trans.h (gfc_build_indirect_ref): Remove declaration.

2005-12-16  Richard Guenther  <rguenther@suse.de>

* trans.h (gfc_build_indirect_ref): Remove declaration.
* trans.c (gfc_build_indirect_ref): Remove.
* trans-array.c (gfc_trans_array_ctor_element): Use
build_fold_indirect_ref instead of gfc_build_indirect_ref.
(gfc_trans_array_constructor_value): Likewise.
(gfc_conv_array_index_offset): Likewise.
(gfc_conv_scalarized_array_ref): Likewise.
(gfc_conv_array_ref): Likewise.
(gfc_trans_dummy_array_bias): Likewise.
(gfc_conv_expr_descriptor): Likewise.
(gfc_conv_array_parameter): Likewise.
* trans-decl.c (gfc_finish_cray_pointee): Likewise.
(gfc_get_symbol_decl): Likewise.
* trans-expr.c (gfc_conv_substring): Likewise.
(gfc_conv_component_ref): Likewise.
(gfc_conv_variable): Likewise.
(gfc_add_interface_mapping): Likewise.
(gfc_conv_function_call): Likewise.
* trans-intrinsic.c (gfc_conv_intrinsic_ichar): Likewise.
(gfc_conv_intrinsic_transfer): Likewise.
* trans-io.c (nml_get_addr_expr): Likewise.
(transfer_namelist_element): Likewise.
(transfer_expr): Likewise.
* trans-stmt.c (gfc_trans_nested_forall_loop): Likewise.
(allocate_temp_for_forall_nest_1): Likewise.
(gfc_trans_forall_1): Likewise.

From-SVN: r108678

re PR objc/25360 (Complex types are not encoded)

2005-12-14  Andrew Pinski  <pinskia@physics.uc.edu>

        PR objc/25360
        * objc/objc-act.c (encode_type): Encode Complex types as 'j' followed
        by the inner type.

2005-12-14  Andrew Pinski  <pinskia@physics.uc.edu>

        PR objc/25360
        * objc/objc-api.c (_C_COMPLEX): New define.
        * encoding.c (objc_sizeof_type): Handle _C_Complex.
        (objc_alignof_type): Likewise.
        (objc_skip_typespec): Likewise.

From-SVN: r108675

re PR fortran/25264 (write to internal unit from the string itself gives wrong result ?)

2005-12-16  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libgfortran/25264
PR libgfortran/25349
* gfortran.dg/tl_editing.f90: Added additional checks.
* gfortran.dg/t_editing.f: New test.
* gfortran.dg/write_padding.f90: New test

From-SVN: r108673

re PR middle-end/25457 (reorg.c:966: warning: unused variable 'insn')

PR middle-end/25457
* reorg.c (mostly_true_jump): Remove unused variable 'insn'.

From-SVN: r108672

re PR fortran/25264 (write to internal unit from the string itself gives wrong result ?)

2005-12-16  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libgfortran/25264
PR libgfortran/25349
* io/unit.c (get_unit): Delete code that cleared the string when the
unit was opened, which is too soon.
* io/transfer.c (next_record_w): Pass done flag in.  Change logic for
setting max_pos.  Add code to position unit and pad record as needed.

From-SVN: r108671

re PR testsuite/25215 (-fpic/-fPIC failure in gcc.dg/20050503-1.c)

PR testsuite/25215
* gcc.dg/20050503-1.c: Skip on x86 ilp32 with -fpic/-fPIC.

From-SVN: r108668

re PR target/25447 (<built-in>:0: ICE: in pa_scalar_mode_supported_p, at config/pa/pa.c:5932)

PR target/25447
* config/pa/pa.c (pa_scalar_mode_supported_p): Handle
MODE_DECIMAL_FLOAT.

From-SVN: r108665

jcf-parse.c (set_source_filename): Set the decl source location even when returning early.

* jcf-parse.c (set_source_filename): Set the decl source location
even when returning early.

From-SVN: r108662

* testsuite/libjava.jacks/jacks.xfail
(8.5.2-accessible-static-member-usage-3): Removed.

From-SVN: r108656

loop-invariant.c (move_loop_invariants): Add missing hunk from the previous patch, to nullify df.

        * loop-invariant.c (move_loop_invariants): Add missing hunk from
        the previous patch, to nullify df.

From-SVN: r108655

reorg.c (mostly_true_jump): Clean up code depending on LABEL_OUTSIDE_LOOP_P and loop notes.

* reorg.c (mostly_true_jump): Clean up code depending on
LABEL_OUTSIDE_LOOP_P and loop notes.  Remove code doing
poor man's branch prediction, instead rely on REG_BR_PROB
notes to be available.

From-SVN: r108653

ChangeLog: Fix missing file names.

2005-12-16  Richard Guenther  <rguenther@suse.de>

        * ChangeLog: Fix missing file names.

From-SVN: r108652

re PR rtl-optimization/24899 (loop.c miscompiles libgnomecanvas)

PR rtl-optimization/24899
* loop.c (strength_reduce): Don't reduce giv that is not always
computable and where add_val or mult_val can trap.

* gcc.c-torture/execute/20051215-1.c: New test.

From-SVN: r108642

* doc/invoke.texi (-fdump-tree-*): Remove note about C/C++ only.

From-SVN: r108639

mt.h (LIB_SPEC, [...]): Correct names of ms1-16-002 files.

* config/mt/mt.h (LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Correct
names of ms1-16-002 files.

From-SVN: r108638

configure.in: Replace ms1 with mt.

* configure.in: Replace ms1 with mt.
* configure: Rebuilt.

From-SVN: r108637

Checkin missing parts of my 2005-12-12 patch

From-SVN: r108636

re PR driver/20425 (-print-search-dirs doesn't honor mutil-os/multilib settings)

PR driver/20425
* gcc.c (for_each_path): New function.
(add_to_obstack, file_at_path): New functions.
(struct file_at_path_info, struct add_to_obstack_info): New.
(build_search_list): Rewrite using for_each_path.  Constify struct
path_prefix pointer.  Add do_multi param.  Adjust all callers.
(find_a_file): Similarly, but just change existing param to bool.
(putenv_from_prefixes): Add do_multi param, make "paths" const.
(do_spec_path): Delete.
(struct spec_path_info): New.
(spec_path): New function.
(do_spec_1): Use for_each_path for %D and %I.
(find_file): Adjust for find_a_file changes.
(main): Search multilibs for specs.  Print multilib lib path for
-print-search-dirs.  Likewise add multilibs to LIBRARY_PATH_ENV.
(read_specs): Search multilibs for %include and %include_noerr.
(is_directory): Remove second string param.  Change last param
to a bool.  Don't use concat.  Remove SMALL_ARG_MAX test, always
check path is a dir.  Update all callers.
* doc/hostconfig.texi (SMALL_ARG_MAX): Remove mention.
* system.h (SMALL_ARG_MAX): Poison.
* config.gcc: Don't define SMALL_ARG_MAX.
* config.host: Likewise.
* config.build: Likewise.

