GCC 3.3 Release Series
Changes, New Features, and Fixes

Caveats

• The preprocessor no longer accepts multi-line string literals. They were deprecated in 3.0, 3.1, and 3.2.
• The preprocessor no longer supports the -A- switch when appearing alone. -A- followed by an assertion is still supported.
• Support for all the systems obsoleted in GCC 3.1 has been removed from GCC 3.3. See below for a list of systems which are obsoleted in this release.
• Checking for null format arguments has been decoupled from the rest of the format checking mechanism. Programs which use the format attribute may regain this functionality by using the new nonnull function attribute. Note that all functions for which GCC has a built-in format attribute, an appropriate built-in nonnull attribute is also applied.
• The DWARF (version 1) debugging format has been deprecated and will be removed in a future version of GCC. Version 2 of the DWARF debugging format will continue to be supported for the foreseeable future.
• The C and Objective-C compilers no longer accept the "Naming Types" extension (typedef foo = bar); it was already unavailable in C++. Code which uses it will need to be changed to use the "typeof" extension instead: typedef typeof(bar) foo. (We have removed this extension without a period of deprecation because it has caused the compiler to crash since version 3.0 and no one noticed until very recently. Thus we conclude it is not in widespread use.)
• The -traditional C compiler option has been removed. It was deprecated in 3.1 and 3.2. (Traditional preprocessing remains available.) The <varargs.h> header, used for writing variadic functions in traditional C, still exists but will produce an error message if used.

General Optimizer Improvements

• A new scheme for accurately describing processor pipelines, the DFA scheduler, has been added.

• Pavel Nejedly, Charles University Prague, has contributed new file format used by the edge coverage profiler (-fprofile-arcs).

  The new format is robust and diagnoses common mistakes where profiles from different versions (or compilations) of the program are combined resulting in nonsensical profiles and slow code to produced with profile feedback. Additionally this format allows extra data to be gathered. Currently, overall statistics are produced helping optimizers to identify hot spots of a program globally replacing the old intra-procedural scheme and resulting in better code. Note that the gcov tool from older GCC versions will not be able to parse the profiles generated by GCC 3.3 and vice versa.

• Jan Hubicka, SuSE Labs, has contributed a new superblock formation pass enabled using -ftracer. This pass simplifies the control flow of functions allowing other optimizations to do better job.

  He also contributed the function reordering pass (-freorder-functions) to optimize function placement using profile feedback.

New Languages and Language specific improvements

C/ObjC/C++

• The preprocessor now accepts directives within macro arguments. It processes them just as if they had not been within macro arguments.
• The separate ISO and traditional preprocessors have been completely removed. The front-end handles either type of preprocessed output if necessary.
• In C99 mode preprocessor arithmetic is done in the precision of the target's intmax_t, as required by that standard.
• The preprocessor can now copy comments inside macros to the output file when the macro is expanded. This feature, enabled using the -CC option, is intended for use by applications which place metadata or directives inside comments, such as lint.
• The method of constructing the list of directories to be searched for header files has been revised. If a directory named by a -I option is a standard system include directory, the option is ignored to ensure that the default search order for system directories and the special treatment of system header files are not defeated.
• A few more ISO C99 features now work correctly.
• A new function attribute, nonnull, has been added which allows pointer arguments to functions to be specified as requiring a non-null value. The compiler currently uses this information to issue a warning when it detects a null value passed in such an argument slot.
• A new type attribute, may_alias, has been added. Accesses to objects with types with this attribute are not subjected to type-based alias analysis, but are instead assumed to be able to alias any other type of objects, just like the char type.

C++

• Type based alias analysis has been implemented for C++ aggregate types.

Objective-C

• Generate an error if Objective-C objects are passed by value in function and method calls.
• When -Wselector is used, check the whole list of selectors at the end of compilation, and emit a warning if a @selector() is not known.
• Define __NEXT_RUNTIME__ when compiling for the NeXT runtime.
• No longer need to include objc/objc-class.h to compile self calls in class methods (NeXT runtime only).
• New -Wundeclared-selector option.
• Removed selector bloating which was causing object files to be 10% bigger on average (GNU runtime only).
• Using at run time @protocol() objects has been fixed in certain situations (GNU runtime only).
• Type checking has been fixed and improved in many situations involving protocols.

