICE due to NRV and inlining 13282 * [175]15890 (c++) ICE in c_expand_expr, in c-common.c 13283 * [176]16180 ICE: segmentation fault in RTL optimization 13284 * [177]16224 (c++) ICE in write_unscoped_name (template/namespace) 13285 * [178]16408 ICE: in delete_insn, in cfgrtl.c 13286 * [179]16529 (c++) ICE for: namespace-alias shall not be declared as 13287 the name of any other entity 13288 * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw 13289 * [181]16706 (c++) ICE in finish_member_declaration, in 13290 cp/semantics.c 13291 * [182]16810 (c++) Legal C++ program with cast gives ICE in 13292 build_ptrmemfunc 13293 * [183]16851 (c++) ICE when throwing a comma expression 13294 * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 13295 * [185]16904 (c++) ICE in finish_class_member_access_expr, in 13296 cp/typeck.c 13297 * [186]16905 (c++) ICE (segfault) with exceptions 13298 * [187]16964 (c++) ICE in cp_parser_class_specifier due to 13299 redefinition 13300 * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x' 13301 (identifier_node) in dependent_template_p, in cp/pt.c 13302 13303 Preprocessor bugs 13304 13305 * [189]16366 Preprocessor option -remap causes memory corruption 13306 13307 Optimization 13308 13309 * [190]15345 unreferenced nested inline functions not optimized away 13310 * [191]16590 Incorrect execution when compiling with -O2 13311 * [192]16693 Bitwise AND is lost when used within a cast to an enum 13312 of the same precision 13313 * [193]17078 Jump into if(0) substatement fails 13314 13315 Problems in generated debug information 13316 13317 * [194]13956 incorrect stabs for nested local variables 13318 13319 C front end bugs 13320 13321 * [195]16684 GCC should not warn about redundant redeclarations of 13322 built-ins 13323 13324 C++ compiler and library 13325 13326 * [196]12658 Thread safety problems in locale::global() and 13327 locale::locale() 13328 * [197]13092 g++ accepts invalid pointer-to-member conversion 13329 * [198]15320 Excessive memory consumption 13330 * [199]16246 Incorrect template argument deduction 13331 * [200]16273 Memory exhausted when using nested classes and virtual 13332 functions 13333 * [201]16401 ostringstream in gcc 3.4.x very slow for big data 13334 * [202]16411 undefined reference to 13335 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 13336 >::file() 13337 * [203]16489 G++ incorrectly rejects use of a null constant integral 13338 expression as a null constant pointer 13339 * [204]16618 offsetof fails with constant member 13340 * [205]16637 syntax error reported for valid input code 13341 * [206]16717 __attribute__((constructor)) broken in C++ 13342 * [207]16813 compiler error in DEBUG version of range insertion 13343 std::map::insert 13344 * [208]16853 pointer-to-member initialization from incompatible one 13345 accepted 13346 * [209]16889 ambiguity is not detected 13347 * [210]16959 Segmentation fault in ios_base::sync_with_stdio 13348 13349 Java compiler and library 13350 13351 * [211]7587 direct threaded interpreter not thread-safe 13352 * [212]16473 ServerSocket accept() leaks file descriptors 13353 * [213]16478 Hash synchronization deadlock with finalizers 13354 13355 Alpha-specific 13356 13357 * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 13358 * [215]16974 could not split insn (ice in final_scan_insn, in 13359 final.c) 13360 13361 x86-specific 13362 13363 * [216]16298 ICE in output_operand 13364 * [217]17113 ICE with SSE2 intrinsics 13365 13366 x86-64 specific 13367 13368 * [218]14697 libstdc++ couldn't find 32bit libgcc_s 13369 13370 MIPS-specific 13371 13372 * [219]15869 [mips64] No NOP after LW (with -mips1 -O0) 13373 * [220]16325 [mips64] value profiling clobbers gp on mips 13374 * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 13375 char[]s 13376 * [222]16380 [mips64] Use of uninitialised register after dbra 13377 conversion 13378 * [223]16407 [mips64] Unaligned access to local variables 13379 * [224]16643 [mips64] verify_local_live_at_start ICE after 13380 crossjumping & cfgcleanup 13381 13382 ARM-specific 13383 13384 * [225]15927 THUMB -O2: strength-reduced iteration variable ends up 13385 off by 1 13386 * [226]15948 THUMB: ICE with non-commutative cbranch 13387 * [227]17019 THUMB: bad switch statement in md code for 13388 addsi3_cbranch_scratch 13389 13390 IA64-specific 13391 13392 * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 13393 (-mtune=merced) 13394 * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 13395 (-mtune=itanium) 13396 * [230]16278 Gcc failed to build Linux kernel with -mtune=merced 13397 * [231]16414 ICE on valid code: typo in comparison of asm_noperands 13398 result 13399 * [232]16445 ICE on valid code: don't count ignored insns 13400 * [233]16490 ICE (segfault) while compiling with -fprofile-use 13401 * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 13402 13403 PowerPC-specific 13404 13405 * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 13406 * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 13407 issue) 13408 13409 SPARC-specific 13410 13411 * [237]16199 ICE while compiling apache 2.0.49 13412 * [238]16416 -m64 doesn't imply -mcpu=v9 anymore 13413 * [239]16430 ICE when returning non-C aggregates larger than 16 bytes 13414 13415 Bugs specific to embedded processors 13416 13417 * [240]16379 [m32r] can't output large model function call of memcpy 13418 * [241]17093 [m32r] ICE with -msdata=use -O0 13419 * [242]17119 [m32r] ICE at switch case 0x8000 13420 13421 DJGPP-specific 13422 13423 * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 13424 13425 Alpha Tru64-specific 13426 13427 * [244]16210 libstdc++ gratuitously omits "long long" I/O 13428 13429 Testsuite, documentation issues (compiler is not affected): 13430 13431 * [245]15488 (libstdc++) possibly insufficient file permissions for 13432 executing test suite 13433 * [246]16250 ada/doctools runs makeinfo even in release tarball 13434 __________________________________________________________________ 13435 13436 GCC 3.4.3 13437 13438 This is the [247]list of problem reports (PRs) from GCC's bug tracking 13439 system that are known to be fixed in the 3.4.3 release. This list might 13440 not be complete (that is, it is possible that some PRs that have been 13441 fixed are not listed here). 13442 13443 Bootstrap failures 13444 13445 * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 13446 * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 13447 when undeclared 13448 13449 Internal compiler errors (ICEs) affecting multiple platforms 13450 13451 * [250]13948 (java) GCJ segmentation fault while compiling GL4Java 13452 .class files 13453 * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 13454 * [252]16301 (c++) ICE when "strong" attribute is attached to a using 13455 directive 13456 * [253]16566 ICE with flexible arrays 13457 * [254]17023 ICE with nested functions in parameter declaration 13458 * [255]17027 ICE with noreturn function in loop at -O2 13459 * [256]17524 ICE in grokdeclarator, in cp/decl.c 13460 * [257]17826 (c++) ICE in cp_tree_equal 13461 13462 C and optimization bugs 13463 13464 * [258]15526 -ftrapv aborts on 0 * (-1) 13465 * [259]16999 #ident stopped working 13466 * [260]17503 quadratic behaviour in invalid_mode_change_p 13467 * [261]17581 Long long arithmetic fails inside a switch/case 13468 statement when compiled with -O2 13469 * [262]18129 -fwritable-strings doesn't work 13470 13471 C++ compiler and library bugs 13472 13473 * [263]10975 incorrect initial ostringstream::tellp() 13474 * [264]11722 Unbuffered filebuf::sgetn is slow 13475 * [265]14534 Unrecognizing static function as a template parameter 13476 when its return value is also templated 13477 * [266]15172 Copy constructor optimization in aggregate 13478 initialization 13479 * [267]15786 Bad error message for frequently occuring error. 13480 * [268]16162 Rejects valid member-template-definition 13481 * [269]16612 empty basic_strings can't live in shared memory 13482 * [270]16715 std::basic_iostream is instantiated when used, even 13483 though instantiations are already contained in libstdc++ 13484 * [271]16848 code in /ext/demangle.h appears broken 13485 * [272]17132 GCC fails to eliminate function template specialization 13486 when argument deduction fails 13487 * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 13488 in ropeimpl.h 13489 * [274]17327 use of `enumeral_type' in template type unification 13490 * [275]17393 "unused variable '._0'" warning with -Wall 13491 * [276]17501 Confusion with member templates 13492 * [277]17537 g++ not passing -lstdc++ to linker when all command line 13493 arguments are libraries 13494 * [278]17585 usage of unqualified name of static member from within 13495 class not allowed 13496 * [279]17821 Poor diagnostic for using "." instead of "->" 13497 * [280]17829 wrong error: call of overloaded function is ambiguous 13498 * [281]17851 Misleading diagnostic for invalid function declarations 13499 with undeclared types 13500 * [282]17976 Destructor is called twice 13501 * [283]18020 rejects valid definition of enum value in template 13502 * [284]18093 bogus conflict in namespace aliasing 13503 * [285]18140 C++ parser bug when using >> in templates 13504 13505 Fortran 13506 13507 * [286]17541 data statements with double precision constants fail 13508 13509 x86-specific 13510 13511 * [287]17853 -O2 ICE for MMX testcase 13512 13513 SPARC-specific 13514 13515 * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c 13516 13517 Darwin-specific 13518 13519 * [289]17167 FATAL:Symbol L_foo$stub already defined. 13520 13521 AIX-specific 13522 13523 * [290]17277 could not catch an exception when specified -maix64 13524 13525 Solaris-specific 13526 13527 * [291]17505 <cmath> calls acosf(), ceilf(), and other functions 13528 missing from system libraries 13529 13530 HP/UX specific: 13531 13532 * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 13533 13534 ARM-specific 13535 13536 * [293]17384 ICE with mode attribute on structures 13537 13538 MIPS-specific 13539 13540 * [294]17770 No NOP after LWL with -mips1 13541 13542 Other embedded target specific 13543 13544 * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 13545 * [296]14064 [avr-elf] -fdata-sections triggers ICE 13546 * [297]14678 [m68hc11-elf] gcc ICE 13547 * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 13548 * [299]15790 [i686-coff] Alignment error building gcc with i686-coff 13549 target 13550 * [300]15886 [SH] Miscompilation with -O2 -fPIC 13551 * [301]16884 [avr-elf] [fweb related] bug while initializing 13552 variables 13553 13554 Bugs relating to debugger support 13555 13556 * [302]13841 missing debug info for _Complex function arguments 13557 * [303]15860 [big-endian targets] No DW_AT_location debug info is 13558 emitted for formal arguments to a function that uses "register" 13559 qualifiers 13560 13561 Testsuite issues (compiler not affected) 13562 13563 * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 13564 * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 13565 * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 13566 testsuite 13567 13568 Documentation 13569 13570 * [307]15498 typo in gcc manual: non-existing locale example en_UK, 13571 should be en_GB 13572 * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 13573 document broken shell 13574 * [309]16406 USE_LD_AS_NEEDED undocumented 13575 __________________________________________________________________ 13576 13577 GCC 3.4.4 13578 13579 This is the [310]list of problem reports (PRs) from GCC's bug tracking 13580 system that are known to be fixed in the 3.4.4 release. This list might 13581 not be complete (that is, it is possible that some PRs that have been 13582 fixed are not listed here). 13583 __________________________________________________________________ 13584 13585 GCC 3.4.5 13586 13587 This is the [311]list of problem reports (PRs) from GCC's bug tracking 13588 system that are known to be fixed in the 3.4.5 release. This list might 13589 not be complete (that is, it is possible that some PRs that have been 13590 fixed are not listed here). 13591 13592 Bootstrap issues 13593 13594 * [312]24688 sco_math fixincl breaks math.h 13595 13596 C compiler bugs 13597 13598 * [313]17188 struct Foo { } redefinition 13599 * [314]20187 wrong code for ((unsigned char)(unsigned long 13600 long)((a?a:1)&(a*b)))?0:1) 13601 * [315]21873 infinite warning loop on bad array initializer 13602 * [316]21899 enum definition accepts values to be overriden 13603 * [317]22061 ICE in find_function_data, in function.c 13604 * [318]22308 Failure to diagnose violation of constraint 6.516p2 13605 * [319]22458 ICE on missing brace 13606 * [320]22589 ICE casting to long long 13607 * [321]24101 Segfault with preprocessed source 13608 13609 C++ compiler and library bugs 13610 13611 * [322]10611 operations on vector mode not recognized in C++ 13612 * [323]13377 unexpected behavior of namespace usage directive 13613 * [324]16002 Strange error message with new parser 13614 * [325]17413 local classes as template argument 13615 * [326]17609 spurious error message after using keyword 13616 * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 13617 * [328]18124 ICE with invalid template template parameter 13618 * [329]18155 typedef in template declaration not rejected 13619 * [330]18177 ICE with const_cast for undeclared variable 13620 * [331]18368 C++ error message regression 13621 * [332]16378 ICE when returning a copy of a packed member 13622 * [333]18466 int ::i; accepted 13623 * [334]18512 ICE on invalid usage of template base class 13624 * [335]18454 ICE when returning undefined type 13625 * [336]18738 typename not allowed with non-dependent qualified name 13626 * [337]18803 rejects access to operator() in template 13627 * [338]19004 ICE in uses_template_parms, in cp/pt.c 13628 * [339]19208 Spurious error about variably modified type 13629 * [340]18253 bad error message / ICE for invalid template parameter 13630 * [341]19608 ICE after friend function definition in local class 13631 * [342]19884 ICE on explicit instantiation of a non-template 13632 constructor 13633 * [343]20153 ICE when C++ template function contains anonymous union 13634 * [344]20563 Infinite loop in diagnostic (and ice after error 13635 message) 13636 * [345]20789 ICE with incomplete type in template 13637 * [346]21336 Internal compiler error when using custom new operators 13638 * [347]21768 ICE in error message due to violation of coding 13639 conventions 13640 * [348]21853 constness of pointer to data member ignored 13641 * [349]21903 Default argument of template function causes a 13642 compile-time error 13643 * [350]21983 multiple diagnostics 13644 * [351]21987 New testsuite failure 13645 g++.dg/warn/conversion-function-1.C 13646 * [352]22153 ICE on invalid template specialization 13647 * [353]22172 Internal compiler error, seg fault. 13648 * [354]21286 filebuf::xsgetn vs pipes 13649 * [355]22233 ICE with wrong number of template parameters 13650 * [356]22508 ICE after invalid operator new 13651 * [357]22545 ICE with pointer to class member & user defined 13652 conversion operator 13653 * [358]23528 Wrong default allocator in ext/hash_map 13654 * [359]23550 char_traits requirements/1.cc test bad math 13655 * [360]23586 Bad diagnostic for invalid namespace-name 13656 * [361]23624 ICE in invert_truthvalue, in fold-const.c 13657 * [362]23639 Bad error message: not a member of '<declaration error>' 13658 * [363]23797 ICE on typename outside template 13659 * [364]23965 Bogus error message: no matching function for call to 13660 'foo(<type error>)' 13661 * [365]24052 &#`label_decl' not supported by dump_expr#<expression 13662 error> 13663 * [366]24580 virtual base class cause exception not to be caught 13664 13665 Problems in generated debug information 13666 13667 * [367]24267 Bad DWARF for altivec vectors 13668 13669 Optimizations issues 13670 13671 * [368]17810 ICE in verify_local_live_at_start 13672 * [369]17860 Wrong generated code for loop with varying bound 13673 * [370]21709 ICE on compile-time complex NaN 13674 * [371]21964 broken tail call at -O2 or more 13675 * [372]22167 Strange optimization bug when using -Os 13676 * [373]22619 Compilation failure for real_const_1.f and 13677 real_const_2.f90 13678 * [374]23241 Invalid code generated for comparison of uchar to 255 13679 * [375]23478 Miscompilation due to reloading of a var that is also 13680 used in EH pad 13681 * [376]24470 segmentation fault in cc1plus when compiling with -O 13682 * [377]24950 ICE in operand_subword_force 13683 13684 Precompiled headers problems 13685 13686 * [378]14400 Cannot compile qt-x11-free-3.3.0 13687 * [379]14940 PCH largefile test fails on various platforms 13688 13689 Preprocessor bugs 13690 13691 * [380]20239 ICE on empty preprocessed input 13692 * [381]15220 "gcc -E -MM -MG" reports missing system headers in 13693 source directory 13694 13695 Testsuite issues 13696 13697 * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 13698 i686-pc-linux-gnu 13699 13700 Alpha specific 13701 13702 * [383]21888 bootstrap failure with linker relaxation enabled 13703 13704 ARM specific 13705 13706 * [384]15342 [arm-linux]: ICE in verify_local_live_at_start 13707 * [385]23985 Memory aliasing information incorrect in inlined memcpy 13708 13709 ColdFile specific 13710 13711 * [386]16719 Illegal move of byte into address register causes 13712 compiler to ICE 13713 13714 HPPA specific 13715 13716 * [387]21723 ICE while building libgfortran 13717 * [388]21841 -mhp-ld/-mgnu-ld documentation 13718 13719 IA-64 specific 13720 13721 * [389]23644 IA-64 hardware models and configuration options 13722 documentation error 13723 * [390]24718 Shared libgcc not used for linking by default 13724 13725 M68000 specific 13726 13727 * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c 13728 13729 MIPS specific 13730 13731 * [392]20621 ICE in change_address_1, in emit-rtl.c 13732 13733 PowerPC and PowerPC64 specific 13734 13735 * [393]18583 error on valid code: const 13736 __attribute__((altivec(vector__))) doesn't work in arrays 13737 * [394]20191 ICE in reload_cse_simplify_operands 13738 * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 13739 * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 13740 * [397]23404 gij trashes args of functions with more than 8 fp args 13741 * [398]23539 C & C++ compiler generating misaligned references 13742 regardless of compiler flags 13743 * [399]24102 floatdisf2_internal2 broken 13744 * [400]24465 -mminimal-toc miscompilation of __thread vars 13745 13746 Solaris specific 13747 13748 * [401]19933 Problem with define of HUGE_VAL in math_c99 13749 * [402]21889 Native Solaris assembler cannot grok DTP-relative debug 13750 symbols 13751 13752 SPARC specific 13753 13754 * [403]19300 PCH failures on sparc-linux 13755 * [404]20301 Assembler labels have a leading "-" 13756 * [405]20673 C PCH testsuite assembly comparison failure 13757 13758 x86 and x86_64 specific 13759 13760 * [406]18582 ICE with arrays of type V2DF 13761 * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 13762 -fsched2-use-traces 13763 * [408]21716 ICE in reg-stack.c's swap_rtx_condition 13764 * [409]24315 amd64 fails -fpeephole2 13765 __________________________________________________________________ 13766 13767 GCC 3.4.6 13768 13769 This is the [410]list of problem reports (PRs) from GCC's bug tracking 13770 system that are known to be fixed in the 3.4.6 release. This list might 13771 not be complete (that is, it is possible that some PRs that have been 13772 fixed are not listed here). 13773 13774 13775 For questions related to the use of GCC, please consult these web 13776 pages and the [411]GCC manuals. If that fails, the 13777 [412]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13778 web pages and the development of GCC are welcome on our developer 13779 list at [413]gcc (a] gcc.gnu.org. All of [414]our lists have public 13780 archives. 13781 13782 Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and 13783 distribution of this entire article is permitted in any medium, 13784 provided this notice is preserved. 13785 13786 These pages are [416]maintained by the GCC team. Last modified 13787 2019-11-28[417]. 13788 13789 References 13790 13791 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 13792 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 13793 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 13794 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 13795 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html 13796 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 13797 7. https://www.boost.org/ 13798 8. https://gcc.gnu.org/PR11953 13799 9. https://gcc.gnu.org/PR8361 13800 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other Builtins 13801 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 13802 12. http://gcc.gnu.org/bugs/#cxx_rvalbind 13803 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 13804 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 13805 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 13806 16. http://www.gnu.org/software/classpath/ 13807 17. http://www.eclipse.org/ 13808 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 13809 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 13810 20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 13811 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 13812 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 13813 23. http://gcc.gnu.org/gcc-3.4/mips-abi.html 13814 24. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html 13815 25. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 13816 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 13817 27. https://gcc.gnu.org/PR10129 13818 28. https://gcc.gnu.org/PR14576 13819 29. https://gcc.gnu.org/PR14760 13820 30. https://gcc.gnu.org/PR14671 13821 31. https://gcc.gnu.org/PR15093 13822 32. https://gcc.gnu.org/PR15178 13823 33. https://gcc.gnu.org/PR12753 13824 34. https://gcc.gnu.org/PR13985 13825 35. https://gcc.gnu.org/PR14810 13826 36. https://gcc.gnu.org/PR14883 13827 37. https://gcc.gnu.org/PR15044 13828 38. https://gcc.gnu.org/PR15057 13829 39. https://gcc.gnu.org/PR15064 13830 40. https://gcc.gnu.org/PR15142 13831 41. https://gcc.gnu.org/PR15159 13832 42. https://gcc.gnu.org/PR15165 13833 43. https://gcc.gnu.org/PR15193 13834 44. https://gcc.gnu.org/PR15209 13835 45. https://gcc.gnu.org/PR15227 13836 46. https://gcc.gnu.org/PR15285 13837 47. https://gcc.gnu.org/PR15299 13838 48. https://gcc.gnu.org/PR15329 13839 49. https://gcc.gnu.org/PR15550 13840 50. https://gcc.gnu.org/PR15554 13841 51. https://gcc.gnu.org/PR15640 13842 52. https://gcc.gnu.org/PR15666 13843 53. https://gcc.gnu.org/PR15696 13844 54. https://gcc.gnu.org/PR15701 13845 55. https://gcc.gnu.org/PR15761 13846 56. https://gcc.gnu.org/PR15829 13847 57. https://gcc.gnu.org/PR14538 13848 58. https://gcc.gnu.org/PR12391 13849 59. https://gcc.gnu.org/PR14649 13850 60. https://gcc.gnu.org/PR15004 13851 61. https://gcc.gnu.org/PR15749 13852 62. https://gcc.gnu.org/PR10646 13853 63. https://gcc.gnu.org/PR12077 13854 64. https://gcc.gnu.org/PR13598 13855 65. https://gcc.gnu.org/PR14211 13856 66. https://gcc.gnu.org/PR14220 13857 67. https://gcc.gnu.org/PR14245 13858 68. https://gcc.gnu.org/PR14340 13859 69. https://gcc.gnu.org/PR14600 13860 70. https://gcc.gnu.org/PR14668 13861 71. https://gcc.gnu.org/PR14775 13862 72. https://gcc.gnu.org/PR14821 13863 73. https://gcc.gnu.org/PR14930 13864 74. https://gcc.gnu.org/PR14932 13865 75. https://gcc.gnu.org/PR14950 13866 76. https://gcc.gnu.org/PR14962 13867 77. https://gcc.gnu.org/PR14975 13868 78. https://gcc.gnu.org/PR15002 13869 79. https://gcc.gnu.org/PR15025 13870 80. https://gcc.gnu.org/PR15046 13871 81. https://gcc.gnu.org/PR15069 13872 82. https://gcc.gnu.org/PR15074 13873 83. https://gcc.gnu.org/PR15083 13874 84. https://gcc.gnu.org/PR15096 13875 85. https://gcc.gnu.org/PR15287 13876 86. https://gcc.gnu.org/PR15317 13877 87. https://gcc.gnu.org/PR15337 13878 88. https://gcc.gnu.org/PR15361 13879 89. https://gcc.gnu.org/PR15412 13880 90. https://gcc.gnu.org/PR15427 13881 91. https://gcc.gnu.org/PR15471 13882 92. https://gcc.gnu.org/PR15503 13883 93. https://gcc.gnu.org/PR15507 13884 94. https://gcc.gnu.org/PR15542 13885 95. https://gcc.gnu.org/PR15565 13886 96. https://gcc.gnu.org/PR15625 13887 97. https://gcc.gnu.org/PR15629 13888 98. https://gcc.gnu.org/PR15742 13889 99. https://gcc.gnu.org/PR15775 13890 100. https://gcc.gnu.org/PR15821 13891 101. https://gcc.gnu.org/PR15862 13892 102. https://gcc.gnu.org/PR15875 13893 103. https://gcc.gnu.org/PR15877 13894 104. https://gcc.gnu.org/PR15947 13895 105. https://gcc.gnu.org/PR16020 13896 106. https://gcc.gnu.org/PR16154 13897 107. https://gcc.gnu.org/PR16174 13898 108. https://gcc.gnu.org/PR14315 13899 109. https://gcc.gnu.org/PR15151 13900 110. https://gcc.gnu.org/PR7993 13901 111. https://gcc.gnu.org/PR15228 13902 112. https://gcc.gnu.org/PR15345 13903 113. https://gcc.gnu.org/PR15945 13904 114. https://gcc.gnu.org/PR15526 13905 115. https://gcc.gnu.org/PR14690 13906 116. https://gcc.gnu.org/PR15112 13907 117. https://gcc.gnu.org/PR15067 13908 118. https://gcc.gnu.org/PR1963 13909 119. https://gcc.gnu.org/PR15717 13910 120. https://gcc.gnu.org/PR14782 13911 121. https://gcc.gnu.org/PR14828 13912 122. https://gcc.gnu.org/PR15202 13913 123. https://gcc.gnu.org/PR14610 13914 124. https://gcc.gnu.org/PR14813 13915 125. https://gcc.gnu.org/PR14857 13916 126. https://gcc.gnu.org/PR15598 13917 127. https://gcc.gnu.org/PR15653 13918 128. https://gcc.gnu.org/PR15189 13919 129. https://gcc.gnu.org/PR15331 13920 130. https://gcc.gnu.org/PR16144 13921 131. https://gcc.gnu.org/PR16176 13922 132. https://gcc.gnu.org/PR11591 13923 133. https://gcc.gnu.org/PR12028 13924 134. https://gcc.gnu.org/PR14478 13925 135. https://gcc.gnu.org/PR14567 13926 136. https://gcc.gnu.org/PR14715 13927 137. https://gcc.gnu.org/PR14902 13928 138. https://gcc.gnu.org/PR14924 13929 139. https://gcc.gnu.org/PR14960 13930 140. https://gcc.gnu.org/PR15106 13931 141. https://gcc.gnu.org/PR16026 13932 142. https://gcc.gnu.org/PR15191 13933 143. https://gcc.gnu.org/PR15662 13934 144. https://gcc.gnu.org/PR15054 13935 145. https://gcc.gnu.org/PR15783 13936 146. https://gcc.gnu.org/PR15626 13937 147. https://gcc.gnu.org/PR14326 13938 148. https://gcc.gnu.org/PR14723 13939 149. https://gcc.gnu.org/PR15290 13940 150. https://gcc.gnu.org/PR15250 13941 151. https://gcc.gnu.org/PR15551 13942 152. https://gcc.gnu.org/PR8309 13943 153. https://gcc.gnu.org/PR13250 13944 154. https://gcc.gnu.org/PR13803 13945 155. https://gcc.gnu.org/PR14093 13946 156. https://gcc.gnu.org/PR14457 13947 157. https://gcc.gnu.org/PR14542 13948 158. https://gcc.gnu.org/PR15100 13949 159. https://gcc.gnu.org/PR15296 13950 160. https://gcc.gnu.org/PR15396 13951 161. https://gcc.gnu.org/PR15782 13952 162. https://gcc.gnu.org/PR11610 13953 163. https://gcc.gnu.org/PR15488 13954 164. https://gcc.gnu.org/PR15489 13955 165. https://gcc.gnu.org/PR13928 13956 166. https://gcc.gnu.org/PR14150 13957 167. https://gcc.gnu.org/PR14949 13958 168. https://gcc.gnu.org/PR15123 13959 169. https://gcc.gnu.org/PR16469 13960 170. https://gcc.gnu.org/PR16344 13961 171. https://gcc.gnu.org/PR16842 13962 172. https://gcc.gnu.org/PR12608 13963 173. https://gcc.gnu.org/PR14492 13964 174. https://gcc.gnu.org/PR15461 13965 175. https://gcc.gnu.org/PR15890 13966 176. https://gcc.gnu.org/PR16180 13967 177. https://gcc.gnu.org/PR16224 13968 178. https://gcc.gnu.org/PR16408 13969 179. https://gcc.gnu.org/PR16529 13970 180. https://gcc.gnu.org/PR16698 13971 181. https://gcc.gnu.org/PR16706 13972 182. https://gcc.gnu.org/PR16810 13973 183. https://gcc.gnu.org/PR16851 13974 184. https://gcc.gnu.org/PR16870 13975 185. https://gcc.gnu.org/PR16904 13976 186. https://gcc.gnu.org/PR16905 13977 187. https://gcc.gnu.org/PR16964 13978 188. https://gcc.gnu.org/PR17068 13979 189. https://gcc.gnu.org/PR16366 13980 190. https://gcc.gnu.org/PR15345 13981 191. https://gcc.gnu.org/PR16590 13982 192. https://gcc.gnu.org/PR16693 13983 193. https://gcc.gnu.org/PR17078 13984 194. https://gcc.gnu.org/PR13956 13985 195. https://gcc.gnu.org/PR16684 13986 196. https://gcc.gnu.org/PR12658 13987 197. https://gcc.gnu.org/PR13092 13988 198. https://gcc.gnu.org/PR15320 13989 199. https://gcc.gnu.org/PR16246 13990 200. https://gcc.gnu.org/PR16273 13991 201. https://gcc.gnu.org/PR16401 13992 202. https://gcc.gnu.org/PR16411 13993 203. https://gcc.gnu.org/PR16489 13994 204. https://gcc.gnu.org/PR16618 13995 205. https://gcc.gnu.org/PR16637 13996 206. https://gcc.gnu.org/PR16717 13997 207. https://gcc.gnu.org/PR16813 13998 208. https://gcc.gnu.org/PR16853 13999 209. https://gcc.gnu.org/PR16889 14000 210. https://gcc.gnu.org/PR16959 14001 211. https://gcc.gnu.org/PR7587 14002 212. https://gcc.gnu.org/PR16473 14003 213. https://gcc.gnu.org/PR16478 14004 214. https://gcc.gnu.org/PR10695 14005 215. https://gcc.gnu.org/PR16974 14006 216. https://gcc.gnu.org/PR16298 14007 217. https://gcc.gnu.org/PR17113 14008 218. https://gcc.gnu.org/PR14697 14009 219. https://gcc.gnu.org/PR15869 14010 220. https://gcc.gnu.org/PR16325 14011 221. https://gcc.gnu.org/PR16357 14012 222. https://gcc.gnu.org/PR16380 14013 223. https://gcc.gnu.org/PR16407 14014 224. https://gcc.gnu.org/PR16643 14015 225. https://gcc.gnu.org/PR15927 14016 226. https://gcc.gnu.org/PR15948 14017 227. https://gcc.gnu.org/PR17019 14018 228. https://gcc.gnu.org/PR16130 14019 229. https://gcc.gnu.org/PR16142 14020 230. https://gcc.gnu.org/PR16278 14021 231. https://gcc.gnu.org/PR16414 14022 232. https://gcc.gnu.org/PR16445 14023 233. https://gcc.gnu.org/PR16490 14024 234. https://gcc.gnu.org/PR16683 14025 235. https://gcc.gnu.org/PR16195 14026 236. https://gcc.gnu.org/PR16239 14027 237. https://gcc.gnu.org/PR16199 14028 238. https://gcc.gnu.org/PR16416 14029 239. https://gcc.gnu.org/PR16430 14030 240. https://gcc.gnu.org/PR16379 14031 241. https://gcc.gnu.org/PR17093 14032 242. https://gcc.gnu.org/PR17119 14033 243. https://gcc.gnu.org/PR15928 14034 244. https://gcc.gnu.org/PR16210 14035 245. https://gcc.gnu.org/PR15488 14036 246. https://gcc.gnu.org/PR16250 14037 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 14038 248. https://gcc.gnu.org/PR17369 14039 249. https://gcc.gnu.org/PR17850 14040 250. https://gcc.gnu.org/PR13948 14041 251. https://gcc.gnu.org/PR14492 14042 252. https://gcc.gnu.org/PR16301 14043 253. https://gcc.gnu.org/PR16566 14044 254. https://gcc.gnu.org/PR17023 14045 255. https://gcc.gnu.org/PR17027 14046 256. https://gcc.gnu.org/PR17524 14047 257. https://gcc.gnu.org/PR17826 14048 258. https://gcc.gnu.org/PR15526 14049 259. https://gcc.gnu.org/PR16999 14050 260. https://gcc.gnu.org/PR17503 14051 261. https://gcc.gnu.org/PR17581 14052 262. https://gcc.gnu.org/PR18129 14053 263. https://gcc.gnu.org/PR10975 14054 264. https://gcc.gnu.org/PR11722 14055 265. https://gcc.gnu.org/PR14534 14056 266. https://gcc.gnu.org/PR15172 14057 267. https://gcc.gnu.org/PR15786 14058 