From-SVN: r108635

combine.c (combine_simplify_rtx <case NOT, [...]): Move simplifications that do not require additional infrastructure...

2005-12-16  Paolo Bonzini  <bonzini@gnu.org>

* combine.c (combine_simplify_rtx <case NOT, NEG, TRUNCATE,
FLOAT_TRUNCATE, FLOAT_EXTEND, PLUS, MINUS, AND, IOR, XOR,
ABS, VEC_SELECT, POPCOUNT, PARITY, FFS, FLOAT>,
simplify_logical): Move simplifications that do not require
additional infrastructure...
* simplify-rtx.c (simplify_unary_operation_1,
simplify_binary_operation_1): ... here.

From-SVN: r108634

(gfc_conv_descriptor_data_addr): Use build_fold_addr_expr where appropriate.

2005-12-16  Richard Guenther  <rguenther@suse.de>

(gfc_conv_descriptor_data_addr): Use build_fold_addr_expr where
appropriate.
(gfc_trans_allocate_array_storage): Likewise.
(gfc_trans_array_constructor_value): Likewise.
(gfc_conv_array_data): Likewise.
(gfc_conv_expr_descriptor): Likewise.
(gfc_conv_array_parameter): Likewise.
(gfc_conv_variable): Likewise.
(gfc_conv_function_val): Likewise.
(gfc_conv_function_call): Likewise.
(gfc_conv_expr_reference): Likewise.
(gfc_conv_intrinsic_ctime): Likewise.
(gfc_conv_intrinsic_fdate): Likewise.
(gfc_conv_intrinsic_ttynam): Likewise.
(gfc_conv_intrinsic_si_kind): Likewise.
(gfc_conv_intrinsic_trim): Likewise.
(set_parameter_ref): Likewise.
(gfc_convert_array_to_string): Likewise.
(gfc_trans_open): Likewise.
(gfc_trans_close): Likewise.
(build_filepos): Likewise.
(gfc_trans_inquire): Likewise.
(nml_get_addr_expr): Likewise.
(transfer_namelist_element): Likewise.
(build_dt): Likewise.
(gfc_trans_dt_end): Likewise.
(transfer_array_component): Likewise.
(transfer_expr): Likewise.
(transfer_array_desc): Likewise.
(gfc_trans_transfer): Likewise.
(gfc_trans_allocate): Likewise.
(gfc_trans_deallocate): Likewise.

From-SVN: r108633

re PR rtl-optimization/24823 (ICE in insert_save, at caller-save.c:719)

2005-12-16  Andreas Krebbel  <krebbel1@de.ibm.com>

PR 24823
* flow.c (mark_used_dest_regs): New function.
(mark_used_regs): Call mark_used_dest_regs.

2005-12-16  Andreas Krebbel  <krebbel1@de.ibm.com>

PR 24823
* gfortran.dg/pr24823.f: New test.

From-SVN: r108632

tree-ssa-dse.c (dse_optimize_stmt): Correctly handle PHI nodes which represent a use and definition of the same...

* tree-ssa-dse.c (dse_optimize_stmt): Correctly handle PHI nodes which
represent a use and definition of the same SSA_NAME.

* gcc.dg/tree-ssa/ssa-dse-8.c: New test.

From-SVN: r108630

target-def.h (TARGET_DECIMAL_FLOAT_SUPPORTED_P): New.

2005-12-16  Jon Grimm  <jgrimm2@us.ibm.com>
    Janis Johnson  <janis187@us.ibm.com>
    Ben Elliston  <bje@au.ibm.com>

* target-def.h (TARGET_DECIMAL_FLOAT_SUPPORTED_P): New.
(TARGET_INITIALIZER): Add TARGET_DECIMAL_FLOAT_P.
* target.h (struct gcc_target): Add decimal_float_supported_p.
* targhooks.c (default_scalar_mode_supported_p): Handle
MODE_DECIMAL_FLOAT.
* builtins.def: Add new builtins for 32, 64 and 128 bit variants
of inf, nan, finite, isinf and isnan.
* builtin-types.def (BT_DFLOAT32, BT_DFLOAT64, BT_DFLOAT128,
BT_DFLOAT32_PTR, BT_DFLOAT64_PTR, BT_DFLOAT128_PTR,
BT_FN_DFLOAT32, BT_FN_DFLOAT64, BT_FN_DFLOAT128,
BT_FN_INT_DFLOAT32, BT_FN_INT_DFLOAT64, BT_FN_INT_DFLOAT128,
BT_FN_DFLOAT32_CONST_STRING, BT_FN_DFLOAT64_CONST_STRING,
BT_FN_DFLOAT32_CONST_STRING, BT_FN_DFLOAT32_DFLOAT32,
BT_FN_DFLOAT64_DFLOAT64, BT_FN_DFLOAT128_DFLOAT128): New.
* c-decl.c (declspecs_add_type): Verify combos on type qualifiers.
Pedwarn if decimal floating point types are used.  Error if
decimal floating point is not supported by the target.
(finish_declspecs): Return type from DFP typespec_word.
* c-typeck.c (c_common_type): Choose the decimal floating point
type with the greater precision when determining a common type.
(convert_arguments): Warn if there is a mismatch between argument
and prototype for decimal float types.  Warn of conversions with
binary float types and of precision narrowing due to prototype.
* c-parser.c (reswords): Add _Decimal32, _Decimal64, _Decimal128.
(c_token_starts_typename): Handle RID_DFLOAT32/64/128.
(c_token_starts_declspecs): Likewise.
(c_parser_attributes): Likewise.
* c-common.h (enum rid): Add new enumeration values RID_DFLOAT32,
RID_DFLOAT64, RID_DFLOAT128.
(T_D32, TEX_D32, T_D64, TEX_D64, T_D128, TEX_D128): New macros.
* c-common.c (c_common_type_for_mode): Handle decimal float modes.
(shorten_compare): Convert DFP/BFP operands to a common type.
(c_common_modes_and_builtins): Register built-in decimal float
types if the target supports them.
(handle_mode_addtribute): Handle MODE_DECIMAL_FLOAT.
* builtins.c (fold_builtin_1): Handle 32, 64 and 128 bit cases of
inf, nan, finite, isinf and isnan builtins.
* c-cppbuiltin.c (builtin_define_decimal_float_constants): New.
(builtin_define_float_constants): Assert non-decimal radix.
(c_cpp_builtins): Register built-in __DEC_EVAL_METHOD__ define.
Call builtin_define_decimal_float_constants for each type.
* c-lex.c (interpret_float): Decode decimal float types from CPP_N
flags.  Use real_from_string3, which can handle binary or decimal
floats.
* c-tree.h (enum c_typespec_keyword): Add cts_dfloat32,
cts_dfloat64, cts_dfloat128.
* tree.c (build_common_tree_nodes_2): Add decimal float types.
* tree.h (enum tree_index): Add new enumeration values
TI_DFLOAT32_TYPE, TI_DFLOAT64_TYPE, TI_DFLOAT128_TYPE,
TI_DFLOAT32_PTR_TYPE, TI_DFLOAT64_PTR_TYPE, TI_DFLOAT128_PTR_TYPE.
(dfloat32_type_node): New macro.
(dfloat64_type_node, dfloat128_type_node): Likewise.
(dfloat32_ptr_type_node, dfloat64_ptr_type_node): Likewise.
(dfloat128_ptr_type_node): Likewise.
* c-pretty-print.c (pp_c_floating_constant): Append 32, 64 and 128
bit decimal floating point types with "df", "dd" and "dl".
* c-format.h (enum format_lengths): Add new enumeration values
FMT_LEN_H, FMT_LEN_D and FMT_LEN_DD.
* c-format.c (printf_length_specs, scanf_length_specs): Add
entries for H, D, DD.
(print_char_table, scan_char_table): Use new entries.
(asm_fprintf_char_table, gcc_diag_char_table,
gcc_cdiag_char_table, gcc_cxxdiag_char_table): Adjust for longer
length arrays.
* defaults.h (DECIMAL32_TYPE_SIZE): Define.
(DECIMAL64_TYPE_SIZE): Likewise.
(DECIMAL128_TYPE_SIZE): Likewise.
(TARGET_DEC_EVAL_METHOD): Likewise.
* doc/extend.texi (Decimal Float): New node.
(Constructing Calls): Document decimal float built-ins.
* doc/tm.texi: Document TARGET_DECIMAL_FLOAT_SUPPORTED_P hook.
* Makefile.in (USER_H): Add $(srcdir)/ginclude/decfloat.h.
* ginclude/decfloat.h: New file.