Java

• The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 1.4) API.
• The JDK 1.4 assert facility has been implemented.
• The bytecode interpreter is now direct threaded and thus faster.

Fortran

• Fortran improvements are listed in the Fortran documentation.

Ada

• Ada tasking now works with glibc 2.3.x threading libraries.

New Targets and Target Specific Improvements

• The following changes have been made to the HP-PA port:
  • The port now defaults to scheduling for the PA8000 series of processors.
  • Scheduling support for the PA7300 processor has been added.
  • The 32-bit port now supports weak symbols under HP-UX 11.
  • The handling of initializers and finalizers has been improved under HP-UX 11. The 64-bit port no longer uses collect2.
  • Dwarf2 EH support has been added to the 32-bit linux port.
  • ABI fixes to correct the passing of small structures by value.
• The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to use the DFA processor pipeline description.
• The following NetBSD configurations for the SuperH processor family have been added:
  • SH3, big-endian, sh-*-netbsdelf*
  • SH3, little-endian, shle-*-netbsdelf*
  • SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd*
  • SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd*
  • SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd*
  • SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd*
• The following changes have been made to the IA-32/x86-64 port:
  • SSE2 and 3dNOW! intrinsics are now supported.
  • Support for thread local storage has been added to the IA-32 and x86-64 ports.
  • The x86-64 port has been significantly improved.
• The following changes have been made to the MIPS port:
  • All configurations now accept the -mabi switch. Note that you will need appropriate multilibs for this option to work properly.
  • ELF configurations will always pass an ABI flag to the assembler, except when the MIPS EABI is selected.
  • -mabi=64 no longer selects MIPS IV code.
  • The -mcpu option, which was deprecated in 3.1 and 3.2, has been removed from this release.
  • -march now changes the core ISA level. In previous releases, it would change the use of processor-specific extensions, but would leave the core ISA unchanged. For example, mips64-elf -march=r8000 will now generate MIPS IV code.
  • Under most configurations, -mipsN now acts as a synonym for -march.
  • There are some new preprocessor macros to describe the -march and -mtune settings. See the documentation of those options for details.
  • Support for the NEC VR-Series processors has been added. This includes the 54xx, 5500, and 41xx series.
  • Support for the Sandcraft sr71k processor has been added.
• The following changes have been made to the S/390 port:
  • Support to build the Java runtime libraries has been added. Java is now enabled by default on s390-*-linux* and s390x-*-linux* targets.
  • Multilib support for the s390x-*-linux* target has been added; this allows to build 31-bit binaries using the -m31 option.
  • Support for thread local storage has been added.
  • Inline assembler code may now use the 'Q' constraint to specify memory operands without index register.
  • Various platform-specific performance improvements have been implemented; in particular, the compiler now uses the BRANCH ON COUNT family of instructions and makes more frequent use of the TEST UNDER MASK family of instructions.
• The following changes have been made to the PowerPC port:
  • Support for IBM Power4 processor added.
  • Support for Motorola e500 SPE added.
  • Support for AIX 5.2 added.
  • Function and Data sections now supported on AIX.
  • Sibcall optimizations added.
• The support for H8 Tiny is added to the H8/300 port with -mn.

Obsolete Systems

Support for a number of older systems has been declared obsolete in GCC 3.3. Unless there is activity to revive them, the next release of GCC will have their sources permanently removed.