268. https://gcc.gnu.org/PR16162 14059 269. https://gcc.gnu.org/PR16612 14060 270. https://gcc.gnu.org/PR16715 14061 271. https://gcc.gnu.org/PR16848 14062 272. https://gcc.gnu.org/PR17132 14063 273. https://gcc.gnu.org/PR17259 14064 274. https://gcc.gnu.org/PR17327 14065 275. https://gcc.gnu.org/PR17393 14066 276. https://gcc.gnu.org/PR17501 14067 277. https://gcc.gnu.org/PR17537 14068 278. https://gcc.gnu.org/PR17585 14069 279. https://gcc.gnu.org/PR17821 14070 280. https://gcc.gnu.org/PR17829 14071 281. https://gcc.gnu.org/PR17851 14072 282. https://gcc.gnu.org/PR17976 14073 283. https://gcc.gnu.org/PR18020 14074 284. https://gcc.gnu.org/PR18093 14075 285. https://gcc.gnu.org/PR18140 14076 286. https://gcc.gnu.org/PR17541 14077 287. https://gcc.gnu.org/PR17853 14078 288. https://gcc.gnu.org/PR17245 14079 289. https://gcc.gnu.org/PR17167 14080 290. https://gcc.gnu.org/PR17277 14081 291. https://gcc.gnu.org/PR17505 14082 292. https://gcc.gnu.org/PR17684 14083 293. https://gcc.gnu.org/PR17384 14084 294. https://gcc.gnu.org/PR17770 14085 295. https://gcc.gnu.org/PR11476 14086 296. https://gcc.gnu.org/PR14064 14087 297. https://gcc.gnu.org/PR14678 14088 298. https://gcc.gnu.org/PR15583 14089 299. https://gcc.gnu.org/PR15790 14090 300. https://gcc.gnu.org/PR15886 14091 301. https://gcc.gnu.org/PR16884 14092 302. https://gcc.gnu.org/PR13841 14093 303. https://gcc.gnu.org/PR15860 14094 304. https://gcc.gnu.org/PR17465 14095 305. https://gcc.gnu.org/PR17469 14096 306. https://gcc.gnu.org/PR18138 14097 307. https://gcc.gnu.org/PR15498 14098 308. https://gcc.gnu.org/PR15747 14099 309. https://gcc.gnu.org/PR16406 14100 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 14101 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 14102 312. https://gcc.gnu.org/PR24688 14103 313. https://gcc.gnu.org/PR17188 14104 314. https://gcc.gnu.org/PR20187 14105 315. https://gcc.gnu.org/PR21873 14106 316. https://gcc.gnu.org/PR21899 14107 317. https://gcc.gnu.org/PR22061 14108 318. https://gcc.gnu.org/PR22208 14109 319. https://gcc.gnu.org/PR22458 14110 320. https://gcc.gnu.org/PR22589 14111 321. https://gcc.gnu.org/PR24101 14112 322. https://gcc.gnu.org/PR10611 14113 323. https://gcc.gnu.org/PR13377 14114 324. https://gcc.gnu.org/PR16002 14115 325. https://gcc.gnu.org/PR17413 14116 326. https://gcc.gnu.org/PR17609 14117 327. https://gcc.gnu.org/PR17618 14118 328. https://gcc.gnu.org/PR18124 14119 329. https://gcc.gnu.org/PR18155 14120 330. https://gcc.gnu.org/PR18177 14121 331. https://gcc.gnu.org/PR18368 14122 332. https://gcc.gnu.org/PR18378 14123 333. https://gcc.gnu.org/PR18466 14124 334. https://gcc.gnu.org/PR18512 14125 335. https://gcc.gnu.org/PR18545 14126 336. https://gcc.gnu.org/PR18738 14127 337. https://gcc.gnu.org/PR18803 14128 338. https://gcc.gnu.org/PR19004 14129 339. https://gcc.gnu.org/PR19208 14130 340. https://gcc.gnu.org/PR19253 14131 341. https://gcc.gnu.org/PR19608 14132 342. https://gcc.gnu.org/PR19884 14133 343. https://gcc.gnu.org/PR20153 14134 344. https://gcc.gnu.org/PR20563 14135 345. https://gcc.gnu.org/PR20789 14136 346. https://gcc.gnu.org/PR21336 14137 347. https://gcc.gnu.org/PR21768 14138 348. https://gcc.gnu.org/PR21853 14139 349. https://gcc.gnu.org/PR21903 14140 350. https://gcc.gnu.org/PR21983 14141 351. https://gcc.gnu.org/PR21987 14142 352. https://gcc.gnu.org/PR22153 14143 353. https://gcc.gnu.org/PR22172 14144 354. https://gcc.gnu.org/PR21286 14145 355. https://gcc.gnu.org/PR22233 14146 356. https://gcc.gnu.org/PR22508 14147 357. https://gcc.gnu.org/PR22545 14148 358. https://gcc.gnu.org/PR23528 14149 359. https://gcc.gnu.org/PR23550 14150 360. https://gcc.gnu.org/PR23586 14151 361. https://gcc.gnu.org/PR23624 14152 362. https://gcc.gnu.org/PR23639 14153 363. https://gcc.gnu.org/PR23797 14154 364. https://gcc.gnu.org/PR23965 14155 365. https://gcc.gnu.org/PR24052 14156 366. https://gcc.gnu.org/PR24580 14157 367. https://gcc.gnu.org/PR24267 14158 368. https://gcc.gnu.org/PR17810 14159 369. https://gcc.gnu.org/PR17860 14160 370. https://gcc.gnu.org/PR21709 14161 371. https://gcc.gnu.org/PR21964 14162 372. https://gcc.gnu.org/PR22167 14163 373. https://gcc.gnu.org/PR22619 14164 374. https://gcc.gnu.org/PR23241 14165 375. https://gcc.gnu.org/PR23478 14166 376. https://gcc.gnu.org/PR24470 14167 377. https://gcc.gnu.org/PR24950 14168 378. https://gcc.gnu.org/PR14400 14169 379. https://gcc.gnu.org/PR14940 14170 380. https://gcc.gnu.org/PR20239 14171 381. https://gcc.gnu.org/PR15220 14172 382. https://gcc.gnu.org/PR19275 14173 383. https://gcc.gnu.org/PR21888 14174 384. https://gcc.gnu.org/PR15342 14175 385. https://gcc.gnu.org/PR23985 14176 386. https://gcc.gnu.org/PR16719 14177 387. https://gcc.gnu.org/PR21723 14178 388. https://gcc.gnu.org/PR21841 14179 389. https://gcc.gnu.org/PR23644 14180 390. https://gcc.gnu.org/PR24718 14181 391. https://gcc.gnu.org/PR18421 14182 392. https://gcc.gnu.org/PR20621 14183 393. https://gcc.gnu.org/PR18583 14184 394. https://gcc.gnu.org/PR20191 14185 395. https://gcc.gnu.org/PR22083 14186 396. https://gcc.gnu.org/PR23070 14187 397. https://gcc.gnu.org/PR23404 14188 398. https://gcc.gnu.org/PR23539 14189 399. https://gcc.gnu.org/PR24102 14190 400. https://gcc.gnu.org/PR24465 14191 401. https://gcc.gnu.org/PR19933 14192 402. https://gcc.gnu.org/PR21889 14193 403. https://gcc.gnu.org/PR19300 14194 404. https://gcc.gnu.org/PR20301 14195 405. https://gcc.gnu.org/PR20673 14196 406. https://gcc.gnu.org/PR18582 14197 407. https://gcc.gnu.org/PR19340 14198 408. https://gcc.gnu.org/PR21716 14199 409. https://gcc.gnu.org/PR24315 14200 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 14201 411. https://gcc.gnu.org/onlinedocs/ 14202 412. mailto:gcc-help (a] gcc.gnu.org 14203 413. mailto:gcc (a] gcc.gnu.org 14204 414. https://gcc.gnu.org/lists.html 14205 415. https://www.fsf.org/ 14206 416. https://gcc.gnu.org/about.html 14207 417. http://validator.w3.org/check/referer 14208 ====================================================================== 14209 http://gcc.gnu.org/gcc-3.3/index.html 14210 GCC 3.3 Release Series 14211 14212 (This release series is no longer supported.) 14213 14214 May 03, 2005 14215 14216 The [1]GNU project and the GCC developers are pleased to announce the 14217 release of GCC 3.3.6. 14218 14219 This release is a bug-fix release, containing fixes for regressions in 14220 GCC 3.3.5 relative to previous releases of GCC. 14221 14222 This release is the last of the series 3.3.x. 14223 14224 The GCC 3.3 release series includes numerous [2]new features, 14225 improvements, bug fixes, and other changes, thanks to an [3]amazing 14226 group of volunteers. 14227 14228 Release History 14229 14230 GCC 3.3.6 14231 May 3, 2005 ([4]changes) 14232 14233 GCC 3.3.5 14234 September 30, 2004 ([5]changes) 14235 14236 GCC 3.3.4 14237 May 31, 2004 ([6]changes) 14238 14239 GCC 3.3.3 14240 February 14, 2004 ([7]changes) 14241 14242 GCC 3.3.2 14243 October 16, 2003 ([8]changes) 14244 14245 GCC 3.3.1 14246 August 8, 2003 ([9]changes) 14247 14248 GCC 3.3 14249 May 14, 2003 ([10]changes) 14250 14251 References and Acknowledgements 14252 14253 GCC used to stand for the GNU C Compiler, but since the compiler 14254 supports several other languages aside from C, it now stands for the 14255 GNU Compiler Collection. 14256 14257 A list of [11]successful builds is updated as new information becomes 14258 available. 14259 14260 The GCC developers would like to thank the numerous people that have 14261 contributed new features, improvements, bug fixes, and other changes as 14262 well as test results to GCC. This [12]amazing group of volunteers is 14263 what makes GCC successful. 14264 14265 For additional information about GCC please refer to the [13]GCC 14266 project web site or contact the [14]GCC development mailing list. 14267 14268 To obtain GCC please use [15]our mirror sites, or our CVS server. 14269 14270 14271 For questions related to the use of GCC, please consult these web 14272 pages and the [16]GCC manuals. If that fails, the 14273 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14274 web pages and the development of GCC are welcome on our developer 14275 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 14276 archives. 14277 14278 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 14279 distribution of this entire article is permitted in any medium, 14280 provided this notice is preserved. 14281 14282 These pages are [21]maintained by the GCC team. Last modified 14283 2019-11-28[22]. 14284 14285 References 14286 14287 1. http://www.gnu.org/ 14288 2. http://gcc.gnu.org/gcc-3.3/changes.html 14289 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14290 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 14291 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 14292 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 14293 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 14294 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 14295 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 14296 10. http://gcc.gnu.org/gcc-3.3/changes.html 14297 11. http://gcc.gnu.org/gcc-3.3/buildstat.html 14298 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14299 13. http://gcc.gnu.org/index.html 14300 14. mailto:gcc (a] gcc.gnu.org 14301 15. http://gcc.gnu.org/mirrors.html 14302 16. https://gcc.gnu.org/onlinedocs/ 14303 17. mailto:gcc-help (a] gcc.gnu.org 14304 18. mailto:gcc (a] gcc.gnu.org 14305 19. https://gcc.gnu.org/lists.html 14306 20. https://www.fsf.org/ 14307 21. https://gcc.gnu.org/about.html 14308 22. http://validator.w3.org/check/referer 14309 ====================================================================== 14310 http://gcc.gnu.org/gcc-3.3/changes.html 14311 GCC 3.3 Release Series 14312 Changes, New Features, and Fixes 14313 14314 The latest release in the 3.3 release series is [1]GCC 3.3.6. 14315 14316 Caveats 14317 14318 * The preprocessor no longer accepts multi-line string literals. They 14319 were deprecated in 3.0, 3.1, and 3.2. 14320 * The preprocessor no longer supports the -A- switch when appearing 14321 alone. -A- followed by an assertion is still supported. 14322 * Support for all the systems [2]obsoleted in GCC 3.1 has been 14323 removed from GCC 3.3. See below for a [3]list of systems which are 14324 obsoleted in this release. 14325 * Checking for null format arguments has been decoupled from the rest 14326 of the format checking mechanism. Programs which use the format 14327 attribute may regain this functionality by using the new [4]nonnull 14328 function attribute. Note that all functions for which GCC has a 14329 built-in format attribute, an appropriate built-in nonnull 14330 attribute is also applied. 14331 * The DWARF (version 1) debugging format has been deprecated and will 14332 be removed in a future version of GCC. Version 2 of the DWARF 14333 debugging format will continue to be supported for the foreseeable 14334 future. 14335 * The C and Objective-C compilers no longer accept the "Naming Types" 14336 extension (typedef foo = bar); it was already unavailable in C++. 14337 Code which uses it will need to be changed to use the "typeof" 14338 extension instead: typedef typeof(bar) foo. (We have removed this 14339 extension without a period of deprecation because it has caused the 14340 compiler to crash since version 3.0 and no one noticed until very 14341 recently. Thus we conclude it is not in widespread use.) 14342 * The -traditional C compiler option has been removed. It was 14343 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 14344 available.) The <varargs.h> header, used for writing variadic 14345 functions in traditional C, still exists but will produce an error 14346 message if used. 14347 * GCC 3.3.1 automatically places zero-initialized variables in the 14348 .bss section on some operating systems. Versions of GNU Emacs up to 14349 (and including) 21.3 will not work correctly when using this 14350 optimization; you can use -fno-zero-initialized-in-bss to disable 14351 it. 14352 14353 General Optimizer Improvements 14354 14355 * A new scheme for accurately describing processor pipelines, the 14356 [5]DFA scheduler, has been added. 14357 * Pavel Nejedly, Charles University Prague, has contributed new file 14358 format used by the edge coverage profiler (-fprofile-arcs). 14359 The new format is robust and diagnoses common mistakes where 14360 profiles from different versions (or compilations) of the program 14361 are combined resulting in nonsensical profiles and slow code to 14362 produced with profile feedback. Additionally this format allows 14363 extra data to be gathered. Currently, overall statistics are 14364 produced helping optimizers to identify hot spots of a program 14365 globally replacing the old intra-procedural scheme and resulting in 14366 better code. Note that the gcov tool from older GCC versions will 14367 not be able to parse the profiles generated by GCC 3.3 and vice 14368 versa. 14369 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 14370 pass enabled using -ftracer. This pass simplifies the control flow 14371 of functions allowing other optimizations to do better job. 14372 He also contributed the function reordering pass 14373 (-freorder-functions) to optimize function placement using profile 14374 feedback. 14375 14376 New Languages and Language specific improvements 14377 14378 C/ObjC/C++ 14379 14380 * The preprocessor now accepts directives within macro arguments. It 14381 processes them just as if they had not been within macro arguments. 14382 * The separate ISO and traditional preprocessors have been completely 14383 removed. The front end handles either type of preprocessed output 14384 if necessary. 14385 * In C99 mode preprocessor arithmetic is done in the precision of the 14386 target's intmax_t, as required by that standard. 14387 * The preprocessor can now copy comments inside macros to the output 14388 file when the macro is expanded. This feature, enabled using the 14389 -CC option, is intended for use by applications which place 14390 metadata or directives inside comments, such as lint. 14391 * The method of constructing the list of directories to be searched 14392 for header files has been revised. If a directory named by a -I 14393 option is a standard system include directory, the option is 14394 ignored to ensure that the default search order for system 14395 directories and the special treatment of system header files are 14396 not defeated. 14397 * A few more [6]ISO C99 features now work correctly. 14398 * A new function attribute, nonnull, has been added which allows 14399 pointer arguments to functions to be specified as requiring a 14400 non-null value. The compiler currently uses this information to 14401 issue a warning when it detects a null value passed in such an 14402 argument slot. 14403 * A new type attribute, may_alias, has been added. Accesses to 14404 objects with types with this attribute are not subjected to 14405 type-based alias analysis, but are instead assumed to be able to 14406 alias any other type of objects, just like the char type. 14407 14408 C++ 14409 14410 * Type based alias analysis has been implemented for C++ aggregate 14411 types. 14412 14413 Objective-C 14414 14415 * Generate an error if Objective-C objects are passed by value in 14416 function and method calls. 14417 * When -Wselector is used, check the whole list of selectors at the 14418 end of compilation, and emit a warning if a @selector() is not 14419 known. 14420 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 14421 * No longer need to include objc/objc-class.h to compile self calls 14422 in class methods (NeXT runtime only). 14423 * New -Wundeclared-selector option. 14424 * Removed selector bloating which was causing object files to be 10% 14425 bigger on average (GNU runtime only). 14426 * Using at run time @protocol() objects has been fixed in certain 14427 situations (GNU runtime only). 14428 * Type checking has been fixed and improved in many situations 14429 involving protocols. 14430 14431 Java 14432 14433 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 14434 1.4) API. 14435 * The JDK 1.4 assert facility has been implemented. 14436 * The bytecode interpreter is now direct threaded and thus faster. 14437 14438 Fortran 14439 14440 * Fortran improvements are listed in [7]the Fortran documentation. 14441 14442 Ada 14443 14444 * Ada tasking now works with glibc 2.3.x threading libraries. 14445 14446 New Targets and Target Specific Improvements 14447 14448 * The following changes have been made to the HP-PA port: 14449 + The port now defaults to scheduling for the PA8000 series of 14450 processors. 14451 + Scheduling support for the PA7300 processor has been added. 14452 + The 32-bit port now supports weak symbols under HP-UX 11. 14453 + The handling of initializers and finalizers has been improved 14454 under HP-UX 11. The 64-bit port no longer uses collect2. 14455 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 14456 + ABI fixes to correct the passing of small structures by value. 14457 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 14458 use the DFA processor pipeline description. 14459 * The following NetBSD configurations for the SuperH processor family 14460 have been added: 14461 + SH3, big-endian, sh-*-netbsdelf* 14462 + SH3, little-endian, shle-*-netbsdelf* 14463 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 14464 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 14465 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 14466 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 14467 * The following changes have been made to the IA-32/x86-64 port: 14468 + SSE2 and 3dNOW! intrinsics are now supported. 14469 + Support for thread local storage has been added to the IA-32 14470 and x86-64 ports. 14471 + The x86-64 port has been significantly improved. 14472 * The following changes have been made to the MIPS port: 14473 + All configurations now accept the -mabi switch. Note that you 14474 will need appropriate multilibs for this option to work 14475 properly. 14476 + ELF configurations will always pass an ABI flag to the 14477 assembler, except when the MIPS EABI is selected. 14478 + -mabi=64 no longer selects MIPS IV code. 14479 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 14480 been removed from this release. 14481 + -march now changes the core ISA level. In previous releases, 14482 it would change the use of processor-specific extensions, but 14483 would leave the core ISA unchanged. For example, mips64-elf 14484 -march=r8000 will now generate MIPS IV code. 14485 + Under most configurations, -mipsN now acts as a synonym for 14486 -march. 14487 + There are some new preprocessor macros to describe the -march 14488 and -mtune settings. See the documentation of those options 14489 for details. 14490 + Support for the NEC VR-Series processors has been added. This 14491 includes the 54xx, 5500, and 41xx series. 14492 + Support for the Sandcraft sr71k processor has been added. 14493 * The following changes have been made to the S/390 port: 14494 + Support to build the Java runtime libraries has been added. 14495 Java is now enabled by default on s390-*-linux* and 14496 s390x-*-linux* targets. 14497 + Multilib support for the s390x-*-linux* target has been added; 14498 this allows to build 31-bit binaries using the -m31 option. 14499 + Support for thread local storage has been added. 14500 + Inline assembler code may now use the 'Q' constraint to 14501 specify memory operands without index register. 14502 + Various platform-specific performance improvements have been 14503 implemented; in particular, the compiler now uses the BRANCH 14504 ON COUNT family of instructions and makes more frequent use of 14505 the TEST UNDER MASK family of instructions. 14506 * The following changes have been made to the PowerPC port: 14507 + Support for IBM Power4 processor added. 14508 + Support for Motorola e500 SPE added. 14509 + Support for AIX 5.2 added. 14510 + Function and Data sections now supported on AIX. 14511 + Sibcall optimizations added. 14512 * The support for H8 Tiny is added to the H8/300 port with -mn. 14513 14514 Obsolete Systems 14515 14516 Support for a number of older systems has been declared obsolete in GCC 14517 3.3. Unless there is activity to revive them, the next release of GCC 14518 will have their sources permanently removed. 14519 14520 All configurations of the following processor architectures have been 14521 declared obsolete: 14522 * Matsushita MN10200, mn10200-*-* 14523 * Motorola 88000, m88k-*-* 14524 * IBM ROMP, romp-*-* 14525 14526 Also, some individual systems have been obsoleted: 14527 * Alpha 14528 + Interix, alpha*-*-interix* 14529 + Linux libc1, alpha*-*-linux*libc1* 14530 + Linux ECOFF, alpha*-*-linux*ecoff* 14531 * ARM 14532 + Generic a.out, arm*-*-aout* 14533 + Conix, arm*-*-conix* 14534 + "Old ABI," arm*-*-oabi 14535 + StrongARM/COFF, strongarm-*-coff* 14536 * HPPA (PA-RISC) 14537 + Generic OSF, hppa1.0-*-osf* 14538 + Generic BSD, hppa1.0-*-bsd* 14539 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 14540 + HiUX, hppa*-*-hiux* 14541 + Mach Lites, hppa*-*-lites* 14542 * Intel 386 family 14543 + Windows NT 3.x, i?86-*-win32 14544 * MC68000 family 14545 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 14546 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 14547 m68k-sun-mach* 14548 + AT&T systems, m68000-att-sysv* 14549 + Atari systems, m68k-atari-sysv* 14550 + Motorola systems, m68k-motorola-sysv* 14551 + NCR systems, m68k-ncr-sysv* 14552 + Plexus systems, m68k-plexus-sysv* 14553 + Commodore systems, m68k-cbm-sysv* 14554 + Citicorp TTI, m68k-tti-* 14555 + Unos, m68k-crds-unos* 14556 + Concurrent RTU, m68k-ccur-rtu* 14557 + Linux a.out, m68k-*-linux*aout* 14558 + Linux libc1, m68k-*-linux*libc1* 14559 + pSOS, m68k-*-psos* 14560 * MIPS 14561 + Generic ECOFF, mips*-*-ecoff* 14562 + SINIX, mips-sni-sysv4 14563 + Orion RTEMS, mips64orion-*-rtems* 14564 * National Semiconductor 32000 14565 + OpenBSD, ns32k-*-openbsd* 14566 * POWER (aka RS/6000) and PowerPC 14567 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 14568 + Bull BOSX, rs6000-bull-bosx 14569 + Generic Mach, rs6000-*-mach* 14570 + Generic SysV, powerpc*-*-sysv* 14571 + Linux libc1, powerpc*-*-linux*libc1* 14572 * Sun SPARC 14573 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 14574 sparclite-*-aout*, and sparc86x-*-aout* 14575 + NetBSD a.out, sparc-*-netbsd*aout* 14576 + Generic BSD, sparc-*-bsd* 14577 + ChorusOS, sparc-*-chorusos* 14578 + Linux a.out, sparc-*-linux*aout* 14579 + Linux libc1, sparc-*-linux*libc1* 14580 + LynxOS, sparc-*-lynxos* 14581 + Solaris on HAL hardware, sparc-hal-solaris2* 14582 + SunOS versions 3 and 4, sparc-*-sunos[34]* 14583 * NEC V850 14584 + RTEMS, v850-*-rtems* 14585 * VAX 14586 + VMS, vax-*-vms* 14587 14588 Documentation improvements 14589 14590 Other significant improvements 14591 14592 * Almost all front-end dependencies in the compiler have been 14593 separated out into a set of language hooks. This should make adding 14594 a new front end clearer and easier. 14595 * One effect of removing the separate preprocessor is a small 14596 increase in the robustness of the compiler in general, and the 14597 maintainability of target descriptions. Previously target-specific 14598 built-in macros and others, such as __FAST_MATH__, had to be 14599 handled with so-called specs that were hard to maintain. Often they 14600 would fail to behave properly when conflicting options were 14601 supplied on the command line, and define macros in the user's 14602 namespace even when strict ISO compliance was requested. 14603 Integrating the preprocessor has cleanly solved these issues. 14604 * The Makefile suite now supports redirection of make install by 14605 means of the variable DESTDIR. 14606 __________________________________________________________________ 14607 14608 GCC 3.3 14609 14610 Detailed release notes for the GCC 3.3 release follow. 14611 14612 Bug Fixes 14613 14614 bootstrap failures 14615 14616 * [8]10140 cross compiler build failures: missing __mempcpy (DUP: 14617 [9]10198,[10]10338) 14618 14619 Internal compiler errors (multi-platform) 14620 14621 * [11]3581 large string causes segmentation fault in cc1 14622 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 14623 * [13]5533 (c++) ICE when processing std::accumulate(begin, end, 14624 init, invalid_op) 14625 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 14626 * [15]6412 (c++) ICE in retrieve_specialization 14627 * [16]6620 (c++) partial template specialization causes an ICE 14628 (segmentation fault) 14629 * [17]6663 (c++) ICE with attribute aligned 14630 * [18]7068 ICE with incomplete types 14631 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation 14632 * [20]7647 (c++) ICE when data member has the name of the enclosing 14633 class 14634 * [21]7675 ICE in fixup_var_refs_1 14635 * [22]7718 'complex' template instantiation causes ICE 14636 * [23]8116 (c++) ICE in member template function 14637 * [24]8358 (ada) Ada compiler accesses freed memory, crashes 14638 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 14639 fault 14640 * [26]8564 (c++) ICE in find_function_data, in function.c 14641 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 14642 * [28]8766 (c++) ICE after failed initialization of static template 14643 variable 14644 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c 14645 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 14646 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class 14647 definition 14648 * [32]9216 (c++) ICE on missing template parameter 14649 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c 14650 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 14651 loop 14652 * [35]9429 (c++) ICE in template instantiation with a pointered new 14653 operator 14654 * [36]9516 Internal error when using a big array 14655 * [37]9600 (c++) ICE with typedefs in template class 14656 * [38]9629 (c++) virtual inheritance segfault 14657 * [39]9672 (c++) ICE: Error reporting routines re-entered 14658 * [40]9749 (c++) ICE in write_expression on invalid function 14659 prototype 14660 * [41]9794 (fortran) ICE: floating point exception during constant 14661 folding 14662 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE 14663 * [43]9916 (c++) ICE with noreturn function in ?: statement 14664 * [44]9936 ICE with local function and variable-length 2d array 14665 * [45]10262 (c++) cc1plus crashes with large generated code 14666 * [46]10278 (c++) ICE in parser for invalid code 14667 * [47]10446 (c++) ICE on definition of nonexistent member function of 14668 nested class in a class template 14669 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable 14670 declaration 14671 * [49]10506 (c++) ICE in build_new at cp/init.c with 14672 -fkeep-inline-functions and multiple inheritance 14673 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the 14674 precision of the declared type 14675 14676 Optimization bugs 14677 14678 * [51]2001 Inordinately long compile times in reload CSE regs 14679 * [52]2391 Exponential compilation time explosion in combine 14680 * [53]2960 Duplicate loop conditions even with -Os 14681 * [54]4046 redundant conditional branch 14682 * [55]6405 Loop-unrolling related performance regressions 14683 * [56]6798 very long compile time with large case-statement 14684 * [57]6871 const objects shouldn't be moved to .bss 14685 * [58]6909 problem w/ -Os on modified loop-2c.c test case 14686 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of 14687 non-void function'' warning 14688 * [60]7642 optimization problem with signbit() 14689 * [61]8634 incorrect code for inlining of memcpy under -O2 14690 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as 14691 regular function call 14692 14693 C front end 14694 14695 * [63]2161 long if-else cascade overflows parser stack 14696 * [64]4319 short accepted on typedef'd char 14697 * [65]8602 incorrect line numbers in warning messages when using 14698 inline functions 14699 * [66]9177 -fdump-translation-unit: C front end deletes function_decl 14700 AST nodes and breaks debugging dumps 14701 * [67]9853 miscompilation of non-constant structure initializer 14702 14703 c++ compiler and library 14704 14705 * [68]45 legal template specialization code is rejected (DUP: 14706 [69]3784) 14707 * [70]764 lookup failure: friend operator and dereferencing a pointer 14708 and templates (DUP: [71]5116) 14709 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP: 14710 2863) 14711 * [73]3663 G++ doesn't check access control during template 14712 instantiation 14713 * [74]3797 gcc fails to emit explicit specialization of a template 14714 member 14715 * [75]3948 Two destructors are called when no copy destructor is 14716 defined (ABI change) 14717 * [76]4137 Conversion operator within template is not accepted 14718 * [77]4361 bogus ambiguity taking the address of a member template 14719 * [78]4802 g++ accepts illegal template code (access to private 14720 member; DUP: [79]5837) 14721 * [80]4803 inline function is used but never defined, and g++ does 14722 not object 14723 * [81]5094 Partial specialization cannot be friend? 14724 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 14725 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 14726 time 14727 * [84]7015 certain __asm__ constructs rejected 14728 * [85]7086 compile time regression (quadratic behavior in 14729 fixup_var_refs) 14730 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and 14731 std::abort 14732 * [87]7247 copy constructor missing when inlining enabled (invalid 14733 optimization?) 14734 * [88]7441 string array initialization compilation time regression 14735 from seconds to minutes 14736 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong 14737 * [90]7804 bad printing of floating point constant in warning message 14738 * [91]8099 Friend classes and template specializations 14739 * [92]8117 member function pointers and multiple inheritance 14740 * [93]8205 using declaration and multiple inheritance 14741 * [94]8645 unnecessary non-zero checks in stl_tree.h 14742 * [95]8724 explicit destructor call for incomplete class allowed 14743 * [96]8805 compile time regression with many member variables 14744 * [97]8691 -O3 and -fno-implicit-templates are incompatible 14745 * [98]8700 unhelpful error message for binding temp to reference 14746 * [99]8724 explicit destructor call for incomplete class allowed 14747 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems 14748 * [101]9016 Failure to consistently constant fold "constant" C++ 14749 objects 14750 * [102]9053 g++ confused about ambiguity of overloaded function 14751 templates 14752 * [103]9152 undefined virtual thunks 14753 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out 14754 * [105]9297 data corruption due to codegen bug (when copying.) 