Co-Authored-By: Ben Elliston <bje@au.ibm.com>
Co-Authored-By: Janis Johnson <janis187@us.ibm.com>
From-SVN: r108629

reload.c (find_reloads): Fix comment typo.

* reload.c (find_reloads): Fix comment typo.
* recog.h (recog_memoized): Correct comment.

From-SVN: r108627

basic-block.h, [...]: Fix comment typos.

* basic-block.h, config/i386/winnt.c, config/pa/pa.c,
config/s390/s390.c, dfp.c, expr.c, fold-const.c, params.def,
reload.c, struct-equiv.c, tree-ssa-ccp.c, tree-ssa-pre.c,
tree-ssa-reassoc.c, tree-ssa-structalias.c: Fix comment typos.
* doc/invoke.texi, doc/tm.texi: Fix typos.

From-SVN: r108626

* parser.c, pt.c: Fix comment typos.

From-SVN: r108625

dependency.c, [...]: Fix comment typos.

* dependency.c, resolve.c, trans-array.c: Fix comment typos.
* gfortran.texi: Fix typos.

From-SVN: r108624

struct-layout-encoding-1_generate.c (base_types): Change C++ style comments to C style comments.

2005-12-16  Andrew Pinski  <pinskia@physics.uc.edu>

        * objc.dg/gnu-encoding/struct-layout-encoding-1_generate.c (base_types):
        Change C++ style comments to C style comments.
        Remove aligned types from the array.
        Keep Enums disabled.
        (complex_types): Remove aligned types from the array.
        Comment out "_Complex long double".
        (attrib_types): Remove.
        (complex_attrib_types): Remove.
        (attrib_array_types): Remove.
        (complex_attrib_array_types): Remove.
        (aligned_bitfld_types): Remove.
        (n_aligned_bitfld_types): Remove.
        (FEATURE_ALIGNEDPACKED): Delete.
        (ALL_FEATURES): Remove FEATURE_ALIGNEDPACKED.
        (singles): Remove support for adding attributes.
        (choose_type): Remove support for attributed types.
        (generate_fields): Remove support for attributed bit-fields.
        Remove support for adding attributes.
        (generate_random_tests): Remove support for adding attributes.
        (features): Add back "[0] :0" and "complex vector [0]".
        (main): Remove support for attributed bit-fields.
        Uncomment the calling of generate_random_tests for the rests of
        the tests.

From-SVN: r108623

* real.c (decimal_quad_format): Correct values for emin, emax.

From-SVN: r108622

predicates.md (m32c_psi_scale): New.

* config/m32c/predicates.md (m32c_psi_scale): New.
* config/m32c/m32c.c (m32c_expand_neg_mulpsi3): New.
* config/m32c/muldiv.md (mulpsi3): Support negative constants.

From-SVN: r108620

Daily bump.

From-SVN: r108614

re PR java/25429 (can't see private static final int CREATE = 1)

gcc/java:
PR java/25429
* parse.y (resolve_expression_name): Don't generate accessor
methods for constant fields.
libjava:
* testsuite/libjava.compile/rh175833.java: New file.
* testsuite/libjava.compile/pr25429.java: New file.

Co-Authored-By: Andrew Haley <aph@redhat.com>
From-SVN: r108608

re PR rtl-optimization/25224 (ICE in initialize_original_copy_tables)

PR rtl-optimization/25224
* tree-ssa-loop-unswitch.c (tree_unswitch_single_loop): Free copy
tables.

From-SVN: r108606

Commit part of Zdenek's larger loop-invariant.c patch.

* loop-invariant.c (df): New global variable.
(find_defs, check_dependencies, find_invariant_insn, record_uses,
find_invariants_bb, find_invariants_body, find_invariants,
find_invariants_to_move, move_invariants, free_inv_motion_data,
move_single_loop_invariants, move_loop_invariants): Do not pass df in
arguments.

Co-Authored-By: Steven Bosscher <stevenb@suse.de>
From-SVN: r108605

varasm.c (default_unique_section_1): Use special section names for SECCAT_DATA_REL{,_RO}{,_LOCAL}.

* varasm.c (default_unique_section_1): Use special section
names for SECCAT_DATA_REL{,_RO}{,_LOCAL}.

From-SVN: r108602

* nl.po: Update.

From-SVN: r108600

invoke.texi (Warning Options): Document -Walways-true being enabled by -Wall.

* doc/invoke.texi (Warning Options): Document -Walways-true being
enabled by -Wall.

From-SVN: r108599

re PR other/25028 (TImode-to-floating conversions broken)

PR other/25028
* libgcc2.h (SF_SIZE, DF_SIZE, XF_SIZE, TF_SIZE): Define.
* libgcc2.c (__floatdixf, __floatundixf, __floatditf,
__floatunditf): Use #error if type sizes don't match requirements
of implementation.
(__floatdisf, __floatdidf): Unify.  Possibly use XFmode or TFmode
as wider floating-point type.  Use #error if type sizes don't
match requirements of implementation.  Avoid overflow in computing
Wtype_MAXp1_F * Wtype_MAXp1_F.  When special casing conversion,
shift one more bit.  Cast 1 to DWtype or UDWtype for shifting.
(__floatundisf, __floatundidf): Likewise.
* config/ia64/hpux.h (XF_SIZE, TF_SIZE): Define.
* config/ia64/ia64.c (ia64_init_libfuncs): Use
_U_Qfcnvfxt_quad_to_quad and _U_Qfcnvxf_quad_to_quad for
TFmode-TImode conversions.
* doc/tm.texi (SF_SIZE, DF_SIZE, XF_SIZE, TF_SIZE): Document.

testsuite:
* gcc.dg/torture/fp-int-convert-timode.c: Only XFAIL for LP64 IA64
HP-UX.

