gcc/ONEWS

   1 This file contains information about GCC releases up to GCC 2.8.1, and
   2 some information about EGCS releases.  For more details of changes in
   3 EGCS releases, and details of changes in GCC 2.95 and more recent
   4 releases, see the release notes on the GCC web site and the file NEWS
   5 which contains the most relevant parts of those release notes in text
   6 form.
   7
   8 Noteworthy changes in GCC for EGCS 1.1.
   9 ---------------------------------------
  10
  11 The compiler now implements global common subexpression elimination (gcse) as
  12 well as global constant/copy propagation.  (link to gcse page).
  13
  14 More major improvements have been made to the alias analysis code.  A new
  15 option to allow front-ends to provide alias information to the optimizers
  16 has also been added (-fstrict-aliasing).  -fstrict-aliasing is off by default
  17 now, but will be enabled by default in the future. (link to alias page)
  18
  19 Major changes continue in the exception handling support.  This release
  20 includes some changes to reduce static overhead for exception handling.  It
  21 also includes some major changes to the setjmp/longjmp based EH mechanism to
  22 make it less pessimistic.  And finally, major infrastructure improvements
  23 to the dwarf2 EH mechanism have been made to make our EH support extensible.
  24
  25 We have fixed the infamous security problems with temporary files.
  26
  27 The "regmove" optimization pass has been nearly completely rewritten.  It now
  28 uses much more information about the target to determine profitability of
  29 transformations.
  30
  31 The compiler now recomputes register usage information immediately before
  32 register allocation.  Previously such information was only not kept up to
  33 date after instruction combination which led to poor register allocation
  34 choices by our priority based register allocator.
  35
  36 The register reloading phase of the compiler has been improved to better
  37 optimize spill code.  This primarily helps targets which generate lots of
  38 spills (like the x86 ports and many register poor embedded ports).
  39
  40 A few changes in the heuristics used by the register allocator and scheduler
  41 have been made which can significantly improve performance for certain
  42 applications.
  43
  44 The compiler's branch shortening algorithms have been significantly improved
  45 to work better on targets which align jump targets.
  46
  47 The compiler now supports the "ADDRESSOF" optimization which can significantly
  48 reduce the overhead for certain inline calls (and inline calls in general).
  49
  50 The compiler now supports a code size optimization switch (-Os).  When enabled
  51 the compiler will prefer optimizations which improve code size over those
  52 which improve code speed.
  53
  54 The compiler has been improved to completely eliminate library calls which
  55 compute constant values.  This is particularly useful on machines which
  56 do not have integer mul/div or floating point support on-chip.
  57
  58 GCC now supports a "--help" option to print detailed help information.
  59
  60 cpplib has been greatly improved.  It is probably usable for some sites now
  61 (major missing feature is trigraphs).
  62
  63 Memory footprint for the compiler has been significantly reduced for certain
  64 pathalogical cases.
  65
  66 Build time improvements for targets which support lots of sched parameters
  67 (alpha and mips primarily).
  68
  69 Compile time for certain programs using large constant initializers has been
  70 improved (affects glibc significantly).
  71
  72 Plus an incredible number of infrastructure changes, warning fixes, bugfixes
  73 and local optimizations.
  74
  75 Various improvements have been made to better support cross compilations.  They
  76 are still not easy, but they are improving.
  77
  78 Target specific NEWS
  79
  80     Sparc: Now includes V8 plus and V9 support, lots of tuning for Ultrasparcs
  81            and uses the Haifa scheduler by default.
  82
  83     Alpha: EV6 tuned, optimized expansion of memcpy/bzero.
  84
  85     x86: Data in the static store is aligned per Intel recommendations.  Jump
  86          targets are aligned per Intel recommendations.  Improved epilogue
  87          sequences for Pentium chips.  Backend improvements which should help
  88          register allocation on all x86 variants.  Support for PPro conditional
  89          move instructions has been fixed and enabled.  Random changes
  90          throughout the port to make generated code more Pentium friendly.
  91          Improved support for 64bit integer operations.
  92          Unixware 7, a System V Release 5 target is now supported.
  93          SCO OpenServer targets can support GAS.  See gcc/INSTALL for details.
  94
  95     RS6000/PowerPC: Includes AIX4.3 support as well as PowerPC64 support.
  96                     Haifa instruction scheduling is enabled by default now.
  97
  98     MIPS: Multiply/Multiply-Add support has been largely rewritten to generate
  99           more efficient code.  Includes mips16 support.
 100
 101     M68K: Various micro-optimizations and Coldfire fixes.
 102
 103     M32r: Major improvements to this port.
 104
 105     Arm: Includes Thumb and super interworking support.
 106
 107 EGCS includes all gcc2 changes up to and including the June 9, 1998 snapshot.
 108
 109
 110 Noteworthy changes in GCC version 2.8.1
 111 ---------------------------------------
 112
 113 Numerous bugs have been fixed and some minor performance
 114 improvements (compilation speed) have been made.
 115
 116 Noteworthy changes in GCC version 2.8.0
 117 ---------------------------------------
 118
 119 A major change in this release is the addition of a framework for
 120 exception handling, currently used by C++.  Many internal changes and
 121 optimization improvements have been made.  These increase the
 122 maintainability and portability of GCC.  GCC now uses autoconf to
 123 compute many host parameters.
 124
 125 The following lists changes that add new features or targets.
 126
 127 See cp/NEWS for new features of C++ in this release.
 128
 129 New tools and features:
 130
 131     The Dwarf 2 debugging information format is supported on ELF systems, and
 132     is the default for -ggdb where available.  It can also be used for C++.