14755 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken 14756 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 14757 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in 14758 local classes 14759 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters 14760 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439) 14761 * [112]9474 GCC freezes in compiling a weird code mixing <iostream> 14762 and <iostream.h> 14763 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1) 14764 [114][DR 231] 14765 * [115]9555 ostream inserters fail to set badbit on exception 14766 * [116]9561 ostream inserters rethrow exception of wrong type 14767 * [117]9563 ostream::sentry returns true after a failed preparation 14768 * [118]9582 one-definition rule violation in std::allocator 14769 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors 14770 * [120]9683 bug in initialization chains for static const variables 14771 from template classes 14772 * [121]9791 -Woverloaded-virtual reports hiding of destructor 14773 * [122]9817 collate::compare doesn't handle nul characters 14774 * [123]9825 filebuf::sputbackc breaks sbumpc 14775 * [124]9826 operator>>(basic_istream, basic_string) fails to compile 14776 with custom traits 14777 * [125]9924 Multiple using statements for builtin functions not 14778 allowed 14779 * [126]9946 destructor is not called for temporary object 14780 * [127]9964 filebuf::close() sometimes fails to close file 14781 * [128]9988 filebuf::overflow writes EOF to file 14782 * [129]10033 optimization breaks polymorphic references w/ typeid 14783 operator 14784 * [130]10097 filebuf::underflow drops characters 14785 * [131]10132 filebuf destructor can throw exceptions 14786 * [132]10180 gcc fails to warn about non-inlined function 14787 * [133]10199 method parametrized by template does not work everywhere 14788 * [134]10300 use of array-new (nothrow) in segfaults on NULL return 14789 * [135]10427 Stack corruption with variable-length automatic arrays 14790 and virtual destructors 14791 * [136]10503 Compilation never stops in fixed_type_or_null 14792 14793 Objective-C 14794 14795 * [137]5956 selectors aren't matched properly when added to the 14796 selector table 14797 14798 Fortran compiler and library 14799 14800 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't 14801 detect 14802 * [139]3924 g77 generates code that is rejected by GAS if COFF debug 14803 info requested 14804 * [140]5634 doc: explain that configure --prefix=~/... does not work 14805 * [141]6367 multiple repeat counts confuse namelist read into array 14806 * [142]6491 Logical operations error on logicals when using 14807 -fugly-logint 14808 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C" 14809 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 14810 on irix6.5 14811 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 14812 assume a direct access file 14813 * [146]7278 g77 "bug"; the executable misbehaves (with -O2 14814 -fno-automatic) 14815 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows 14816 * [148]7388 Incorrect output with 0-based array of characters 14817 * [149]8587 Double complex zero ** double precision number -> NaN 14818 instead of zero 14819 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 14820 unknown register name line-length-none 14821 * [151]10197 Direct access files not unformatted by default 14822 14823 Java compiler and library 14824 14825 * [152]6005 gcj fails to build rhug on alpha 14826 * [153]6389 System.getProperty("") should always throw an 14827 IllegalArgumentException 14828 * [154]6576 java.util.ResourceBundle.getResource ignores locale 14829 * [155]6652 new java.io.File("").getCanonicalFile() throws exception 14830 * [156]7060 getMethod() doesn't search super interface 14831 * [157]7073 bytecode interpreter gives wrong answer for interface 14832 getSuperclass() 14833 * [158]7180 possible bug in 14834 javax.naming.spi.NamingManager.getPlusPath() 14835 * [159]7416 java.security startup refs "GNU libgcj.security" 14836 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent 14837 env (DUP: [161]7578) 14838 * [162]7611 Internal error while compiling libjava with -O 14839 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry 14840 * [164]7766 ZipInputStream.available returns 0 immediately after 14841 construction 14842 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 14843 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 14844 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens' 14845 * [168]8234 ZipInputStream chokes when InputStream.read() returns 14846 small chunks 14847 * [169]8415 reflection bug: exception info for Method 14848 * [170]8481 java.Random.nextInt(int) may return negative 14849 * [171]8593 Error reading GZIPped files with BufferedReader 14850 * [172]8759 java.beans.Introspector has no flushCaches() or 14851 flushFromCaches() methods 14852 * [173]8997 spin() calls Thread.sleep 14853 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 14854 instead of the root content of C: 14855 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns 14856 wrong return codes 14857 * [176]9271 Severe bias in java.security.SecureRandom 14858 14859 Ada compiler and library 14860 14861 * [177]6767 make gnatlib-shared fails on -laddr2line 14862 * [178]9911 gnatmake fails to link when GCC configured with 14863 --with-sjlj-exceptions=yes 14864 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled 14865 * [180]10546 Ada tasking not working on Red Hat 9 14866 14867 preprocessor 14868 14869 * [181]7029 preprocessor should ignore #warning with -M 14870 14871 ARM-specific 14872 14873 * [182]2903 [arm] Optimization bug with long long arithmetic 14874 * [183]7873 arm-linux-gcc fails when assigning address to a bit field 14875 14876 FreeBSD-specific 14877 14878 * [184]7680 float functions undefined in math.h/cmath with #define 14879 _XOPEN_SOURCE 14880 14881 HP-UX or HP-PA-specific 14882 14883 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 14884 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 14885 fputc_unlocked 14886 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 14887 14888 m68hc11-specific 14889 14890 * [188]6744 Bad assembler code generated: reference to pseudo 14891 register z 14892 * [189]7361 Internal compiler error in reload_cse_simplify_operands, 14893 in reload1.c 14894 14895 MIPS-specific 14896 14897 * [190]9496 [mips-linux] bug in optimizer? 14898 14899 PowerPC-specific 14900 14901 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 14902 space 14903 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux 14904 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg 14905 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c 14906 14907 SPARC-specific 14908 14909 * [195]10267 (documentation) Wrong build instructions for 14910 *-*-solaris2* 14911 14912 x86-specific (Intel/AMD) 14913 14914 * [196]7916 ICE in instantiate_virtual_register_1 14915 * [197]7926 (c++) i486 instructions in header files make c++ programs 14916 crash on i386 14917 * [198]8555 ICE in gen_split_1231 14918 * [199]8994 ICE with -O -march=pentium4 14919 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs 14920 * [201]9806 ICE in inline assembly with -fPIC flag 14921 * [202]10077 gcc -msse2 generates movd to move dwords between xmm 14922 regs 14923 * [203]10233 64-bit comparison only comparing bottom 32-bits 14924 * [204]10286 type-punning doesn't work with __m64 and -O 14925 * [205]10308 [x86] ICE with -O -fgcse or -O2 14926 __________________________________________________________________ 14927 14928 GCC 3.3.1 14929 14930 Bug Fixes 14931 14932 This section lists the problem reports (PRs) from GCC's bug tracking 14933 system that are known to be fixed in the 3.3.1 release. This list might 14934 not be complete (that is, it is possible that some PRs that have been 14935 fixed are not listed here). 14936 14937 Bootstrap failures 14938 14939 * [206]11272 [Solaris] make bootstrap fails while building libstdc++ 14940 14941 Internal compiler errors (multi-platform) 14942 14943 * [207]5754 ICE on invalid nested template class 14944 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 14945 and --enable-checking 14946 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c 14947 * [210]7053 (c++) ICE when declaring a function already defined as a 14948 friend method of a template class 14949 * [211]8164 (c++) ICE when using different const expressions as 14950 template parameter 14951 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c 14952 * [213]9559 (c++) ICE with invalid initialization of a static const 14953 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 14954 when redeclaring a static member variable 14955 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in 14956 dwarfout.c with -g -O -finline-functions 14957 * [216]10432 (c++) ICE in poplevel, in cp/decl.c 14958 * [217]10475 ICE in subreg_highpart_offset for code with long long 14959 * [218]10635 (c++) ICE when dereferencing an incomplete type casted 14960 from a void pointer 14961 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 14962 instantiating static member variables 14963 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets 14964 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 14965 * [222]10796 (c++) ICE when defining an enum with two values: -1 and 14966 MAX_INT_64BIT 14967 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 14968 sched.c 14969 * [224]10939 (c++) ICE with template code 14970 * [225]10956 (c++) ICE when specializing a template member function 14971 of a template class, in tsubst, in cp/pt.c 14972 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*() 14973 defined) 14974 * [227]11059 (c++) ICE with empty union 14975 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 14976 -O2 -fnon-call-exceptions 14977 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type 14978 * [230]11149 (c++) ICE on error when instantiation with call function 14979 of a base type 14980 * [231]11228 (c++) ICE on new-expression using array operator new and 14981 default-initialization 14982 * [232]11282 (c++) Infinite memory usage after syntax error 14983 * [233]11301 (fortran) ICE with -fno-globals 14984 * [234]11308 (c++) ICE when using an enum type name as if it were a 14985 class or namespace 14986 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from 14987 an empty struct 14988 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR 14989 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 14990 template member functions 14991 14992 Optimization bugs 14993 14994 * [238]11198 -O2 -frename-registers generates wrong code (aliasing 14995 problem) 14996 * [239]11304 Wrong code production with -fomit-frame-pointer 14997 * [240]11381 volatile memory access optimized away 14998 * [241]11536 [strength-reduce] -O2 optimization produces wrong code 14999 * [242]11557 constant folding bug generates wrong code 15000 15001 C front end 15002 15003 * [243]5897 No warning for statement after return 15004 * [244]11279 DWARF-2 output mishandles large enums 15005 15006 Preprocessor bugs 15007 15008 * [245]11022 no warning for non-compatible macro redefinition 15009 15010 C++ compiler and library 15011 15012 * [246]2330 static_cast<>() to a private base is allowed 15013 * [247]5388 Incorrect message "operands to ?: have different types" 15014 * [248]5390 Libiberty fails to demangle multi-digit template 15015 parameters 15016 * [249]7877 Incorrect parameter passing to specializations of member 15017 function templates 15018 * [250]9393 Anonymous namespaces and compiling the same file twice 15019 * [251]10032 -pedantic converts some errors to warnings 15020 * [252]10468 const typeof(x) is non-const, but only in templates 15021 * [253]10527 confused error message with "new int()" parameter 15022 initializer 15023 * [254]10679 parameter MIN_INLINE_INSNS is not honored 15024 * [255]10682 gcc chokes on a typedef for an enum inside a class 15025 template 15026 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 15027 0. 15028 * [257]10845 template member function (with nested template as 15029 parameter) cannot be called anymore if another unrelated template 15030 member function is defined 15031 * [258]10849 Cannot define an out-of-class specialization of a 15032 private nested template class 15033 * [259]10888 Suppress -Winline warnings for system headers 15034 * [260]10929 -Winline warns about functions for which no definition 15035 is visible 15036 * [261]10931 valid conversion static_cast<const unsigned 15037 int&>(lvalue-of-type-int) is rejected 15038 * [262]10940 Bad code with explicit specialization 15039 * [263]10968 If member function implicitly instantiated, explicit 15040 instantiation of class fails to instantiate it 15041 * [264]10990 Cannot convert with dynamic_cast<> to a private base 15042 class from within a member function 15043 * [265]11039 Bad interaction between implicit typename deprecation 15044 and friendship 15045 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say 15046 "__unused__" instead 15047 * [267]11095 C++ iostream manipulator causes segfault when called 15048 with negative argument 15049 * [268]11098 g++ doesn't emit complete debugging information for 15050 local variables in destructors 15051 * [269]11137 GNU/Linux shared library constructors not called unless 15052 there's one global object 15053 * [270]11154 spurious ambiguity report for template class 15054 specialization 15055 * [271]11329 Compiler cannot find user defined implicit typecast 15056 * [272]11332 Spurious error with casts in ?: expression 15057 * [273]11431 static_cast behavior with subclasses when default 15058 constructor available 15059 * [274]11528 money_get facet does not accept "$.00" as valid 15060 * [275]11546 Type lookup problems in out-of-line definition of a 15061 class doubly nested from a template class 15062 * [276]11567 C++ code containing templated member function with same 15063 name as pure virtual member function results in linking failure 15064 * [277]11645 Failure to deal with using and private inheritance 15065 15066 Java compiler and library 15067 15068 * [278]5179 Qualified static field access doesn't initialize its 15069 class 15070 * [279]8204 gcj -O2 to native reorders certain instructions 15071 improperly 15072 * [280]10838 java.io.ObjectInputStream syntax error 15073 * [281]10886 The RMI registry that comes with GCJ does not work 15074 correctly 15075 * [282]11349 JNDI URL context factories not located correctly 15076 15077 x86-specific (Intel/AMD) 15078 15079 * [283]4823 ICE on inline assembly code 15080 * [284]8878 miscompilation with -O and SSE 15081 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3 15082 -masm=intel 15083 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 15084 in reload1.c 15085 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 15086 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source 15087 * [289]11044 [x86] out of range loop instructions for FP code on K6 15088 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE 15089 built-ins 15090 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 15091 is used 15092 15093 SPARC- or Solaris- specific 15094 15095 * [292]9362 solaris 'as' dies when fed .s and "-gstabs" 15096 * [293]10142 [SPARC64] gcc produces wrong code when passing 15097 structures by value 15098 * [294]10663 New configure check aborts with Sun tools. 15099 * [295]10835 combinatorial explosion in scheduler on HyperSPARC 15100 * [296]10876 ICE in calculate_giv_inc when building KDE 15101 * [297]10955 wrong code at -O3 for structure argument in context of 15102 structure return 15103 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25 15104 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 15105 Linux kernel 15106 15107 ia64 specific 15108 15109 * [300]10907 gcc violates the ia64 ABI (GP must be preserved) 15110 * [301]11320 scheduler bug (in machine depended reorganization pass) 15111 * [302]11599 bug with conditional and __builtin_prefetch 15112 15113 PowerPC specific 15114 15115 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 15116 during loop) 15117 * [304]10871 error in rs6000_stack_info save_size computation 15118 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 15119 cures it 15120 15121 m68k-specific 15122 15123 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx 15124 * [307]10557 [m68k] ICE in subreg_offset_representable_p 15125 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p 15126 15127 ARM-specific 15128 15129 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for 15130 functions with __attribute__ ((interrupt ("IRQ"))) 15131 * [310]10842 [arm] Clobbered link register is copied to pc under 15132 certain circumstances 15133 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes 15134 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 15135 (3.4) 15136 15137 MIPS-specific 15138 15139 * [313]11084 ICE in propagate_one_insn, in flow.c 15140 15141 SH-specific 15142 15143 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf 15144 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 15145 * [316]11096 i686-linux to sh-linux cross compiler fails to compile 15146 C++ files 15147 15148 GNU/Linux (or Hurd?) specific 15149 15150 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 15151 15152 UnixWare specific 15153 15154 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 15155 7.1.1 15156 15157 Cygwin (or mingw) specific 15158 15159 * [319]5287 ICE with dllimport attribute 15160 * [320]10148 [MingW/CygWin] Compiler dumps core 15161 15162 DJGPP specific 15163 15164 * [321]8787 GCC fails to emit .intel_syntax when invoked with 15165 -masm=intel on DJGPP 15166 15167 Darwin (and MacOS X) specific 15168 15169 * [322]10900 trampolines crash 15170 15171 Documentation 15172 15173 * [323]1607 (c++) Format attributes on methods undocumented 15174 * [324]4252 Invalid option `-fdump-translation-unit' 15175 * [325]4490 Clarify restrictions on -m96bit-long-double, 15176 -m128bit-long-double 15177 * [326]10355 document an issue with regparm attribute on some systems 15178 (e.g. Solaris) 15179 * [327]10726 (fortran) Documentation for function "IDate Intrinsic 15180 (Unix)" is wrong 15181 * [328]10805 document bug in old version of Sun assembler 15182 * [329]10815 warn against GNU binutils on AIX 15183 * [330]10877 document need for newer binutils on i?86-*-linux-gnu 15184 * [331]11280 Manual incorrect with respect to -freorder-blocks 15185 * [332]11466 Document -mlittle-endian and its restrictions for the 15186 sparc64 port 15187 15188 Testsuite bugs (compiler itself is not affected) 15189 15190 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 15191 report failure 15192 * [334]10810 gcc-3.3 fails make check: buffer overrun in 15193 test_demangle.c 15194 __________________________________________________________________ 15195 15196 GCC 3.3.2 15197 15198 Bug Fixes 15199 15200 This section lists the problem reports (PRs) from GCC's bug tracker 15201 that are known to be fixed in the 3.3.2 release. This list might not be 15202 complete (that is, it is possible that some PRs that have been fixed 15203 are not listed here). 15204 15205 Bootstrap failures and problems 15206 15207 * [335]8336 [SCO5] bootstrap config still tries to use COFF options 15208 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 15209 --enable-threads=posix 15210 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 15211 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 15212 7.1.1) 15213 * [339]11687 xstormy16-elf build fails in libf2c 15214 * [340]12263 [SGI IRIX] bootstrap fails during compile of 15215 libf2c/libI77/backspace.c 15216 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9 15217 fix-header processing) 15218 15219 Internal compiler errors (multi-platform) 15220 15221 * [342]7277 Casting integers to vector types causes ICE 15222 * [343]7939 (c++) ICE on invalid function template specialization 15223 * [344]11063 (c++) ICE on parsing initialization list of const array 15224 member 15225 * [345]11207 ICE with negative index in array element designator 15226 * [346]11522 (fortran) g77 dwarf-2 ICE in 15227 add_abstract_origin_attribute 15228 * [347]11595 (c++) ICE on duplicate label definition 15229 * [348]11646 (c++) ICE in commit_one_edge_insertion with 15230 -fnon-call-exceptions -fgcse -O 15231 * [349]11665 ICE in struct initializer when taking address 15232 * [350]11852 (c++) ICE with bad struct initializer. 15233 * [351]11878 (c++) ICE in cp_expr_size 15234 * [352]11883 ICE with any -O on mercury-generated C code 15235 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 15236 cp/typeck2.c when applying typeid operator to template template 15237 parameter 15238 * [354]12146 ICE in lookup_template_function, in cp/pt.c 15239 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions 15240 -fno-gcse -O2 15241 * [356]12369 (c++) ICE with templates and friends 15242 * [357]12446 ICE in emit_move_insn on complicated array reference 15243 * [358]12510 ICE in final_scan_insn 15244 * [359]12544 ICE with large parameters used in nested functions 15245 15246 C and optimization bugs 15247 15248 * [360]9862 spurious warnings with -W -finline-functions 15249 * [361]10962 lookup_field is a linear search on a linked list (can be 15250 slow if large struct) 15251 * [362]11370 -Wunreachable-code gives false complaints 15252 * [363]11637 invalid assembly with -fnon-call-exceptions 15253 * [364]11885 Problem with bitfields in packed structs 15254 * [365]12082 Inappropriate unreachable code warnings 15255 * [366]12180 Inline optimization fails for variadic function 15256 * [367]12340 loop unroller + gcse produces wrong code 15257 15258 C++ compiler and library 15259 15260 * [368]3907 nested template parameter collides with member name 15261 * [369]5293 confusing message when binding a temporary to a reference 15262 * [370]5296 [DR115] Pointers to functions and to template functions 15263 behave differently in deduction 15264 * [371]7939 ICE on function template specialization 15265 * [372]8656 Unable to assign function with __attribute__ and pointer 15266 return type to an appropriate variable 15267 * [373]10147 Confusing error message for invalid template function 15268 argument 15269 * [374]11400 std::search_n() makes assumptions about Size parameter 15270 * [375]11409 issues with using declarations, overloading, and 15271 built-in functions 15272 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 15273 multiple bits in mask 15274 * [377]11786 operator() call on variable in other namespace not 15275 recognized 15276 * [378]11867 static_cast ignores ambiguity 15277 * [379]11928 bug with conversion operators that are typedefs 15278 * [380]12114 Uninitialized memory accessed in dtor 15279 * [381]12163 static_cast + explicit constructor regression 15280 * [382]12181 Wrong code with comma operator and c++ 15281 * [383]12236 regparm and fastcall messes up parameters 15282 * [384]12266 incorrect instantiation of unneeded template during 15283 overload resolution 15284 * [385]12296 istream::peek() doesn't set eofbit 15285 * [386]12298 [sjlj exceptions] Stack unwind destroys 15286 not-yet-constructed object 15287 * [387]12369 ICE with templates and friends 15288 * [388]12337 apparently infinite loop in g++ 15289 * [389]12344 stdcall attribute ignored if function returns a pointer 15290 * [390]12451 missing(late) class forward declaration in cxxabi.h 15291 * [391]12486 g++ accepts invalid use of a qualified name 15292 15293 x86 specific (Intel/AMD) 15294 15295 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX 15296 builtins 15297 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 15298 -O2 15299 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture 15300 * [395]12116 [k6] Invalid assembly output values with X-MAME code 15301 * [396]12070 ICE converting between double and long double with 15302 -msoft-float 15303 15304 ia64-specific 15305 15306 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 15307 * [398]11535 __builtin_return_address may not work on ia64 15308 * [399]11693 [ia64] ICE in gen_nop_type 15309 * [400]12224 [ia64] Thread-local storage doesn't work 15310 15311 PowerPC-specific 15312 15313 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 15314 kernel 15315 * [402]11319 loop miscompiled on ppc32 15316 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code 15317 15318 SPARC-specific 15319 15320 * [404]11662 wrong code for expr. with cast to long long and 15321 exclusive or 15322 * [405]11965 invalid assembler code for a shift < 32 operation 15323 * [406]12301 (c++) stack corruption when a returned expression throws 15324 an exception 15325 15326 Alpha-specific 15327 15328 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of 15329 kernel 2.4.22-pre8 15330 15331 HPUX-specific 15332 15333 * [408]11313 problem with #pragma weak and static inline functions 15334 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore? 15335 15336 Solaris specific 15337 15338 * [410]12166 Profiled programs crash if PROFDIR is set 15339 15340 Solaris-x86 specific 15341 15342 * [411]12101 i386 Solaris no longer works with GNU as? 15343 15344 Miscellaneous embedded target-specific bugs 15345 15346 * [412]10988 [m32r-elf] wrong blockmove code with -O3 15347 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 15348 -O2 15349 * [414]11902 [sh4] spec file improperly inserts rpath even when none 15350 needed 15351 * [415]11903 [sh4] -pthread fails to link due to error in spec file 15352 on sh4 15353 __________________________________________________________________ 15354 15355 GCC 3.3.3 15356 15357 Minor features 15358 15359 In addition to the bug fixes documented below, this release contains 15360 few minor features such as: 15361 * Support for --with-sysroot 15362 * Support for automatic detection of executable stacks 15363 * Support for SSE3 instructions 15364 * Support for thread local storage debugging under GDB on S390 15365 15366 Bug Fixes 15367 15368 This section lists the problem reports (PRs) from GCC's bug tracker 15369 that are known to be fixed in the 3.3.3 release. This list might not be 15370 complete (that is, it is possible that some PRs that have been fixed 15371 are not listed here). 15372 15373 Bootstrap failures and issues 15374 15375 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 15376 * [417]12399 boehm-gc fails (when building a cross compiler): libtool 15377 unable to infer tagged configuration 15378 * [418]13068 mklibgcc.in doesn't handle multi-level multilib 15379 subdirectories properly 15380 15381 Internal compiler errors (multi-platform) 15382 15383 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to 15384 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 15385 * [420]10555 (c++) ICE on undefined template argument 15386 * [421]10706 (c++) ICE in mangle_class_name_for_template 15387 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops 15388 active 15389 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c 15390 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0 15391 * [425]12632 (fortran) -fbounds-check ICE 15392 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc 15393 3.3.2 15394 * [427]12726 (c++) ICE (segfault) on trivial code 15395 * [428]12890 (c++) ICE on compilation of class with throwing method 15396 * [429]12900 (c++) ICE in rtl_verify_flow_info_1 15397 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 15398 correct code with -O2 -fno-force-mem 15399 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive 15400 template 15401 * [432]13318 ICE: floating point exception in the loop optimizer 15402 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 15403 except.c 15404 * [434]13574 (c++) invalid array default initializer in class lets 15405 gcc consume all memory and die 15406 * [435]13475 ICE on SIMD variables with partial value initialization 15407 * [436]13797 (c++) ICE on invalid template parameter 15408 * [437]13824 (java) gcj SEGV with simple .java program 15409 15410 C and optimization bugs 15411 15412 * [438]8776 loop invariants are not removed (most likely) 15413 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 15414 strncmp by memcmp 15415 * [440]11350 undefined labels with -Os -fPIC 15416 * [441]12826 Optimizer removes reference through volatile pointer 15417 * [442]12500 stabs debug info: void no longer a predefined / builtin 15418 type 15419 * [443]12941 builtin-bitops-1.c miscompilation (latent bug) 15420 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix 15421 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled 15422 * [446]13507 spurious printf format warning 15423 * [447]13382 Type information for const pointer disappears during 15424 optimization. 15425 * [448]13394 noreturn attribute ignored on recursive invokation 15426 * [449]13400 Compiled code crashes storing to read-only location 15427 * [450]13521 Endless loop in calculate_global_regs_live 15428 15429 C++ compiler and library 15430 15431 Some of the bug fixes in this list were made to implement decisions 15432 that the ISO C++ standards committee has made concerning several defect 15433 reports (DRs). Links in the list below point to detailed discussion of 15434 the relevant defect report. 15435 * [451]2094 unimplemented: use of `ptrmem_cst' in template type 15436 unification 15437 * [452]2294 using declaration confusion 15438 * [453]5050 template instantiation depth exceeds limit: recursion 15439 problem? 15440 * [454]9371 Bad exception handling in 15441 i/ostream::operator>>/<<(streambuf*) 15442 * [455]9546 bad exception handling in ostream members 15443 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the 15444 face of unknown locales 15445 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work 15446 * [459]10095 istream::operator>>(int&) sets ios::badbit when 15447 ios::failbit is set. 15448 * [460]11554 Warning about reordering of initializers doesn't mention 15449 location of constructor 15450 * [461]12297 istream::sentry::sentry() handles eof() incorrectly. 