GCC 4.5 Release Series
Changes, New Features, and Fixes

Caveats

• Support has been removed for all the configurations obsoleted in GCC 4.4.
• Support has been removed for the protoize and unprotoize utilities, obsoleted in GCC 4.4.

General Optimizer Improvements

• The -save-temps now takes an optional argument. The -save-temps and -save-temps=cwd switches write the temporary files in the current working directory based on the original source file. The -save-temps=obj switch will write files into the directory specified with the -o option, and the intermediate filenames are based on the output file. This will allow the user to get the compiler intermediate files when doing parallel builds without two builds of the same filename located in different directories from interfering with each other.
• Debugging dumps are now created in the same directory as the object file rather than in the current working directory. This allows the user to get debugging dumps when doing parallel builds without two builds of the same filename interfering with each other.
• The Multiple Precision Complex Library (MPC) is now used if found at configure time. This allows GCC to evaluate calls to complex built-in math functions having constant arguments at compile-time and replace them with their mathematically equivalent results. Some complex functions were supported before; for those a fall back is used if MPC is not available.
• A new link-time optimizer has been added (-flto). When this flag is used, GCC generates a bytecode representation of each input file and writes it to special ELF sections in each object file. When the object files are linked together, all the function bodies are read from these ELF sections and instantiated as if they had been part of the same translation unit. This enables interprocedural optimizations to work across different files (and even different languages), potentially improving the performance of the generated code. To use the link-timer optimizer, -flto needs to be specified at compile time and during the final link. If the program does not require any symbols to be exported, it is possible to combine -flto and -fwhopr with -fwhole-program to allow the interprocedural optimizers to use more aggressive assumptions.
• The automatic parallelization pass was enhanced to support parallelization of outer loops.
• Automatic parallelization can be enabled as part of Graphite. In addition to -ftree-parallelize-loops=, specify -floop-parallelize-all to enable the Graphite-based optimization.

New Languages and Language specific improvements

All languages

• The -fshow-column option is now on by default. This means error messages now have a column associated with them.

Ada

• Compilation of programs heavily using discriminated record types with variant parts has been sped up and generates more compact code.

C family

• If a header named in a #include directive is not found, the compiler exits immediately. This avoids a cascade of errors arising from declarations expected to be found in that header being missing.
• GCC now implements C90- and C99-conforming rules for constant expressions. This may cause warnings or errors for some code using expressions that can be folded to a constant but are not constant expressions as defined by ISO C.
• A new built-in function __builtin_unreachable() has been added that tells the compiler that control will never reach that point. It may be used after asm statements that terminate by transferring control elsewhere, and in other places that are known to be unreachable.
• The -Wlogical-op option now warns for logical expressions such as (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be mistakes. This option is disabled by default.

C

• The -Wenum-compare option, which warns when comparing values of different enum types, now works for C. It formerly only worked for C++. This warning is enabled by -Wall. It may be avoided by using a type cast.
• The -Wcast-qual option now warns about casts which are unsafe in that they permit const-correctness to be violated without further warnings. Specifically, it warns about cases where a qualifier is added when all the lower types are not const. For example, it warns about a cast from char ** to const char **.
• The -Wc++-compat option is significantly improved. It issues new warnings for:
  • Using C++ reserved operator names as identifiers.
  • Conversions to enum types without explicit casts.
  • Using va_arg with an enum type.
  • Using different enum types in the two branches of ?:.
  • Using ++ or -- on a variable of enum type.
  • Using the same name as both a struct, union or enum tag and a typedef, unless the typedef refers to the tagged type itself.
  • Using a struct, union, or enum which is defined within another struct or union.
  • A struct field defined using a typedef if there is a field in the struct, or an enclosing struct, whose name is the typedef name.
  • Duplicate definitions at file scope.
  • Uninitialized const variables.
  • A global variable with an anonymous struct, union, or enum type.
  • Using a string constant to initialize a char array whose size is the length of the string.
• The new -Wjump-misses-init option warns about cases where a goto or switch skips the initialization of a variable. This sort of branch is an error in C++ but not in C. This warning is enabled by -Wc++-compat.