From-SVN: r108598

freebsd-spec.h (FBSD_TARGET_OS_CPP_BUILTINS): Use builtin_define_with_int_value() instead of adding a new check for every new...

        * freebsd-spec.h (FBSD_TARGET_OS_CPP_BUILTINS):
        Use builtin_define_with_int_value() instead of
        adding a new check for every new major FreeBSD version.

From-SVN: r108597

MAINTAINERS: Update my email address.

2005-12-15  Bryce McKinlay  <mckinlay@redhat.com>

        * MAINTAINERS: Update my email address.

From-SVN: r108596

t-darwin (darwin-c.o): Depend on $(TREE_H) instead of tree.h.

2005-12-14  Andrew Pinski  <pinskia@physics.uc.edu>

        * config/t-darwin (darwin-c.o): Depend on $(TREE_H) instead of tree.h.

From-SVN: r108585

re PR libobjc/14382 ([libobjc] Calling +initialize on a per category basis)

2005-12-15  David Ayers  <d.ayers@inode.at>

PR libobjc/14382
* README (+load,+initialize): Fix documentation to reflect
intended and implemented semantics for +load and +initialize.

From-SVN: r108584

predicates.md (call_memory_operand): New.

* config/arm/predicates.md (call_memory_operand): New.
* config/arm/arm.md (*call_mem, *call_value_mem): Use it.

testsuite/
* gcc.dg/20051215-1.c: New file.

From-SVN: r108583

unwind-dw2-fde-glibc.c (_Unwind_IteratePhdrCallback): Guard against prev_cache_entry == NULL when moving a cache entry to the head of...

2005-12-15  Andrew Haley  <aph@redhat.com>

        * unwind-dw2-fde-glibc.c (_Unwind_IteratePhdrCallback): Guard
        against prev_cache_entry == NULL when moving a cache entry to the
        head of the list.

From-SVN: r108582

mt.c (mt_override_options): Require lower case architectures.

* config/mt/mt.c (mt_override_options): Require lower case
architectures.  Revert accidental default to ms2.
* config/mt/mt.h (ASM_SPEC, LIBSPEC, STARTFILE_SPEC,
ENDFILE_SPEC): Require lower case architectures.  Revert default
to ms2.

* doc/invoke.texi (MT Options): Rename and update.
* doc/md.texi (MorphoTech family): Rename and update.

From-SVN: r108581

Makefile.tpl (all, [...]): Don't invoke $(stage) at the end.

2005-12-15  Paolo Bonzini  <bonzini@gnu.org>

* Makefile.tpl (all, do-[+make_target+], do-check, install,
install-host-nogcc): Don't invoke $(stage) at the end.
* Makefile.in: Regenerate.

From-SVN: r108580

stor-layout.c (layout_type): Don't complain about a too small an array element type size compared to its...

* stor-layout.c (layout_type): Don't complain about a too small
an array element type size compared to its alignment when the
size overflows.

From-SVN: r108575

re PR target/24969 (tmpdir-gcc.dg-struct-layout-1/t026 fails execution)

PR target/24969
* i386.c (classify_argument): Properly adjust offset of bitfield for
substructures.

From-SVN: r108573

final.c (final_scan_insn): Flip in_cold_section_p when changing between the hot and cold sections.

* final.c (final_scan_insn): Flip in_cold_section_p when changing
between the hot and cold sections.  Use current_function_section
to get the new section.
* dwarf2out.c (output_line_info): Use in_cold_section_p to
determine whether we are assembling hot or cold code.
(secname_for_decl, dwarf2out_var_location): Likewise.
(dwarf2out_init, dwarf2out_finish): Use switch_to_section.
* varasm.c (last_text_section): Delete.
(in_cold_section_p): New variable.
(hot_function_section): New function.
(current_function_section): Pass in_cold_section_p as the
reloc argument to select_section.  Use it to decide between
unlikely_function_section and hot_function_section.
(assemble_start_function): Use switch_to_section.  Set
in_cold_section_p instead of last_text_section.
(assemble_end_function): Use switch_to_section.
(switch_to_section): Don't set last_text_section.
* config/darwin/darwin.c (machopic_select_section): Trust the reloc
argument to make the right choice between hot and cold sections.

From-SVN: r108572

stl_vector.h (vector<>::_M_get_Tp_allocator): Change to return by ref and add non const version.

2005-12-15  Paolo Carlini  <pcarlini@suse.de>

* include/bits/stl_vector.h (vector<>::_M_get_Tp_allocator): Change
to return by ref and add non const version.
* include/bits/stl_deque.h (deque<>::_M_get_Tp_allocator): Likewise.