All configurations of the following processor architectures have been declared obsolete:

• Matsushita MN10200, mn10200-*-*
• Motorola 88000, m88k-*-*
• IBM ROMP, romp-*-*

Also, some individual systems have been obsoleted:

• Alpha
  • Interix, alpha*-*-interix*
  • Linux libc1, alpha*-*-linux*libc1*
  • Linux ECOFF, alpha*-*-linux*ecoff*
• ARM
  • Generic a.out, arm*-*-aout*
  • Conix, arm*-*-conix*
  • "Old ABI," arm*-*-oabi
  • StrongARM/COFF, strongarm-*-coff*
• HPPA (PA-RISC)
  • Generic OSF, hppa1.0-*-osf*
  • Generic BSD, hppa1.0-*-bsd*
  • HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]*
  • HiUX, hppa*-*-hiux*
  • Mach Lites, hppa*-*-lites*
• Intel 386 family
  • Windows NT 3.x, i?86-*-win32
• MC68000 family
  • HP systems, m68000-hp-bsd* and m68k-hp-bsd*
  • Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and m68k-sun-mach*
  • AT&T systems, m68000-att-sysv*
  • Atari systems, m68k-atari-sysv*
  • Motorola systems, m68k-motorola-sysv*
  • NCR systems, m68k-ncr-sysv*
  • Plexus systems, m68k-plexus-sysv*
  • Commodore systems, m68k-cbm-sysv*
  • Citicorp TTI, m68k-tti-*
  • Unos, m68k-crds-unos*
  • Concurrent RTU, m68k-ccur-rtu*
  • Linux a.out, m68k-*-linux*aout*
  • Linux libc1, m68k-*-linux*libc1*
  • pSOS, m68k-*-psos*
• MIPS
  • Generic ECOFF, mips*-*-ecoff*
  • SINIX, mips-sni-sysv4
  • Orion RTEMS, mips64orion-*-rtems*
• National Semiconductor 32000
  • OpenBSD, ns32k-*-openbsd*
• POWER (aka RS/6000) and PowerPC
  • AIX versions 1, 2, and 3, rs6000-ibm-aix[123]*
  • Bull BOSX, rs6000-bull-bosx
  • Generic Mach, rs6000-*-mach*
  • Generic SysV, powerpc*-*-sysv*
  • Linux libc1, powerpc*-*-linux*libc1*
• Sun SPARC
  • Generic a.out, sparc-*-aout*, sparclet-*-aout*, sparclite-*-aout*, and sparc86x-*-aout*
  • NetBSD a.out, sparc-*-netbsd*aout*
  • Generic BSD, sparc-*-bsd*
  • ChorusOS, sparc-*-chorusos*
  • Linux a.out, sparc-*-linux*aout*
  • Linux libc1, sparc-*-linux*libc1*
  • LynxOS, sparc-*-lynxos*
  • Solaris on HAL hardware, sparc-hal-solaris2*
  • SunOS versions 3 and 4, sparc-*-sunos[34]*
• NEC V850
  • RTEMS, v850-*-rtems*
• VAX
  • VMS, vax-*-vms*

Documentation improvements

Other significant improvements

• Almost all front-end dependencies in the compiler have been separated out into a set of language hooks. This should make adding a new front end clearer and easier.
• One effect of removing the separate preprocessor is a small increase in the robustness of the compiler in general, and the maintainability of target descriptions. Previously target-specific built-in macros and others, such as __FAST_MATH__, had to be handled with so-called specs that were hard to maintain. Often they would fail to behave properly when conflicting options were supplied on the command line, and define macros in the user's namespace even when strict ISO compliance was requested. Integrating the preprocessor has cleanly solved these issues.
• The Makefile suite now supports redirection of make install by means of the variable DESTDIR.

GCC 3.3

Detailed release notes for the GCC 3.3 release follow.