 133     The Dwarf version 1 debugging format is also permitted for C++, but
 134     does not work well.
 135
 136     gcov.c is provided for test coverage analysis and branch profiling
 137     analysis is also supported; see -fprofile-arcs, -ftest-coverage,
 138     and -fbranch-probabilities.
 139
 140     Support for the Checker memory checking tool.
 141
 142     New switch, -fstack-check, to check for stack overflow on systems that
 143     don't have such built into their ABI.
 144
 145     New switches, -Wundef and -Wno-undef to warn if an undefined identifier
 146     is evaluated in an #if directive.
 147
 148     Options -Wall and -Wimplicit now cause GCC to warn about implicit int
 149     in declarations (e.g. `register i;'), since the C Standard committee
 150     has decided to disallow this in the next revision of the standard;
 151     -Wimplicit-function-declarations and -Wimplicit-int are subsets of
 152     this.
 153
 154     Option -Wsign-compare causes GCC to warn about comparison of signed and
 155     unsigned values.
 156
 157     Add -dI option of cccp for cxref.
 158
 159 New features in configuration, installation and specs file handling:
 160
 161     New option --enable-c-cpplib to configure script.
 162
 163     You can use --with-cpu on the configure command to specify the default
 164     CPU that GCC should generate code for.
 165
 166     The -specs=file switch allows you to override default specs used in
 167     invoking programs like cc1, as, etc.
 168
 169     Allow including one specs file from another and renaming a specs
 170     variable.
 171
 172     You can now relocate all GCC files with a single environment variable
 173     or a registry entry under Windows 95 and Windows NT.
 174
 175 Changes in Objective-C:
 176
 177     The Objective-C Runtime Library has been made thread-safe.
 178
 179     The Objective-C Runtime Library contains an interface for creating
 180     mutexes, condition mutexes, and threads; it requires a back-end
 181     implementation for the specific platform and/or thread package.
 182     Currently supported are DEC/OSF1, IRIX, Mach, OS/2, POSIX, PCThreads,
 183     Solaris, and Windows32.  The --enable-threads parameter can be used
 184     when configuring GCC to enable and select a thread back-end.
 185
 186     Objective-C is now configured as separate front-end language to GCC,
 187     making it more convenient to conditionally build it.
 188
 189     The internal structures of the Objective-C Runtime Library have
 190     changed sufficiently to warrant a new version number; now version 8.
 191     Programs compiled with an older version must be recompiled.
 192
 193     The Objective-C Runtime Library can be built as a DLL on Windows 95
 194     and Windows NT systems.
 195
 196     The Objective-C Runtime Library implements +load.
 197
 198 The following new targets are supported (see also list under each
 199 individual CPU below):
 200
 201     Embedded target m32r-elf.
 202     Embedded Hitachi Super-H using ELF.
 203     RTEMS real-time system on various CPU targets.
 204     ARC processor.
 205     NEC V850 processor.
 206     Matsushita MN10200 processor.
 207     Matsushita MN10300 processor.
 208     Sparc and PowerPC running on VxWorks.
 209     Support both glibc versions 1 and 2 on Linux-based GNU systems.
 210
 211 New features for DEC Alpha systems:
 212
 213     Allow detailed specification of IEEE fp support:
 214       -mieee, -mieee-with-inexact, and -mieee-conformant
 215       -mfp-trap-mode=xxx, -mfp-round-mode=xxx, -mtrap-precision=xxx
 216     -mcpu=xxx for CPU selection
 217     Support scheduling parameters for EV5.
 218     Add support for BWX, CIX, and MAX instruction set extensions.
 219     Support Linux-based GNU systems.
 220     Support VMS.
 221
 222 Additional supported processors and systems for MIPS targets:
 223
 224     MIPS4 instruction set.
 225     R4100, R4300 and R5000 processors.
 226     N32 and N64 ABI.
 227     IRIX 6.2.
 228     SNI SINIX.
 229
 230 New features for Intel x86 family:
 231
 232     Add scheduling parameters for Pentium and Pentium Pro.
 233     Support stabs on Solaris-x86.
 234     Intel x86 processors running the SCO OpenServer 5 family.
 235     Intel x86 processors running DG/UX.
 236     Intel x86 using Cygwin32 or Mingw32 on Windows 95 and Windows NT.
 237
 238 New features for Motorola 68k family:
 239
 240     Support for 68060 processor.
 241     More consistent switches to specify processor.
 242     Motorola 68k family running AUX.
 243     68040 running pSOS, ELF object files, DBX debugging.
 244     Coldfire variant of Motorola m68k family.
 245
 246 New features for the HP PA RISC:
 247
 248     -mspace and -mno-space
 249     -mlong-load-store and -mno-long-load-store
 250     -mbig-switch -mno-big-switch
 251
 252     GCC on the PA requires either gas-2.7 or the HP assembler; for best
 253     results using GAS is highly recommended.  GAS is required for -g and
 254     exception handling support.
 255
 256 New features for SPARC-based systems:
 257
 258     The ultrasparc cpu.
 259     The sparclet cpu, supporting only a.out file format.
 260     Sparc running SunOS 4 with the GNU assembler.
 261     Sparc running the Linux-based GNU system.
 262     Embedded Sparc processors running the ELF object file format.
 263     -mcpu=xxx
 264     -mtune=xxx
 265     -malign-loops=xxx
 266     -malign-jumps=xxx
 267     -malign-functions=xxx
 268     -mimpure-text and -mno-impure-text
 269
 270     Options -mno-v8 and -mno-sparclite are no longer supported on SPARC
 271     targets.  Options -mcypress, -mv8, -msupersparc, -msparclite, -mf930,
 272     and -mf934 are deprecated and will be deleted in GCC 2.9.  Use
 273     -mcpu=xxx instead.