From-SVN: r108570

Second part of MS1->MT renaming

Second part of MS1->MT renaming
* config/mt/mt.md (doloop_end): Call mt_add_loop.
(movqi, movhi): Call mt_set_memflags.
(*movdf_internal): Call mt_split_words.
(reload_inqi, reload_outqi, reload_inhi, reload_outhi): Call
mt_set_memflags.
(cmpsi): Store to mt_compare_op[01].
(beq, bne, bge, bgt, ble, blt, bgeu, bgtu, bleu, bltu, bunge,
bungt, bunle, bunlt): Call mt_emit_cbranch.
(prologue): Call mt_expand_prologue.
(epilogu): Call mt_expend_epilogue.
(eh_return): Call mt_expand_eh_return.
(eh_epilogue): Call me_emit_eh_epilogue.
(conditional_trap): Store to mt_compare_op[01].
* config/mt/crti.asm: Adjust comment.
* config/mt/crtn.asm: Adjust comment.
* config/mt/mt.c (MT_INT_ARG_FIRST): Renamed.
(mt_compare_op0, mt_compare_op1): Renamed.
(current_frame_info, zero_frame_info): Adjust type.
(mt_ucmpsi3_libcall): Renamed.
(mt_flag_delayed_branch): Renamed.
(mt_struct_value_rtx, mt_return_addr_rtx): Renamed.
(mt_nops_required, mt_nop_reasons): Renamed.
(mt_asm_output_opcode, mt_get_attr_type, mt_final_prescan_insn,
mt_debug_stack, mt_print_operand_simple_address,
mt_print_operand_address, mt_print_operand): Renamed, adjusted.
(mt_init_cumulative_args, mt_function_arg_slotno, mt_function_arg,
mt_function_arg_advance, mt_arg_partial_bytes,
mt_pass_by_reference, mt_function_arg_boundary,
mt_reg_ok_for_base_p, mt_legitimate_simple_address_p,
mt_legitimate_address_p): Renamed, adjusted.
(mt_cpu): Renamed.
(mt_init_machine_status, mt_override_options, mt_builtin_saveregs,
mt_va_start, mt_compute_frame_size, mt_emit_save_restore,
mt_emit_save_fp, mt_emit_save_regs, mt_interrupt_function_p,
mt_expand_prologue, mt_epilogue_uses, mt_expand_epilogue,
mt_expand_eh_return, mt_emit_eh_prologue,
mt_handl_interrupt_attribute): Renamed, adjusted.
(mt_attribute_table): Renamed, adjusted.
(mt_initial_elimination_offset, mt_generate_compare,
mt_emit_cbranch, mt_set_memflags_1, mt_set_memflags,
mt_secondary_reload_class, mt_function_value, mt_split_words,
mt_pass_in_stack, mt_add_loop, mt_loop_nesting, mt_block_length,
mt_scan_loop, mt_reorg_loops): Renamed, adjusted.
(mt_labels): Renamed.
(mt_add_branches, mt_check_delay_slot, mt_reorg_hazard,
mt_machine_reorg): Renamed, adjusted.
(mt_attribute_table: Renamed.
(TARGET_ATTRIBUTE_TABLE, TARGET_STRUCT_VALUE_RTX,
TARGET_PASS_BY_REFERENCE, TARGET_MUST_PASS_IN_STACK,
TARGET_ARG_PARTIAL_BYTES, TARGET_MACHINE_DEPENDENT_REORG): Adjust.
* config/mt/mt.opt (march): Set mt_cpu_string.
* config/mt/mt.h (mt_ucmpsi3_libcall, mt_cpu): Renamed.
(TARGET_CPU_CPP_BUILTINS): Adjust.
(TARGET_MS1_64_001, TARGET_MS1_16_002, TARGET_MS1_16_003,
TARGET_MS2): Adjust.
(TARGET_VERSION, OVERRIDE_OPTIONS, SECONDARY_RELOAD_CLASS): Adjust.
(MT_MIN_INT): Renamed.
(RETURN_ADDR_RTX): Adjust.
(struct mt_frame_info): Renamed.
(current_frame_info): Adjust type.
(INITIAL_ELIMINATION_OFFSET): Adjust.
(MT_NUM_ARG_REGS): Renamed.
(REG_PARM_STACK_SPACE, FUNCTION_ARG, INIT_CUMULATVE_ARGS,
FUNCTION_ARG_ADVANCE, FUNCTION_ARG_BOUNDARY, FUNCTION_VALUE,
LIBCALL_VALUE, EPILOGUE_USES, EXPAND_BUILTIN_VA_START,
GO_IF_LEGITIMATE_ADDRESS, REG_OK_FOR_BASE_P, ASM_OUTPUT_OPCODE,
FINAL_REPSCAN_INSN, PRINT_OPERAND, PRINT_OPERAND_ADDRESS): Adjust.
(mt_compare_op0, mt_compare_op1): Renamed.
* config/mt/mt-protos.h: Rename prototypes.

From-SVN: r108569

tree-flow.h (okay_component_ref_for_subvars): Remove.

2005-12-15  Richard Guenther  <rguenther@suse.de>

* tree-flow.h (okay_component_ref_for_subvars): Remove.
(get_ref_base_and_extent): Declare.
* tree-dfa.c (okay_component_ref_for_subvars): Remove.
(get_ref_base_and_extent): New function.
* tree-ssa-alias.c (find_used_portions): Use it.
* tree-ssa-structalias.c (get_constraint_for_component_ref):
Likewise.
* tree-ssa-operands.c (get_expr_operands): Likewise.

From-SVN: r108568

re PR libstdc++/25421 (catching exception from codecvt_byname() segfaults)

2005-12-15  Paolo Carlini  <pcarlini@suse.de>

PR libstdc++/25421
* config/locale/gnu/c_locale.cc (_S_destroy_c_locale): Check
for null argument.
* testsuite/22_locale/facet/25421.cc: New.

From-SVN: r108565

combine.c: Remove force_to_mode's fourth parameter.

2005-12-15  Paolo Bonzini  <bonzini@gnu.org>

        * combine.c: Remove force_to_mode's fourth parameter.
        (force_to_mode): Do not handle REG.
        (canon_reg_for_combine): New.
        (make_field_assignment): Use it where the removed argument was non-NULL.

From-SVN: r108564

libjava.exp: Revert 2005-12-01.

* testsuite/lib/libjava.exp: Revert 2005-12-01.
* testsuite/libjava.lang/Array_3.xfail: Delete.

From-SVN: r108563

re PR target/25406 (gcc.dg/20030625-1.c, gcc.dg/20050620-1.c, gcc.dg/940510-1.c, gcc.dg/c99-flex-array-1.c, gcc.dg/pr14475.c, and gcc.dg/noncompile/incomplete-1.c fail on powerpc-darwin)

PR target/25406
* config/rs6000/rs6000.c (rs6000_special_round_type_align): Handle
error_mark_node.  Make alignment params unsigned.
* config/rs6000/rs6000-protos.h
(rs6000_special_round_type_align): Update prototype.
(rs6000_machopic_legitimize_pic_address): Remove arg names.

From-SVN: r108562

repo1.C: Call cleanup-repo-files.

* g++.dg/rtti/repo1.C: Call cleanup-repo-files.
* gcc.dg/20050330-2.c: Call cleanup-coverage-files.

From-SVN: r108558

re PR fortran/18197 (bus error on returning from a function)

fortran/
2005-12-14  Erik Edelmann  <eedelman@gcc.gnu.org>

        PR fortran/18197
        * resolve.c (resolve_formal_arglist): Remove code to set
        the type of a function symbol from it's result symbol.

testsuite/
2005-12-14  Erik Edelmann  <eedelman@gcc.gnu.org>

        PR fortran/18197
        * gfortran.dg/dummy_functions_1.f90: New.

From-SVN: r108555

Daily bump.

From-SVN: r108552

re PR rtl-optimization/25310 (ICE in reload_cse_simplify_operands, at postreload.c:393)

PR rtl-optimization/25310
* reload1.c (eliminate_regs_in_insn): Handle lowpart SUBREGs
of the eliminable register when substituting into a PLUS.

PR rtl-optimization/25310
* gcc.c-torture/compile/pr25310.c: New test.

From-SVN: r108543

Make-lang.in (objcp/objcp-decl.o): Add depends on $(EXPR_H).

2005-12-14  Andrew Pinski  <pinskia@physics.uc.edu>

        * Make-lang.in (objcp/objcp-decl.o): Add depends on $(EXPR_H).

From-SVN: r108538

re PR debug/25023 (ICE in def_cfa_1, at dwarf2out.c:792)

PR debug/25023
* config/i386/i386.c (ix86_force_to_memory): Always use
SImode push for HImode in -m32.
(ix86_free_from_memory): Likewise.