15451 * [462]12352 Exception safety problems in src/localename.cc 15452 * [463]12438 Memory leak in locale::combine() 15453 * [464]12540 Memory leak in locale::locale(const char*) 15454 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented 15455 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented 15456 * [470]12696 memory eating infinite loop in diagnostics (error 15457 recovery problem) 15458 * [471]12815 Code compiled with optimization behaves unexpectedly 15459 * [472]12862 Conflicts between typedefs/enums and namespace member 15460 declarations 15461 * [473]12926 Wrong value after assignment in initialize list using 15462 bit-fields 15463 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented 15464 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented 15465 * [478]13007 basic_streambuf::pubimbue, imbue wrong 15466 * [479]13009 Implicitly-defined assignment operator writes to wrong 15467 memory 15468 * [480]13057 regparm attribute not applied to destructor 15469 * [481]13070 -Wformat option ignored in g++ 15470 * [482]13081 forward template declarations in <complex> let inlining 15471 fail 15472 * [483]13239 Assertion does not seem to work correctly anymore 15473 * [484]13262 "xxx is private within this context" when initializing a 15474 self-contained template class 15475 * [485]13290 simple typo in concept checking for std::generate_n 15476 * [486]13323 Template code does not compile in presence of typedef 15477 * [487]13369 __verify_grouping (and __add_grouping?) not correct 15478 * [488]13371 infinite loop with packed struct and inlining 15479 * [489]13445 Template argument replacement "dereferences" a typedef 15480 * [490]13461 Fails to access protected-ctor from public constant 15481 * [491]13462 Non-standard-conforming type set::pointer 15482 * [492]13478 gcc uses wrong constructor to initialize a const 15483 reference 15484 * [493]13544 "conflicting types" for enums in different scopes 15485 * [494]13650 string::compare should not (always) use 15486 traits_type::length() 15487 * [495]13683 bogus warning about passing non-PODs through ellipsis 15488 * [496]13688 Derived class is denied access to protected base class 15489 member class 15490 * [497]13774 Member variable cleared in virtual multiple inheritance 15491 class 15492 * [498]13884 Protect sstream.tcc from extern template use 15493 15494 Java compiler and library 15495 15496 * [499]10746 [win32] garbage collection crash in GCJ 15497 15498 Objective-C compiler and library 15499 15500 * [500]11433 Crash due to dereferencing null pointer when querying 15501 protocol 15502 15503 Fortran compiler and library 15504 15505 * [501]12633 logical expression gives incorrect result with 15506 -fugly-logint option 15507 * [502]13037 [gcse-lm] g77 generates incorrect code 15508 * [503]13213 Hex constant problem when compiling with -fugly-logint 15509 and -ftypeless-boz 15510 15511 x86-specific (Intel/AMD) 15512 15513 * [504]4490 ICE with -m128bit-long-double 15514 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have 15515 `reg' in make_field_assignment, in combine.c 15516 * [506]12441 ICE: can't find a register to spill 15517 * [507]12943 array static-init failure under -fpic, -fPIC 15518 * [508]13608 Incorrect code with -O3 -ffast-math 15519 15520 PowerPC-specific 15521 15522 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of 15523 __attribute__((aligned(16))) 15524 * [510]11793 ICE in extract_insn, in recog.c (const_vector's) 15525 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in 15526 altivec.md) 15527 * [512]12537 g++ generates writeable text sections 15528 15529 SPARC-specific 15530 15531 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0 15532 -m64 15533 * [514]12865 mprotect call to make trampoline executable may fail 15534 * [515]13354 ICE in sparc_emit_set_const32 15535 15536 ARM-specific 15537 15538 * [516]10467 [arm] ICE in pre_insert_copy_insn, 15539 15540 ia64-specific 15541 15542 * [517]11226 ICE passing struct arg with two floats 15543 * [518]11227 ICE for _Complex float, _Complex long double args 15544 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64 15545 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn 15546 * Various fixes for libunwind 15547 15548 Alpha-specific 15549 15550 * [521]12654 Incorrect comparison code generated for Alpha 15551 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 15552 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 15553 15554 HPPA-specific 15555 15556 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 15557 * [525]12158 [hppa] compilation does not terminate at -O1 15558 15559 S390-specific 15560 15561 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only 15562 (1<<24)-1 possible for CLCL-Instruction 15563 15564 SH-specific 15565 15566 * [527]9365 segfault in gen_far_branch (config/sh/sh.c) 15567 * [528]10392 optimizer generates faulty array indexing 15568 * [529]11322 SH profiler outputs multiple definitions of symbol 15569 * [530]13069 gcc/config/sh/rtems.h broken 15570 * [531]13302 Putting a va_list in a struct causes seg fault 15571 * [532]13585 Incorrect optimization of call to sfunc 15572 * Fix inappropriately exported libgcc functions from the shared 15573 library 15574 15575 Other embedded target specific 15576 15577 * [533]8916 [mcore] unsigned char assign gets hosed. 15578 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c 15579 * [535]13122 [h8300] local variable gets corrupted by function call 15580 when -fomit-frame-pointer is given 15581 * [536]13256 [cris] strict_low_part mistreated in delay slots 15582 * [537]13373 [mcore] optimization with -frerun-cse-after-loop 15583 -fexpensive-optimizations produces wrong code on mcore 15584 15585 GNU HURD-specific 15586 15587 * [538]12561 gcc/config/t-gnu needs updating to work with 15588 --with-sysroot 15589 15590 Tru64 Unix specific 15591 15592 * [539]6243 testsuite fails almost all tests due to no libintl in 15593 LD_LIBRARY_PATH during test. 15594 * [540]11397 weak aliases broken on Tru64 UNIX 15595 15596 AIX-specific 15597 15598 * [541]12505 build failure due to defines of uchar in cpphash.h and 15599 sys/types.h 15600 * [542]13150 WEAK symbols not exported by collect2 15601 15602 IRIX-specific 15603 15604 * [543]12666 fixincludes problem on IRIX 6.5.19m 15605 15606 Solaris-specific 15607 15608 * [544]12969 Including sys/byteorder.h breaks configure checks 15609 15610 Testsuite problems (compiler is not affected) 15611 15612 * [545]10819 testsuite creates CR+LF on compiler version lines in 15613 test summary files 15614 * [546]11612 abi_check not finding correct libgcc_s.so.1 15615 15616 Miscellaneous 15617 15618 * [547]13211 using -###, incorrect warnings about unused linker file 15619 are produced 15620 __________________________________________________________________ 15621 15622 GCC 3.3.4 15623 15624 This is the [548]list of problem reports (PRs) from GCC's bug tracking 15625 system that are known to be fixed in the 3.3.4 release. This list might 15626 not be complete (that is, it is possible that some PRs that have been 15627 fixed are not listed here). 15628 __________________________________________________________________ 15629 15630 GCC 3.3.5 15631 15632 This is the [549]list of problem reports (PRs) from GCC's bug tracking 15633 system that are known to be fixed in the 3.3.5 release. This list might 15634 not be complete (that is, it is possible that some PRs that have been 15635 fixed are not listed here). 15636 __________________________________________________________________ 15637 15638 GCC 3.3.6 15639 15640 This is the [550]list of problem reports (PRs) from GCC's bug tracking 15641 system that are known to be fixed in the 3.3.6 release. This list might 15642 not be complete (that is, it is possible that some PRs that have been 15643 fixed are not listed here). 15644 15645 15646 For questions related to the use of GCC, please consult these web 15647 pages and the [551]GCC manuals. If that fails, the 15648 [552]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15649 web pages and the development of GCC are welcome on our developer 15650 list at [553]gcc (a] gcc.gnu.org. All of [554]our lists have public 15651 archives. 15652 15653 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and 15654 distribution of this entire article is permitted in any medium, 15655 provided this notice is preserved. 15656 15657 These pages are [556]maintained by the GCC team. Last modified 15658 2019-11-28[557]. 15659 15660 References 15661 15662 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 15663 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 15664 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 15665 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 15666 5. http://gcc.gnu.org/news/dfa.html 15667 6. http://gcc.gnu.org/c99status.html 15668 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 15669 8. https://gcc.gnu.org/PR10140 15670 9. https://gcc.gnu.org/PR10198 15671 10. https://gcc.gnu.org/PR10338 15672 11. https://gcc.gnu.org/PR3581 15673 12. https://gcc.gnu.org/PR4382 15674 13. https://gcc.gnu.org/PR5533 15675 14. https://gcc.gnu.org/PR6387 15676 15. https://gcc.gnu.org/PR6412 15677 16. https://gcc.gnu.org/PR6620 15678 17. https://gcc.gnu.org/PR6663 15679 18. https://gcc.gnu.org/PR7068 15680 19. https://gcc.gnu.org/PR7083 15681 20. https://gcc.gnu.org/PR7647 15682 21. https://gcc.gnu.org/PR7675 15683 22. https://gcc.gnu.org/PR7718 15684 23. https://gcc.gnu.org/PR8116 15685 24. https://gcc.gnu.org/PR8358 15686 25. https://gcc.gnu.org/PR8511 15687 26. https://gcc.gnu.org/PR8564 15688 27. https://gcc.gnu.org/PR8660 15689 28. https://gcc.gnu.org/PR8766 15690 29. https://gcc.gnu.org/PR8803 15691 30. https://gcc.gnu.org/PR8846 15692 31. https://gcc.gnu.org/PR8906 15693 32. https://gcc.gnu.org/PR9216 15694 33. https://gcc.gnu.org/PR9261 15695 34. https://gcc.gnu.org/PR9263 15696 35. https://gcc.gnu.org/PR9429 15697 36. https://gcc.gnu.org/PR9516 15698 37. https://gcc.gnu.org/PR9600 15699 38. https://gcc.gnu.org/PR9629 15700 39. https://gcc.gnu.org/PR9672 15701 40. https://gcc.gnu.org/PR9749 15702 41. https://gcc.gnu.org/PR9794 15703 42. https://gcc.gnu.org/PR9829 15704 43. https://gcc.gnu.org/PR9916 15705 44. https://gcc.gnu.org/PR9936 15706 45. https://gcc.gnu.org/PR10262 15707 46. https://gcc.gnu.org/PR10278 15708 47. https://gcc.gnu.org/PR10446 15709 48. https://gcc.gnu.org/PR10451 15710 49. https://gcc.gnu.org/PR10506 15711 50. https://gcc.gnu.org/PR10549 15712 51. https://gcc.gnu.org/PR2001 15713 52. https://gcc.gnu.org/PR2391 15714 53. https://gcc.gnu.org/PR2960 15715 54. https://gcc.gnu.org/PR4046 15716 55. https://gcc.gnu.org/PR6405 15717 56. https://gcc.gnu.org/PR6798 15718 57. https://gcc.gnu.org/PR6871 15719 58. https://gcc.gnu.org/PR6909 15720 59. https://gcc.gnu.org/PR7189 15721 60. https://gcc.gnu.org/PR7642 15722 61. https://gcc.gnu.org/PR8634 15723 62. https://gcc.gnu.org/PR8750 15724 63. https://gcc.gnu.org/PR2161 15725 64. https://gcc.gnu.org/PR4319 15726 65. https://gcc.gnu.org/PR8602 15727 66. https://gcc.gnu.org/PR9177 15728 67. https://gcc.gnu.org/PR9853 15729 68. https://gcc.gnu.org/PR45 15730 69. https://gcc.gnu.org/PR3784 15731 70. https://gcc.gnu.org/PR764 15732 71. https://gcc.gnu.org/PR5116 15733 72. https://gcc.gnu.org/PR2862 15734 73. https://gcc.gnu.org/PR3663 15735 74. https://gcc.gnu.org/PR3797 15736 75. https://gcc.gnu.org/PR3948 15737 76. https://gcc.gnu.org/PR4137 15738 77. https://gcc.gnu.org/PR4361 15739 78. https://gcc.gnu.org/PR4802 15740 79. https://gcc.gnu.org/PR5837 15741 80. https://gcc.gnu.org/PR4803 15742 81. https://gcc.gnu.org/PR5094 15743 82. https://gcc.gnu.org/PR5730 15744 83. https://gcc.gnu.org/PR6713 15745 84. https://gcc.gnu.org/PR7015 15746 85. https://gcc.gnu.org/PR7086 15747 86. https://gcc.gnu.org/PR7099 15748 87. https://gcc.gnu.org/PR7247 15749 88. https://gcc.gnu.org/PR7441 15750 89. https://gcc.gnu.org/PR7768 15751 90. https://gcc.gnu.org/PR7804 15752 91. https://gcc.gnu.org/PR8099 15753 92. https://gcc.gnu.org/PR8117 15754 93. https://gcc.gnu.org/PR8205 15755 94. https://gcc.gnu.org/PR8645 15756 95. https://gcc.gnu.org/PR8724 15757 96. https://gcc.gnu.org/PR8805 15758 97. https://gcc.gnu.org/PR8691 15759 98. https://gcc.gnu.org/PR8700 15760 99. https://gcc.gnu.org/PR8724 15761 100. https://gcc.gnu.org/PR8949 15762 101. https://gcc.gnu.org/PR9016 15763 102. https://gcc.gnu.org/PR9053 15764 103. https://gcc.gnu.org/PR9152 15765 104. https://gcc.gnu.org/PR9182 15766 105. https://gcc.gnu.org/PR9297 15767 106. https://gcc.gnu.org/PR9318 15768 107. https://gcc.gnu.org/PR9320 15769 108. https://gcc.gnu.org/PR9400 15770 109. https://gcc.gnu.org/PR9424 15771 110. https://gcc.gnu.org/PR9425 15772 111. https://gcc.gnu.org/PR9439 15773 112. https://gcc.gnu.org/PR9474 15774 113. https://gcc.gnu.org/PR9548 15775 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 15776 115. https://gcc.gnu.org/PR9555 15777 116. https://gcc.gnu.org/PR9561 15778 117. https://gcc.gnu.org/PR9563 15779 118. https://gcc.gnu.org/PR9582 15780 119. https://gcc.gnu.org/PR9622 15781 120. https://gcc.gnu.org/PR9683 15782 121. https://gcc.gnu.org/PR9791 15783 122. https://gcc.gnu.org/PR9817 15784 123. https://gcc.gnu.org/PR9825 15785 124. https://gcc.gnu.org/PR9826 15786 125. https://gcc.gnu.org/PR9924 15787 126. https://gcc.gnu.org/PR9946 15788 127. https://gcc.gnu.org/PR9964 15789 128. https://gcc.gnu.org/PR9988 15790 129. https://gcc.gnu.org/PR10033 15791 130. https://gcc.gnu.org/PR10097 15792 131. https://gcc.gnu.org/PR10132 15793 132. https://gcc.gnu.org/PR10180 15794 133. https://gcc.gnu.org/PR10199 15795 134. https://gcc.gnu.org/PR10300 15796 135. https://gcc.gnu.org/PR10427 15797 136. https://gcc.gnu.org/PR10503 15798 137. https://gcc.gnu.org/PR5956 15799 138. https://gcc.gnu.org/PR1832 15800 139. https://gcc.gnu.org/PR3924 15801 140. https://gcc.gnu.org/PR5634 15802 141. https://gcc.gnu.org/PR6367 15803 142. https://gcc.gnu.org/PR6491 15804 143. https://gcc.gnu.org/PR6742 15805 144. https://gcc.gnu.org/PR7113 15806 145. https://gcc.gnu.org/PR7236 15807 146. https://gcc.gnu.org/PR7278 15808 147. https://gcc.gnu.org/PR7384 15809 148. https://gcc.gnu.org/PR7388 15810 149. https://gcc.gnu.org/PR8587 15811 150. https://gcc.gnu.org/PR9038 15812 151. https://gcc.gnu.org/PR10197 15813 152. https://gcc.gnu.org/PR6005 15814 153. https://gcc.gnu.org/PR6389 15815 154. https://gcc.gnu.org/PR6576 15816 155. https://gcc.gnu.org/PR6652 15817 156. https://gcc.gnu.org/PR7060 15818 157. https://gcc.gnu.org/PR7073 15819 158. https://gcc.gnu.org/PR7180 15820 159. https://gcc.gnu.org/PR7416 15821 160. https://gcc.gnu.org/PR7570 15822 161. https://gcc.gnu.org/PR7578 15823 162. https://gcc.gnu.org/PR7611 15824 163. https://gcc.gnu.org/PR7709 15825 164. https://gcc.gnu.org/PR7766 15826 165. https://gcc.gnu.org/PR7785 15827 166. https://gcc.gnu.org/PR7786 15828 167. https://gcc.gnu.org/PR8142 15829 168. https://gcc.gnu.org/PR8234 15830 169. https://gcc.gnu.org/PR8415 15831 170. https://gcc.gnu.org/PR8481 15832 171. https://gcc.gnu.org/PR8593 15833 172. https://gcc.gnu.org/PR8759 15834 173. https://gcc.gnu.org/PR8997 15835 174. https://gcc.gnu.org/PR9253 15836 175. https://gcc.gnu.org/PR9254 15837 176. https://gcc.gnu.org/PR9271 15838 177. https://gcc.gnu.org/PR6767 15839 178. https://gcc.gnu.org/PR9911 15840 179. https://gcc.gnu.org/PR10020 15841 180. https://gcc.gnu.org/PR10546 15842 181. https://gcc.gnu.org/PR7029 15843 182. https://gcc.gnu.org/PR2903 15844 183. https://gcc.gnu.org/PR7873 15845 184. https://gcc.gnu.org/PR7680 15846 185. https://gcc.gnu.org/PR8705 15847 186. https://gcc.gnu.org/PR9986 15848 187. https://gcc.gnu.org/PR10056 15849 188. https://gcc.gnu.org/PR6744 15850 189. https://gcc.gnu.org/PR7361 15851 190. https://gcc.gnu.org/PR9496 15852 191. https://gcc.gnu.org/PR7067 15853 192. https://gcc.gnu.org/PR8480 15854 193. https://gcc.gnu.org/PR8784 15855 194. https://gcc.gnu.org/PR10315 15856 195. https://gcc.gnu.org/PR10267 15857 196. https://gcc.gnu.org/PR7916 15858 197. https://gcc.gnu.org/PR7926 15859 198. https://gcc.gnu.org/PR8555 15860 199. https://gcc.gnu.org/PR8994 15861 200. https://gcc.gnu.org/PR9426 15862 201. https://gcc.gnu.org/PR9806 15863 202. https://gcc.gnu.org/PR10077 15864 203. https://gcc.gnu.org/PR10233 15865 204. https://gcc.gnu.org/PR10286 15866 205. https://gcc.gnu.org/PR10308 15867 206. https://gcc.gnu.org/PR11272 15868 207. https://gcc.gnu.org/PR5754 15869 208. https://gcc.gnu.org/PR6597 15870 209. https://gcc.gnu.org/PR6949 15871 210. https://gcc.gnu.org/PR7053 15872 211. https://gcc.gnu.org/PR8164 15873 212. https://gcc.gnu.org/PR8384 15874 213. https://gcc.gnu.org/PR9559 15875 214. https://gcc.gnu.org/PR9649 15876 215. https://gcc.gnu.org/PR9864 15877 216. https://gcc.gnu.org/PR10432 15878 217. https://gcc.gnu.org/PR10475 15879 218. https://gcc.gnu.org/PR10635 15880 219. https://gcc.gnu.org/PR10661 15881 220. https://gcc.gnu.org/PR10700 15882 221. https://gcc.gnu.org/PR10712 15883 222. https://gcc.gnu.org/PR10796 15884 223. https://gcc.gnu.org/PR10890 15885 224. https://gcc.gnu.org/PR10939 15886 225. https://gcc.gnu.org/PR10956 15887 226. https://gcc.gnu.org/PR11041 15888 227. https://gcc.gnu.org/PR11059 15889 228. https://gcc.gnu.org/PR11083 15890 229. https://gcc.gnu.org/PR11105 15891 230. https://gcc.gnu.org/PR11149 15892 231. https://gcc.gnu.org/PR11228 15893 232. https://gcc.gnu.org/PR11282 15894 233. https://gcc.gnu.org/PR11301 15895 234. https://gcc.gnu.org/PR11308 15896 235. https://gcc.gnu.org/PR11473 15897 236. https://gcc.gnu.org/PR11503 15898 237. https://gcc.gnu.org/PR11513 15899 238. https://gcc.gnu.org/PR11198 15900 239. https://gcc.gnu.org/PR11304 15901 240. https://gcc.gnu.org/PR11381 15902 241. https://gcc.gnu.org/PR11536 15903 242. https://gcc.gnu.org/PR11557 15904 243. https://gcc.gnu.org/PR5897 15905 244. https://gcc.gnu.org/PR11279 15906 245. https://gcc.gnu.org/PR11022 15907 246. https://gcc.gnu.org/PR2330 15908 247. https://gcc.gnu.org/PR5388 15909 248. https://gcc.gnu.org/PR5390 15910 249. https://gcc.gnu.org/PR7877 15911 250. https://gcc.gnu.org/PR9393 15912 251. https://gcc.gnu.org/PR10032 15913 252. https://gcc.gnu.org/PR10468 15914 253. https://gcc.gnu.org/PR10527 15915 254. https://gcc.gnu.org/PR10679 15916 255. https://gcc.gnu.org/PR10682 15917 256. https://gcc.gnu.org/PR10689 15918 257. https://gcc.gnu.org/PR10845 15919 258. https://gcc.gnu.org/PR10849 15920 259. https://gcc.gnu.org/PR10888 15921 260. https://gcc.gnu.org/PR10929 15922 261. https://gcc.gnu.org/PR10931 15923 262. https://gcc.gnu.org/PR10940 15924 263. https://gcc.gnu.org/PR10968 15925 264. https://gcc.gnu.org/PR10990 15926 265. https://gcc.gnu.org/PR11039 15927 266. https://gcc.gnu.org/PR11062 15928 267. https://gcc.gnu.org/PR11095 15929 268. https://gcc.gnu.org/PR11098 15930 269. https://gcc.gnu.org/PR11137 15931 270. https://gcc.gnu.org/PR11154 15932 271. https://gcc.gnu.org/PR11329 15933 272. https://gcc.gnu.org/PR11332 15934 273. https://gcc.gnu.org/PR11431 15935 274. https://gcc.gnu.org/PR11528 15936 275. https://gcc.gnu.org/PR11546 15937 276. https://gcc.gnu.org/PR11567 15938 277. https://gcc.gnu.org/PR11645 15939 278. https://gcc.gnu.org/PR5179 15940 279. https://gcc.gnu.org/PR8204 15941 280. https://gcc.gnu.org/PR10838 15942 281. https://gcc.gnu.org/PR10886 15943 282. https://gcc.gnu.org/PR11349 15944 283. https://gcc.gnu.org/PR4823 15945 284. https://gcc.gnu.org/PR8878 15946 285. https://gcc.gnu.org/PR9815 15947 286. https://gcc.gnu.org/PR10402 15948 287. https://gcc.gnu.org/PR10504 15949 288. https://gcc.gnu.org/PR10673 15950 289. https://gcc.gnu.org/PR11044 15951 290. https://gcc.gnu.org/PR11089 15952 291. https://gcc.gnu.org/PR11420 15953 292. https://gcc.gnu.org/PR9362 15954 293. https://gcc.gnu.org/PR10142 15955 294. https://gcc.gnu.org/PR10663 15956 295. https://gcc.gnu.org/PR10835 15957 296. https://gcc.gnu.org/PR10876 15958 297. https://gcc.gnu.org/PR10955 15959 298. https://gcc.gnu.org/PR11018 15960 299. https://gcc.gnu.org/PR11556 15961 300. https://gcc.gnu.org/PR10907 15962 301. https://gcc.gnu.org/PR11320 15963 302. https://gcc.gnu.org/PR11599 15964 303. https://gcc.gnu.org/PR9745 15965 304. https://gcc.gnu.org/PR10871 15966 305. https://gcc.gnu.org/PR11440 15967 306. https://gcc.gnu.org/PR7594 15968 307. https://gcc.gnu.org/PR10557 15969 308. https://gcc.gnu.org/PR11054 15970 309. https://gcc.gnu.org/PR10834 15971 310. https://gcc.gnu.org/PR10842 15972 311. https://gcc.gnu.org/PR11052 15973 312. https://gcc.gnu.org/PR11183 15974 313. https://gcc.gnu.org/PR11084 15975 314. https://gcc.gnu.org/PR10331 15976 315. https://gcc.gnu.org/PR10413 15977 316. https://gcc.gnu.org/PR11096 15978 317. https://gcc.gnu.org/PR2873 15979 318. https://gcc.gnu.org/PR3163 15980 319. https://gcc.gnu.org/PR5287 15981 320. https://gcc.gnu.org/PR10148 15982 321. https://gcc.gnu.org/PR8787 15983 322. https://gcc.gnu.org/PR10900 15984 323. https://gcc.gnu.org/PR1607 15985 324. https://gcc.gnu.org/PR4252 15986 325. https://gcc.gnu.org/PR4490 15987 326. https://gcc.gnu.org/PR10355 15988 327. https://gcc.gnu.org/PR10726 15989 328. https://gcc.gnu.org/PR10805 15990 329. https://gcc.gnu.org/PR10815 15991 330. https://gcc.gnu.org/PR10877 15992 331. https://gcc.gnu.org/PR11280 15993 332. https://gcc.gnu.org/PR11466 15994 333. https://gcc.gnu.org/PR10737 15995 334. https://gcc.gnu.org/PR10810 15996 335. https://gcc.gnu.org/PR8336 15997 336. https://gcc.gnu.org/PR9330 15998 337. https://gcc.gnu.org/PR9631 15999 338. https://gcc.gnu.org/PR9877 16000 339. https://gcc.gnu.org/PR11687 16001 340. https://gcc.gnu.org/PR12263 16002 341. https://gcc.gnu.org/PR12490 16003 342. https://gcc.gnu.org/PR7277 16004 343. https://gcc.gnu.org/PR7939 16005 344. https://gcc.gnu.org/PR11063 16006 345. https://gcc.gnu.org/PR11207 16007 346. https://gcc.gnu.org/PR11522 16008 347. https://gcc.gnu.org/PR11595 16009 348. https://gcc.gnu.org/PR11646 16010 349. https://gcc.gnu.org/PR11665 16011 350. https://gcc.gnu.org/PR11852 16012 351. https://gcc.gnu.org/PR11878 16013 352. https://gcc.gnu.org/PR11883 16014 353. https://gcc.gnu.org/PR11991 16015 354. https://gcc.gnu.org/PR12146 16016 355. https://gcc.gnu.org/PR12215 16017 356. https://gcc.gnu.org/PR12369 16018 357. https://gcc.gnu.org/PR12446 16019 358. https://gcc.gnu.org/PR12510 16020 359. https://gcc.gnu.org/PR12544 16021 360. https://gcc.gnu.org/PR9862 16022 361. https://gcc.gnu.org/PR10962 16023 362. https://gcc.gnu.org/PR11370 16024 363. https://gcc.gnu.org/PR11637 16025 364. https://gcc.gnu.org/PR11885 16026 365. https://gcc.gnu.org/PR12082 16027 366. https://gcc.gnu.org/PR12180 16028 367. https://gcc.gnu.org/PR12340 16029 368. https://gcc.gnu.org/PR3907 16030 369. https://gcc.gnu.org/PR5293 16031 370. https://gcc.gnu.org/PR5296 16032 371. https://gcc.gnu.org/PR7939 16033 372. https://gcc.gnu.org/PR8656 16034 373. https://gcc.gnu.org/PR10147 16035 374. https://gcc.gnu.org/PR11400 16036 375. https://gcc.gnu.org/PR11409 16037 376. https://gcc.gnu.org/PR11740 16038 377. https://gcc.gnu.org/PR11786 16039 378. https://gcc.gnu.org/PR11867 16040 379. https://gcc.gnu.org/PR11928 16041 380. https://gcc.gnu.org/PR12114 16042 381. https://gcc.gnu.org/PR12163 16043 382. https://gcc.gnu.org/PR12181 16044 383. https://gcc.gnu.org/PR12236 16045 384. https://gcc.gnu.org/PR12266 16046 385. https://gcc.gnu.org/PR12296 16047 386. https://gcc.gnu.org/PR12298 16048 387. https://gcc.gnu.org/PR12369 16049 388. https://gcc.gnu.org/PR12337 16050 389. https://gcc.gnu.org/PR12344 16051 390. https://gcc.gnu.org/PR12451 16052 391. https://gcc.gnu.org/PR12486 16053 392. https://gcc.gnu.org/PR8869 16054 393. https://gcc.gnu.org/PR9786 16055 394. https://gcc.gnu.org/PR11689 16056 395. https://gcc.gnu.org/PR12116 16057 396. https://gcc.gnu.org/PR12070 16058 397. https://gcc.gnu.org/PR11184 16059 398. https://gcc.gnu.org/PR11535 16060 399. https://gcc.gnu.org/PR11693 16061 400. https://gcc.gnu.org/PR12224 16062 401. https://gcc.gnu.org/PR11087 16063 402. https://gcc.gnu.org/PR11319 16064 403. https://gcc.gnu.org/PR11949 16065 404. https://gcc.gnu.org/PR11662 16066 405. https://gcc.gnu.org/PR11965 16067 406. https://gcc.gnu.org/PR12301 16068 407. https://gcc.gnu.org/PR11717 16069 408. https://gcc.gnu.org/PR11313 16070 409. https://gcc.gnu.org/PR11712 16071 410. https://gcc.gnu.org/PR12166 16072 411. https://gcc.gnu.org/PR12101 16073 412. https://gcc.gnu.org/PR10988 16074 413. https://gcc.gnu.org/PR11805 16075 414. https://gcc.gnu.org/PR11902 16076 415. https://gcc.gnu.org/PR11903 16077 416. https://gcc.gnu.org/PR11890 16078 417. https://gcc.gnu.org/PR12399 16079 418. https://gcc.gnu.org/PR13068 16080 419. https://gcc.gnu.org/PR10060 16081 420. https://gcc.gnu.org/PR10555 16082 421. https://gcc.gnu.org/PR10706 16083 422. https://gcc.gnu.org/PR11496 16084 423. https://gcc.gnu.org/PR11741 16085 424. https://gcc.gnu.org/PR12440 16086 425. https://gcc.gnu.org/PR12632 16087 426. https://gcc.gnu.org/PR12712 16088 427. https://gcc.gnu.org/PR12726 16089 428. https://gcc.gnu.org/PR12890 16090 429. https://gcc.gnu.org/PR12900 16091 430. https://gcc.gnu.org/PR13060 16092 431. https://gcc.gnu.org/PR13289 16093 432. https://gcc.gnu.org/PR13318 16094 433. https://gcc.gnu.org/PR13392 16095 434. https://gcc.gnu.org/PR13574 16096 435. https://gcc.gnu.org/PR13475 16097 436. https://gcc.gnu.org/PR13797 16098 437. https://gcc.gnu.org/PR13824 16099 438. https://gcc.gnu.org/PR8776 16100 439. https://gcc.gnu.org/PR10339 16101 440. https://gcc.gnu.org/PR11350 16102 441. https://gcc.gnu.org/PR12826 16103 442. https://gcc.gnu.org/PR12500 16104 443. https://gcc.gnu.org/PR12941 16105 444. https://gcc.gnu.org/PR12953 16106 445. https://gcc.gnu.org/PR13041 16107 446. https://gcc.gnu.org/PR13507 16108 447. https://gcc.gnu.org/PR13382 16109 448. https://gcc.gnu.org/PR13394 16110 449. https://gcc.gnu.org/PR13400 16111 450. https://gcc.gnu.org/PR13521 16112 451. https://gcc.gnu.org/PR2094 16113 452. https://gcc.gnu.org/PR2294 16114 453. https://gcc.gnu.org/PR5050 16115 454. https://gcc.gnu.org/PR9371 16116 455. https://gcc.gnu.org/PR9546 16117 456. https://gcc.gnu.org/PR10081 16118 457. https://gcc.gnu.org/PR10093 16119 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 16120 459. https://gcc.gnu.org/PR10095 16121 460. https://gcc.gnu.org/PR11554 16122 461. https://gcc.gnu.org/PR12297 16123 462. https://gcc.gnu.org/PR12352 16124 463. https://gcc.gnu.org/PR12438 16125 464. https://gcc.gnu.org/PR12540 16126 465. https://gcc.gnu.org/PR12594 16127 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 16128 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 16129 468. https://gcc.gnu.org/PR12657 16130 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 16131 470. https://gcc.gnu.org/PR12696 16132 471. https://gcc.gnu.org/PR12815 16133 472. https://gcc.gnu.org/PR12862 16134 473. https://gcc.gnu.org/PR12926 16135 474. https://gcc.gnu.org/PR12967 16136 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 16137 476. https://gcc.gnu.org/PR12971 16138 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 16139 478. https://gcc.gnu.org/PR13007 16140 479. https://gcc.gnu.org/PR13009 16141 480. https://gcc.gnu.org/PR13057 16142 481. https://gcc.gnu.org/PR13070 16143 482. https://gcc.gnu.org/PR13081 16144 483. https://gcc.gnu.org/PR13239 16145 484. https://gcc.gnu.org/PR13262 16146 485. https://gcc.gnu.org/PR13290 16147 486. https://gcc.gnu.org/PR13323 16148 487. https://gcc.gnu.org/PR13369 16149 488. https://gcc.gnu.org/PR13371 16150 489. https://gcc.gnu.org/PR13445 16151 490. https://gcc.gnu.org/PR13461 16152 491. https://gcc.gnu.org/PR13462 16153 492. https://gcc.gnu.org/PR13478 16154 493. https://gcc.gnu.org/PR13544 16155 494. https://gcc.gnu.org/PR13650 16156 495. https://gcc.gnu.org/PR13683 16157 496. https://gcc.gnu.org/PR13688 16158 497. https://gcc.gnu.org/PR13774 16159 498. https://gcc.gnu.org/PR13884 16160 499. https://gcc.gnu.org/PR10746 16161 500. https://gcc.gnu.org/PR11433 16162 501. https://gcc.gnu.org/PR12633 16163 502. https://gcc.gnu.org/PR13037 16164 503. https://gcc.gnu.org/PR13213 16165 504. https://gcc.gnu.org/PR4490 16166 505. https://gcc.gnu.org/PR12292 16167 506. https://gcc.gnu.org/PR12441 16168 507. https://gcc.gnu.org/PR12943 16169 508. https://gcc.gnu.org/PR13608 16170 509. https://gcc.gnu.org/PR11598 16171 510. https://gcc.gnu.org/PR11793 16172 511. https://gcc.gnu.org/PR12467 16173 512. https://gcc.gnu.org/PR12537 16174 513. https://gcc.gnu.org/PR12496 16175 514. https://gcc.gnu.org/PR12865 16176 515. https://gcc.gnu.org/PR13354 16177 516. https://gcc.gnu.org/PR10467 16178 517. https://gcc.gnu.org/PR11226 16179 518. https://gcc.gnu.org/PR11227 16180 519. https://gcc.gnu.org/PR12644 16181 520. https://gcc.gnu.org/PR13149 16182 521. https://gcc.gnu.org/PR12654 16183 522. https://gcc.gnu.org/PR12965 16184 523. https://gcc.gnu.org/PR13031 16185 524. https://gcc.gnu.org/PR11634 16186 525. https://gcc.gnu.org/PR12158 16187 526. https://gcc.gnu.org/PR11992 16188 527. https://gcc.gnu.org/PR9365 16189 528. https://gcc.gnu.org/PR10392 16190 529. https://gcc.gnu.org/PR11322 16191 530. https://gcc.gnu.org/PR13069 16192 531. https://gcc.gnu.org/PR13302 16193 532. https://gcc.gnu.org/PR13585 16194 533. https://gcc.gnu.org/PR8916 16195 534. https://gcc.gnu.org/PR11576 16196 535. https://gcc.gnu.org/PR13122 16197 536. https://gcc.gnu.org/PR13256 16198 537. https://gcc.gnu.org/PR13373 16199 538. https://gcc.gnu.org/PR12561 16200 539. https://gcc.gnu.org/PR6243 16201 540. https://gcc.gnu.org/PR11397 16202 541. https://gcc.gnu.org/PR12505 16203 542. https://gcc.gnu.org/PR13150 16204 543. https://gcc.gnu.org/PR12666 16205 544. https://gcc.gnu.org/PR12969 16206 545. https://gcc.gnu.org/PR10819 16207 546. https://gcc.gnu.org/PR11612 16208 547. https://gcc.gnu.org/PR13211 16209 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 16210 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 16211 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 16212 551. https://gcc.gnu.org/onlinedocs/ 16213 552. mailto:gcc-help (a] gcc.gnu.org 16214 553. mailto:gcc (a] gcc.gnu.org 16215 554. https://gcc.gnu.org/lists.html 16216 555. https://www.fsf.org/ 16217 556. https://gcc.gnu.org/about.html 16218 557. http://validator.w3.org/check/referer 16219 ====================================================================== 16220 http://gcc.gnu.org/gcc-3.2/index.html 16221 GCC 3.2 Release Series 16222 16223 (This release series is no longer supported.) 16224 16225 April 25, 2003 16226 16227 The [1]GNU project and the GCC developers are pleased to announce the 16228 release of GCC 3.2.3. 16229 16230 The purpose of the GCC 3.2 release series is to provide a stable 16231 platform for OS distributors to use building their next releases. A 16232 primary objective was to stabilize the C++ ABI; we believe that the 16233 interface to the compiler and the C++ standard library are now 16234 relatively stable. 16235 16236 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 16237 interoperate with code compiled by GCC 3.1.1 or earlier. 16238 16239 Please refer to our [2]detailed list of news, caveats, and bug-fixes 16240 for further information. 16241 16242 Release History 16243 16244 GCC 3.2.3 16245 April 25, 2003 ([3]changes) 16246 16247 GCC 3.2.2 16248 February 5, 2003 ([4]changes) 16249 16250 GCC 3.2.1 16251 November 19, 2002 ([5]changes) 16252 16253 GCC 3.2 16254 August 14, 2002 ([6]changes) 16255 16256 References and Acknowledgements 16257 16258 GCC used to stand for the GNU C Compiler, but since the compiler 16259 supports several other languages aside from C, it now stands for the 16260 GNU Compiler Collection. 16261 16262 A list of [7]successful builds is updated as new information becomes 16263 available. 16264 16265 The GCC developers would like to thank the numerous people that have 16266 contributed new features, improvements, bug fixes, and other changes as 16267 well as test results to GCC. This [8]amazing group of volunteers is 16268 what makes GCC successful. 16269 16270 For additional information about GCC please refer to the [9]GCC project 16271 web site or contact the [10]GCC development mailing list. 16272 16273 To obtain GCC please use [11]our mirror sites, or our CVS server. 