 274
 275 New features for rs6000 and PowerPC systems:
 276
 277     Solaris 2.51 running on PowerPC's.
 278     The Linux-based GNU system running on PowerPC's.
 279     -mcpu=604e,602,603e,620,801,823,mpc505,821,860,power2
 280     -mtune=xxx
 281     -mrelocatable-lib, -mno-relocatable-lib
 282     -msim, -mmve, -memb
 283     -mupdate, -mno-update
 284     -mfused-madd, -mno-fused-madd
 285
 286     -mregnames
 287     -meabi
 288     -mcall-linux, -mcall-solaris, -mcall-sysv-eabi, -mcall-sysv-noeabi
 289     -msdata, -msdata=none, -msdata=default, -msdata=sysv, -msdata=eabi
 290     -memb, -msim, -mmvme
 291     -myellowknife, -mads
 292     wchar_t is now of type long as per the ABI, not unsigned short.
 293     -p/-pg support
 294     -mcpu=403 now implies -mstrict-align.
 295     Implement System V profiling.
 296
 297     Aix 4.1 GCC targets now default to -mcpu=common so that programs
 298     compiled can be moved between rs6000 and powerpc based systems.  A
 299     consequence of this is that -static won't work, and that some programs
 300     may be slightly slower.
 301
 302     You can select the default value to use for -mcpu=xxx on rs6000 and
 303     powerpc targets by using the --with-cpu=xxx option when configuring the
 304     compiler.  In addition, a new options, -mtune=xxx was added that
 305     selects the machine to schedule for but does not select the
 306     architecture level.
 307
 308     Directory names used for storing the multilib libraries on System V
 309     and embedded PowerPC systems have been shortened to work with commands
 310     like tar that have fixed limits on pathname size.
 311
 312 New features for the Hitachi H8/300(H):
 313
 314     -malign-300
 315     -ms (for the Hitachi H8/S processor)
 316     -mint32
 317
 318 New features for the ARM:
 319
 320     -march=xxx, -mtune=xxx, -mcpu=xxx
 321     Support interworking with Thumb code.
 322     ARM processor with a.out object format, COFF, or AOF assembler.
 323     ARM on "semi-hosted" platform.
 324     ARM running NetBSD.
 325     ARM running the Linux-based GNU system.
 326
 327 New feature for Solaris systems:
 328
 329     GCC installation no longer makes a copy of system include files,
 330     thus insulating GCC better from updates to the operating system.
 331
 332 \f
 333 Noteworthy changes in GCC version 2.7.2
 334 ---------------------------------------
 335
 336 A few bugs have been fixed (most notably the generation of an
 337 invalid assembler opcode on some RS/6000 systems).
 338
 339 Noteworthy changes in GCC version 2.7.1
 340 ---------------------------------------
 341
 342 This release fixes numerous bugs (mostly minor) in GCC 2.7.0, but
 343 also contains a few new features, mostly related to specific targets.
 344
 345 Major changes have been made in code to support Windows NT.
 346
 347 The following new targets are supported:
 348
 349         2.9 BSD on PDP-11
 350         Linux on m68k
 351         HP/UX version 10 on HP PA RISC (treated like version 9)
 352         DEC Alpha running Windows NT
 353
 354 When parsing C, GCC now recognizes C++ style `//' comments unless you
 355 specify `-ansi' or `-traditional'.
 356
 357 The PowerPC System V targets (powerpc-*-sysv, powerpc-*-eabi) now use the
 358 calling sequence specified in the System V Application Binary Interface
 359 Processor Supplement (PowerPC Processor ABI Supplement) rather than the calling
 360 sequence used in GCC version 2.7.0.  That calling sequence was based on the AIX
 361 calling sequence without function descriptors.  To compile code for that older
 362 calling sequence, either configure the compiler for powerpc-*-eabiaix or use
 363 the -mcall-aix switch when compiling and linking.
 364 \f
 365 Noteworthy changes in GCC version 2.7.0
 366 ---------------------------------------
 367
 368 GCC now works better on systems that use ".obj" and ".exe" instead of
 369 ".o" and no extension.  This involved changes to the driver program,
 370 gcc.c, to convert ".o" names to ".obj" and to GCC's Makefile to use
 371 ".obj" and ".exe" in filenames that are not targets.  In order to
 372 build GCC on such systems, you may need versions of GNU make and/or
 373 compatible shells.  At this point, this support is preliminary.
 374
 375 Object file extensions of ".obj" and executable file extensions of
 376 ".exe" are allowed when using appropriate version of GNU Make.
 377
 378 Numerous enhancements were made to the __attribute__ facility including
 379 more attributes and more places that support it.  We now support the
 380 "packed", "nocommon", "noreturn", "volatile", "const", "unused",
 381 "transparent_union", "constructor", "destructor", "mode", "section",
 382 "align", "format", "weak", and "alias" attributes.  Each of these
 383 names may also be specified with added underscores, e.g., "__packed__".
 384 __attribute__ may now be applied to parameter definitions, function
 385 definitions, and structure, enum, and union definitions.
 386
 387 GCC now supports returning more structures in registers, as specified by
 388 many calling sequences (ABIs), such as on the HP PA RISC.
 389
 390 A new option '-fpack-struct' was added to automatically pack all structure
 391 members together without holes.
 392
 393 There is a new library (cpplib) and program (cppmain) that at some
 394 point will replace cpp (aka cccp).  To use cppmain as cpp now, pass
 395 the option CCCP=cppmain to make.  The library is already used by the
 396 fix-header program, which should speed up the fixproto script.
 397
 398 New options for supported targets:
 399
 400     GNU on many targets.
 401     NetBSD on MIPS, m68k, VAX, and x86.
 402     LynxOS on x86, m68k, Sparc, and RS/6000.
 403     VxWorks on many targets.
 404
 405     Windows/NT on x86 architecture.  Initial support for Windows/NT on Alpha
 406     (not fully working).
 407
 408     Many embedded targets, specifically UDI on a29k, aout, coff, elf,
 409     and vsta "operating systems" on m68k, m88k, mips, sparc, and x86.
 410
 411 Additional support for x86 (i386, i486, and Pentium):
 412
 413     Work with old and new linkers for Linux-based GNU systems,
 414         supporting both a.out and ELF.
 415     FreeBSD on x86.
 416     Stdcall convention.
 417     -malign-double, -mregparm=, -malign-loops= and -malign-jumps=  switches.
 418     On ISC systems, support -Xp like -posix.
 419
 420 Additions for RS/6000:
 421
 422     Instruction scheduling information for PowerPC 403.
 423     AIX 4.1 on PowerPC.
 424     -mstring and -mno-string.
 425     -msoft-float and floating-point emulation included.
 426     Preliminary support for PowerPC System V.4 with or without the GNU as.
 427     Preliminary support for EABI.
 428     Preliminary support for 64-bit systems.
 429     Both big and little endian systems.
 430
 431 New features for MIPS-based systems:
 432
 433     r4650.
 434     mips4 and R8000.
 435     Irix 6.0.
 436     64-bit ABI.
 437     Allow dollar signs in labels on SGI/Irix 5.x.
 438
 439 New support for HP PA RISC:
 440
 441     Generation of PIC (requires binutils-2.5.2.u6 or later).
 442     HP-UX version 9 on HP PA RISC (dynamically links even with -g).
 443     Processor variants for HP PA RISC: 700, 7100, and 7100LC.
 444     Automatic generation of long calls when needed.
 445     -mfast-indirect-calls for kernels and static binaries.
 446
 447     The called routine now copies arguments passed by invisible reference,
 448     as required by the calling standard.
 449
 450 Other new miscellaneous target-specific support:
 451
 452     -mno-multm on a29k.
 453     -mold-align for i960.
 454     Configuration for "semi-hosted" ARM.
 455     -momit-leaf-frame-pointer for M88k.
 456     SH3 variant of Hitachi Super-H and support both big and little endian.
 457
 458 Changes to Objective-C:
 459
 460     Bare-bones implementation of NXConstantString has been added,
 461     which is invoked by the @"string" directive.
 462
 463     Class * has been changed to Class to conform to the NextSTEP and
 464     OpenStep runtime.
 465
 466     Enhancements to make dynamic loading easier.
 467
 468     The module version number has been updated to Version 7, thus existing
 469     code will need to be recompiled to use the current run-time library.
 470
 471 GCC now supports the ISO Normative Addendum 1 to the C Standard.
 472 As a result:
 473
 474     The header <iso646.h> defines macros for C programs written
 475     in national variants of ISO 646.
 476
 477     The following digraph tokens are supported:
 478         <:      :>      <%      %>      %:      %:%:
 479     These behave like the following, respectively:
 480         [       ]       {       }       #       ##
 481
 482     Digraph tokens are supported unless you specify the `-traditional'
 483     option; you do not need to specify `-ansi' or `-trigraphs'.  Except
 484     for contrived and unlikely examples involving preprocessor
 485     stringizing, digraph interpretation doesn't change the meaning of
 486     programs; this is unlike trigraph interpretation, which changes the
 487     meanings of relatively common strings.
 488
 489     The macro __STDC_VERSION__ has the value 199409L.
 490
 491   As usual, for full conformance to the standard, you also need a
 492   C library that conforms.
 493
 494 The following lists changes that have been made to g++.  If some
 495 features mentioned below sound unfamiliar, you will probably want to
 496 look at the recently-released public review copy of the C++ Working
 497 Paper.  For PostScript and PDF (Adobe Acrobat) versions, see the
 498 archive at ftp://research.att.com/dist/stdc++/WP.  For HTML and ASCII
 499 versions, see ftp://ftp.cygnus.com/pub/g++.  On the web, see
 500 http://www.cygnus.com/~mrs/wp-draft.
 501
 502 The scope of variables declared in the for-init-statement has been changed
 503 to conform to http://www.cygnus.com/~mrs/wp-draft/stmt.html#stmt.for; as a
 504 result, packages such as groff 1.09 will not compile unless you specify the
 505 -fno-for-scope flag.  PLEASE DO NOT REPORT THIS AS A BUG; this is a change
 506 mandated by the C++ standardization committee.
 507
 508 Binary incompatibilities:
 509
 510     The builtin 'bool' type is now the size of a machine word on RISC targets,
 511     for code efficiency; it remains one byte long on CISC targets.
 512
 513     Code that does not use #pragma interface/implementation will most
 514     likely shrink dramatically, as g++ now only emits the vtable for a
 515     class in the translation unit where its first non-inline, non-abstract
 516     virtual function is defined.
 517
 518     Classes that do not define the copy constructor will sometimes be
 519     passed and returned in registers.  This may illuminate latent bugs in
 520     your code.