C++

• Improved experimental support for the upcoming ISO C++ standard, C++0x. Including support for raw strings, lambda expressions and explicit type conversion operators.
• When printing the name of a class template specialization, G++ will now omit any template arguments which match the default template argument for that parameter. This behavior (and the pretty-printing of function template specializations as template signature and arguments) can be disabled with the -fno-pretty-templates option.
• Access control is now applied to typedef names used in a template, which may cause G++ to reject some ill-formed code that was accepted by earlier releases. The -fno-access-control flag can be used as a temporary workaround until the code is corrected.
• Compilation time for code that uses templates should now scale linearly with the number of instantiations rather than quadratically, as template instantiations are now looked up using hash tables.
• Declarations of functions that look like builtin declarations of library functions are only considered to be redeclarations if they are declared with extern "C". This may cause problems with code that omits extern "C" on hand-written declarations of C library functions such as abort or memcpy. Such code is ill-formed, but was accepted by earlier releases.
• Diagnostics that used to complain about passing non-POD types to ... or jumping past the declaration of a non-POD variable now check for triviality rather than PODness, as per C++0x.
• In C++0x mode local and anonymous classes are now allowed as template arguments, and in declarations of variables and functions with linkage, so long as any such declaration that is used is also defined (DR 757).
• Labels may now have attributes, as has been permitted for a while in C. This is only permitted when the label definition and the attribute specifier is followed by a semicolon—i.e., the label applies to an empty statement. The only useful attribute for a label is unused.

Runtime Library (libstdc++)

• The new -static-libstdc++ option directs g++ to link the C++ library statically, even if the default would normally be to link it dynamically.

Fortran

• The COMMON default padding has been changed – instead of adding the padding before a variable it is now added afterwards, which increases the compatibility with other vendors and helps to obtain the correct output in some cases. Cf. also the -falign-commons flag (added in 4.4).
• The -finit-real= flag now also supports the value snan for signalling not-a-number; to be effective, one additionally needs to enable trapping (e.g. via -ffpe-trap=). Note: Compile-time optimizations can turn a signalling NaN into a quiet one.
• The new flag -fcheck= has been added with the options bounds, array-temps, do, pointer, and recursive. The bounds and array-temps options are equivalent to -fbounds-check and -fcheck-array-temporaries. The do option checks for invalid modification of loop iteration variables, and the recursive option tests for recursive calls to subroutines/functions which are not marked as recursive. With pointer pointer association checks in calls are performed; however, neither undefined pointers nor pointers in expressions are handled. Using -fcheck=all enables all these run-time checks.
• The run-time checking -fcheck=bounds now warns about invalid string lengths of character dummy arguments. Additionally, more compile-time checks have been added.
• GNU Fortran no longer links against libgfortranbegin. As before, MAIN__ (assembler symbol name) is the actual Fortran main program, which is invoked by the main function. However, main is now generated and put in the same object file as MAIN__. For the time being, libgfortranbegin still exists for backward compatibility. For details see the new Mixed-Language Programming chapter in the manual.
• The I/O library was restructured for performance and cleaner code.
• Array assignments and WHERE are now run in parallel when OpenMP's WORKSHARE is used.
• The experimental flag -fwhole-file was added. The flag allows whole-file checking of procedure arguments and allows for better optimizations. It can also be used with -fwhole-program, which is now also supported in gfortran.
• More Fortran 2003 and Fortran 2008 mathematical functions can now be used as initialization expressions.
• Some extended attributes such as STDCALL are now supported via the GCC$ compiler directive.
• For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN intrinsic behaves now as if zero were always positive.
• For legacy compatibiliy: On Cygwin and MinGW, the special files CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now supported.
• Fortran 2003 support has been extended:
  • Procedure-pointer function results and procedure-pointer components (including PASS),
  • allocatable scalars,
  • DEFERRED type-bound procedures,
  • the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements have been implemented.
  • The ALLOCATE statement supports type-specs and the SOURCE= argument.
  • OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC type-bound procedure (i.e. as type-bound operators).
  • Rounding (ROUND=, RZ, ...) for output is now supported.
  • The INT_FAST{8,16,32,64,128}_T kind type parameters of the intrinsic module ISO_C_BINDING are now supported.
  • Extensible derived types with type-bound procedure or procedure pointer with PASS attribute now have to use CLASS in line with the Fortran 2003 standard; the workaround to use TYPE is no longer supported.
  • Experimental, incomplete support for polymorphism, including CLASS (as allocatable or dummy), SELECT TYPE and dynamic dispatch of type-bound procedure calls. Some features do not work yet such as unlimited polymorphism (CODE(*)), CLASS IS, EXTENDS_TYPE_OF.
• Fortran 2008 support has been extended:
  • The OPEN statement now supports the NEWUNIT= option, which returns a unique file unit, thus preventing inadvertent use of the same unit in different parts of the program.
  • Support for unlimited format items has been added.
  • The INT{8,16,32} and REAL{32,64,128} kind type parameters of the intrinsic module ISO_FORTRAN_ENV are now supported.
  • Using complex arguments with TAN, SINH, COSH, TANH, ASIN, ACOS, and ATAN is now possible; the functions ASINH, ACOSH, and ATANH have been added (for real and complex arguments) and ATAN(Y,X) is now an alias for ATAN2(Y,X).
  • The BLOCK construct has been implemented.