* gcc.dg/pr25023.c: New test.

From-SVN: r108537

CompilerProcess.java: Use a new thread to handle stdout from the child process.

2005-12-14  Andrew Haley  <aph@redhat.com>

        * gnu/java/rmi/rmic/CompilerProcess.java: Use a new thread to
        handle stdout from the child process.

From-SVN: r108536

re PR classpath/25389 (File(new URI("file:./")) -> java.lang.NullPointerException)

PR classpath/25389:
* java/io/File.java (File): Throw IllegalArgumentException if URI is
non-hierarchical.

From-SVN: r108528

redecl1.C, [...]: Remove xfail.

* g++.old-deja/g++.brendan/redecl1.C,
g++.old-deja/g++.eh/inline1.C, g++.old-deja/g++.ns/crash2.C,
g++.old-deja/g++.ns/template7.C, g++.old-deja/g++.oliva/ns3.C,
g++.old-deja/g++.oliva/typeof1.C,
g++.old-deja/g++.other/crash16.C,
g++.old-deja/g++.other/crash34.C,
g++.old-deja/g++.other/lookup5.C, g++.old-deja/g++.pt/crash65.C,
g++.old-deja/g++.pt/decl2.C, g++.old-deja/g++.pt/explicit76.C,
g++.old-deja/g++.pt/memtemp85.C, g++.old-deja/g++.pt/memtemp86.C,
g++.old-deja/g++.pt/memtemp87.C, g++.old-deja/g++.pt/memtemp88.C,
g++.old-deja/g++.pt/memtemp91.C, g++.old-deja/g++.pt/spec14.C,
g++.old-deja/g++.pt/static10.C, treelang/compile/exit.tree:
Remove xfail.

From-SVN: r108526

configure.in: Flip the top-level bootstrap switch.

2005-12-14  Paolo Bonzini  <bonzini@gnu.org>

* configure.in: Flip the top-level bootstrap switch.
* configure: Regenerate.

From-SVN: r108525

arm.md: Provide a big-endian version of the (zero_extend:SI (subreg:QI ...)) splitter.

* gcc/config/arm/arm.md: Provide a big-endian version of the
(zero_extend:SI (subreg:QI ...)) splitter.

From-SVN: r108524

sibcall-6.c: Skip in -fPIC mode on X86 ilp32.

* gcc.dg/sibcall-6.c: Skip in -fPIC mode on X86 ilp32.
* gcc.target/i386/20020523-1.c: Ditto.
* gcc.target/i386/tailcall-1.c: Ditto.
* gcc.target/i386/ssetype-3.c: Make magic* static to work with -fPIC.
* gcc.target/i386/ssetype-5.c: Ditto.

Co-Authored-By: Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
From-SVN: r108521

struct-equiv.c (note_local_live): Handle hard regs with different hard_regno_nregs.

* struct-equiv.c (note_local_live): Handle hard regs with different
hard_regno_nregs.

From-SVN: r108520

tree-ssa-ccp.c (fold_stmt_r): DATA argument is now a pointer to a structure containing state rather than a...

* tree-ssa-ccp.c (fold_stmt_r): DATA argument is now a pointer
to a structure containing state rather than a pointer to bool.
(case ARRAY_REF): New code to handle folding some array references.
(case ADDR_EXPR): Note when we are processing expressions found
within an ADDRE_EXPR.
(fold_stmt, fold_stmt_inplace): Pass in a structure to fold_stmt_r
for state variables rather than just a pointer to a boolean.
* tree-ssa-dom.c (simplify_rhs_and_lookup_avail_expr): Remove
handling of constant string references.

* gcc.dg/tree-ssa/foldstring-1.c: New test.

From-SVN: r108519

s390.md ("*tstdi_extimm", [...]): Merge.

2005-12-14  Adrian Straetling  <straetling@de.ibm.com>

* config/s390/s390.md ("*tstdi_extimm", "*tstsi_extimm"): Merge.
("*tstdi_cconly_extimm", "*tstsi_cconly_extimm"): Merge.
("*tstdi_cconly2", "*tstsi_cconly2"): Merge.
Move other tst* patterns to retain partial order.

From-SVN: r108517

s390.md ("d0", "1"): New mode attributes.

2005-12-14  Adrian Straetling  <straetling@de.ibm.com>

* config/s390/s390.md ("d0", "1"): New mode attributes.
("<shift>di3", "<shift>si3"): Merge.
("*<shift>di3_and", "*<shift>si3_and"): Merge.
("ashrdi3", "ashrsi3"): Merge.
("*ashrdi3_cc", "*ashrsi3_cc"): Merge.
("*ashrdi3_cc_and", "*ashrsi3_cc_and"): Merge.
("*ashrdi3_cconly", "*ashrsi3_cconly"): Merge.
("*ashrdi3_cconly_and", "*ashrsi3_cconly_and"): Merge.
("*ashrdi3", "*ashrsi3"): Merge.
("*ashrdi3_and", "*ashrsi3_and"): Merge.

From-SVN: r108516

s390.md ("extend<mode>di2", [...]): Merge.

2005-12-14  Adrian Straetling  <straetling@de.ibm.com>

* config/s390/s390.md ("extend<mode>di2", "extend<mode>si2"): Merge.
("*extendqidi2_extimm", "*extendqisi2_extimm"): Merge.
("*extendqidi2", "*extendqisi2") Merge.
("*extendqidi2_short_displ", "*extendqisi2_short_displ"): Merge.
("zero_extendhidi2", "zero_extendqidi2"): Merge.
("*zero_extend<mode>di2_extimm", "*zero_extend<mode>si2_extimm"): Merge.
Move some patterns to retain partial ordering.

From-SVN: r108515

s390.md ("atype", "length"): Rewrite.

2005-12-14  Adrian Straetling  <straetling@de.ibm.com>

* config/s390/s390.md ("atype", "length"): Rewrite.
("*insv<mode>_reg_imm", "*insv<mode>_reg_extimm"): Add mode.

From-SVN: r108514

Makefile.tpl: Throughout the file, use : $(MAKE) along with $(stage) and $(unstage).

* Makefile.tpl: Throughout the file, use : $(MAKE) along with
$(stage) and $(unstage).
(EXTRA_TARGET_FLAGS): Correct double-quoting.
(all): Remove stray semicolon.
(local-distclean): Don't handle multilib.tmp and multilib.out.
(install.all): Set $s for consistency.
(configure-[+prefix+][+module+]): Instead of [+deps+], handle
check_multilibs setting.  Always make the install directory.
(configure-stage[+id+]-[+prefix+][+module+]): Likewise.
Correct @if/@endif.
(all-stage[+id+]-[+prefix+][+module+]): Correct @if/@endif.
($(TARGET_SUBDIR)/[+module+]/multilib.out): Remove.
(stage[+id+]-start, stage[+id+]-end): Stage $(TARGET_SUBDIR).
(multilib.out): Remove.
* Makefile.in: Regenerated.