 521
 522 Support for automatic template instantiation has *NOT* been added, due
 523 to a disagreement over design philosophies.
 524
 525 Support for exception handling has been improved; more targets are now
 526 supported, and throws will use the RTTI mechanism to match against the
 527 catch parameter type.  Optimization is NOT SUPPORTED with
 528 -fhandle-exceptions; no need to report this as a bug.
 529
 530 Support for Run-Time Type Identification has been added with -frtti.
 531 This support is still in alpha; one major restriction is that any file
 532 compiled with -frtti must include <typeinfo.h>.
 533
 534 Preliminary support for namespaces has been added.  This support is far
 535 from complete, and probably not useful.
 536
 537 Synthesis of compiler-generated constructors, destructors and
 538 assignment operators is now deferred until the functions are used.
 539
 540 The parsing of expressions such as `a ? b : c = 1' has changed from
 541 `(a ? b : c) = 1' to `a : b ? (c = 1)'.
 542
 543 The code generated for testing conditions, especially those using ||
 544 and &&, is now more efficient.
 545
 546 The operator keywords and, and_eq, bitand, bitor, compl, not, not_eq,
 547 or, or_eq, xor and xor_eq are now supported.  Use -ansi or
 548 -foperator-names to enable them.
 549
 550 The 'explicit' keyword is now supported.  'explicit' is used to mark
 551 constructors and type conversion operators that should not be used
 552 implicitly.
 553
 554 g++ now accepts the typename keyword, though it currently has no
 555 semantics; it can be a no-op in the current template implementation.
 556 You may want to start using it in your code, however, since the
 557 pending rewrite of the template implementation to compile STL properly
 558 (perhaps for 2.8.0, perhaps not) will require you to use it as
 559 indicated by the current draft.
 560
 561 Handling of user-defined type conversion has been overhauled so that
 562 type conversion operators are now found and used properly in
 563 expressions and function calls.
 564
 565 -fno-strict-prototype now only applies to function declarations with
 566 "C" linkage.
 567
 568 g++ now warns about 'if (x=0)' with -Wparentheses or -Wall.
 569
 570 #pragma weak and #pragma pack are supported on System V R4 targets, as
 571 are various other target-specific #pragmas supported by gcc.
 572
 573 new and delete of const types is now allowed (with no additional
 574 semantics).
 575
 576 Explicit instantiation of template methods is now supported.  Also,
 577 'inline template class foo<int>;' can be used to emit only the vtable
 578 for a template class.
 579
 580 With -fcheck-new, g++ will check the return value of all calls to
 581 operator new, and not attempt to modify a returned null pointer.
 582
 583 The template instantiation code now handles more conversions when
 584 passing to a parameter that does not depend on template arguments.
 585 This means that code like 'string s; cout << s;' now works.
 586
 587 Invalid jumps in a switch statement past declarations that require
 588 initializations are now caught.
 589
 590 Functions declared 'extern inline' now have the same linkage semantics
 591 as inline member functions.  On supported targets, where previously
 592 these functions (and vtables, and template instantiations) would have
 593 been defined statically, they will now be defined as weak symbols so
 594 that only one out-of-line definition is used.
 595
 596 collect2 now demangles linker output, and c++filt has become part of
 597 the gcc distribution.
 598 \f
 599 Noteworthy changes in GCC version 2.6.3:
 600
 601 A few more bugs have been fixed.
 602
 603 Noteworthy changes in GCC version 2.6.2:
 604
 605 A few bugs have been fixed.
 606
 607 Names of attributes can now be preceded and followed by double underscores.
 608
 609 Noteworthy changes in GCC version 2.6.1:
 610
 611 Numerous (mostly minor) bugs have been fixed.
 612
 613 The following new configurations are supported:
 614
 615         GNU on x86 (instead of treating it like MACH)
 616         NetBSD on Sparc and Motorola 68k
 617         AIX 4.1 on RS/6000 and PowerPC systems
 618         Sequent DYNIX/ptx 1.x and 2.x.
 619         Both COFF and ELF configurations on AViiON without using /bin/gcc
 620         Windows/NT on x86 architecture; preliminary
 621         AT&T DSP1610 digital signal processor chips
 622         i960 systems on bare boards using COFF
 623         PDP11; target only and not extensively tested
 624
 625 The -pg option is now supported for Alpha under OSF/1 V3.0 or later.
 626
 627 Files with an extension of ".c++" are treated as C++ code.
 628
 629 The -Xlinker and -Wl arguments are now passed to the linker in the
 630 position they were specified on the command line.  This makes it
 631 possible, for example, to pass flags to the linker about specific
 632 object files.
 633
 634 The use of positional arguments to the configure script is no longer
 635 recommended.  Use --target= to specify the target; see the GCC manual.
 636
 637 The 386 now supports two new switches: -mreg-alloc=<string> changes
 638 the default register allocation order used by the compiler, and
 639 -mno-wide-multiply disables the use of the mul/imul instructions that
 640 produce 64 bit results in EAX:EDX from 32 bit operands to do long long
 641 multiplies and 32-bit division by constants.
 642 \f
 643 Noteworthy changes in GCC version 2.6.0:
 644
 645 Numerous bugs have been fixed, in the C and C++ front-ends, as
 646 well as in the common compiler code.
 647
 648 This release includes the C, Objective-C, and C++ compilers.  However,
 649 we have moved the files for the C++ compiler (G++) files to a
 650 subdirectory, cp.  Subsequent releases of GCC will split these files
 651 to a separate TAR file.