16274 16275 16276 For questions related to the use of GCC, please consult these web 16277 pages and the [12]GCC manuals. If that fails, the 16278 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16279 web pages and the development of GCC are welcome on our developer 16280 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 16281 archives. 16282 16283 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 16284 distribution of this entire article is permitted in any medium, 16285 provided this notice is preserved. 16286 16287 These pages are [17]maintained by the GCC team. Last modified 16288 2019-11-28[18]. 16289 16290 References 16291 16292 1. http://www.gnu.org/ 16293 2. http://gcc.gnu.org/gcc-3.2/changes.html 16294 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 16295 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 16296 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 16297 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2 16298 7. http://gcc.gnu.org/gcc-3.2/buildstat.html 16299 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16300 9. http://gcc.gnu.org/index.html 16301 10. mailto:gcc (a] gcc.gnu.org 16302 11. http://gcc.gnu.org/mirrors.html 16303 12. https://gcc.gnu.org/onlinedocs/ 16304 13. mailto:gcc-help (a] gcc.gnu.org 16305 14. mailto:gcc (a] gcc.gnu.org 16306 15. https://gcc.gnu.org/lists.html 16307 16. https://www.fsf.org/ 16308 17. https://gcc.gnu.org/about.html 16309 18. http://validator.w3.org/check/referer 16310 ====================================================================== 16311 http://gcc.gnu.org/gcc-3.2/changes.html 16312 GCC 3.2 Release Series 16313 Changes, New Features, and Fixes 16314 16315 The latest release in the 3.2 release series is [1]GCC 3.2.3. 16316 16317 Caveats and New Features 16318 16319 Caveats 16320 16321 * The C++ compiler does not correctly zero-initialize 16322 pointers-to-data members. You must explicitly initialize them. For 16323 example: int S::*m(0); will work, but depending on 16324 default-initialization to zero will not work. This bug cannot be 16325 fixed in GCC 3.2 without inducing unacceptable risks. It will be 16326 fixed in GCC 3.3. 16327 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 16328 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 16329 a number of C++ ABI fixes which make its C++ compiler generate 16330 binary code which is incompatible with the C++ compilers found in 16331 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 16332 16333 Frontend Enhancements 16334 16335 C/C++/Objective-C 16336 16337 * The method of constructing the list of directories to be searched 16338 for header files has been revised. If a directory named by a -I 16339 option is a standard system include directory, the option is 16340 ignored to ensure that the default search order for system 16341 directories and the special treatment of system header files are 16342 not defeated. 16343 * The C and Objective-C compilers no longer accept the "Naming Types" 16344 extension (typedef foo = bar); it was already unavailable in C++. 16345 Code which uses it will need to be changed to use the "typeof" 16346 extension instead: typedef typeof(bar) foo. (We have removed this 16347 extension without a period of deprecation because it has caused the 16348 compiler to crash since version 3.0 and no one noticed until very 16349 recently. Thus we conclude it is not in widespread use.) 16350 16351 C++ 16352 16353 * GCC 3.2 fixed serveral differences between the C++ ABI implemented 16354 in GCC and the multi-vendor standard, but more have been found 16355 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 16356 code which is affected by these bugs. We will fix these bugs in 16357 some future release, once we are confident that all have been 16358 found; until then, it is our intention to make changes to the ABI 16359 only if they are necessary for correct compilation of C++, as 16360 opposed to conformance to the ABI documents. 16361 * For details on how to build an ABI compliant compiler for GNU/Linux 16362 systems, check the [3]common C++ ABI page. 16363 16364 New Targets and Target Specific Improvements 16365 16366 IA-32 16367 16368 * Fixed a number of bugs in SSE and MMX intrinsics. 16369 * Fixed common compiler crashes with SSE instruction set enabled 16370 (implied by -march=pentium3, pentium4, athlon-xp) 16371 * __m128 and __m128i is not 128bit aligned when used in structures. 16372 16373 x86-64 16374 16375 * A bug whereby the compiler could generate bad code for bzero has 16376 been fixed. 16377 * ABI fixes (implying ABI incompatibilities with previous version in 16378 some corner cases) 16379 * Fixed prefetch code generation 16380 __________________________________________________________________ 16381 16382 GCC 3.2.3 16383 16384 3.2.3 is a bug fix release only; there are no new features that were 16385 not present in GCC 3.2.2. 16386 16387 Bug Fixes 16388 16389 This section lists the problem reports (PRs) from GCC's bug tracking 16390 system that are known to be fixed in the 3.2.3 release. This list might 16391 not be complete (that is, it is possible that some PRs that have been 16392 fixed are not listed here), and some of the titles have been changed to 16393 make them more clear. 16394 16395 Internal Compiler Errors (multi-platform) 16396 16397 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in 16398 cc1plus 16399 * [5]6440: (c++) template specializations cause ICE 16400 * [6]7050: (c++) ICE on: (i ? get_string() : throw) 16401 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 16402 * [8]7982: (c++) ICE due to infinite recursion (using STL set) 16403 * [9]8068: exceedingly high (infinite) memory usage 16404 * [10]8178: ICE with __builtin_ffs 16405 * [11]8396: ICE in copy_to_mode_reg, in explow.c 16406 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 16407 * [13]9768: ICE when optimizing inline code at -O2 16408 * [14]9798: (c++) Infinite recursion (segfault) in 16409 cp/decl.c:push_using_directive with recursive using directives 16410 * [15]9799: mismatching structure initializer with nested flexible 16411 array member: ICE 16412 * [16]9928: ICE on duplicate enum declaration 16413 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 16414 sparc, alpha) 16415 * [18]10352: ICE in find_reloads_toplev 16416 * [19]10336: ICE with -Wunreachable-code 16417 16418 C/optimizer bugs: 16419 16420 * [20]8224: Incorrect joining of signed and unsigned division 16421 * [21]8613: -O2 produces wrong code with builtin strlen and 16422 postincrements 16423 * [22]8828: gcc reports some code is unreachable when it is not 16424 * [23]9226: GCSE breaking argument passing 16425 * [24]9853: miscompilation of non-constant structure initializer 16426 * [25]9797: C99-style struct initializers are miscompiled 16427 * [26]9967: Some standard C function calls should not be replaced 16428 when optimizing for size 16429 * [27]10116: ce2: invalid merge of join_bb in the context of switch 16430 statements 16431 * [28]10171: wrong code for inlined function 16432 * [29]10175: -Wunreachable-code doesn't work for single lines 16433 16434 C++ compiler and library: 16435 16436 * [30]8316: Confusing diagnostic for code that misuses conversion 16437 operators 16438 * [31]9169: filebuf output fails if codecvt<>::out returns noconv 16439 * [32]9420: incomplete type incorrectly reported 16440 * [33]9459: typeof in return type specification of template not 16441 supported 16442 * [34]9507: filebuf::open handles ios_base::ate incorrectly 16443 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc 16444 * [36]9602: Total confusion about template/friend/virtual/abstract 16445 * [37]9993: destructor not called for local object created within and 16446 returned from infinite loop 16447 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 16448 system 16449 16450 Java compiler and library: 16451 16452 * [39]9652: libgcj build fails on irix6.5.1[78] 16453 * [40]10144: gas on solaris complains about bad .stabs lines for 16454 java, native as unaffected 16455 16456 x86-specific (Intel/AMD): 16457 16458 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86 16459 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions 16460 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 16461 failed 16462 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 16463 failed 16464 16465 SPARC-specific: 16466 16467 * [45]7784: [Sparc] ICE in extract_insn, in recog.c 16468 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in 16469 unroll.c 16470 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 16471 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 16472 execute/loop-2d.c 16473 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 16474 * [50]9414: Scheduling bug on Ultrasparc 16475 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64 16476 16477 m68k-specific: 16478 16479 * [52]7248: broken "inclusive or" code 16480 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 16481 16482 PowerPC-specific: 16483 16484 * [54]9732: Wrong code with -O2 -fPIC 16485 * [55]10073: ICE: powerpc cannot split insn 16486 16487 Alpha-specific: 16488 16489 * [56]7702: optimization problem on a DEC alpha under OSF1 16490 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 16491 16492 HP-specific: 16493 16494 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 16495 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 16496 (missing symbol) 16497 * [60]10271: Floating point args don't get reloaded across function 16498 calls with -O2 16499 16500 MIPS specific: 16501 16502 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 16503 compile/920501-4.c 16504 16505 CRIS specific: 16506 16507 * [62]10377: gcc-3.2.2 creates bad assembler code for cris 16508 16509 Miscellaneous and minor bugs: 16510 16511 * [63]6955: collect2 says "core dumped" when there is no core 16512 __________________________________________________________________ 16513 16514 GCC 3.2.2 16515 16516 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 16517 install by means of the DESTDIR variable. Parts of the GCC tree have 16518 featured that support long before, but now it is available even from 16519 the top level. 16520 16521 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 16522 features that were not present in GCC 3.2.1. 16523 16524 Bug Fixes 16525 16526 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 16527 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 16528 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 16529 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 16530 change, and thus restores ABI-compatibility with previous releases 16531 (except GCC 3.2.1) on these platforms. 16532 16533 This section lists the problem reports (PRs) from GCC's bug tracking 16534 system that are known to be fixed in the 3.2.2 release. This list might 16535 not be complete (that is, it is possible that some PRs that have been 16536 fixed are not listed here) and some of the titles have been changed to 16537 make them more clear. 16538 16539 Internal Compiler Errors (multi-platform) 16540 16541 * [64]5919: (c++) ICE when passing variable array to template 16542 function 16543 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 16544 * [66]7507: ICE with -O2 when address of called function is a 16545 complicated expression 16546 * [67]7622: ICE with nested inline functions if function's address is 16547 taken 16548 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 16549 [69]9258) 16550 * [70]8031: (c++) ICE in code comparing typeids and casting from 16551 virtual base 16552 * [71]8275: ICE in simplify_subreg 16553 * [72]8332: (c++) builtin strlen/template interaction causes ICE 16554 * [73]8372: (c++) ICE on explicit call of destructor 16555 * [74]8439: (c, not c++) empty struct causes ICE 16556 * [75]8442: (c++) ICE with nested template classes 16557 * [76]8518: ICE when compiling mplayer ("extern inline" issue) 16558 * [77]8615: (c++) ICE with out-of-range character constant template 16559 argument 16560 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 16561 * [79]8799: (c++) ICE: error reporting routines re-entered 16562 * [80]9328: (c++) ICE with typeof(X) for overloaded X 16563 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes 16564 16565 C++ (compiler and library) bugs 16566 16567 * [82]47: scoping in nested classes is broken 16568 * [83]6745: problems with iostream rdbuf() member function 16569 * [84]8214: conversion from const char* const to char* sometimes 16570 accepted illegally 16571 * [85]8493: builtin strlen and overload resolution (same bug as 16572 [86]8332) 16573 * [87]8503: strange behaviour of function types 16574 * [88]8727: compiler confused by inheritance from an anonymous struct 16575 * [89]7445: poor performance of std::locale::classic() in 16576 multi-threaded applications 16577 * [90]8230: mishandling of overflow in vector<T>::resize 16578 * [91]8399: sync_with_stdio(false) breaks unformatted input 16579 * [92]8662: illegal access of private member of unnamed class is 16580 accepted 16581 * [93]8707: "make distclean" fails in libstdc++-v3 directory 16582 * [94]8708: __USE_MALLOC doesn't work 16583 * [95]8790: Use of non-thread-safe strtok in src/localename.cc 16584 * [96]8887: Bug in date formats with --enable-clocale=generic 16585 * [97]9076: Call Frame Instructions are not handled correctly during 16586 unwind operation 16587 * [98]9151: std::setprecision limited to 16 digits when outputting a 16588 double to a stream 16589 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers 16590 * [100]9269: libstdc++ headers: explicit specialization of function 16591 must precede its first use 16592 * [101]9322: return value of basic_streambuf<>::getloc affected by 16593 locale::global 16594 * [102]9433: segfault in runtime support for dynamic_cast 16595 16596 C and optimizer bugs 16597 16598 * [103]8032: GCC incorrectly initializes static structs that have 16599 flexible arrays 16600 * [104]8639: simple arithmetic expression broken 16601 * [105]8794: optimization improperly eliminates certain expressions 16602 * [106]8832: traditional "asm volatile" code is illegally optimized 16603 * [107]8988: loop optimizer bug: with -O2, code is generated that 16604 segfaults (found on i386, bug present for all platforms) 16605 * [108]9492: structure copy clobbers subsequent stores to structure 16606 16607 Objective-C bugs 16608 16609 * [109]9267: Objective-C parser won't build with newer bison versions 16610 (e.g. 1.875) 16611 16612 Ada bugs 16613 16614 * [110]8344: Ada build problem due to conflict between gcc/final.o, 16615 gcc/ada/final.o 16616 16617 Preprocessor bugs 16618 16619 * [111]8524: _Pragma within macros is improperly expanded 16620 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 16621 -fshort-wchar 16622 16623 ARM-specific 16624 16625 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95 16626 16627 x86-specific (Intel/AMD) 16628 16629 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 16630 * [115]8599: loop unroll bug with -march=k6-3 16631 * [116]9506: ABI breakage in structure return (affects BSD and 16632 Cygwin, but not GNU/Linux) 16633 16634 FreeBSD 5.0 specific 16635 16636 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 16637 16638 RTEMS-specific 16639 16640 * [118]9292: hppa1.1-rtems configurery problems 16641 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 16642 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue 16643 * [121]9296: gthr-rtems regression 16644 * [122]9316: powerpc-rtems: extending multilibs 16645 16646 HP-PA specific 16647 16648 * [123]9493: ICE with -O2 when building a simple function 16649 16650 Documentation 16651 16652 * [124]7341: hyperlink to gcov in GCC documentation doesn't work 16653 * [125]8947: Please add a warning about "-malign-double" in docs 16654 * [126]7448, [127]8882: typo cleanups 16655 __________________________________________________________________ 16656 16657 GCC 3.2.1 16658 16659 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 16660 generates code that is known not to be binary-compatible with the 16661 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 16662 in the distribution, for details. 16663 16664 This release also removes an old GCC extension, "naming types", and the 16665 documentation now directs users to use a different GCC extension, 16666 __typeof__, instead. The feature had evidently been broken for a while. 16667 16668 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 16669 the new warning there are no new features that were not present in GCC 16670 3.2. 16671 16672 In addition, the previous fix for [128]PR 7445 (poor performance of 16673 std::locale::classic() in multi-threaded applications) was reverted 16674 ("unfixed"), because the "fix" was not thread-safe. 16675 16676 Bug Fixes 16677 16678 This section lists the problem reports (PRs) from GCC's bug tracking 16679 system that are known to be fixed in the 3.2.1 release. This list might 16680 not be complete (that is, it is possible that some PRs that have been 16681 fixed are not listed here). As you can see, the number of bug fixes is 16682 quite large, so it is strongly recommended that users of earlier GCC 16683 3.x releases upgrade to GCC 3.2.1. 16684 16685 Internal Compiler Errors (multi-platform) 16686 16687 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 16688 * [130]5661: (c++) ICE instantiating template on array of unknown 16689 size (bad code) 16690 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 16691 64-bit platforms 16692 * [132]6994: (c++) ICE in find_function_data 16693 * [133]7150: preprocessor: GCC -dM -E gives an ICE 16694 * [134]7160: ICE when optimizing branches without a return value 16695 * [135]7228: (c++) ICE when using member template and template 16696 function 16697 * [136]7266: (c++) ICE with -pedantic on missing typename 16698 * [137]7353: ICE from use of "Naming Types" extension, see above 16699 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c 16700 * [139]7478: (c++) ICE on static_cast inside template 16701 * [140]7526: preprocessor core dump when _Pragma implies #pragma 16702 dependency 16703 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803 16704 is a duplicate) 16705 * [143]7754: (c++) ICE on union with template parameter 16706 * [144]7788: (c++) redeclaring a definition as an incomplete class 16707 causes ICE 16708 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c 16709 * [146]8055: preprocessor dies with SIG11 when building FreeBSD 16710 kernel 16711 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 16712 variables 16713 * [148]8134: (c++) ICE in force_store_init_value on legal code 16714 * [149]8149: (c++) ICE on incomplete type 16715 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 16716 initialization 16717 16718 C++ (compiler and library) bugs 16719 16720 * [151]5607: No pointer adjustment in covariant return types 16721 * [152]6579: Infinite loop with statement expressions in member 16722 initialization 16723 * [153]6803: Default copy constructor bug in GCC 3.1 16724 * [154]7176: g++ confused by friend and static member with same name 16725 * [155]7188: Segfault with template class and recursive (incorrect) 16726 initializer list 16727 * [156]7306: Regression: GCC 3.x fails to compile code with virtual 16728 inheritance if a method has a variable number of arguments 16729 * [157]7461: ctype<char>::classic_table() returns offset array on 16730 Cygwin 16731 * [158]7524: f(const float arg[3]) fails 16732 * [159]7584: Erroneous ambiguous base error on using declaration 16733 * [160]7676: Member template overloading problem 16734 * [161]7679: infinite loop when a right parenthesis is missing 16735 * [162]7811: default locale not taken from environment 16736 * [163]7961: compare( char *) implemented incorrectly in 16737 basic_string<> 16738 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if 16739 streambuf::underflow() leaves gptr() NULL (dups: [165]8127, 16740 [166]6745) 16741 * [167]8096: deque::at() throws std::range_error instead of 16742 std::out_of_range 16743 * [168]8127: cout << cin.rdbuf() infinite loop 16744 * [169]8218: Excessively large memory consumed for classes with large 16745 array members 16746 * [170]8287: GCC 3.2: Destructor called for non-constructed local 16747 object 16748 * [171]8347: empty vector range used in string construction causes 16749 core dump 16750 * [172]8348: fail() flag is set in istringstream when eof() flag is 16751 set 16752 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file) 16753 16754 C and optimizer bugs 16755 16756 * [174]6627: -fno-align-functions doesn't seem to disable function 16757 alignment 16758 * [175]6631: life_analysis misoptimizes code to initialize fields of 16759 a structure 16760 * [176]7102: unsigned char division results in floating exception 16761 * [177]7120: Run once loop should *always* be unrolled 16762 (pessimization) 16763 * [178]7209: Bug involving array referencing and ?: operator 16764 * [179]7515: invalid inlining of global function with -O3 16765 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test 16766 * [181]8467: bug in sibling call optimization 16767 16768 Preprocessor bugs 16769 16770 * [182]4890: incorrect line markers from the traditional preprocessor 16771 * [183]7357: -M option omits system headers files (making it the same 16772 as -MM) 16773 * [184]7358: Changes to Sun's make Dependencies 16774 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 16775 C headers 16776 * [186]7862: Interrupting GCC -MD removes .d file but not .o 16777 * [187]8190: Failed compilation deletes -MD dependency file 16778 * [188]8524: _Pragma within macro is improperly expanded 16779 16780 x86 specific (Intel/AMD) 16781 16782 * [189]5351: (i686-only) function pass-by-value structure copy 16783 corrupts stack ([190]7591 is a duplicate) 16784 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with 16785 -march=pentium3/pentium2/athlon (these are all the same underlying 16786 bug, in MMX register use) 16787 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same 16788 as above?) 16789 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 16790 * [199]6981: wrong code in 64-bit manipulation on x86 16791 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 16792 macro 16793 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 16794 intrinsics are broken 16795 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 16796 -march=pentium4 16797 * [203]7693: Typo in i386 mmintrin.h header 16798 * [204]7723: ICE - Pentium3 sse - GCC 3.2 16799 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 16800 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 16801 16802 PowerPC specific 16803 16804 * [207]5967: GCC bug when profiling nested functions on powerpc 16805 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while 16806 loop on PowerPC 16807 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 16808 * [210]7130: miscompiled code for GCC-3.1 on 16809 powerpc-unknown-linux-gnu with -funroll-all-loops 16810 * [211]7133: PowerPC ICE: unrecognizable insn 16811 * [212]7380: ICE in extract_insn, at recog.c:2148 16812 * [213]8252: ICE on Altivec code with optimization turned on 16813 * [214]8451: Altivec ICE in GCC 3.2 16814 16815 HP/PA specific 16816 16817 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa 16818 16819 SPARC specific 16820 16821 * [216]6668: when using --disable-multilib, libgcc_s.so is installed 16822 in the wrong place on sparc-solaris 16823 * [217]7151: ICE when compiling for UltraSPARC 16824 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 16825 double and -O1 16826 * [219]7842: [REGRESSION] SPARC code gen bug 16827 16828 ARM specific 16829 16830 * [220]7856: [arm] invalid offset in constant pool reference 16831 * [221]7967: optimization produces wrong code (ARM) 16832 16833 Alpha specific 16834 16835 * [222]7374: __builtin_fabsl broken on alpha 16836 16837 IBM s390 specific 16838 16839 * [223]7370: ICE in fixup_var_refs_1 on s390x 16840 * [224]7409: loop optimization bug on s390x-linux-gnu 16841 * [225]8232: s390x: ICE when using bcmp with int length argument 16842 16843 SCO specific 16844 16845 * [226]7623: SCO OpenServer build fails with machmode.def: undefined 16846 symbol: BITS_PER_UNIT 16847 16848 m68k/Coldfire specific 16849 16850 * [227]8314: crtbegin, crtend need to be multilib'ed for this 16851 platform 16852 16853 Documentation 16854 16855 * [228]761: Document some undocumented options 16856 * [229]5610: Fix documentation about invoking SSE instructions 16857 (-mfpmath=sse) 16858 * [230]7484: List -Wmissing-declarations as C-only option 16859 * [231]7531: -mcmodel not documented for x86-64 16860 * [232]8120: Update documentation of bad use of ## 16861 __________________________________________________________________ 16862 16863 GCC 3.2 16864 16865 3.2 is a small bug fix release, but there is a change to the 16866 application binary interface (ABI), hence the change to the second part 16867 of the version number. 16868 16869 The main purpose of the 3.2 release is to correct a couple of problems 16870 in the C++ ABI, with the intention of providing a stable interface 16871 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 16872 16873 Bug Fixes 16874 16875 C++ 16876 16877 * [233]7320: g++ 3.2 relocation problem 16878 * [234]7470: vtable: virtual function pointers not in declaration 16879 order 16880 16881 libstdc++ 16882 16883 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t 16884 * [236]6503, [237]6642, [238]7186: Problems with comparing or 16885 subtracting various types of const and non-const iterators 16886 * [239]7216: ambiguity with basic_iostream::traits_type 16887 * [240]7220: problem with basic_istream::ignore(0,delimiter) 16888 * [241]7222: locale::operator==() doesn't work on std::locale("") 16889 * [242]7286: placement operator delete issue 16890 * [243]7442: cxxabi.h does not match the C++ ABI 16891 * [244]7445: poor performance of std::locale::classic() in 16892 multi-threaded applications 16893 16894 x86-64 specific 16895 16896 * [245]7291: off-by-one in generated inline bzero code for x86-64 16897 16898 16899 For questions related to the use of GCC, please consult these web 16900 pages and the [246]GCC manuals. If that fails, the 16901 [247]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16902 web pages and the development of GCC are welcome on our developer 16903 list at [248]gcc (a] gcc.gnu.org. All of [249]our lists have public 16904 archives. 16905 16906 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and 16907 distribution of this entire article is permitted in any medium, 16908 provided this notice is preserved. 16909 16910 These pages are [251]maintained by the GCC team. Last modified 16911 2019-11-28[252]. 