Bug Fixes

bootstrap failures

• 10140 cross compiler build failures: missing __mempcpy (DUP: 10198,10338)

Internal compiler errors (multi-platform)

• 3581 large string causes segmentation fault in cc1
• 4382 __builtin_{set,long}jmp with -O3 can crash the compiler
• 6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out
• 6412 (c++) ICE in retrieve_specialization
• 6620 (c++) partial template specialization causes an ICE (segmentation fault)
• 6663 (c++) ICE with attribute aligned
• 7068 ICE with incomplete types
• 7647 (c++) ICE when data member has the name of the enclosing class
• 7675 ICE in fixup_var_refs_1
• 7718 'complex' template instantiation causes ICE
• 8116 (c++) ICE in member template function
• 8358 (ada) Ada compiler accesses freed memory, crashes
• 8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation fault
• 8564 (c++) ICE in find_function_data, in function.c
• 8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c
• 8766 (c++) ICE after failed initialization of static template variable
• 8803 ICE in instantiate_virtual_regs_1, in function.c
• 8846 (c++) ICE after diagnostic if fr_FR@euro locale is set
• 8906 (c++) ICE (Segmentation fault) when parsing nested-class definition
• 9216 (c++) ICE on missing template parameter
• 9261 (c++) ICE in arg_assoc, in cp/decl2.c
• 9263 (fortran) ICE caused by invalid PARAMETER in implied DO loop
• 9429 (c++) ICE in template instantiation with a pointered new operator
• 9516 Internal error when using a big array
• 9600 (c++) ICE with typedefs in template class
• 9629 (c++) virtual inheritance segfault
• 9672 (c++) ICE: Error reporting routines re-entered
• 9749 (c++) ICE in write_expression on invalid function prototype
• 9794 (fortran) ICE: floating point exception during constant folding
• 9829 (c++) Missing colon in nested namespace usage causes ICE
• 9916 (c++) ICE with noreturn function in ?: statement
• 9936 ICE with local function and variable-length 2d array
• 10262 (c++) cc1plus crashes with large generated code
• 10278 (c++) ICE in parser for invalid code
• 10446 (c++) ICE on definition of nonexistent member function of nested class in a class template
• 10451 (c++) ICE in grokdeclarator on spurious mutable declaration
• 10506 (c++) ICE in build_new at cp/init.c with -fkeep-inline-functions and multiple inheritance
• 10549 (c++) ICE in store_bit_field on bitfields that exceed the precision of the declared type

Optimization bugs

• 2001 Inordinately long compile times in reload CSE regs
• 2391 Exponential compilation time explosion in combine
• 2960 Duplicate loop conditions even with -Os
• 4046 redundant conditional branch
• 6405 Loop-unrolling related performance regressions
• 6798 very long compile time with large case-statement
• 6871 const objects shouldn't be moved to .bss
• 6909 problem w/ -Os on modified loop-2c.c test case
• 7189 gcc -O2 -Wall does not print ``control reaches end of non-void function'' warning
• 7642 optimization problem with signbit()
• 8634 incorrect code for inlining of memcpy under -O2
• 8750 Cygwin prolog generation erroneously emitting __alloca as regular function call

c front end

• 2161 long if-else cascade overflows parser stack
• 4319 short accepted on typedef'd char
• 8602 incorrect line numbers in warning messages when using inline functions
• 9177 -fdump-translation-unit: C front end deletes function_decl AST nodes and breaks debugging dumps
• 9853 miscompilation of non-constant structure initializer