 652
 653 The G++ team has been tracking the development of the ANSI standard for C++.
 654 Here are some new features added from the latest working paper:
 655
 656         * built-in boolean type 'bool', with constants 'true' and 'false'.
 657         * array new and delete (operator new [] and delete []).
 658         * WP-conforming lifetime of temporaries.
 659         * explicit instantiation of templates (template class A<int>;),
 660           along with an option (-fno-implicit-templates) to disable emission
 661           of implicitly instantiated templates, obsoletes -fexternal-templates.
 662         * static member constants (static const int foo = 4; within the
 663           class declaration).
 664
 665 Many error messages have been improved to tell the user more about the
 666 problem.  Conformance checking with -pedantic-errors has been
 667 improved.  G++ now compiles Fresco.
 668
 669 There is now an experimental implementation of virtual functions using
 670 thunks instead of Cfront-style vtables, enabled with -fvtable-thunks.
 671 This option also enables a heuristic which causes the compiler to only
 672 emit the vtable in the translation unit where its first non-inline
 673 virtual function is defined; using this option and
 674 -fno-implicit-templates, users should be able to avoid #pragma
 675 interface/implementation altogether.
 676
 677 Signatures have been added as a GNU C++ extension.  Using the option
 678 -fhandle-signatures, users are able to turn on recognition of
 679 signatures.  A short introduction on signatures is in the section
 680 `Extension to the C++ Language' in the manual.
 681
 682 The `g++' program is now a C program, rather than a shell script.
 683
 684 Lots and lots and lots of bugs fixes, in nested types, access control,
 685 pointers to member functions, the parser, templates, overload
 686 resolution, etc, etc.
 687
 688 There have been two major enhancements to the Objective-C compiler:
 689
 690 1) Added portability.  It now runs on Alpha, and some problems with
 691    message forwarding have been addressed on other platforms.
 692
 693 2) Selectors have been redefined to be pointers to structs like:
 694    { void *sel_id, char *sel_types }, where the sel_id is the unique
 695    identifier, the selector itself is no longer unique.
 696
 697    Programmers should use the new function sel_eq to test selector
 698    equivalence.
 699
 700 The following major changes have been made to the base compiler and
 701 machine-specific files.
 702
 703 - The MIL-STD-1750A is a new port, but still preliminary.
 704
 705 - The h8/300h is now supported; both the h8/300 and h8/300h ports come
 706   with 32 bit IEEE 754 software floating point support.
 707
 708 - The 64-bit Sparc (v9) and 64-bit MIPS chips are supported.
 709
 710 - NetBSD is supported on m68k, Intel x86, and pc523 systems and FreeBSD
 711   on x86.
 712
 713 - COFF is supported on x86, m68k, and Sparc systems running LynxOS.
 714
 715 - 68K systems from Bull and Concurrent are supported and System V
 716   Release 4 is supported on the Atari.
 717
 718 - GCC supports GAS on the Motorola 3300 (sysV68) and debugging
 719   (assuming GAS) on the Plexus 68K system.  (However, GAS does not yet
 720   work on those systems).
 721
 722 - System V Release 4 is supported on MIPS (Tandem).
 723
 724 - For DG/UX, an ELF configuration is now supported, and both the ELF
 725   and BCS configurations support ELF and COFF object file formats.
 726
 727 - OSF/1 V2.0 is supported on Alpha.
 728
 729 - Function profiling is also supported on Alpha.
 730
 731 - GAS and GDB is supported for Irix 5 (MIPS).
 732
 733 - "common mode" (code that will run on both POWER and PowerPC
 734   architectures) is now supported for the RS/6000 family; the
 735   compiler knows about more PPC chips.
 736
 737 - Both NeXTStep 2.1 and 3 are supported on 68k-based architectures.
 738
 739 - On the AMD 29k, the -msoft-float is now supported, as well as
 740   -mno-sum-in-toc for RS/6000, -mapp-regs and -mflat for Sparc, and
 741   -membedded-pic for MIPS.
 742
 743 - GCC can now convert division by integer constants into the equivalent
 744   multiplication and shift operations when that is faster than the
 745   division.
 746
 747 - Two new warning options, -Wbad-function-cast and
 748   -Wmissing-declarations have been added.
 749
 750 - Configurations may now add machine-specific __attribute__ options on
 751   type; many machines support the `section' attribute.
 752
 753 - The -ffast-math flag permits some optimization that violate strict
 754   IEEE rules, such as converting X * 0.0 to 0.0.
 755 \f
 756 Noteworthy changes in GCC version 2.5.8:
 757
 758 This release only fixes a few serious bugs.  These include fixes for a
 759 bug that prevented most programs from working on the RS/6000, a bug
 760 that caused invalid assembler code for programs with a `switch'
 761 statement on the NS32K, a G++ problem that caused undefined names in
 762 some configurations, and several less serious problems, some of which
 763 can affect most configuration.
 764
 765 Noteworthy change in GCC version 2.5.7:
 766
 767 This release only fixes a few bugs, one of which was causing bootstrap
 768 compare errors on some systems.
 769
 770 Noteworthy change in GCC version 2.5.6:
 771
 772 A few backend bugs have been fixed, some of which only occur on one
 773 machine.