16912 16913 References 16914 16915 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 16916 2. http://gcc.gnu.org/gcc-3.1/changes.html 16917 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html 16918 4. https://gcc.gnu.org/PR3782 16919 5. https://gcc.gnu.org/PR6440 16920 6. https://gcc.gnu.org/PR7050 16921 7. https://gcc.gnu.org/PR7741 16922 8. https://gcc.gnu.org/PR7982 16923 9. https://gcc.gnu.org/PR8068 16924 10. https://gcc.gnu.org/PR8178 16925 11. https://gcc.gnu.org/PR8396 16926 12. https://gcc.gnu.org/PR8674 16927 13. https://gcc.gnu.org/PR9768 16928 14. https://gcc.gnu.org/PR9798 16929 15. https://gcc.gnu.org/PR9799 16930 16. https://gcc.gnu.org/PR9928 16931 17. https://gcc.gnu.org/PR10114 16932 18. https://gcc.gnu.org/PR10352 16933 19. https://gcc.gnu.org/PR10336 16934 20. https://gcc.gnu.org/PR8224 16935 21. https://gcc.gnu.org/PR8613 16936 22. https://gcc.gnu.org/PR8828 16937 23. https://gcc.gnu.org/PR9226 16938 24. https://gcc.gnu.org/PR9853 16939 25. https://gcc.gnu.org/PR9797 16940 26. https://gcc.gnu.org/PR9967 16941 27. https://gcc.gnu.org/PR10116 16942 28. https://gcc.gnu.org/PR10171 16943 29. https://gcc.gnu.org/PR10175 16944 30. https://gcc.gnu.org/PR8316 16945 31. https://gcc.gnu.org/PR9169 16946 32. https://gcc.gnu.org/PR9420 16947 33. https://gcc.gnu.org/PR9459 16948 34. https://gcc.gnu.org/PR9507 16949 35. https://gcc.gnu.org/PR9538 16950 36. https://gcc.gnu.org/PR9602 16951 37. https://gcc.gnu.org/PR9993 16952 38. https://gcc.gnu.org/PR10167 16953 39. https://gcc.gnu.org/PR9652 16954 40. https://gcc.gnu.org/PR10144 16955 41. https://gcc.gnu.org/PR8746 16956 42. https://gcc.gnu.org/PR9888 16957 43. https://gcc.gnu.org/PR9638 16958 44. https://gcc.gnu.org/PR9954 16959 45. https://gcc.gnu.org/PR7784 16960 46. https://gcc.gnu.org/PR7796 16961 47. https://gcc.gnu.org/PR8281 16962 48. https://gcc.gnu.org/PR8366 16963 49. https://gcc.gnu.org/PR8726 16964 50. https://gcc.gnu.org/PR9414 16965 51. https://gcc.gnu.org/PR10067 16966 52. https://gcc.gnu.org/PR7248 16967 53. https://gcc.gnu.org/PR8343 16968 54. https://gcc.gnu.org/PR9732 16969 55. https://gcc.gnu.org/PR10073 16970 56. https://gcc.gnu.org/PR7702 16971 57. https://gcc.gnu.org/PR9671 16972 58. https://gcc.gnu.org/PR8694 16973 59. https://gcc.gnu.org/PR9953 16974 60. https://gcc.gnu.org/PR10271 16975 61. https://gcc.gnu.org/PR6362 16976 62. https://gcc.gnu.org/PR10377 16977 63. https://gcc.gnu.org/PR6955 16978 64. https://gcc.gnu.org/PR5919 16979 65. https://gcc.gnu.org/PR7129 16980 66. https://gcc.gnu.org/PR7507 16981 67. https://gcc.gnu.org/PR7622 16982 68. https://gcc.gnu.org/PR7681 16983 69. https://gcc.gnu.org/PR9528 16984 70. https://gcc.gnu.org/PR8031 16985 71. https://gcc.gnu.org/PR8275 16986 72. https://gcc.gnu.org/PR8332 16987 73. https://gcc.gnu.org/PR8372 16988 74. https://gcc.gnu.org/PR8439 16989 75. https://gcc.gnu.org/PR8442 16990 76. https://gcc.gnu.org/PR8518 16991 77. https://gcc.gnu.org/PR8615 16992 78. https://gcc.gnu.org/PR8663 16993 79. https://gcc.gnu.org/PR8799 16994 80. https://gcc.gnu.org/PR9328 16995 81. https://gcc.gnu.org/PR9465 16996 82. https://gcc.gnu.org/PR47 16997 83. https://gcc.gnu.org/PR6745 16998 84. https://gcc.gnu.org/PR8214 16999 85. https://gcc.gnu.org/PR8493 17000 86. https://gcc.gnu.org/PR8332 17001 87. https://gcc.gnu.org/PR8503 17002 88. https://gcc.gnu.org/PR8727 17003 89. https://gcc.gnu.org/PR7445 17004 90. https://gcc.gnu.org/PR8230 17005 91. https://gcc.gnu.org/PR8399 17006 92. https://gcc.gnu.org/PR8662 17007 93. https://gcc.gnu.org/PR8707 17008 94. https://gcc.gnu.org/PR8708 17009 95. https://gcc.gnu.org/PR8790 17010 96. https://gcc.gnu.org/PR8887 17011 97. https://gcc.gnu.org/PR9076 17012 98. https://gcc.gnu.org/PR9151 17013 99. https://gcc.gnu.org/PR9168 17014 100. https://gcc.gnu.org/PR9269 17015 101. https://gcc.gnu.org/PR9322 17016 102. https://gcc.gnu.org/PR9433 17017 103. https://gcc.gnu.org/PR8032 17018 104. https://gcc.gnu.org/PR8639 17019 105. https://gcc.gnu.org/PR8794 17020 106. https://gcc.gnu.org/PR8832 17021 107. https://gcc.gnu.org/PR8988 17022 108. https://gcc.gnu.org/PR9492 17023 109. https://gcc.gnu.org/PR9267 17024 110. https://gcc.gnu.org/PR8344 17025 111. https://gcc.gnu.org/PR8524 17026 112. https://gcc.gnu.org/PR8880 17027 113. https://gcc.gnu.org/PR9090 17028 114. https://gcc.gnu.org/PR8588 17029 115. https://gcc.gnu.org/PR8599 17030 116. https://gcc.gnu.org/PR9506 17031 117. https://gcc.gnu.org/PR9484 17032 118. https://gcc.gnu.org/PR9292 17033 119. https://gcc.gnu.org/PR9293 17034 120. https://gcc.gnu.org/PR9295 17035 121. https://gcc.gnu.org/PR9296 17036 122. https://gcc.gnu.org/PR9316 17037 123. https://gcc.gnu.org/PR9493 17038 124. https://gcc.gnu.org/PR7341 17039 125. https://gcc.gnu.org/PR8947 17040 126. https://gcc.gnu.org/PR7448 17041 127. https://gcc.gnu.org/PR8882 17042 128. https://gcc.gnu.org/PR7445 17043 129. https://gcc.gnu.org/PR2521 17044 130. https://gcc.gnu.org/PR5661 17045 131. https://gcc.gnu.org/PR6419 17046 132. https://gcc.gnu.org/PR6994 17047 133. https://gcc.gnu.org/PR7150 17048 134. https://gcc.gnu.org/PR7160 17049 135. https://gcc.gnu.org/PR7228 17050 136. https://gcc.gnu.org/PR7266 17051 137. https://gcc.gnu.org/PR7353 17052 138. https://gcc.gnu.org/PR7411 17053 139. https://gcc.gnu.org/PR7478 17054 140. https://gcc.gnu.org/PR7526 17055 141. https://gcc.gnu.org/PR7721 17056 142. https://gcc.gnu.org/PR7803 17057 143. https://gcc.gnu.org/PR7754 17058 144. https://gcc.gnu.org/PR7788 17059 145. https://gcc.gnu.org/PR8031 17060 146. https://gcc.gnu.org/PR8055 17061 147. https://gcc.gnu.org/PR8067 17062 148. https://gcc.gnu.org/PR8134 17063 149. https://gcc.gnu.org/PR8149 17064 150. https://gcc.gnu.org/PR8160 17065 151. https://gcc.gnu.org/PR5607 17066 152. https://gcc.gnu.org/PR6579 17067 153. https://gcc.gnu.org/PR6803 17068 154. https://gcc.gnu.org/PR7176 17069 155. https://gcc.gnu.org/PR7188 17070 156. https://gcc.gnu.org/PR7306 17071 157. https://gcc.gnu.org/PR7461 17072 158. https://gcc.gnu.org/PR7524 17073 159. https://gcc.gnu.org/PR7584 17074 160. https://gcc.gnu.org/PR7676 17075 161. https://gcc.gnu.org/PR7679 17076 162. https://gcc.gnu.org/PR7811 17077 163. https://gcc.gnu.org/PR7961 17078 164. https://gcc.gnu.org/PR8071 17079 165. https://gcc.gnu.org/PR8127 17080 166. https://gcc.gnu.org/PR6745 17081 167. https://gcc.gnu.org/PR8096 17082 168. https://gcc.gnu.org/PR8127 17083 169. https://gcc.gnu.org/PR8218 17084 170. https://gcc.gnu.org/PR8287 17085 171. https://gcc.gnu.org/PR8347 17086 172. https://gcc.gnu.org/PR8348 17087 173. https://gcc.gnu.org/PR8391 17088 174. https://gcc.gnu.org/PR6627 17089 175. https://gcc.gnu.org/PR6631 17090 176. https://gcc.gnu.org/PR7102 17091 177. https://gcc.gnu.org/PR7120 17092 178. https://gcc.gnu.org/PR7209 17093 179. https://gcc.gnu.org/PR7515 17094 180. https://gcc.gnu.org/PR7814 17095 181. https://gcc.gnu.org/PR8467 17096 182. https://gcc.gnu.org/PR4890 17097 183. https://gcc.gnu.org/PR7357 17098 184. https://gcc.gnu.org/PR7358 17099 185. https://gcc.gnu.org/PR7602 17100 186. https://gcc.gnu.org/PR7862 17101 187. https://gcc.gnu.org/PR8190 17102 188. https://gcc.gnu.org/PR8524 17103 189. https://gcc.gnu.org/PR5351 17104 190. https://gcc.gnu.org/PR7591 17105 191. https://gcc.gnu.org/PR6845 17106 192. https://gcc.gnu.org/PR7034 17107 193. https://gcc.gnu.org/PR7124 17108 194. https://gcc.gnu.org/PR7174 17109 195. https://gcc.gnu.org/PR7134 17110 196. https://gcc.gnu.org/PR7375 17111 197. https://gcc.gnu.org/PR7390 17112 198. https://gcc.gnu.org/PR6890 17113 199. https://gcc.gnu.org/PR6981 17114 200. https://gcc.gnu.org/PR7242 17115 201. https://gcc.gnu.org/PR7396 17116 202. https://gcc.gnu.org/PR7630 17117 203. https://gcc.gnu.org/PR7693 17118 204. https://gcc.gnu.org/PR7723 17119 205. https://gcc.gnu.org/PR7951 17120 206. https://gcc.gnu.org/PR8146 17121 207. https://gcc.gnu.org/PR5967 17122 208. https://gcc.gnu.org/PR6984 17123 209. https://gcc.gnu.org/PR7114 17124 210. https://gcc.gnu.org/PR7130 17125 211. https://gcc.gnu.org/PR7133 17126 212. https://gcc.gnu.org/PR7380 17127 213. https://gcc.gnu.org/PR8252 17128 214. https://gcc.gnu.org/PR8451 17129 215. https://gcc.gnu.org/PR7250 17130 216. https://gcc.gnu.org/PR6668 17131 217. https://gcc.gnu.org/PR7151 17132 218. https://gcc.gnu.org/PR7335 17133 219. https://gcc.gnu.org/PR7842 17134 220. https://gcc.gnu.org/PR7856 17135 221. https://gcc.gnu.org/PR7967 17136 222. https://gcc.gnu.org/PR7374 17137 223. https://gcc.gnu.org/PR7370 17138 224. https://gcc.gnu.org/PR7409 17139 225. https://gcc.gnu.org/PR8232 17140 226. https://gcc.gnu.org/PR7623 17141 227. https://gcc.gnu.org/PR8314 17142 228. https://gcc.gnu.org/PR761 17143 229. https://gcc.gnu.org/PR5610 17144 230. https://gcc.gnu.org/PR7484 17145 231. https://gcc.gnu.org/PR7531 17146 232. https://gcc.gnu.org/PR8120 17147 233. https://gcc.gnu.org/PR7320 17148 234. https://gcc.gnu.org/PR7470 17149 235. https://gcc.gnu.org/PR6410 17150 236. https://gcc.gnu.org/PR6503 17151 237. https://gcc.gnu.org/PR6642 17152 238. https://gcc.gnu.org/PR7186 17153 239. https://gcc.gnu.org/PR7216 17154 240. https://gcc.gnu.org/PR7220 17155 241. https://gcc.gnu.org/PR7222 17156 242. https://gcc.gnu.org/PR7286 17157 243. https://gcc.gnu.org/PR7442 17158 244. https://gcc.gnu.org/PR7445 17159 245. https://gcc.gnu.org/PR7291 17160 246. https://gcc.gnu.org/onlinedocs/ 17161 247. mailto:gcc-help (a] gcc.gnu.org 17162 248. mailto:gcc (a] gcc.gnu.org 17163 249. https://gcc.gnu.org/lists.html 17164 250. https://www.fsf.org/ 17165 251. https://gcc.gnu.org/about.html 17166 252. http://validator.w3.org/check/referer 17167 ====================================================================== 17168 http://gcc.gnu.org/gcc-3.1/index.html 17169 GCC 3.1 17170 17171 (This release series is no longer supported.) 17172 17173 July 27, 2002 17174 17175 The [1]GNU project and the GCC developers are pleased to announce the 17176 release of GCC 3.1.1. 17177 17178 The links below still apply to GCC 3.1.1. 17179 17180 May 15, 2002 17181 17182 The [2]GNU project and the GCC developers are pleased to announce the 17183 release of GCC 3.1. 17184 17185 GCC used to stand for the GNU C Compiler, but since the compiler 17186 supports several other languages aside from C, it now stands for the 17187 GNU Compiler Collection. 17188 17189 A list of [3]successful builds is updated as new information becomes 17190 available. 17191 17192 The GCC developers would like to thank the numerous people that have 17193 contributed [4]new features, improvements, bug fixes, and other changes 17194 as well as test results to GCC. This [5]amazing group of volunteers is 17195 what makes GCC successful. 17196 17197 For additional information about GCC please refer to the [6]GCC project 17198 web site or contact the [7]GCC development mailing list. 17199 17200 To obtain GCC please use [8]our mirror sites, or our CVS server. 17201 __________________________________________________________________ 17202 17203 17204 For questions related to the use of GCC, please consult these web 17205 pages and the [9]GCC manuals. If that fails, the 17206 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17207 web pages and the development of GCC are welcome on our developer 17208 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 17209 archives. 17210 17211 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 17212 distribution of this entire article is permitted in any medium, 17213 provided this notice is preserved. 17214 17215 These pages are [14]maintained by the GCC team. Last modified 17216 2019-11-28[15]. 17217 17218 References 17219 17220 1. http://www.gnu.org/ 17221 2. http://www.gnu.org/ 17222 3. http://gcc.gnu.org/gcc-3.1/buildstat.html 17223 4. http://gcc.gnu.org/gcc-3.1/changes.html 17224 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17225 6. http://gcc.gnu.org/index.html 17226 7. mailto:gcc (a] gcc.gnu.org 17227 8. http://gcc.gnu.org/mirrors.html 17228 9. https://gcc.gnu.org/onlinedocs/ 17229 10. mailto:gcc-help (a] gcc.gnu.org 17230 11. mailto:gcc (a] gcc.gnu.org 17231 12. https://gcc.gnu.org/lists.html 17232 13. https://www.fsf.org/ 17233 14. https://gcc.gnu.org/about.html 17234 15. http://validator.w3.org/check/referer 17235 ====================================================================== 17236 http://gcc.gnu.org/gcc-3.1/changes.html 17237 GCC 3.1 Release Series 17238 Changes, New Features, and Fixes 17239 17240 Additional changes in GCC 3.1.1 17241 17242 * A bug related to how structures and unions are returned has been 17243 fixed for powerpc-*-netbsd*. 17244 * An important bug in the implementation of -fprefetch-loop-arrays 17245 has been fixed. Previously the optimization prefetched random 17246 blocks of memory for most targets except for i386. 17247 * The Java compiler now compiles Java programs much faster and also 17248 works with parallel make. 17249 * Nested functions have been fixed for mips*-*-netbsd*. 17250 * Some missing floating point support routines have beed added for 17251 mips*-*-netbsd*. 17252 * This [1]message gives additional information about the bugs fixed 17253 in this release. 17254 17255 Caveats 17256 17257 * The -traditional C compiler option has been deprecated and will be 17258 removed in GCC 3.3. (It remains possible to preprocess non-C code 17259 with the traditional preprocessor.) 17260 * The default debugging format for most ELF platforms (including 17261 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 17262 from stabs to DWARF2. This requires GDB 5.1.1 or later. 17263 17264 General Optimizer Improvements 17265 17266 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 17267 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 17268 for profile driven optimizations. 17269 Options -fprofile-arcs and -fbranch-probabilities can now be used 17270 to improve speed of the generated code by profiling the actual 17271 program behaviour on typical runs. In the absence of profile info 17272 the compiler attempts to guess the profile statically. 17273 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 17274 monitor performance of the generated code. 17275 According to the SPECInt2000 results on an AMD Athlon CPU, the code 17276 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 17277 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 17278 is about 2.1% faster compared to 2.95.3. Tests were done using the 17279 -O2 -march=athlon command-line options. 17280 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 17281 infrastructure developed by CodeSourcery, LLC for the C++ front 17282 end, so that it is now used in the C front end too. Inlining 17283 functions as trees exposes them earlier to the compiler, giving it 17284 more opportunities for optimization. 17285 * Support for data prefetching instructions has been added to the GCC 17286 back end and several targets. A new __builtin_prefetch intrinsic is 17287 available to explicitly insert prefetch instructions and 17288 experimental support for loop array prefetching has been added (see 17289 -fprefetch-loop-array documentation). 17290 * Support for emitting debugging information for macros has been 17291 added for DWARF2. It is activated using -g3. 17292 17293 New Languages and Language specific improvements 17294 17295 C/C++ 17296 17297 * A few more [4]ISO C99 features. 17298 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 17299 * The preprocessor's symbol table has been merged with the symbol 17300 table of the C, C++ and Objective-C front ends. 17301 * The preprocessor consumes less memory than the preprocessor in GCC 17302 3.0, often significantly so. On normal input files, it typically 17303 consumes less memory than pre-3.0 cccp-based GCC, too. 17304 17305 C++ 17306 17307 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 17308 was a workaround to allow std compliant code to work with the 17309 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 17310 * The C++ ABI has been fixed so that void (A::*)() const is mangled 17311 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 17312 affects pointer to cv-qualified member function types. 17313 * The C++ ABI has been changed to correctly handle this code: 17314 struct A { 17315 void operator delete[] (void *, size_t); 17316 }; 17317 17318 struct B : public A { 17319 }; 17320 17321 new B[10]; 17322 17323 The amount of storage allocated for the array will be greater than 17324 it was in 3.0, in order to store the number of elements in the 17325 array, so that the correct size can be passed to operator delete[] 17326 when the array is deleted. Previously, the value passed to operator 17327 delete[] was unpredictable. 17328 This change will only affect code that declares a two-argument 17329 operator delete[] with a second parameter of type size_t in a base 17330 class, and does not override that definition in a derived class. 17331 * The C++ ABI has been changed so that: 17332 struct A { 17333 void operator delete[] (void *, size_t); 17334 void operator delete[] (void *); 17335 }; 17336 17337 does not cause unnecessary storage to be allocated when an array of 17338 A objects is allocated. 17339 This change will only affect code that declares both of these forms 17340 of operator delete[], and declared the two-argument form before the 17341 one-argument form. 17342 * The C++ ABI has been changed so that when a parameter is passed by 17343 value, any cleanup for that parameter is performed in the caller, 17344 as specified by the ia64 C++ ABI, rather than the called function 17345 as before. As a result, classes with a non-trivial destructor but a 17346 trivial copy constructor will be passed and returned by invisible 17347 reference, rather than by bitwise copy as before. 17348 * G++ now supports the "named return value optimization": for code 17349 like 17350 A f () { 17351 A a; 17352 ... 17353 return a; 17354 } 17355 17356 G++ will allocate a in the return value slot, so that the return 17357 becomes a no-op. For this to work, all return statements in the 17358 function must return the same variable. 17359 * Improvements to the C++ library are listed in [5]the libstdc++-v3 17360 FAQ. 17361 17362 Objective-C 17363 17364 * Annoying linker warnings (due to incorrect code being generated) 17365 have been fixed. 17366 * If a class method cannot be found, the compiler no longer issues a 17367 warning if a corresponding instance method exists in the root 17368 class. 17369 * Forward @protocol declarations have been fixed. 17370 * Loading of categories has been fixed in certain situations (GNU run 17371 time only). 17372 * The class lookup in the run-time library has been rewritten so that 17373 class method dispatch is more than twice as fast as it used to be 17374 (GNU run time only). 17375 17376 Java 17377 17378 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 17379 javax.transaction. 17380 * Property files and other system resources can be compiled into 17381 executables which use libgcj using the new gcj --resource feature. 17382 * libgcj has been ported to more platforms. In particular there is 17383 now a mostly-functional mingw32 (Windows) target port. 17384 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 17385 Java code can now be called from a C/C++ application. 17386 * gcj can now use builtin functions for certain known methods, for 17387 instance Math.cos. 17388 * gcj can now automatically remove redundant array-store checks in 17389 some common cases. 17390 * The --no-store-checks optimization option was added. This can be 17391 used to omit runtime store checks for code which is known not to 17392 throw ArrayStoreException 17393 * The following third party interface standards were added to libgcj: 17394 org.w3c.dom and org.xml.sax. 17395 * java.security has been merged with GNU Classpath. The new package 17396 is now JDK 1.2 compliant, and much more complete. 17397 * A bytecode verifier was added to the libgcj interpreter. 17398 * java.lang.Character was rewritten to comply with the Unicode 3.0 17399 standard, and improve performance. 17400 * Partial support for many more locales was added to libgcj. 17401 * Socket timeouts have been implemented. 17402 * libgcj has been merged into a single shared library. There are no 17403 longer separate shared libraries for the garbage collector and 17404 zlib. 17405 * Several performance improvements were made to gcj and libgcj: 17406 + Hash synchronization (thin locks) 17407 + A special allocation path for finalizer-free objects 17408 + Thread-local allocation 17409 + Parallel GC, and other GC tweaks 17410 17411 Fortran 17412 17413 Fortran improvements are listed in [6]the Fortran documentation. 17414 17415 Ada 17416 17417 [7]AdaCore, has contributed its GNAT Ada 95 front end and associated 17418 tools. The GNAT compiler fully implements the Ada language as defined 17419 by the ISO/IEC 8652 standard. 17420 17421 Please note that the integration of the Ada front end is still work in 17422 progress. 17423 17424 New Targets and Target Specific Improvements 17425 17426 * Hans-Peter Nilsson has contributed a port to MMIX, the CPU 17427 architecture used in new editions of Donald E. Knuth's The Art of 17428 Computer Programming. 17429 * Axis Communications has contributed its port to the CRIS CPU 17430 architecture, used in the ETRAX system-on-a-chip series. 17431 * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH 17432 SH5 64-bit RISC microprocessor architecture, extending the existing 17433 SH port. 17434 * UltraSPARC is fully supported in 64-bit mode. The option -m64 17435 enables it. 17436 * For compatibility with the Sun compiler #pragma redefine_extname 17437 has been implemented on Solaris. 17438 * The x86 back end has had some noticeable work done to it. 17439 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 17440 Jaeger have contributed a port to the AMD x86-64 architecture. 17441 For more information on x86-64 see http://www.x86-64.org. 17442 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 17443 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 17444 enable the respective instruction sets. Intel C++ compatible 17445 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 17446 will be added in next major release. 17447 + Following those improvements, targets for Pentium MMX, K6-2, 17448 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 17449 added. Refer to the documentation on -march= and -mcpu= 17450 options for details. 17451 + For those targets that support it, -mfpmath=sse will cause the 17452 compiler to generate SSE/SSE2 instructions for floating point 17453 math instead of x87 instructions. Usually, this will lead to 17454 quicker code especially on the Pentium 4. Note that only 17455 scalar floating point instructions are used and GCC does not 17456 exploit SIMD features yet. 17457 + Prefetch support has been added to the Pentium III, Pentium 4, 17458 K6-2, K6-3, and Athlon series. 17459 + Code generated for floating point to integer conversions has 17460 been improved leading to better performance of many 3D 17461 applications. 17462 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 17463 * C++ support for AIX has been improved. 17464 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 17465 PowerPC port supporting the AltiVec programming model (SIMD). The 17466 support, though presently useful, is experimental and is expected 17467 to stabilize for 3.2. The support is written to conform to 17468 Motorola's AltiVec specs. See -maltivec. 17469 17470 Obsolete Systems 17471 17472 Support for a number of older systems has been declared obsolete in GCC 17473 3.1. Unless there is activity to revive them, the next release of GCC 17474 will have their sources permanently removed. 17475 17476 All configurations of the following processor architectures have been 17477 declared obsolete: 17478 * MIL-STD-1750A, 1750a-*-* 17479 * AMD A29k, a29k-*-* 17480 * Convex, c*-convex-* 17481 * Clipper, clipper-*-* 17482 * Elxsi, elxsi-*-* 17483 * Intel i860, i860-*-* 17484 * Sun picoJava, pj-*-* and pjl-*-* 17485 * Western Electric 32000, we32k-*-* 17486 17487 Most configurations of the following processor architectures have been 17488 declared obsolete, but we are preserving a few systems which may have 17489 active developers. It is unlikely that the remaining systems will 17490 survive much longer unless we see definite signs of port activity. 17491 * Motorola 88000 except 17492 + Generic a.out, m88k-*-aout* 17493 + Generic SVR4, m88k-*-sysv4 17494 + OpenBSD, m88k-*-openbsd* 17495 * NS32k except 17496 + NetBSD, ns32k-*-netbsd* 17497 + OpenBSD, ns32k-*-openbsd*. 17498 * ROMP except 17499 + OpenBSD, romp-*-openbsd*. 17500 17501 Finally, only some configurations of these processor architectures are 17502 being obsoleted. 17503 * Alpha: 17504 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 17505 alpha*-*-osf[45], are still supported.) 17506 * ARM: 17507 + RISCiX, arm-*-riscix*. 17508 * i386: 17509 + 386BSD, i?86-*-bsd* 17510 + Chorus, i?86-*-chorusos* 17511 + DG/UX, i?86-*-dgux* 17512 + FreeBSD 1.x, i?86-*-freebsd1.* 17513 + IBM AIX, i?86-*-aix* 17514 + ISC UNIX, i?86-*-isc* 17515 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 17516 + NEXTstep, i?86-next-* 17517 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 17518 + RTEMS/coff, i?86-*-rtemscoff* 17519 + RTEMS/go32, i?86-go32-rtems* 17520 + Sequent/BSD, i?86-sequent-bsd* 17521 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 17522 i?86-sequent-sysv3* 17523 + SunOS, i?86-*-sunos* 17524 * Motorola 68000: 17525 + Altos, m68[k0]*-altos-* 17526 + Apollo, m68[k0]*-apollo-* 17527 + Apple A/UX, m68[k0]*-apple-* 17528 + Bull, m68[k0]*-bull-* 17529 + Convergent, m68[k0]*-convergent-* 17530 + Generic SVR3, m68[k0]*-*-sysv3* 17531 + ISI, m68[k0]*-isi-* 17532 + LynxOS, m68[k0]*-*-lynxos* 17533 + NEXT, m68[k0]*-next-* 17534 + RTEMS/coff, m68[k0]*-*-rtemscoff* 17535 + Sony, m68[k0]*-sony-* 17536 * MIPS: 17537 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 17538 + Generic BSD, mips-*-bsd* 17539 + Generic System V, mips-*-sysv* 17540 + IRIX before version 5, mips-sgi-irix[1234]* 17541 + RiscOS, mips-*-riscos* 17542 + Sony, mips-sony-* 17543 + Tandem, mips-tandem-* 17544 * SPARC: 17545 + RTEMS/a.out, sparc-*-rtemsaout*. 17546 17547 Documentation improvements 17548 17549 * The old manual ("Using and Porting the GNU Compiler Collection") 17550 has been replaced by a users manual ("Using the GNU Compiler 17551 Collection") and a separate internals reference manual ("GNU 17552 Compiler Collection Internals"). 17553 * More complete and much improved documentation about GCC's internal 17554 representation used by the C and C++ front ends. 17555 * Many cleanups and improvements in general. 17556 17557 17558 For questions related to the use of GCC, please consult these web 17559 pages and the [8]GCC manuals. If that fails, the 17560 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17561 web pages and the development of GCC are welcome on our developer 17562 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 17563 archives. 17564 17565 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 17566 distribution of this entire article is permitted in any medium, 17567 provided this notice is preserved. 17568 17569 These pages are [13]maintained by the GCC team. Last modified 17570 2019-11-28[14]. 17571 17572 References 17573 17574 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 17575 2. http://gcc.gnu.org/news/profiledriven.html 17576 3. http://gcc.gnu.org/benchmarks/ 17577 4. http://gcc.gnu.org/c99status.html 17578 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 17579 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 17580 7. https://www.adacore.com/ 17581 8. https://gcc.gnu.org/onlinedocs/ 17582 9. mailto:gcc-help (a] gcc.gnu.org 17583 10. mailto:gcc (a] gcc.gnu.org 17584 11. https://gcc.gnu.org/lists.html 17585 12. https://www.fsf.org/ 17586 13. https://gcc.gnu.org/about.html 17587 14. http://validator.w3.org/check/referer 17588 ====================================================================== 17589 http://gcc.gnu.org/gcc-3.0/index.html 17590 GCC 3.0.4 17591 17592 (This release series is no longer supported.) 17593 17594 February 20, 2002 17595 17596 The [1]GNU project and the GCC developers are pleased to announce the 17597 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 17598 series. 17599 17600 GCC used to stand for the GNU C Compiler, but since the compiler 17601 supports several other languages aside from C, it now stands for the 17602 GNU Compiler Collection. 17603 17604 GCC 3.0.x has several new optimizations, new targets, new languages and 17605 many other new features, relative to GCC 2.95.x. See the [2]new 17606 features page for a more complete list. 17607 17608 A list of [3]successful builds is updated as new information becomes 17609 available. 17610 17611 The GCC developers would like to thank the numerous people that have 17612 contributed new features, test results, bug fixes, etc to GCC. This 17613 [4]amazing group of volunteers is what makes GCC successful. 17614 17615 And finally, we can't in good conscience fail to mention some 17616 [5]caveats to using GCC 3.0.x. 17617 17618 For additional information about GCC please refer to the [6]GCC project 17619 web site or contact the [7]GCC development mailing list. 17620 17621 To obtain GCC please use [8]our mirror sites, or our CVS server. 17622 __________________________________________________________________ 17623 17624 Previous 3.0.x Releases 17625 17626 December 20, 2001: GCC 3.0.3 has been released. 17627 October 25, 2001: GCC 3.0.2 has been released. 17628 August 20, 2001: GCC 3.0.1 has been released. 17629 June 18, 2001: GCC 3.0 has been released. 17630 17631 17632 For questions related to the use of GCC, please consult these web 17633 pages and the [9]GCC manuals. If that fails, the 17634 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17635 web pages and the development of GCC are welcome on our developer 17636 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 17637 archives. 17638 17639 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 17640 distribution of this entire article is permitted in any medium, 17641 provided this notice is preserved. 17642 17643 These pages are [14]maintained by the GCC team. Last modified 17644 2019-11-28[15]. 17645 17646 References 17647 17648 1. http://www.gnu.org/ 17649 2. http://gcc.gnu.org/gcc-3.0/features.html 17650 3. http://gcc.gnu.org/gcc-3.0/buildstat.html 17651 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17652 5. http://gcc.gnu.org/gcc-3.0/caveats.html 17653 6. http://gcc.gnu.org/index.html 17654 7. mailto:gcc (a] gcc.gnu.org 17655 8. http://gcc.gnu.org/mirrors.html 17656 9. https://gcc.gnu.org/onlinedocs/ 17657 10. mailto:gcc-help (a] gcc.gnu.org 17658 11. mailto:gcc (a] gcc.gnu.org 17659 12. https://gcc.gnu.org/lists.html 17660 13. https://www.fsf.org/ 17661 14. https://gcc.gnu.org/about.html 17662 15. http://validator.w3.org/check/referer 17663 ====================================================================== 17664 http://gcc.gnu.org/gcc-3.0/features.html 17665 GCC 3.0 New Features 17666 17667 Additional changes in GCC 3.0.4 17668 17669 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 17670 system, which use the ELF object file format, on x86 processors. 17671 * Correct debugging information is generated from functions that have 17672 lines from multiple files (e.g. yacc output). 17673 * A fix for whitespace handling in the -traditional preprocessor, 17674 which can affect Fortran. 17675 * Fixes to the exception handling runtime. 17676 * More fixes for bad code generation in C++. 17677 * A fix for shared library generation under AIX 4.3. 17678 * Documentation updates. 17679 * Port of GCC to Tensilica's Xtensa processor contributed. 17680 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 17681 17682 Additional changes in GCC 3.0.3 17683 17684 * A fix to correct an accidental change to the PowerPC ABI. 17685 * Fixes for bad code generation on a variety of architectures. 17686 * Improvements to the debugging information generated for C++ 17687 classes. 17688 * Fixes for bad code generation in C++. 17689 * A fix to avoid crashes in the C++ demangler. 17690 * A fix to the C++ standard library to avoid buffer overflows. 17691 * Miscellaneous improvements for a variety of architectures. 17692 17693 Additional changes in GCC 3.0.2 17694 17695 * Fixes for bad code generation during loop unrolling. 17696 * Fixes for bad code generation by the sibling call optimization. 17697 * Minor improvements to x86 code generation. 17698 * Implementation of function descriptors in C++ vtables for IA64. 17699 * Numerous minor bug-fixes. 17700 17701 Additional changes in GCC 3.0.1 17702 17703 * C++ fixes for incorrect code-generation. 17704 * Improved cross-compiling support for the C++ standard library. 17705 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 17706 in GCC 3.0. 17707 * Fixes for various exception-handling bugs. 17708 * A port to the S/390 architecture. 17709 17710 General Optimizer Improvements 17711 17712 * [2]Basic block reordering pass. 17713 * New if-conversion pass with support for conditional (predicated) 17714 execution. 17715 * New tail call and sibling call elimination optimizations. 17716 * New register renaming pass. 17717 * New (experimental) [3]static single assignment (SSA) representation 17718 support. 17719 * New dead-code elimination pass implemented using the SSA 17720 representation. 17721 * [4]Global null pointer test elimination. 17722 * [5]Global code hoisting/unification. 17723 * More builtins and optimizations for stdio.h, string.h and old BSD 17724 functions, as well as for ISO C99 functions. 17725 * New builtin __builtin_expect for giving hints to the branch 17726 predictor. 17727 17728 New Languages and Language specific improvements 17729 17730 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 17731 and supported, including the run-time library containing most 17732 common non-GUI Java classes, a bytecode interpreter, and the Boehm 17733 conservative garbage collector. Many bugs have been fixed. GCJ can 17734 compile Java source or Java bytecodes to either native code or Java 17735 class files, and supports native methods written in either the 17736 standard JNI or the more efficient and convenient CNI. 17737 * Here is a [6]partial list of C++ improvements, both new features 17738 and those no longer supported. 17739 * New C++ ABI. On the IA-64 platform GCC is capable of 17740 inter-operating with other IA-64 compilers. 17741 * The new ABI also significantly reduces the size of symbol and debug 17742 information. 17743 * New C++ support library and many C++ bug fixes, vastly improving 17744 our conformance to the ISO C++ standard. 17745 * New [7]inliner for C++. 17746 * Rewritten C preprocessor, integrated into the C, C++ and Objective 17747 C compilers, with very many improvements including ISO C99 support 17748 and [8]improvements to dependency generation. 17749 * Support for more [9]ISO C99 features. 