From-SVN: r108513

tree-ssa-operands.c (build_ssa_operands): Update leading comment.

* tree-ssa-operands.c (build_ssa_operands): Update leading comment.

From-SVN: r108512

struct-equiv.c (rtx_equiv_p): Allow arbitrary RVALUE values for PARALLELs with a mode.

* struct-equiv.c (rtx_equiv_p): Allow arbitrary RVALUE values for
PARALLELs with a mode.

From-SVN: r108511

cfgloopmanip.c (lv_adjust_loop_entry_edge): Adjust the type of cond_expr parameter.

* cfgloopmanip.c (lv_adjust_loop_entry_edge): Adjust the type of
cond_expr parameter.

From-SVN: r108510

loop-unswitch.c (unswitch_loop): Free zero_bitmap before returning.

        * loop-unswitch.c (unswitch_loop): Free zero_bitmap before returning.

From-SVN: r108509

re PR bootstrap/25397 (Bootstrap failed)

2005-12-14  J"orn Rennecke <joern.rennecke@st.com>

PR bootstrap/25397:

* struct-equiv.c (struct_equiv_init): Fix off-by-one error in clearing
of STACK_REGS bits.

* struct-euiv.c (rtx_equiv_p): Remove SUBREG case.

From-SVN: r108508

re PR target/25254 (ICE with -mcmodel=medium -mlarge-data-threshold=1)

PR target/25254
PR target/24188
* config/i386/i386.c (x86_64_elf_select_section): If DECL is not
DECL_P, call get_section rather than get_named_section.  Supply
section flags to it.

* gcc.target/i386/pr25254.c: New test.
* gfortran.dg/PR24188.f: New test.

From-SVN: r108506

fortran-torture.exp: Catch remaining uses of remote_file build delete.

2005-12-14  David Billinghurst  (David.Billinghurst@riotinto.com)

* lib/fortran-torture.exp:  Catch remaining uses of
remote_file build delete.

From-SVN: r108503

stkalign.C: Remove xfail.

* g++.old-deja/g++.oliva/stkalign.C: Remove xfail.

From-SVN: r108500

Daily bump.

From-SVN: r108495

c-common.c (c_common_truthvalue_conversion): Generalise warning for addresses converted to booleans; not just function addresses.

* c-common.c (c_common_truthvalue_conversion): Generalise warning
for addresses converted to booleans; not just function addresses.
* c-typeck.c (build_binary_op): Warn for address comparisons which
can never be NULL (eg. func == NULL or &var == NULL).
* common.opt (Walways-true): New option.
* c-opts.c (c_common_handle_option): Set it with -Wall.
* doc/invoke.texi: Document it.
testsuite/
* gcc.dg/warn-addr-cmp.c: New test.

From-SVN: r108489

fpgnulib.c (__unordsf2, [...]): New functions.

2005-12-13  Paul Brook  <paul@codesourcery.com>

* config/m68k/fpgnulib.c (__unordsf2, __unorddf2, __unordxf2,
__floatunsidf, __floatunsisf, __floatunsixf): New functions.

From-SVN: r108487

t-mt (crti.o, crtn.o): Add multilib options.

* config/mt/t-mt (crti.o, crtn.o): Add multilib options.
(EXTRA_MULTILIB_PARTS): Define.

From-SVN: r108481

PR rtl-optimization/20070 / part1

PR rtl-optimization/20070 / part1
* flow.c (update_life_info): If PROP_POST_REGSTACK is set, call
count_or_remove_death_notes with kill == -1.
(mark_set_1): Don't add REG_DEAD / REG_UNUSED notes for stack
registers if PROP_POST_REGSTACK is set.
(mark_used_reg): Likewise.
(count_or_remove_death_notes): If kill is -1, don't remove REG_DEAD /
REG_UNUSED notes for stack regs.
* cfgcleanup.c (condjump_equiv_p): Change parameters and processing
to match rtx_equiv_p machinery.  Change caller.
(outgoing_edges_match): Likewise.
(try_crossjump_to_edge): Use struct_equiv_block_eq
instead of flow_find_cross_jump.
* basic-block.h (PROP_POST_REGSTACK, STRUCT_EQUIV_START): Define.
(STRUCT_EQUIV_RERUN, STRUCT_EQUIV_FINAL): Likewise.
(STRUCT_EQUIV_NEED_FULL_BLOCK, STRUCT_EQUIV_MATCH_JUMPS): Likewise.
(STRUCT_EQUIV_MAX_LOCAL): Likewise.
(struct struct_equiv_checkpoint, struct equiv_info): Likewise.
(insns_match_p): Update prototype.
(flow_find_cross_jump): Remove prototype.
(struct_equiv_block_eq, struct_equiv_init): Declare.
(rtx_equiv_p, condjump_equiv_p): Likewise.
* struct-equiv.c: Include reload.h.
(IMPOSSIBLE_MOVE_FACTOR): Define.
(assign_reg_reg_set, struct_equiv_make_checkpoint): New functions.
(struct_equiv_improve_checkpoint): Likewise.
(struct_equiv_restore_checkpoint, rtx_equiv_p): Likewise.
(set_dest_equiv_p, set_dest_addr_equiv_p, struct_equiv_init): Likewise.
(struct_equiv_merge, find_dying_input): Likewise.
(resolve_input_conflict, note_local_live): Likewise.
(death_notes_match_p): Change parameters and processing
to match rtx_equiv_p machinery.  Change caller.
(insns_match_p): Likewise.
(flow_find_cross_jump): Replace with:
(struct_equiv_block_eq).

Back out this change:
2005-03-07  Kazu Hirata  <kazu@cs.umass.edu>
          * recog.c (verify_changes): Make it static.
          * recog.h: Remove the corresponding prototype.

From-SVN: r108480

rtlhooks.c (gen_lowpart_general): Handle SUBREGs of floating point values.

* rtlhooks.c (gen_lowpart_general): Handle SUBREGs of floating point
values.

From-SVN: r108479

t-mt (MULTILIB_OPTIONS): Add ms2

* config/mt/t-mt (MULTILIB_OPTIONS): Add ms2
(MULTILIB_DIRNAMES): Add ms2.  Prefix ms1 dirs with 'ms1'.
* config/mt/mt.c (ms1_final_prescan): Use TARGET_MS2,
TARGET_MS1_64_001 appropriately.
(ms1_machine_reorg): Use TARGET_MS2.

From-SVN: r108477

trans-expr.c (gfc_conv_substring): Use fold_build2 and build_int_cst.

2005-12-13  Richard Guenther  <rguenther@suse.de>

* trans-expr.c (gfc_conv_substring): Use fold_build2 and
build_int_cst.