 774
 775 The C++ compiler in 2.5.6 includes:
 776
 777  * fixes for some common crashes
 778  * correct handling of nested types that are referenced as `foo::bar'
 779  * spurious warnings about friends being declared static and never
 780    defined should no longer appear
 781  * enums that are local to a method in a class, or a class that's
 782    local to a function, are now handled correctly.  For example:
 783        class foo { void bar () { enum { x, y } E; x; } };
 784        void bar () { class foo { enum { x, y } E; E baz; }; }
 785
 786 Noteworthy change in GCC version 2.5.5:
 787
 788 A large number of C++ bugs have been fixed.
 789
 790 The fixproto script adds prototypes conditionally on __cplusplus.
 791
 792 Noteworthy change in GCC version 2.5.4:
 793
 794 A bug fix in passing of structure arguments for the HP-PA architecture
 795 makes code compiled with GCC 2.5.4 incompatible with code compiled
 796 with earlier versions (if it passes struct arguments of 33 to 64 bits,
 797 interspersed with other types of arguments).
 798
 799 Noteworthy change in gcc version 2.5.3:
 800
 801 The method of "mangling" C++ function names has been changed.  So you
 802 must recompile all C++ programs completely when you start using GCC
 803 2.5.  Also, GCC 2.5 requires libg++ version 2.5.  Earlier libg++
 804 versions won't work with GCC 2.5.  (This is generally true--GCC
 805 version M.N requires libg++ version M.N.)
 806 \f
 807 Noteworthy GCC changes in version 2.5:
 808
 809 * There is now support for the IBM 370 architecture as a target.
 810 Currently the only operating system supported is MVS; GCC does not run
 811 on MVS, so you must produce .s files using GCC as a cross compiler,
 812 then transfer them to MVS to assemble them.  This port is not reliable
 813 yet.
 814
 815 * The Power PC is now supported.
 816
 817 * The i860-based Paragon machine is now supported.
 818
 819 * The Hitachi 3050 (an HP-PA machine) is now supported.
 820
 821 * The variable __GNUC_MINOR__ holds the minor version number of GCC, as
 822 an integer.  For version 2.5.X, the value is 5.
 823
 824 * In C, initializers for static and global variables are now processed
 825 an element at a time, so that they don't need a lot of storage.
 826
 827 * The C syntax for specifying which structure field comes next in an
 828 initializer is now `.FIELDNAME='.  The corresponding syntax for
 829 array initializers is now `[INDEX]='.  For example,
 830
 831   char whitespace[256]
 832     = { [' '] = 1, ['\t'] = 1, ['\n'] = 1 };
 833
 834 This was changed to accord with the syntax proposed by the Numerical
 835 C Extensions Group (NCEG).
 836
 837 * Complex numbers are now supported in C.  Use the keyword __complex__
 838 to declare complex data types.  See the manual for details.
 839
 840 * GCC now supports `long double' meaningfully on the Sparc (128-bit
 841 floating point) and on the 386 (96-bit floating point).  The Sparc
 842 support is enabled on Solaris 2.x because earlier system versions
 843 (SunOS 4) have bugs in the emulation.
 844
 845 * All targets now have assertions for cpu, machine and system.  So you
 846 can now use assertions to distinguish among all supported targets.
 847
 848 * Nested functions in C may now be inline.  Just declare them inline
 849 in the usual way.
 850
 851 * Packed structure members are now supported fully; it should be possible
 852 to access them on any supported target, no matter how little alignment
 853 they have.
 854
 855 * To declare that a function does not return, you must now write
 856 something like this (works only in 2.5):
 857
 858     void fatal () __attribute__ ((noreturn));
 859
 860 or like this (works in older versions too):
 861
 862     typedef void voidfn ();
 863
 864     volatile voidfn fatal;
 865
 866 It used to be possible to do so by writing this:
 867
 868     volatile void fatal ();
 869
 870 but it turns out that ANSI C requires that to mean something
 871 else (which is useless).
 872
 873 Likewise, to declare that a function is side-effect-free
 874 so that calls may be deleted or combined, write
 875 something like this (works only in 2.5):
 876
 877     int computation () __attribute__ ((const));
 878
 879 or like this (works in older versions too):
 880
 881     typedef int intfn ();
 882
 883     const intfn computation;
 884
 885 * The new option -iwithprefixbefore specifies a directory to add to
 886 the search path for include files in the same position where -I would
 887 put it, but uses the specified prefix just like -iwithprefix.
 888
 889 * Basic block profiling has been enhanced to record the function the
 890 basic block comes from, and if the module was compiled for debugging,
 891 the line number and filename.  A default version of the basic block
 892 support module has been added to libgcc2 that appends the basic block
 893 information to a text file 'bb.out'.  Machine descriptions can now
 894 override the basic block support module in the target macro file.
 895
 896 New features in g++:
 897
 898 * The new flag `-fansi-overloading' for C++.  Use a newly implemented
 899 scheme of argument matching for C++.  It makes g++ more accurately
 900 obey the rules set down in Chapter 13 of the Annotated C++ Reference
 901 Manual (the ARM).  This option will be turned on by default in a
 902 future release.
 903
 904 * The -finline-debug flag is now gone (it was never really used by the
 905   compiler).
 906
 907 * Recognizing the syntax for pointers to members, e.g., "foo::*bar", has been
 908   dramatically improved.  You should not get any syntax errors or incorrect
 909   runtime results while using pointers to members correctly; if you do, it's
 910   a definite bug.
 911
 912 * Forward declaration of an enum is now flagged as an error.
 913
 914 * Class-local typedefs are now working properly.
 915
 916 * Nested class support has been significantly improved.  The compiler
 917   will now (in theory) support up to 240 nested classes before hitting
 918   other system limits (like memory size).
 919
 920 * There is a new C version of the `g++' driver, to replace the old
 921   shell script.  This should significantly improve the performance of
 922   executing g++ on a system where a user's PATH environment variable
 923   references many NFS-mounted filesystems.  This driver also works
 924   under MS-DOS and OS/2.