17750 * Many improvements to support for checking calls to format functions 17751 such as printf and scanf, including support for ISO C99 format 17752 features, extensions from the Single Unix Specification and GNU 17753 libc 2.2, checking of strfmon formats and features to assist in 17754 auditing for format string security bugs. 17755 * New warnings for C code that may have undefined semantics because 17756 of violations of sequence point rules in the C standard (such as a 17757 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 17758 * Additional warning option -Wfloat-equal. 17759 * Improvements to -Wtraditional. 17760 * Fortran improvements are listed in [10]the Fortran documentation. 17761 17762 New Targets and Target Specific Improvements 17763 17764 * New x86 back end, generating much improved code. 17765 * Support for a generic i386-elf target contributed. 17766 * New option to emit x86 assembly code using Intel style syntax 17767 (-mintel-syntax). 17768 * HPUX 11 support contributed. 17769 * Improved PowerPC code generation, including scheduled prologue and 17770 epilogue. 17771 * Port of GCC to Intel's IA-64 processor contributed. 17772 * Port of GCC to Motorola's MCore 210 and 340 contributed. 17773 * New unified back-end for Arm, Thumb and StrongArm contributed. 17774 * Port of GCC to Intel's XScale processor contributed. 17775 * Port of GCC to Atmel's AVR microcontrollers contributed. 17776 * Port of GCC to Mitsubishi's D30V processor contributed. 17777 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 17778 processor family) contributed. 17779 * Port of GCC to Fujitsu's FR30 processor contributed. 17780 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 17781 * Port of GCC to Sun's picoJava processor core contributed. 17782 17783 Documentation improvements 17784 17785 * Substantially rewritten and improved C preprocessor manual. 17786 * Many improvements to other documentation. 17787 * Manpages for gcc, cpp and gcov are now generated automatically from 17788 the master Texinfo manual, eliminating the problem of manpages 17789 being out of date. (The generated manpages are only extracts from 17790 the full manual, which is provided in Texinfo form, from which 17791 info, HTML, other formats and a printed manual can be generated.) 17792 * Generated info files are included in the release tarballs alongside 17793 their Texinfo sources, avoiding problems on some platforms with 17794 building makeinfo as part of the GCC distribution. 17795 17796 Other significant improvements 17797 17798 * Garbage collection used internally by the compiler for most memory 17799 allocation instead of obstacks. 17800 * Lengauer and Tarjan algorithm used for computing dominators in the 17801 CFG. This algorithm can be significantly faster and more space 17802 efficient than our older algorithm. 17803 * gccbug script provided to assist in submitting bug reports to our 17804 bug tracking system. (Bug reports previously submitted directly to 17805 our mailing lists, for which you received no bug tracking number, 17806 should be submitted again using gccbug if you can reproduce the 17807 problem with GCC 3.0.) 17808 * The internal libgcc library is [11]built as a shared library on 17809 systems that support it. 17810 * Extensive testsuite included with GCC, with many new tests. In 17811 addition to tests for GCC bugs that have been fixed, many tests 17812 have been added for language features, compiler warnings and 17813 builtin functions. 17814 * Additional language-independent warning options -Wpacked, -Wpadded, 17815 -Wunreachable-code and -Wdisabled-optimization. 17816 * Target-independent options -falign-functions, -falign-loops and 17817 -falign-jumps. 17818 17819 Plus a great many bug fixes and almost all the [12]features found in 17820 GCC 2.95. 17821 17822 17823 For questions related to the use of GCC, please consult these web 17824 pages and the [13]GCC manuals. If that fails, the 17825 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17826 web pages and the development of GCC are welcome on our developer 17827 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 17828 archives. 17829 17830 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 17831 distribution of this entire article is permitted in any medium, 17832 provided this notice is preserved. 17833 17834 These pages are [18]maintained by the GCC team. Last modified 17835 2019-11-28[19]. 17836 17837 References 17838 17839 1. http://www.netbsd.org/ 17840 2. http://gcc.gnu.org/news/reorder.html 17841 3. http://gcc.gnu.org/news/ssa.html 17842 4. http://gcc.gnu.org/news/null.html 17843 5. http://gcc.gnu.org/news/unify.html 17844 6. http://gcc.gnu.org/gcc-3.0/c++features.html 17845 7. http://gcc.gnu.org/news/inlining.html 17846 8. http://gcc.gnu.org/news/dependencies.html 17847 9. http://gcc.gnu.org/c99status.html 17848 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 17849 11. http://gcc.gnu.org/gcc-3.0/libgcc.html 17850 12. http://gcc.gnu.org/gcc-2.95/features.html 17851 13. https://gcc.gnu.org/onlinedocs/ 17852 14. mailto:gcc-help (a] gcc.gnu.org 17853 15. mailto:gcc (a] gcc.gnu.org 17854 16. https://gcc.gnu.org/lists.html 17855 17. https://www.fsf.org/ 17856 18. https://gcc.gnu.org/about.html 17857 19. http://validator.w3.org/check/referer 17858 ====================================================================== 17859 http://gcc.gnu.org/gcc-3.0/caveats.html 17860 GCC 3.0 Caveats 17861 17862 * -fstrict-aliasing is now part of -O2 and higher optimization 17863 levels. This allows the compiler to assume the strictest aliasing 17864 rules applicable to the language being compiled. For C and C++, 17865 this activates optimizations based on the type of expressions. This 17866 optimization may thus break old, non-compliant code. 17867 * Enumerations are now properly promoted to int in function 17868 parameters and function returns. Normally this change is not 17869 visible, but when using -fshort-enums this is an ABI change. 17870 * The undocumented extension that allowed C programs to have a label 17871 at the end of a compound statement has been deprecated and may be 17872 removed in a future version. Programs that now generate a warning 17873 about this may be fixed by adding a null statement (a single 17874 semicolon) after the label. 17875 * The poorly documented extension that allowed string constants in C, 17876 C++ and Objective C to contain unescaped newlines has been 17877 deprecated and may be removed in a future version. Programs using 17878 this extension may be fixed in several ways: the bare newline may 17879 be replaced by \n, or preceded by \n\, or string concatenation may 17880 be used with the bare newline preceded by \n" and " placed at the 17881 start of the next line. 17882 * The Chill compiler is not included in GCC 3.0, because of the lack 17883 of a volunteer to convert it to use garbage collection. 17884 * Certain non-standard iostream methods from earlier versions of 17885 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 17886 ostream::form, and istream::gets. 17887 * The new C++ ABI is not yet fully supported by current (as of 17888 2001-07-01) releases and development versions of GDB, or any 17889 earlier versions. There is a problem setting breakpoints by line 17890 number, and other related issues that have been fixed in GCC 3.0 17891 but not yet handled in GDB: 17892 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 17893 17894 17895 For questions related to the use of GCC, please consult these web 17896 pages and the [2]GCC manuals. If that fails, the 17897 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17898 web pages and the development of GCC are welcome on our developer 17899 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 17900 17901 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 17902 distribution of this entire article is permitted in any medium, 17903 provided this notice is preserved. 17904 17905 These pages are [7]maintained by the GCC team. Last modified 17906 2019-11-28[8]. 17907 17908 References 17909 17910 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 17911 2. https://gcc.gnu.org/onlinedocs/ 17912 3. mailto:gcc-help (a] gcc.gnu.org 17913 4. mailto:gcc (a] gcc.gnu.org 17914 5. https://gcc.gnu.org/lists.html 17915 6. https://www.fsf.org/ 17916 7. https://gcc.gnu.org/about.html 17917 8. http://validator.w3.org/check/referer 17918 ====================================================================== 17919 http://gcc.gnu.org/gcc-2.95/index.html 17920 GCC 2.95 17921 17922 (This release series is no longer supported.) 17923 17924 March 16, 2001: The GNU project and the GCC developers are pleased to 17925 announce the release of GCC version 2.95.3. 17926 17927 Release History 17928 17929 GCC 2.95.3 17930 March 16, 2001 17931 17932 GCC 2.95.2 17933 October 27, 1999 17934 17935 GCC 2.95.1 17936 August 19, 1999 17937 17938 GCC 2.95 17939 July 31, 1999. This is the first release of GCC since the April 17940 1999 GCC/EGCS reunification and includes nearly a year's worth 17941 of new development and bugfixes. 17942 17943 References and Acknowledgements 17944 17945 GCC used to stand for the GNU C Compiler, but since the compiler 17946 supports several other languages aside from C, it now stands for the 17947 GNU Compiler Collection. 17948 17949 The whole suite has been extensively [1]regression tested and 17950 [2]package tested. It should be reliable and suitable for widespread 17951 use. 17952 17953 The compiler has several new optimizations, new targets, new languages 17954 and other new features. See the [3]new features page for a more 17955 complete list of new features found in the GCC 2.95 releases. 17956 17957 The sources include installation instructions in both HTML and 17958 plaintext forms in the install directory in the distribution. However, 17959 the most up to date installation instructions and [4]build/test status 17960 are on the web pages. We will update those pages as new information 17961 becomes available. 17962 17963 The GCC developers would like to thank the numerous people that have 17964 contributed new features, test results, bugfixes, etc to GCC. This 17965 [5]amazing group of volunteers is what makes GCC successful. 17966 17967 And finally, we can't in good conscience fail to mention some 17968 [6]caveats to using GCC 2.95. 17969 17970 Download GCC 2.95 from one of our many [7]mirror sites. 17971 17972 For additional information about GCC please see the [8]GCC project web 17973 server or contact the [9]GCC development mailing list. 17974 17975 17976 For questions related to the use of GCC, please consult these web 17977 pages and the [10]GCC manuals. If that fails, the 17978 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17979 web pages and the development of GCC are welcome on our developer 17980 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 17981 archives. 17982 17983 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 17984 distribution of this entire article is permitted in any medium, 17985 provided this notice is preserved. 17986 17987 These pages are [15]maintained by the GCC team. Last modified 17988 2019-11-28[16]. 17989 17990 References 17991 17992 1. http://gcc.gnu.org/gcc-2.95/regress.html 17993 2. http://gcc.gnu.org/gcc-2.95/othertest.html 17994 3. http://gcc.gnu.org/gcc-2.95/features.html 17995 4. http://gcc.gnu.org/gcc-2.95/buildstat.html 17996 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17997 6. http://gcc.gnu.org/gcc-2.95/caveats.html 17998 7. http://gcc.gnu.org/mirrors.html 17999 8. http://gcc.gnu.org/index.html 18000 9. mailto:gcc (a] gcc.gnu.org 18001 10. https://gcc.gnu.org/onlinedocs/ 18002 11. mailto:gcc-help (a] gcc.gnu.org 18003 12. mailto:gcc (a] gcc.gnu.org 18004 13. https://gcc.gnu.org/lists.html 18005 14. https://www.fsf.org/ 18006 15. https://gcc.gnu.org/about.html 18007 16. http://validator.w3.org/check/referer 18008 ====================================================================== 18009 http://gcc.gnu.org/gcc-2.95/features.html 18010 GCC 2.95 New Features 18011 18012 * General Optimizer Improvements: 18013 + [1]Localized register spilling to improve speed and code 18014 density especially on small register class machines. 18015 + [2]Global CSE using lazy code motion algorithms. 18016 + [3]Improved global constant/copy propagation. 18017 + [4]Improved control flow graph analysis and manipulation. 18018 + [5]Local dead store elimination. 18019 + [6]Memory Load hoisting/store sinking in loops. 18020 + [7]Type based alias analysis is enabled by default. Note this 18021 feature will expose bugs in the Linux kernel. Please refer to 18022 the FAQ (as shipped with GCC 2.95) for additional information 18023 on this issue. 18024 + Major revamp of GIV detection, combination and simplification 18025 to improve loop performance. 18026 + Major improvements to register allocation and reloading. 18027 * New Languages and Language specific improvements 18028 + [8]Many C++ improvements. 18029 + [9]Many Fortran improvements. 18030 + [10]Java front-end has been integrated. A [11]runtime library 18031 is available separately. 18032 + [12]ISO C99 support 18033 + [13]Chill front-end and runtime has been integrated. 18034 + Boehm garbage collector support in libobjc. 18035 + More support for various pragmas which appear in vendor 18036 include files 18037 * New Targets and Target Specific Improvements 18038 + [14]SPARC backend rewrite. 18039 + -mschedule=8000 will optimize code for PA8000 class 18040 processors; -mpa-risc-2-0 will generate code for PA2.0 18041 processors 18042 + Various micro-optimizations for the ia32 port. K6 18043 optimizations 18044 + Compiler will attempt to align doubles in the stack on the 18045 ia32 port 18046 + Alpha EV6 support 18047 + PowerPC 750 18048 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 18049 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 18050 + c3x, c4x 18051 + HyperSPARC 18052 + SparcLite86x 18053 + sh4 18054 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 18055 arm-linux) 18056 + vxWorks targets include support for vxWorks threads 18057 + StrongARM 110 and ARM9 support added. ARM Scheduling 18058 parameters rewritten. 18059 + Various changes to the MIPS port to avoid assembler macros, 18060 which in turn improves performance 18061 + Various performance improvements to the i960 port. 18062 + Major rewrite of ns32k port 18063 * Other significant improvements 18064 + [15]Ability to dump cfg information and display it using vcg. 18065 + The new faster scheme for fixing vendor header files is 18066 enabled by default. 18067 + Experimental internationalization support. 18068 + multibyte character support 18069 + Some compile-time speedups for pathological problems 18070 + Better support for complex types 18071 * Plus the usual mountain of bugfixes 18072 * Core compiler is based on the gcc2 development tree from Sept 30, 18073 1998, so we have all of the [16]features found in GCC 2.8. 18074 18075 Additional Changes in GCC 2.95.1 18076 18077 * Generic bugfixes and improvements 18078 + Various documentation fixes related to the GCC/EGCS merger. 18079 + Fix memory management bug which could lead to spurious aborts, 18080 core dumps or random parsing errors in the compiler. 18081 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 18082 support. 18083 + Fix infinite loop in the CSE optimizer. 18084 + Avoid undefined behavior in compiler FP emulation code 18085 + Fix install problem when prefix is overridden on the make 18086 install command. 18087 + Fix problem with unwanted installation of assert.h on some 18088 systems. 18089 + Fix problem with finding the wrong assembler in a single tree 18090 build. 18091 + Avoid increasing the known alignment of a register that is 18092 already known to be a pointer. 18093 * Platform specific bugfixes and improvements 18094 + Codegen bugfix for prologue/epilogue for cpu32 target. 18095 + Fix long long code generation bug for the Coldfire target. 18096 + Fix various aborts in the SH compiler. 18097 + Fix bugs in libgcc support library for the SH. 18098 + Fix alpha ev6 code generation bug. 18099 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 18100 AIX platforms. 18101 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 18102 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 18103 targets. 18104 + Fix weak symbol handling for rs6000/ppc svr4 targets. 18105 + Fix various problems with 64bit code generation for the 18106 rs6000/ppc port. 18107 + Fix codegen bug which caused tetex to be mis-compiled on the 18108 x86. 18109 + Fix compiler abort in new cfg code exposed by x86 port. 18110 + Fix out of range array reference in code convert flat 18111 registers to the x87 stacked FP register file. 18112 + Fix minor vxworks configuration bug. 18113 + Fix return type of bsearch for SunOS 4.x. 18114 * Language & Runtime specific fixes. 18115 + The G++ signature extension has been deprecated. It will be 18116 removed in the next major release of G++. Use of signatures 18117 will result in a warning from the compiler. 18118 + Several bugs relating to templates and namespaces were fixed. 18119 + A bug that caused crashes when combining templates with -g on 18120 DWARF1 platforms was fixed. 18121 + Pointers-to-members, virtual functions, and multiple 18122 inheritance should now work together correctly. 18123 + Some code-generation bugs relating to function try blocks were 18124 fixed. 18125 + G++ is a little bit more lenient with certain archaic 18126 constructs than in GCC 2.95. 18127 + Fix to prevent shared library version #s from bring truncated 18128 to 1 digit 18129 + Fix missing std:: in the libstdc++ library. 18130 + Fix stream locking problems in libio. 18131 + Fix problem in java compiler driver. 18132 18133 Additional Changes in GCC 2.95.2 18134 18135 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 18136 the optimizations performed by -fstrict-aliasing are valid according to 18137 the C and C++ standards, the optimization have caused some problems, 18138 particularly with old non-conforming code. 18139 18140 The GCC developers are experimenting with ways to warn users about code 18141 which violates the C/C++ standards, but those warnings are not ready 18142 for widespread use at this time. Rather than wait for those warnings 18143 the GCC developers have chosen to disable -fstrict-aliasing by default 18144 for the GCC 2.95.2 release. 18145 18146 We strongly encourage developers to find and fix code which violates 18147 the C/C++ standards as -fstrict-aliasing may be enabled by default in 18148 future releases. Use the option -fstrict-aliasing to re-enable these 18149 optimizations. 18150 * Generic bugfixes and improvements 18151 + Fix incorrectly optimized memory reference in global common 18152 subexpression elimination (GCSE) optimization pass. 18153 + Fix code generation bug in regmove.c in which it could 18154 incorrectly change a "const" value. 18155 + Fix bug in optimization of conditionals involving volatile 18156 memory references. 18157 + Avoid over-allocation of stack space for some procedures. 18158 + Fixed bug in the compiler which caused incorrect optimization 18159 of an obscure series of bit manipulations, shifts and 18160 arithmetic. 18161 + Fixed register allocator bug which caused teTeX to be 18162 mis-compiled on SPARC targets. 18163 + Avoid incorrect optimization of degenerate case statements for 18164 certain targets such as the ARM. 18165 + Fix out of range memory reference in the jump optimizer. 18166 + Avoid dereferencing null pointer in fix-header. 18167 + Fix test for GCC specific features so that it is possible to 18168 bootstrap with gcc-2.6.2 and older versions of GCC. 18169 + Fix typo in scheduler which could potentially cause out of 18170 range memory accesses. 18171 + Avoid incorrect loop reversal which caused incorrect code for 18172 certain loops on PowerPC targets. 18173 + Avoid incorrect optimization of switch statements on certain 18174 targets (for example the ARM). 18175 * Platform specific bugfixes and improvements 18176 + Work around bug in Sun V5.0 compilers which caused bootstrap 18177 comparison failures on SPARC targets. 18178 + Fix SPARC backend bug which caused aborts in final.c. 18179 + Fix sparc-hal-solaris2* configuration fragments. 18180 + Fix bug in sparc block profiling. 18181 + Fix obscure code generation bug for the PARISC targets. 18182 + Define __STDC_EXT__ for HPUX configurations. 18183 + Various POWERPC64 code generation bugfixes. 18184 + Fix abort for PPC targets using ELF (ex GNU/Linux). 18185 + Fix collect2 problems for AIX targets. 18186 + Correct handling of .file directive for PPC targets. 18187 + Fix bug in fix_trunc x86 patterns. 18188 + Fix x86 port to correctly pop the FP stack for functions that 18189 return structures in memory. 18190 + Fix minor bug in strlen x86 pattern. 18191 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 18192 + Fix template repository code to handle leading underscore in 18193 mangled names. 18194 + Fix weak/weak alias support for OpenBSD. 18195 + GNU/Linux for the ARM has C++ compatible include files. 18196 * Language & Runtime specific fixes. 18197 + Fix handling of constructor attribute in the C front-end which 18198 caused problems building the Chill runtime library on some 18199 targets. 18200 + Fix minor problem merging type qualifiers in the C front-end. 18201 + Fix aliasing bug for pointers and references (C/C++). 18202 + Fix incorrect "non-constant initializer bug" when -traditional 18203 or -fwritable-strings is enabled. 18204 + Fix build error for Chill front-end on SunOS. 18205 + Do not complain about duplicate instantiations when using 18206 -frepo (C++). 18207 + Fix array bounds handling in C++ front-end which caused 18208 problems with dwarf debugging information in some 18209 circumstances. 18210 + Fix minor namespace problem. 18211 + Fix problem linking java programs. 18212 18213 Additional Changes in GCC 2.95.3 18214 18215 * Generic bugfixes and improvements 18216 + Fix numerous problems that caused incorrect optimization in 18217 the register reloading code. 18218 + Fix numerous problems that caused incorrect optimization in 18219 the loop optimizer. 18220 + Fix aborts in the functions build_insn_chain and scan_loops 18221 under some circumstances. 18222 + Fix an alias analysis bug. 18223 + Fix an infinite compilation bug in the combiner. 18224 + A few problems with complex number support have been fixed. 18225 + It is no longer possible for gcc to act as a fork bomb when 18226 installed incorrectly. 18227 + The -fpack-struct option should be recognized now. 18228 + Fixed a bug that caused incorrect code to be generated due to 18229 a lost stack adjustment. 18230 * Platform specific bugfixes and improvements 18231 + Support building ARM toolchains hosted on Windows. 18232 + Fix attribute calculations in ARM toolchains. 18233 + arm-linux support has been improved. 18234 + Fix a PIC failure on sparc targets. 18235 + On ix86 targets, the regparm attribute should now work 18236 reliably. 18237 + Several updates for the h8300 port. 18238 + Fix problem building libio with glibc 2.2. 18239 18240 18241 For questions related to the use of GCC, please consult these web 18242 pages and the [17]GCC manuals. If that fails, the 18243 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18244 web pages and the development of GCC are welcome on our developer 18245 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 18246 archives. 18247 18248 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 18249 distribution of this entire article is permitted in any medium, 18250 provided this notice is preserved. 18251 18252 These pages are [22]maintained by the GCC team. Last modified 18253 2019-11-28[23]. 18254 18255 References 18256 18257 1. http://gcc.gnu.org/news/spill.html 18258 2. http://gcc.gnu.org/news/lcm.html 18259 3. http://gcc.gnu.org/news/cprop.html 18260 4. http://gcc.gnu.org/news/cfg.html 18261 5. http://gcc.gnu.org/news/dse.html 18262 6. http://gcc.gnu.org/news/hoist.html 18263 7. http://gcc.gnu.org/news/alias.html 18264 8. http://gcc.gnu.org/gcc-2.95/c++features.html 18265 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 18266 10. http://gcc.gnu.org/news/gcj-announce.txt 18267 11. http://gcc.gnu.org/news/javaannounce.html 18268 12. http://gcc.gnu.org/c99status.html 18269 13. http://gcc.gnu.org/news/chill.html 18270 14. http://gcc.gnu.org/news/sparc.html 18271 15. http://gcc.gnu.org/news/egcs-vcg.html 18272 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html 18273 17. https://gcc.gnu.org/onlinedocs/ 18274 18. mailto:gcc-help (a] gcc.gnu.org 18275 19. mailto:gcc (a] gcc.gnu.org 18276 20. https://gcc.gnu.org/lists.html 18277 21. https://www.fsf.org/ 18278 22. https://gcc.gnu.org/about.html 18279 23. http://validator.w3.org/check/referer 18280 ====================================================================== 18281 http://gcc.gnu.org/gcc-2.95/caveats.html 18282 GCC 2.95 Caveats 18283 18284 * GCC 2.95 will issue an error for invalid asm statements that had 18285 been silently accepted by earlier versions of the compiler. This is 18286 particularly noticeable when compiling older versions of the Linux 18287 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 18288 for more information on this issue. 18289 * GCC 2.95 implements type based alias analysis to disambiguate 18290 memory references. Some programs, particularly the Linux kernel 18291 violate ANSI/ISO aliasing rules and therefore may not operate 18292 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 18293 shipped with GCC 2.95) for more information on this issue. 18294 * GCC 2.95 has a known bug in its handling of complex variables for 18295 64bit targets. Instead of silently generating incorrect code, GCC 18296 2.95 will issue a fatal error for situations it can not handle. 18297 This primarily affects the Fortran community as Fortran makes more 18298 use of complex variables than C or C++. 18299 * GCC 2.95 has an integrated libstdc++, but does not have an 18300 integrated libg++. Furthermore old libg++ releases will not work 18301 with GCC 2.95. You can retrieve a recent copy of libg++ from the 18302 [1]GCC ftp server. 18303 Note most C++ programs only need libstdc++. 18304 * Exception handling may not work with shared libraries, particularly 18305 on alphas, hppas, rs6000/powerpc and mips based platforms. 18306 Exception handling is known to work on x86 GNU/Linux platforms with 18307 shared libraries. 18308 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 18309 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 18310 or EGCS 1.1. As a result it may be necessary to fix C++ code before 18311 it will compile with GCC 2.95. 18312 * G++ is also converting toward the ISO C++ standard; as a result 18313 code which was previously valid (and thus accepted by other 18314 compilers and older versions of g++) may no longer be accepted. The 18315 flag -fpermissive may allow some non-conforming code to compile 18316 with GCC 2.95. 18317 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 18318 1.1.x, EGCS 1.0.x or GCC 2.8.x. 18319 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 18320 between Sept 30, 1998 and April 30, 1999 (the official end of the 18321 GCC 2.8 project). Future GCC releases will include all the changes 18322 from the defunct GCC 2.8 sources. 18323 18324 18325 For questions related to the use of GCC, please consult these web 18326 pages and the [2]GCC manuals. If that fails, the 18327 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18328 web pages and the development of GCC are welcome on our developer 18329 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 18330 18331 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 18332 distribution of this entire article is permitted in any medium, 18333 provided this notice is preserved. 18334 18335 These pages are [7]maintained by the GCC team. Last modified 18336 2019-11-28[8]. 18337 18338 References 18339 18340 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 18341 2. https://gcc.gnu.org/onlinedocs/ 18342 3. mailto:gcc-help (a] gcc.gnu.org 18343 4. mailto:gcc (a] gcc.gnu.org 18344 5. https://gcc.gnu.org/lists.html 18345 6. https://www.fsf.org/ 18346 7. https://gcc.gnu.org/about.html 18347 8. http://validator.w3.org/check/referer 18348 ====================================================================== 18349 http://gcc.gnu.org/egcs-1.1/index.html 18350 EGCS 1.1 18351 18352 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 18353 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 18354 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 18355 18356 EGCS is a free software project to further the development of the GNU 18357 compilers using an open development environment. 18358 18359 EGCS 1.1 is a major new release of the EGCS compiler system. It has 18360 been [1]extensively tested and is believed to be stable and suitable 18361 for widespread use. 18362 18363 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 18364 development sources; it contains all of the new features found in GCC 18365 2.8.1 as well as all new development from GCC up to June 6, 1998. 18366 18367 EGCS 1.1 also contains many improvements and features not found in GCC 18368 or in older versions of EGCS: 18369 * Global common subexpression elimination and global constant/copy 18370 propagation (aka [2]gcse) 18371 * Ongoing improvements to the [3]alias analysis support to allow for 18372 better optimizations throughout the compiler. 18373 * Vastly improved [4]C++ compiler and integrated C++ runtime 18374 libraries. 18375 * Fixes for the /tmp symlink race security problems. 18376 * New targets including mips16, arm-thumb and 64 bit PowerPC. 18377 * Improvements to GNU Fortran (g77) compiler and runtime library made 18378 since g77 version 0.5.23. 18379 18380 See the [5]new features page for a more complete list of new features 18381 found in EGCS 1.1 releases. 18382 18383 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 18384 1.1: 18385 * General improvements and fixes 18386 + Avoid some stack overflows when compiling large functions. 18387 + Avoid incorrect loop invariant code motions. 18388 + Fix some core dumps on Linux kernel code. 18389 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 18390 + Fix code generation problem in gcse. 18391 + Various documentation related fixes. 18392 * g++/libstdc++ improvements and fixes 18393 + MT safe EH fix for setjmp/longjmp based exception handling. 18394 + Fix a few bad interactions between optimization and exception 18395 handling. 18396 + Fixes for demangling of template names starting with "__". 18397 + Fix a bug that would fail to run destructors in some cases 18398 with -O2. 18399 + Fix 'new' of classes with virtual bases. 18400 + Fix crash building Qt on the Alpha. 18401 + Fix failure compiling WIFEXITED macro on GNU/Linux. 18402 + Fix some -frepo failures. 18403 * g77 and libf2c improvements and fixes 18404 + Various documentation fixes. 18405 + Avoid compiler crash on RAND intrinsic. 18406 + Fix minor bugs in makefiles exposed by BSD make programs. 