c++ compiler and library

• 45 legal template specialization code is rejected (DUP: 3784)
• 764 lookup failure: friend operator and dereferencing a pointer and templates (DUP: 5116)
• 2862 gcc accepts invalid explicit instantiation syntax (DUP: 2863)
• 3663 G++ doesn't check access control during template instantiation
• 3797 gcc fails to emit explicit specialization of a template member
• 3948 Two destructors are called when no copy destructor is defined (ABI change)
• 4361 bogus ambiguity taking the address of a member template
• 4802 g++ accepts illegal template code (access to private member; DUP: 5837)
• 4803 inline function is used but never defined, and g++ does not object
• 5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66
• 6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run time
• 7015 certain __asm__ constructs rejected
• 7086 compile time regression (quadratic behavior in fixup_var_refs)
• 7099 G++ doesn't set the noreturn attribute on std::exit and std::abort
• 7247 copy constructor missing when inlining enabled (invalid optimization?)
• 7441 string array initialization compilation time regression from seconds to minutes
• 7768 __PRETTY_FUNCTION__ for template destructor is wrong
• 7804 bad printing of floating point constant in warning message
• 8099 Friend classes and template specializations
• 8117 member function pointers and multiple inheritance
• 8205 using declaration and multiple inheritance
• 8645 unnecessary non-zero checks in stl_tree.h
• 8724 explicit destructor call for incomplete class allowed
• 8805 compile time regression with many member variables
• 8691 -O3 and -fno-implicit-templates are incompatible
• 8700 unhelpful error message for binding temp to reference
• 8724 explicit destructor call for incomplete class allowed
• 8949 numeric_limits<>::denorm_min() and is_iec559 problems
• 9016 Failure to consistently constant fold "constant" C++ objects
• 9053 g++ confused about ambiguity of overloaded function templates
• 9152 undefined virtual thunks
• 9182 basic_filebuf<> does not report errors in codecvt<>::out
• 9297 data corruption due to codegen bug (when copying.)
• 9318 i/ostream::operator>>/<<(streambuf*) broken
• 9320 Incorrect usage of traits_type::int_type in stdio_filebuf
• 9400 bogus -Wshadow warning: shadowed declaration of this in local classes
• 9424 i/ostream::operator>>/<<(streambuf*) drops characters
• 9425 filebuf::pbackfail broken (DUP: 9439)
• 9474 GCC freezes in compiling a weird code mixing <iostream> and <iostream.h>
• 9548 Incorrect results from setf(ios::fixed) and precision(-1) [DR231]
• 9555 ostream inserters fail to set badbit on exception
• 9561 ostream inserters rethrow exception of wrong type
• 9563 ostream::sentry returns true after a failed preparation
• 9582 one-definition rule violation in std::allocator
• 9622 __PRETTY_FUNCTION__ incorrect in template destructors
• 9683 bug in initialization chains for static const variables from template classes
• 9791 -Woverloaded-virtual reports hiding of destructor
• 9817 collate::compare doesn't handle nul characters
• 9825 filebuf::sputbackc breaks sbumpc
• 9826 operator>>(basic_istream, basic_string) fails to compile with custom traits
• 9924 Multiple using statements for builtin functions not allowed
• 9946 destructor is not called for temporary object
• 9964 filebuf::close() sometimes fails to close file
• 9988 filebuf::overflow writes EOF to file
• 10033 optimization breaks polymorphic references w/ typeid operator
• 10097 filebuf::underflow drops characters
• 10132 filebuf destructor can throw exceptions
• 10180 gcc fails to warn about non-inlined function
• 10199 method parametrized by template does not work everywhere
• 10300 use of array-new (nothrow) in segfaults on NULL return
• 10427 Stack corruption with variable-length automatic arrays and virtual destructors
• 10503 Compilation never stops in fixed_type_or_null

Objective-C

• 5956 selectors aren't matched properly when added to the selector table

Fortran compiler and library

• 1832 list directed i/o overflow hangs, -fbounds-check doesn't detect
• 3924 g77 generates code that is rejected by GAS if COFF debug info requested
• 5634 doc: explain that configure --prefix=~/... does not work
• 6367 multiple repeat counts confuse namelist read into array
• 6491 Logical operations error on logicals when using -fugly-logint
• 6742 Generation of C++ Prototype for FORTRAN and extern "C"
• 7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os on irix6.5
• 7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should assume a direct access file
• 7278 g77 "bug"; the executable misbehaves (with -O2 -fno-automatic)
• 7384 DATE_AND_TIME milliseconds field inactive on Windows
• 7388 Incorrect output with 0-based array of characters
• 8587 Double complex zero ** double precision number -> NaN instead of zero
• 9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: unknown register name line-length-none
• 10197 Direct access files not unformatted by default