Java (GCJ)

New Targets and Target Specific Improvements

AIX

• Full cross-toolchain support now available with GNU Binutils

AVR

• The -mno-tablejump option has been removed because it has the same effect as the -fno-jump-tables option.
• Added support for these new AVR devices:
  • ATmega8U2
  • ATmega16U2
  • ATmega32U2

IA-32/x86-64

• GCC now supports handling floating-point excess precision arising from use of the x87 floating-point unit in a way that conforms to ISO C99. This is enabled with -fexcess-precision=standard and with standards conformance options such as -std=c99, and may be disabled using -fexcess-precision=fast.
• Support for the Intel Atom processor is now available through the -march=atom and -mtune=atom options.

MeP

Support has been added for the Toshiba Media embedded Processor (MeP, or mep-elf) embedded target.

MIPS

picochip

RS/6000 (POWER/PowerPC)

• GCC now supports the Power ISA 2.06, which includes the VSX instructions that add vector 64-bit floating point support, new population count instructions, and conversions between floating point and unsigned types.
• Support for the power7 processor is now available through the -mcpu=power7 and -mtune=power7.
• GCC will now vectorize loops that contain simple math functions like copysign when generating code for altivec or VSX targets.
• Support for the A2 processor is now available through the -mcpu=a2 and -mtune=a2 options.
• Support for the 476 processor is now available through the -mcpu={476,476fp} and -mtune={476,476fp} options.

RX

Support has been added for the Renesas RX Processor (rx-elf) target.

MIPS

Documentation improvements

Other significant improvements

Plugins

• It is now possible to extend the compiler without having to modify its source code. A new option -fplugin=file.so tells GCC to load the shared object file.so and execute it as part of the compiler. The internal documentation describes the details on how plugins can interact with the compiler.

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.

For questions related to the use of GCC, please consult these web pages and the GCC manuals. If that fails, the gcc-help@gcc.gnu.org mailing list might help.
Please send comments on these web pages and the development of GCC to our developer mailing list at gcc@gnu.org or gcc@gcc.gnu.org. All of our lists have public archives.

Copyright (C) Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110, USA.

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

Last modified 2009-11-08