18407 + Define _XOPEN_SOURCE for libI77 build to avoid potential 18408 problems on some 64-bit systems. 18409 + Fix problem with implicit endfile on rewind. 18410 + Fix spurious recursive I/O errors. 18411 * platform specific improvements and fixes 18412 + Match all versions of UnixWare7. 18413 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 18414 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 18415 from unsigned ints to double precision floats. 18416 + Fix ARM ABI issue with NetBSD. 18417 + Fix a few arm code generation bugs. 18418 + Fixincludes will fix additional broken SCO OpenServer header 18419 files. 18420 + Fix a m68k backend bug which caused invalid offsets in reg+d 18421 addresses. 18422 + Fix problems with 64bit AIX 4.3 support. 18423 + Fix handling of long longs for varargs/stdarg functions on the 18424 ppc. 18425 + Minor fixes to CPP predefines for Windows. 18426 + Fix code generation problems with gpr<->fpr copies for 64bit 18427 ppc. 18428 + Fix a few coldfire code generation bugs. 18429 + Fix some more header file problems on SunOS 4.x. 18430 + Fix assert.h handling for RTEMS. 18431 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 18432 + Fix x86 compiler abort in reg-stack pass. 18433 + Fix cygwin/windows problem with section attributes. 18434 + Fix Alpha code generation problem exposed by SMP Linux 18435 kernels. 18436 + Fix typo in m68k 32->64bit integer conversion. 18437 + Make sure target libraries build with -fPIC for PPC & Alpha 18438 targets. 18439 18440 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 18441 1.1.1: 18442 * General improvements and fixes 18443 + Fix bug in loop optimizer which caused the SPARC (and 18444 potentially other) ports to segfault. 18445 + Fix infinite recursion in alias analysis and combiner code. 18446 + Fix bug in regclass preferencing. 18447 + Fix incorrect loop reversal which caused incorrect code to be 18448 generated for several targets. 18449 + Fix return value for builtin memcpy. 18450 + Reduce compile time for certain loops which exposed quadratic 18451 behavior in the loop optimizer. 18452 + Fix bug which caused volatile memory to be written multiple 18453 times when only one write was needed/desired. 18454 + Fix compiler abort in caller-save.c 18455 + Fix combiner bug which caused incorrect code generation for 18456 certain division by constant operations. 18457 + Fix incorrect code generation due to a bug in range check 18458 optimizations. 18459 + Fix incorrect code generation due to mis-handling of clobbered 18460 values in CSE. 18461 + Fix compiler abort/segfault due to incorrect register 18462 splitting when unrolling loops. 18463 + Fix code generation involving autoincremented addresses with 18464 ternary operators. 18465 + Work around bug in the scheduler which caused qt to be 18466 mis-compiled on some platforms. 18467 + Fix code generation problems with -fshort-enums. 18468 + Tighten security for temporary files. 18469 + Improve compile time for codes which make heavy use of 18470 overloaded functions. 18471 + Fix multiply defined constructor/destructor symbol problems. 18472 + Avoid setting bogus RPATH environment variable during 18473 bootstrap. 18474 + Avoid GNU-make dependencies in the texinfo subdir. 18475 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 18476 --enable-cpp=<dirname> can be used to specify an additional 18477 install directory for the cpp wrapper script. 18478 + Fix CSE bug which caused incorrect label-label refs to appear 18479 on some platforms. 18480 + Avoid linking in EH routines from libgcc if they are not 18481 needed. 18482 + Avoid obscure bug in aliasing code. 18483 + Fix bug in weak symbol handling. 18484 * Platform-specific improvements and fixes 18485 + Fix detection of PPro/PII on Unixware 7. 18486 + Fix compiler segfault when building spec99 and other programs 18487 for SPARC targets. 18488 + Fix code-generation bugs for integer and floating point 18489 conditional move instructions on the PPro/PII. 18490 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 18491 + Fix build failure for the arc port. 18492 + Fix floating point format configuration for i?86-gnu port. 18493 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 18494 threads are enabled. 18495 + Fix coldfire code generation bugs. 18496 + Fix "unrecognized insn" problems for Alpha and PPC ports. 18497 + Fix h8/300 code generation problem with floating point values 18498 in memory. 18499 + Fix unrecognized insn problems for the m68k port. 18500 + Fix namespace-pollution problem for the x86 port. 18501 + Fix problems with old assembler on x86 NeXT systems. 18502 + Fix PIC code-generation problems for the SPARC port. 18503 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 18504 + Fix minor ISO namespace violation in Alpha varargs/stdarg 18505 support. 18506 + Fix incorrect "braf" instruction usage for the SH port. 18507 + Fix minor bug in va-sh which prevented its use with -ansi. 18508 + Fix problems recognizing and supporting FreeBSD. 18509 + Handle OpenBSD systems correctly. 18510 + Minor fixincludes fix for Digital UNIX 4.0B. 18511 + Fix problems with ctors/dtors in SCO shared libraries. 18512 + Abort instead of generating incorrect code for PPro/PII 18513 floating point conditional moves. 18514 + Avoid multiply defined symbols on GNU/Linux systems using 18515 libc-5.4.xx. 18516 + Fix abort in alpha compiler. 18517 * Fortran-specific fixes 18518 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 18519 is in the documented, non-Y2K-compliant range of 0-99, instead 18520 of being returned as 100 in the year 2000. 18521 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 18522 milliseconds value properly in Values(8). 18523 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 18524 information properly in SArray(7). 18525 18526 Each release includes installation instructions in both HTML and 18527 plaintext forms (see the INSTALL directory in the toplevel directory of 18528 the distribution). However, we also keep the most up to date 18529 installation instructions and [6]build/test status on our web page. We 18530 will update those pages as new information becomes available. 18531 18532 The EGCS project would like to thank the numerous people that have 18533 contributed new features, test results, bugfixes, etc. This [7]amazing 18534 group of volunteers is what makes EGCS successful. 18535 18536 And finally, we can't in good conscience fail to mention some 18537 [8]caveats to using EGCS 1.1. 18538 18539 Download EGCS from egcs.cygnus.com (USA California). 18540 18541 The EGCS 1.1 release is also available on many mirror sites. 18542 [9]Goto mirror list to find a closer site. 18543 18544 18545 For questions related to the use of GCC, please consult these web 18546 pages and the [10]GCC manuals. If that fails, the 18547 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18548 web pages and the development of GCC are welcome on our developer 18549 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 18550 archives. 18551 18552 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 18553 distribution of this entire article is permitted in any medium, 18554 provided this notice is preserved. 18555 18556 These pages are [15]maintained by the GCC team. Last modified 18557 2019-11-28[16]. 18558 18559 References 18560 18561 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 18562 2. http://gcc.gnu.org/news/gcse.html 18563 3. http://gcc.gnu.org/news/alias.html 18564 4. http://gcc.gnu.org/egcs-1.1/c++features.html 18565 5. http://gcc.gnu.org/egcs-1.1/features.html 18566 6. http://gcc.gnu.org/egcs-1.1/buildstat.html 18567 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 18568 8. http://gcc.gnu.org/egcs-1.1/caveats.html 18569 9. http://gcc.gnu.org/mirrors.html 18570 10. https://gcc.gnu.org/onlinedocs/ 18571 11. mailto:gcc-help (a] gcc.gnu.org 18572 12. mailto:gcc (a] gcc.gnu.org 18573 13. https://gcc.gnu.org/lists.html 18574 14. https://www.fsf.org/ 18575 15. https://gcc.gnu.org/about.html 18576 16. http://validator.w3.org/check/referer 18577 ====================================================================== 18578 http://gcc.gnu.org/egcs-1.1/features.html 18579 EGCS 1.1 new features 18580 18581 * Integrated GNU Fortran (g77) compiler and runtime library with 18582 improvements, based on g77 version 0.5.23. 18583 * Vast improvements in the C++ compiler; so many they have [1]page of 18584 their own! 18585 * Compiler implements [2]global common subexpression elimination and 18586 global copy/constant propagation. 18587 * More major improvements in the [3]alias analysis code. 18588 * More major improvements in the exception handling code to improve 18589 performance, lower static overhead and provide the infrastructure 18590 for future improvements. 18591 * The infamous /tmp symlink race security problems have been fixed. 18592 * The regmove optimization pass has been nearly completely rewritten 18593 to improve performance of generated code. 18594 * The compiler now recomputes register usage information before local 18595 register allocation. By providing more accurate information to the 18596 priority based allocator, we get better register allocation. 18597 * The register reloading phase of the compiler optimizes spill code 18598 much better than in previous releases. 18599 * Some bad interactions between the register allocator and 18600 instruction scheduler have been fixed, resulting in much better 18601 code for certain programs. Additionally, we have tuned the 18602 scheduler in various ways to improve performance of generated code 18603 for some architectures. 18604 * The compiler's branch shortening algorithms have been significantly 18605 improved to work better on targets which align jump targets. 18606 * The compiler now supports -Os to prefer optimizing for code space 18607 over optimizing for code speed. 18608 * The compiler will now totally eliminate library calls which compute 18609 constant values. This primarily helps targets with no integer 18610 div/mul support and targets without floating point support. 18611 * The compiler now supports an extensive "--help" option. 18612 * cpplib has been greatly improved and may be suitable for limited 18613 use. 18614 * Memory footprint for the compiler has been significantly reduced 18615 for some pathological cases. 18616 * The time to build EGCS has been improved for certain targets 18617 (particularly the alpha and mips platforms). 18618 * Many infrastructure improvements throughout the compiler, plus the 18619 usual mountain of bugfixes and minor improvements. 18620 * Target dependent improvements: 18621 + SPARC port now includes V8 plus and V9 support as well as 18622 performance tuning for Ultra class machines. The SPARC port 18623 now uses the Haifa scheduler. 18624 + Alpha port has been tuned for the EV6 processor and has an 18625 optimized expansion of memcpy/bzero. The Alpha port now uses 18626 the Haifa scheduler. 18627 + RS6000/PowerPC: support for the Power64 architecture and AIX 18628 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 18629 + x86: Alignment of static store data and jump targets is per 18630 Intel recommendations now. Various improvements throughout the 18631 x86 port to improve performance on Pentium processors 18632 (including improved epilogue sequences for Pentium chips and 18633 backend improvements which should help register allocation on 18634 all x86 variants. Conditional move support has been fixed and 18635 enabled for PPro processors. The x86 port also better supports 18636 64bit operations now. Unixware 7, a System V Release 5 target, 18637 is now supported and SCO OpenServer targets can support GAS. 18638 + MIPS has improved multiply/multiply-add support and now 18639 includes mips16 ISA support. 18640 + M68k has many micro-optimizations and Coldfire fixes. 18641 * Core compiler is based on the GCC development tree from June 9, 18642 1998, so we have all of the [4]features found in GCC 2.8. 18643 18644 18645 For questions related to the use of GCC, please consult these web 18646 pages and the [5]GCC manuals. If that fails, the 18647 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18648 web pages and the development of GCC are welcome on our developer 18649 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 18650 18651 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 18652 distribution of this entire article is permitted in any medium, 18653 provided this notice is preserved. 18654 18655 These pages are [10]maintained by the GCC team. Last modified 18656 2019-11-28[11]. 18657 18658 References 18659 18660 1. http://gcc.gnu.org/egcs-1.1/c++features.html 18661 2. http://gcc.gnu.org/news/gcse.html 18662 3. http://gcc.gnu.org/news/alias.html 18663 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html 18664 5. https://gcc.gnu.org/onlinedocs/ 18665 6. mailto:gcc-help (a] gcc.gnu.org 18666 7. mailto:gcc (a] gcc.gnu.org 18667 8. https://gcc.gnu.org/lists.html 18668 9. https://www.fsf.org/ 18669 10. https://gcc.gnu.org/about.html 18670 11. http://validator.w3.org/check/referer 18671 ====================================================================== 18672 http://gcc.gnu.org/egcs-1.1/caveats.html 18673 EGCS 1.1 Caveats 18674 18675 * EGCS has an integrated libstdc++, but does not have an integrated 18676 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 18677 Lu has made a libg++-2.8.1.2 snapshot available which may work with 18678 EGCS. 18679 Note most C++ programs only need libstdc++. 18680 * Exception handling may not work with shared libraries, particularly 18681 on alphas, hppas, rs6000/powerpc and mips based platforms. 18682 Exception handling is known to work on x86-linux platforms with 18683 shared libraries. 18684 * Some versions of the Linux kernel have bugs which prevent them from 18685 being compiled or from running when compiled by EGCS. See the FAQ 18686 (as shipped with EGCS 1.1) for additional information. 18687 * In general, EGCS is more rigorous about rejecting invalid C++ code 18688 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 18689 a result it may be necessary to fix C++ code before it will compile 18690 with EGCS. 18691 * G++ is also converting toward the ISO C++ standard; as a result 18692 code which was previously valid (and thus accepted by other 18693 compilers and older versions of g++) may no longer be accepted. 18694 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 18695 or GCC 2.8.x due to changes necessary to support thread safe 18696 exception handling. 18697 18698 18699 For questions related to the use of GCC, please consult these web 18700 pages and the [1]GCC manuals. If that fails, the 18701 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18702 web pages and the development of GCC are welcome on our developer 18703 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 18704 18705 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 18706 distribution of this entire article is permitted in any medium, 18707 provided this notice is preserved. 18708 18709 These pages are [6]maintained by the GCC team. Last modified 18710 2019-11-28[7]. 18711 18712 References 18713 18714 1. https://gcc.gnu.org/onlinedocs/ 18715 2. mailto:gcc-help (a] gcc.gnu.org 18716 3. mailto:gcc (a] gcc.gnu.org 18717 4. https://gcc.gnu.org/lists.html 18718 5. https://www.fsf.org/ 18719 6. https://gcc.gnu.org/about.html 18720 7. http://validator.w3.org/check/referer 18721 ====================================================================== 18722 http://gcc.gnu.org/egcs-1.0/index.html 18723 EGCS 1.0 18724 18725 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 18726 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 18727 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 18728 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 18729 18730 EGCS is a collaborative effort involving several groups of hackers 18731 using an open development model to accelerate development and testing 18732 of GNU compilers and runtime libraries. 18733 18734 An important goal of EGCS is to allow wide scale testing of 18735 experimental features and optimizations; therefore, EGCS contains some 18736 features and optimizations which are still under development. However, 18737 EGCS has been carefully tested and should be comparable in quality to 18738 most GCC releases. 18739 18740 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 18741 development sources; it contains nearly all of the new features found 18742 in GCC 2.8. 18743 18744 EGCS 1.0 also contains many improvements and features not found in GCC 18745 2.7 and even the GCC 2.8 series (which was released after the original 18746 EGCS 1.0 release). 18747 * Integrated C++ runtime libraries, including support for most major 18748 GNU/Linux systems! 18749 * The integrated libstdc++ library includes a verbatim copy of SGI's 18750 STL release. 18751 * Integrated GNU Fortran compiler. 18752 * New instruction scheduler. 18753 * New alias analysis code. 18754 18755 See the [1]new features page for a more complete list of new features. 18756 18757 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 18758 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 18759 EGCS 1.0 release: 18760 * Add support for Red Hat 5.0 Linux and better support for Linux 18761 systems using glibc2. 18762 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 18763 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 18764 fix these problems. 18765 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 18766 handling interfaces. 18767 To avoid future compatibility problems, we strongly urge anyone who 18768 is planning on distributing shared libraries that contain C++ code 18769 to upgrade to EGCS 1.0.1 first. 18770 Soon after EGCS 1.0 was released, the GCC developers made some 18771 incompatible changes in libgcc's exception handling interfaces. 18772 These changes were needed to solve problems on some platforms. This 18773 means that GCC 2.8.0, when released, will not be seamlessly 18774 compatible with shared libraries built by EGCS 1.0. The reason is 18775 that the libgcc.a in GCC 2.8.0 will not contain a function needed 18776 by the old interface. 18777 The result of this is that there may be compatibility problems with 18778 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 18779 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 18780 and libgcc.a has the support routines for both the old and the new 18781 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 18782 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 18783 The maintainers of GCC 2.x have decided against including seamless 18784 support for the old interface in 2.8.0, since it was never 18785 "official", so to avoid future compatibility problems we recommend 18786 against distributing any shared libraries built by EGCS 1.0 that 18787 contain C++ code (upgrade to 1.0.1 and use that). 18788 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 18789 The x86 changes fix code generation errors exposed when building 18790 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 18791 The hppa change fixes a compiler abort when configured for use with 18792 RTEMS. 18793 The MIPS changes fix problems with the definition of LONG_MAX on 18794 newer systems, allow for command line selection of the target ABI, 18795 and fix one code generation problem. 18796 The rs6000/ppc change fixes some problems with passing structures 18797 to varargs/stdarg functions. 18798 * A few machine independent bugfixes, mostly to fix code generation 18799 errors when building Linux kernels or glibc. 18800 * Fix a few critical exception handling and template bugs in the C++ 18801 compiler. 18802 * Fix Fortran namelist bug on alphas. 18803 * Fix build problems on x86-solaris systems. 18804 18805 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 18806 serious problems in EGCS 1.0.1. 18807 * General improvements and fixes 18808 + Memory consumption significantly reduced, especially for 18809 templates and inline functions. 18810 + Fix various problems with glibc2.1. 18811 + Fix loop optimization bug exposed by rs6000/ppc port. 18812 + Fix to avoid potential code generation problems in jump.c. 18813 + Fix some undefined symbol problems in dwarf1 debug support. 18814 * g++/libstdc++ improvements and fixes 18815 + libstdc++ in the EGCS release has been updated and should be 18816 link compatible with libstdc++-2.8. 18817 + Various fixes in libio/libstdc++ to work better on GNU/Linux 18818 systems. 18819 + Fix problems with duplicate symbols on systems that do not 18820 support weak symbols. 18821 + Memory corruption bug and undefined symbols in bastring have 18822 been fixed. 18823 + Various exception handling fixes. 18824 + Fix compiler abort for very long thunk names. 18825 * g77 improvements and fixes 18826 + Fix compiler crash for omitted bound in Fortran CASE 18827 statement. 18828 + Add missing entries to g77 lang-options. 18829 + Fix problem with -fpedantic in the g77 compiler. 18830 + Fix "backspace" problem with g77 on alphas. 18831 + Fix x86 backend problem with Fortran literals and -fpic. 18832 + Fix some of the problems with negative subscripts for g77 on 18833 alphas. 18834 + Fixes for Fortran builds on cygwin32/mingw32. 18835 * platform specific improvements and fixes 18836 + Fix long double problems on x86 (exposed by glibc). 18837 + x86 ports define i386 again to keep imake happy. 18838 + Fix exception handling support on NetBSD ports. 18839 + Several changes to collect2 to fix many problems with AIX. 18840 + Define __ELF__ for GNU/Linux on rs6000. 18841 + Fix -mcall-linux problem on GNU/Linux on rs6000. 18842 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 18843 + Allow autoconf to select a proper install problem on AIX 3.1. 18844 + m68k port support includes -mcpu32 option as well as cpu32 18845 multilibs. 18846 + Fix stdarg bug for irix6. 18847 + Allow EGCS to build on irix5 without the gnu assembler. 18848 + Fix problem with static linking on sco5. 18849 + Fix bootstrap on sco5 with native compiler. 18850 + Fix for abort building newlib on H8 target. 18851 + Fix fixincludes handling of math.h on SunOS. 18852 + Minor fix for Motorola 3300 m68k systems. 18853 18854 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 18855 problems reported by Red Hat for builds of Red Hat 5.1. 18856 * Generic bugfixes: 18857 + Fix a typo in the libio library which resulted in incorrect 18858 behavior of istream::get. 18859 + Fix the Fortran negative array index problem. 18860 + Fix a major problem with the ObjC runtime thread support 18861 exposed by glibc2. 18862 + Reduce memory consumption of the Haifa scheduler. 18863 * Target specific bugfixes: 18864 + Fix one x86 floating point code generation bug exposed by 18865 glibc2 builds. 18866 + Fix one x86 internal compiler error exposed by glibc2 builds. 18867 + Fix profiling bugs on the Alpha. 18868 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 18869 + Fix rs6000/ppc bug when converting values from integer types 18870 to floating point types. 18871 18872 The EGCS 1.0 releases include installation instructions in both HTML 18873 and plaintext forms (see the INSTALL directory in the toplevel 18874 directory of the distribution). However, we also keep the most up to 18875 date installation instructions and [2]build/test status on our web 18876 page. We will update those pages as new information becomes available. 18877 18878 And, we can't in good conscience fail to mention some [3]caveats to 18879 using EGCS. 18880 18881 Update: Big thanks to Stanford for providing a high speed link for 18882 downloading EGCS (go.cygnus.com)! 18883 18884 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 18885 (USA California -- High speed link provided by Stanford). 18886 18887 The EGCS 1.0 release is also available many mirror sites. 18888 [4]Goto mirror list to find a closer site 18889 18890 We'd like to thank the numerous people that have contributed new 18891 features, test results, bugfixes, etc. Unfortunately, they're far too 18892 numerous to mention by name. 18893 18894 18895 For questions related to the use of GCC, please consult these web 18896 pages and the [5]GCC manuals. If that fails, the 18897 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18898 web pages and the development of GCC are welcome on our developer 18899 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 18900 18901 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 18902 distribution of this entire article is permitted in any medium, 18903 provided this notice is preserved. 18904 18905 These pages are [10]maintained by the GCC team. Last modified 18906 2019-11-28[11]. 18907 18908 References 18909 18910 1. http://gcc.gnu.org/egcs-1.0/features.html 18911 2. http://gcc.gnu.org/egcs-1.0/buildstat.html 18912 3. http://gcc.gnu.org/egcs-1.0/caveats.html 18913 4. http://gcc.gnu.org/mirrors.html 18914 5. https://gcc.gnu.org/onlinedocs/ 18915 6. mailto:gcc-help (a] gcc.gnu.org 18916 7. mailto:gcc (a] gcc.gnu.org 18917 8. https://gcc.gnu.org/lists.html 18918 9. https://www.fsf.org/ 18919 10. https://gcc.gnu.org/about.html 18920 11. http://validator.w3.org/check/referer 18921 ====================================================================== 18922 http://gcc.gnu.org/egcs-1.0/features.html 18923 EGCS 1.0 features 18924 18925 * Core compiler is based on the gcc2 development tree from Aug 2, 18926 1997, so we have most of the [1]features found in GCC 2.8. 18927 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 18928 * Vast improvements in the C++ compiler; so many they have [2]page of 18929 their own! 18930 * Integrated C++ runtime libraries, including support for most major 18931 GNU/Linux systems! 18932 * New instruction scheduler from IBM Haifa which includes support for 18933 function wide instruction scheduling as well as superscalar 18934 scheduling. 18935 * Significantly improved alias analysis code. 18936 * Improved register allocation for two address machines. 18937 * Significant code generation improvements for Fortran code on 18938 Alphas. 18939 * Various optimizations from the g77 project as well as improved loop 18940 optimizations. 18941 * Dwarf2 debug format support for some targets. 18942 * egcs libstdc++ includes the SGI STL implementation without changes. 18943 * As a result of these and other changes, egcs libstc++ is not binary 18944 compatible with previous releases of libstdc++. 18945 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 18946 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 18947 1.1), Support for RTEMS on several embedded targets, Support for 18948 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 18949 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 18950 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 18951 * RS6000/PowerPC ports generate code which can run on all 18952 RS6000/PowerPC variants by default. 18953 * -mcpu= and -march= switches for the x86 port to allow better 18954 control over how the x86 port generates code. 18955 * Includes the template repository patch (aka repo patch); note the 18956 new template code makes repo obsolete for ELF systems using gnu-ld 18957 such as GNU/Linux. 18958 * Plus the usual assortment of bugfixes and improvements. 18959 18960 18961 For questions related to the use of GCC, please consult these web 18962 pages and the [3]GCC manuals. If that fails, the 18963 [4]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18964 web pages and the development of GCC are welcome on our developer 18965 list at [5]gcc (a] gcc.gnu.org. All of [6]our lists have public archives. 18966 18967 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 18968 distribution of this entire article is permitted in any medium, 18969 provided this notice is preserved. 18970 18971 These pages are [8]maintained by the GCC team. Last modified 18972 2019-11-28[9]. 18973 18974 References 18975 18976 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html 18977 2. http://gcc.gnu.org/egcs-1.0/c++features.html 18978 3. https://gcc.gnu.org/onlinedocs/ 18979 4. mailto:gcc-help (a] gcc.gnu.org 18980 5. mailto:gcc (a] gcc.gnu.org 18981 6. https://gcc.gnu.org/lists.html 18982 7. https://www.fsf.org/ 18983 8. https://gcc.gnu.org/about.html 18984 9. http://validator.w3.org/check/referer 18985 ====================================================================== 18986 http://gcc.gnu.org/egcs-1.0/caveats.html 18987 EGCS 1.0 Caveats 18988 18989 * EGCS has an integrated libstdc++, but does not have an integrated 18990 libg++. Furthermore old libg++ releases will not work with egc; HJ 18991 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 18992 Note most C++ programs only need libstdc++. 18993 * Note that using -pedantic or -Wreturn-type can cause an explosion 18994 in the amount of memory needed for template-heavy C++ code, such as 18995 code that uses STL. Also note that -Wall includes -Wreturn-type, so 18996 if you use -Wall you will need to specify -Wno-return-type to turn 18997 it off. 18998 * Exception handling may not work with shared libraries, particularly 18999 on alphas, hppas, and mips based platforms. Exception handling is 19000 known to work on x86-linux platforms with shared libraries. 19001 * Some versions of the Linux kernel have bugs which prevent them from 19002 being compiled or from running when compiled by EGCS. See the FAQ 19003 (as shipped with EGCS 1.0) for additional information. 19004 * In general, EGCS is more rigorous about rejecting invalid C++ code 19005 or deprecated C++ constructs than G++ 2.7. As a result it may be 19006 necessary to fix C++ code before it will compile with EGCS. 19007 * G++ is also aggressively tracking the C++ standard; as a result 19008 code which was previously valid (and thus accepted by other 19009 compilers and older versions of G++) may no longer be accepted. 19010 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 19011 1.0.x and later releases should work with Red Hat Linux 5.0. 19012 19013 19014 For questions related to the use of GCC, please consult these web 19015 pages and the [1]GCC manuals. If that fails, the 19016 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 19017 web pages and the development of GCC are welcome on our developer 19018 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 19019 19020 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 19021 distribution of this entire article is permitted in any medium, 19022 provided this notice is preserved. 19023 19024 These pages are [6]maintained by the GCC team. Last modified 19025 2019-11-28[7]. 19026 19027 References 19028 19029 1. https://gcc.gnu.org/onlinedocs/ 19030 2. mailto:gcc-help (a] gcc.gnu.org 19031 3. mailto:gcc (a] gcc.gnu.org 19032 4. https://gcc.gnu.org/lists.html 19033 5. https://www.fsf.org/ 19034 6. https://gcc.gnu.org/about.html 19035 7. http://validator.w3.org/check/referer 19036 ====================================================================== 19037
Indexes created Tue Oct 27 00:25:24 UTC 2026