 925
 926 * The ANSI committee working on the C++ standard has adopted a new
 927   keyword `mutable'.  This will allow you to make a specific member be
 928   modifiable in an otherwise const class.
 929 \f
 930 Noteworthy GCC changes in version 2.4.4:
 931
 932   A crash building g++ on various hosts (including m68k) has been
 933   fixed.  Also the g++ compiler no longer reports incorrect
 934   ambiguities in some situations where they do not exist, and
 935   const template member functions are now being found properly.
 936
 937 Noteworthy GCC changes in version 2.4:
 938
 939 * On each target, the default is now to return short structures
 940 compatibly with the "usual" compiler on that target.
 941
 942 For most targets, this means the default is to return all structures
 943 in memory, like long structures, in whatever way is used on that
 944 target.  Use -freg-struct-return to enable returning short structures
 945 (and unions) in registers.
 946
 947 This change means that newly compiled binaries are incompatible with
 948 binaries compiled with previous versions of GCC.
 949
 950 On some targets, GCC is itself the usual compiler.  On these targets,
 951 the default way to return short structures is still in registers.
 952 Use -fpcc-struct-return to tell GCC to return them in memory.
 953
 954 * There is now a floating point emulator which can imitate the way all
 955 supported target machines do floating point arithmetic.
 956
 957 This makes it possible to have cross compilation to and from the VAX,
 958 and between machines of different endianness.  However, this works
 959 only when the target machine description is updated to use the new
 960 facilities, and not all have been updated.
 961
 962 This also makes possible support for longer floating point types.
 963 GCC 2.4 supports extended format on the 68K if you use `long double',
 964 for targets that have a 68881.  (When we have run time library
 965 routines for extended floating point, then `long double' will use
 966 extended format on all 68K targets.)
 967
 968 We expect to support extended floating point on the i386 and Sparc in
 969 future versions.
 970
 971 * Building GCC now automatically fixes the system's header files.
 972 This should require no attention.
 973
 974 * GCC now installs an unsigned data type as size_t when it fixes the
 975 header files (on all but a handful of old target machines).
 976 Therefore, the bug that size_t failed to be unsigned is fixed.
 977
 978 * Building and installation are now completely separate.
 979 All new files are constructed during the build process;
 980 installation just copies them.
 981
 982 * New targets supported: Clipper, Hitachi SH, Hitachi 8300, and Sparc
 983 Lite.
 984
 985 * A totally new and much better Objective C run time system is included.
 986
 987 * Objective C supports many new features.  Alas, I can't describe them
 988 since I don't use that language; however, they are the same ones
 989 supported in recent versions of the NeXT operating system.
 990
 991 * The builtin functions __builtin_apply_args, __builtin_apply and
 992 __builtin_return let you record the arguments and returned
 993 value of a function without knowing their number or type.
 994
 995 * The builtin string variables __FUNCTION__ and __PRETTY_FUNCTION__
 996 give the name of the function in the source, and a pretty-printed
 997 version of the name.  The two are the same in C, but differ in C++.
 998
 999 * Casts to union types do not yield lvalues.
1000
1001 * ## before an empty rest argument discards the preceding sequence
1002 of non-whitespace characters from the macro definition.
1003 (This feature is subject to change.)
1004
1005
1006 New features specific to C++:
1007
1008 * The manual contains a new section ``Common Misunderstandings with
1009 GNU C++'' that C++ users should read.
1010
1011 * #pragma interface and #pragma implementation let you use the same
1012 C++ source file for both interface and implementation.
1013 However, this mechanism is still in transition.
1014
1015 * Named returned values let you avoid an extra constructor call
1016 when a function result has a class type.
1017
1018 * The C++ operators <? and >? yield min and max, respectively.
1019
1020 * C++ gotos can exit a block safely even if the block has
1021 aggregates that require destructors.
1022
1023 * gcc defines the macro __GNUG__ when compiling C++ programs.
1024
1025 * GNU C++ now correctly distinguishes between the prefix and postfix
1026 forms of overloaded operator ++ and --.  To avoid breaking old
1027 code, if a class defines only the prefix form, the compiler
1028 accepts either ++obj or obj++, unless -pedantic is used.
1029
1030 * If you are using version 2.3 of libg++, you need to rebuild it with
1031 `make CC=gcc' to avoid mismatches in the definition of `size_t'.
1032
1033 Newly documented compiler options:
1034
1035 -fnostartfiles
1036         Omit the standard system startup files when linking.
1037
1038 -fvolatile-global
1039         Consider memory references to extern and global data items to
1040         be volatile.
1041
1042 -idirafter DIR
1043         Add DIR to the second include path.
1044
1045 -iprefix PREFIX
1046         Specify PREFIX for later -iwithprefix options.
1047
1048 -iwithprefix DIR
1049         Add PREFIX/DIR to the second include path.
1050
1051 -mv8
1052         Emit Sparc v8 code (with integer multiply and divide).
1053 -msparclite
1054         Emit Sparclite code (roughly v7.5).
1055
1056 -print-libgcc-file-name
1057         Search for the libgcc.a file, print its absolute file name, and exit.
1058
1059 -Woverloaded-virtual
1060         Warn when a derived class function declaration may be an error
1061         in defining a C++ virtual function.
1062
1063 -Wtemplate-debugging
1064         When using templates in a C++ program, warn if debugging is
1065         not yet fully available.
1066
1067 +eN
1068         Control how C++ virtual function definitions are used
1069         (like cfront 1.x).
1070