boost_shared_ptr.h (_Sp_counted_base::release, [...]): Protect barriers with __GTHREADS.

2005-11-10  Paolo Carlini  <pcarlini@suse.de>

* include/tr1/boost_shared_ptr.h (_Sp_counted_base::release,
weak_release): Protect barriers with __GTHREADS.

From-SVN: r106733

re PR ada/23995 (ICE on misaligned rep clause)

PR ada/23995
* trans.c (call_to_gnu): Restore statement lost in translation.

From-SVN: r106731

libgfortran.h: Add missing prototypes for internal_pack functions.

        * libgfortran.h: Add missing prototypes for internal_pack
        functions.

From-SVN: r106730

boost_shared_ptr.h (_Sp_counted_base::release): Optimize by manually inlining weak_release.

2005-11-10  Paolo Carlini  <pcarlini@suse.de>
    Peter Dimov  <pdimov@mmltd.net>

* include/tr1/boost_shared_ptr.h (_Sp_counted_base::release):
Optimize by manually inlining weak_release.

* include/tr1/boost_shared_ptr.h (_Sp_counted_base::release,
weak_release): Use explicit memory barriers.

Co-Authored-By: Peter Dimov <pdimov@mmltd.net>
From-SVN: r106729

re PR libstdc++/22203 (std::numeric_limits<int>::traps is wrong on PPC)

2005-11-09  Benjamin Kosnik  <bkoz@redhat.com>
    Paolo Carlini  <pcarlini@suse.de>
            Gabriel Dos Reis  <gdr@integrable-solutions.net>

PR libstdc++/22203
* numeric_limits.cc: Split into...
* numeric_limits/specialization.cc: ..this.
* numeric_limits/is_iec559.cc: Same.
* numeric_limits/sign.cc: Same.
* numeric_limits/quiet_NaN.cc: Same.
* numeric_limits/digits10.cc: Same.
* numeric_limits/denorm_min.cc: Same.
* numeric_limits/min_max.cc: Same.
* numeric_limits/epsilon.cc: Same.
* numeric_limits/infinity.cc: Same.
* numeric_limits/traps.cc: New.

Co-Authored-By: Gabriel Dos Reis <gdr@integrable-solutions.net>
Co-Authored-By: Paolo Carlini <pcarlini@suse.de>
From-SVN: r106727

Daily bump.

From-SVN: r106723

function.c (assign_stack_local_1): Issue an error message if the frame size overflows in the signed target arithmetics.

* function.c (assign_stack_local_1): Issue an error message if
the frame size overflows in the signed target arithmetics.

From-SVN: r106717

tree.c (build_qualified_type): Chain the new type to the original type's TYPE_NEXT_PTR_TO or TYPE_NEXT_REF_TO...

* tree.c (build_qualified_type): Chain the new type to the original
type's TYPE_NEXT_PTR_TO or TYPE_NEXT_REF_TO linked lists if it is
a POINTER_TYPE or a REFERENCE_TYPE respectively.
(build_pointer_type_for_mode): Only return unqualified types.
(build_reference_type_for_mode): Likewise.

From-SVN: r106716

Fix typo in comment.

From-SVN: r106709

Makefile.in (gnucompare): Do comparison of all files using one of the chosen methods and only afterwards...

* