2004-02-09 Albert Lee <lee@gnat.com>
* errno.c: define _SGI_MP_SOURCE for task-safe errno on IRIX
2004-02-09 Ed Schonberg <schonberg@gnat.com>
* exp_ch3.adb (Build_Slice_Assignment): Handle properly case of null
slices.
* exp_ch6.adb (Expand_Call): Do not inline a call when the subprogram
is nested in an instance that is not frozen yet, to avoid
order-of-elaboration problems in gigi.
* sem_attr.adb (Analyze_Attribute, case 'Access): Within an inlined
body the attribute is legal.
2004-02-09 Robert Dewar <dewar@gnat.com>
* s-rident.ads: Minor comment correction
* targparm.adb: Remove dependence on uintp completely. There was
always a bug in Make in that it called Targparm before initializing
the Uint package. The old code appeared to get away with this, but
the new code did not! This caused an assertion error in gnatmake.
* targparm.ads: Fix bad comment, restriction pragmas with parameters
are indeed fully supported.
From-SVN: r77531
2004-02-09 Paolo Carlini <pcarlini@suse.de>
PR libstdc++/14072
* include/bits/basic_ios.tcc (basic_ios<>::_M_cache_locale):
Don't leave dangling pointers.
* testsuite/27_io/basic_ios/imbue/14072.cc: New.
* testsuite/22_locale/numpunct/members/pod/2.cc: Tweak, the num_put
facet is needed in the final test.
From-SVN: r77526
* parse.y (resolve_qualified_expression_name): In case of inaccessible
class don't use not_accessible_field_error, which can get confused.
From-SVN: r77525
* config/m68k/m68k.h (REGISTER_NAMES): Prefix each name with
REGISTER_PREFIX.
* (M68K_FP_REG_NAME): New macro to specify an alternate name for the
frame pointer register, overridable by OS targets.
* (M68K_REGNAME): Macro to obtain register name for asm output,
eventually replacing %a6 with M68K_FP_REG_NAME.
* config/m68k/coff.h (REGISTER_NAMES): Don't redefine.
* config/m68k/linux.h (REGISTER_NAMES): Likewise.
* config/m68k/m68kelf.h (REGISTER_NAMES): Likewise.
* config/m68k/netbsd-elf.h (REGISTER_NAMES): Likewise.
* config/m68k/m68k.c: Use M68K_REGNAME(x) in place of reg_names[x].
From-SVN: r77511
* config/m68k/m68k.h (REGISTER_NAMES): Prefix each name with
REGISTER_PREFIX.
* (M68K_FP_REG_NAME): New macro to specify an alternate name for the
frame pointer register, overridable by OS targets.
* (M68K_REGNAME): Macro to obtain register name for asm output,
eventually replacing %a6 with M68K_FP_REG_NAME.
* config/m68k/coff.h (REGISTER_NAMES): Don't redefine.
* config/m68k/linux.h (REGISTER_NAMES): Likewise.
* config/m68k/m68kelf.h (REGISTER_NAMES): Likewise.
* config/m68k/netbsd-elf.h (REGISTER_NAMES): Likewise.
* config/m68k/m68k.c: Use M68K_REGNAME(x) in place of reg_names[x].
Co-Authored-By: Bernardo Innocenti <bernie@develer.com>
From-SVN: r77510
* real.c (encode_ibm_extended): Normalize the input value before
converting it to a double. Handle the case where a normal value
rounds to infinity.
From-SVN: r77498
* c-objc-common.c (): Fix a typo in a warning.
* cse.c (preferrable): Change to preferable. Update all of its
callers.
* genautomata.c (ainsn): Change
first_ainsn_with_given_equialence_num to
first_ainsn_with_given_equivalence_num. Update all of its
references.
From-SVN: r77497
2004-02-08 Paolo Carlini <pcarlini@suse.de>
* include/bits/basic_string.tcc (assign(const _CharT*, size_type)):
When working in place remember to set the state to sharable
(otherwise, _M_mutate does it).
From-SVN: r77487
* src/powerpc/ffi.c (ffi_prep_cif_machdep <FFI_LINUX64>): Correct
long double function return and long double arg handling.
(ffi_closure_helper_LINUX64): Formatting. Delete unused "ng" var.
Use "end_pfr" instead of "nf". Correct long double handling.
Localise "temp".
* src/powerpc/linux64.S (ffi_call_LINUX64): Save f2 long double
return value.
* src/powerpc/linux64_closure.S (ffi_closure_LINUX64): Allocate
space for long double return value. Adjust stack frame and offsets.
Load f2 long double return.
From-SVN: r77481
* Makefile.in, config/t-slibgcc-darwin, config/t-slibgcc-elf-ver,
config/t-slibgcc-sld, config/mips/t-iris5-6, config/sh/t-linux:
Use the top level mkinstalldirs, not the one in the gcc subdir.
* mkinstalldirs: Remove (from the gcc subdir).
From-SVN: r77466
PR middle-end/13696
* fold-const.c (fold_convert): New function to provide type
conversion to the middle-end without using convert.
(negate_expr, associate_trees, size_diffop, omit_one_operand,
operand_equal_for_comparison_p, pedantic_omit_one_operand,
invert_truthvalue, optimize_bit_field_compare, range_binop,
decode_field_reference, make_range, build_range_check, unextend,
fold_truthop, extract_muldiv_1, fold_mathfn_compare,
fold_binary_op_with_conditional_arg, fold_inf_compare,
fold_single_bit_test, fold, multiple_of_p): Replace all calls to
convert with calls to fold_convert.
* g++.dg/opt/fold1.C: New test case.
From-SVN: r77464
* genrecog.c (find_operand): add extra argument stop.
(validate_pattern): Verify that mach_dup is duplicating operand
defined lexically earlier.
From-SVN: r77461
* fold-const.c (negate_expr_p, negate_expr): Optimize -(A+B) into
either (-A)-B or (-B)-A, if A or B is easily negated respectively.
(fold) <MINUS_EXPR>: Optimize (A*C) - (B*C) -> (A-B)*C for both
integer types and floating point with unsafe_math_optimizations.
Add similar optimization for (A*C1) - (A*C2) -> A*(C1-C2).
Optimize A - B as A + (-B), if B is easily negated.
From-SVN: r77453
* config/rs6000/t-linux64 (LIB2FUNCS_EXTRA): Add darwin-ldouble.c.
(SHLIB_MAPFILES): Add libgcc-ppc64.ver.
(SHLIB_MKMAP_OPTS): Delete.
(TARGET_LIBGCC2_CFLAGS): Add -specs.
(bispecs): Add rule.
* config/rs6000/libgcc-ppc64.ver: New file.
* config/rs6000/ppc64-fp.c (__fixtfdi, __floatditf): New functions.
(__floatdidf, __floatdisf): Optimize multiply.
(__fixunstfdi): New function.
* config/rs6000/rs6000.c (rs6000_complex_function_value): Allow for
real and imag parts larger than one register.
(function_arg): Correct type of reg used when fp arg split partially
to stack.
* config/rs6000/darwin-ldouble.c: Protect with #if !_SOFT_FLOAT
and __MACH__ or __powerpc64__.
From-SVN: r77440
* builtins.c (expand_builtin_signbit): Use extract_bit_field instead
of gen_highpart or gen_lowpart when the floating point format is
wider than the result mode.
Co-Authored-By: Ulrich Weigand <uweigand@de.ibm.com>
From-SVN: r77439