From-SVN: r108476

re PR java/25366 (ICE in do_resolve_class)

2005-12-13  Andrew Haley  <aph@redhat.com>

        PR java/25366
        PR java/25368
        * class.c (maybe_layout_super_class): Update current_class before
        calling do_resolve_class.

From-SVN: r108474

re PR libfortran/24991 (gfortran build fails with - error:gthr-default.h: No such file or directory)

PR libfortran/24991
* acinclude.m4 (LIBGFOR_CHECK_PRAGMA_WEAK): Rename to...
(LIBGFOR_GTHREAD_WEAK): ... this.  Define SUPPORTS_WEAK rather
than HAVE_PRAGMA_WEAK.  Define GTHREAD_USE_WEAK to 0 on hosts
that shouldn't use weak in gthr.h.
* configure.ac: Use LIBGFOR_GTHREAD_WEAK instead of
LIBGFOR_CHECK_PRAGMA_WEAK.
* config.h.in: Regenerated.
* configure: Regenerated.
* io/io.h (SUPPORTS_WEAK): Don't define here.

* intrinsics/ftell.c (ftell, FTELL_SUB): Add unlock_unit call.
* intrinsics/fget.c (fgetc, fputs): Likewise.
* intrinsics/tty.c (ttynam): Likewise.

From-SVN: r108471

Update.

From-SVN: r108465

re PR debug/25023 (ICE in def_cfa_1, at dwarf2out.c:792)

PR debug/25023
PR target/25293
* expr.c (emit_move_resolve_push): Handle PRE_MODIFY
and POST_MODIFY with CONST_INT adjustment equal to PUSH_ROUNDING.
Fix POST_INC/POST_DEC handling if PUSH_ROUNDING is not identity.
* config/i386/i386.md (pushhi2, pushqi2): Use pushl instead of pushw.
Set mode to SI, adjust constraints.
(pushhi2_rex64, pushqi2_rex64): Set mode to DI.
* config/i386/i386.h (PUSH_ROUNDING): Round up to 4 instead of 2 for
32-bit code.

* gcc.target/i386/pr25293.c: New test.

From-SVN: r108463

re PR c++/24907 ("int x, ;" accepted)

2005-12-13  Petr Machata  <machata@post.cz>

PR c++/24907
* parser.c (cp_parser_simple_declaration): Require comma at the
beginning of processing second and later declarators, instead of
allowing the comma at the end of each iteration.

* g++.dg/parse/comma2.C: New test.

From-SVN: r108462

struct-layout-1.exp: Do not link with libiberty.

* g++.dg/compat/struct-layout-1.exp: Do not link with libiberty.
* g++.dg/compat/struct-layout-1_generate.c (config.h): Do not include.
(limits.h): Include unconditionally.
(stdlib.h): Likewise.
(hashtab.h): Do not include.
(getopt.h): Likewise.
(stddef.h): Include.
(hashval_t): Define.
(struct entry): Add "next" field.
(HASH_SIZE): New macro.
(hash_table): New variable.
(switchfiles): Do not use xmalloc.
(mix): New macro.
(iterative_hash): New function.
(hasht): Remove.
(e_exists): New function.
(e_insert): Likewise.
(output): Use, instead of libiberty hashtable functions.
(main): Do not use getopt.  Do not call htab_create.

Co-Authored-By: Jakub Jelinek <jakub@redhat.com>
From-SVN: r108461

re PR c++/25331 (FAIL: tmpdir-g++.dg-struct-layout-1/t028 cp_compat_[xy]_tst.o compile)

PR c++/25331
* gcc.dg/compat/struct-layout-1_generate.c (subfield): Don't
put atal* and atpaal* attributes on array elements.
(generate_fields): Revert 2005-10-07 and 2005-10-11 changes.
* g++.dg/compat/struct-layout-1_generate.c (subfield): Don't
put atal* and atpaal* attributes on array elements.
(generate_fields): Don't prevent atal* attribute on ETYPE_ARRAY.

From-SVN: r108460

Make-lang.in (fortran/trans-resolve.o): Depend on fortran/dependency.h.

gcc/fortran/
* Make-lang.in (fortran/trans-resolve.o): Depend on
fortran/dependency.h.
* gfortran.h (gfc_expr): Add an "inline_noncopying_intrinsic" flag.
* dependency.h (gfc_get_noncopying_intrinsic_argument): Declare.
(gfc_check_fncall_dependency): Change prototype.
* dependency.c (gfc_get_noncopying_intrinsic_argument): New function.
(gfc_check_argument_var_dependency): New function, split from
gfc_check_fncall_dependency.
(gfc_check_argument_dependency): New function.
(gfc_check_fncall_dependency): Replace the expression parameter with
separate symbol and argument list parameters.  Generalize the function
to handle dependencies for any type of expression, not just variables.
Accept a further argument giving the intent of the expression being
tested.  Ignore intent(in) arguments if that expression is also
intent(in).
* resolve.c: Include dependency.h.
(find_noncopying_intrinsics): New function.
(resolve_function, resolve_call): Call it on success.
* trans-array.h (gfc_conv_array_transpose): Declare.
(gfc_check_fncall_dependency): Remove prototype.
* trans-array.c (gfc_conv_array_transpose): New function.
* trans-intrinsic.c (gfc_conv_intrinsic_function): Don't use the
libcall handling if the expression is to be evaluated inline.
Add a case for handling inline transpose()s.
* trans-expr.c (gfc_trans_arrayfunc_assign): Adjust for the new
interface provided by gfc_check_fncall_dependency.

libgfortran/
* m4/matmul.m4: Use a different order in the special case of a
transposed first argument.
* generated/matmul_c4.c, generated/matmul_c8.c, generated/matmul_c10.c,
* generated/matmul_c16.c, generated/matmul_i4.c, generated/matmul_i8.c,
* generated/matmul_i10.c, generated/matmul_r4.c, generated/matmul_r8.c
* generated/matmul_r10.c, generated/matmul_r16.c: Regenerated.

Co-Authored-By: Victor Leikehman <LEI@il.ibm.com>
From-SVN: r108459

c-cppbuiltin.c (builtin_define_float_constants): Add fp_cast parameter, pass to builtin_define_with_hex_fp_value.

2005-12-13  Carlos O'Donell <carlos@codesourcery.com>

gcc/

* c-cppbuiltin.c (builtin_define_float_constants): Add
fp_cast parameter, pass to builtin_define_with_hex_fp_value.
Define __FLT_HAS_DENORM__, __DBL_HAS_DENORM__, __LDBL_HAS_DENORM__.
(builtin_define_with_hex_fp_value): Use fp_cast when building macro.
(c_cpp_builtins): If flag_single_precision_constant then set fp_cast
to "((double)%sL)" otherwise "%s".

gcc/testsuite/

* gcc.dg/single-precision-constant.c: New test.

libstdc++-v3/

* include/std/std_limits.h (struct numeric_limits):
Use __DBL_HAS_DENORM__, __FLT_HAS_DENORM__, __LDBL_HAS_DENORM__.

From-SVN: r108458