Java compiler and library

• 6005 gcj fails to build rhug on alpha
• 6389 System.getProperty("") should always throw an IllegalArgumentException
• 6576 java.util.ResourceBundle.getResource ignores locale
• 6652 new java.io.File("").getCanonicalFile() throws exception
• 7060 getMethod() doesn't search super interface
• 7073 bytecode interpreter gives wrong answer for interface getSuperclass()
• 7180 possible bug in javax.naming.spi.NamingManager.getPlusPath()
• 7416 java.security startup refs "GNU libgcj.security"
• 7570 Runtime.exec with null envp: child doesn't inherit parent env (DUP: 7578)
• 7611 Internal error while compiling libjava with -O
• 7709 NullPointerException in _Jv_ResolvePoolEntry
• 7766 ZipInputStream.available returns 0 immediately after construction
• 7785 Calendar.getTimeInMillis/setTimeInMillis should be public
• 7786 TimeZone.getDSTSavings() from JDK1.4 not implemented
• 8142 '$' in class names vs. dlopen 'dynamic string tokens'
• 8234 ZipInputStream chokes when InputStream.read() returns small chunks
• 8415 reflection bug: exception info for Method
• 8481 java.Random.nextInt(int) may return negative
• 8593 Error reading GZIPped files with BufferedReader
• 8759 java.beans.Introspector has no flushCaches() or flushFromCaches() methods
• 8997 spin() calls Thread.sleep
• 9253 on win32, java.io.File.listFiles("C:\\") returns pwd instead of the root content of C:
• 9254 java::lang::Object::wait(), threads-win32.cc returns wrong return codes
• 9271 Severe bias in java.security.SecureRandom

Ada compiler and library

• 6767 make gnatlib-shared fails on -laddr2line
• 9911 gnatmake fails to link when GCC configured with --with-sjlj-exceptions=yes
• 10020 Can't bootstrap gcc on AIX with Ada enabled
• 10546 Ada tasking not working on Red Hat 9

preprocessor

• 7029 preprocessor should ignore #warning with -M

ARM-specific

• 2903 [arm] Optimization bug with long long arithmetic
• 7873 arm-linux-gcc fails when assigning address to a bit field

FreeBSD-specific

• 7680 float functions undefined in math.h/cmath with #define _XOPEN_SOURCE

HP-UX or HP-PA-specific

• 8705 [HP-PA] ICE in emit_move_insn_1, in expr.c
• 9986 [HP-UX] Incorrect transformation of fputs_unlocked to fputc_unlocked
• 10056 [HP-PA] ICE at -O2 when building c++ code from doxygen

m68hc11-specific

• 6744 Bad assembler code generated: reference to pseudo register z
• 7361 Internal compiler error in reload_cse_simplify_operands, in reload1.c

MIPS-specific

• 9496 [mips-linux] bug in optimizer?

PowerPC-specific

• 7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of space
• 8480 reload ICEs for LAPACK code on powerpc64-linux
• 8784 [AIX] Internal compiler error in simplify_gen_subreg
• 10315 [powerpc] ICE: in extract_insn, in recog.c

Sparc-specific

• 10267 (documentation) Wrong build instructions for *-*-solaris2*

x86-specific (Intel/AMD)

• 7916 ICE in instantiate_virtual_register_1
• 7926 (c++) i486 instructions in header files make c++ programs crash on i386
• 8555 ICE in gen_split_1231
• 8994 ICE with -O -march=pentium4
• 9426 ICE with -fssa -funroll-loops -fprofile-arcs
• 9806 ICE in inline assembly with -fPIC flag
• 10077 gcc -msse2 generates movd to move dwords between xmm regs
• 10233 64-bit comparison only comparing bottom 32-bits
• 10286 type-punning doesn't work with __m64 and -O
• 10308 [x86] ICE with -O -fgcse or -O2

Please send FSF & GNU inquiries & questions to gnu@gnu.org. There are also other ways to contact the FSF.

These pages are maintained by The GCC team.

Copyright (C) Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.

Verbatim copying and distribution of this entire article is permitted in any medium, provided this notice is preserved.

Last modified 2003-05-15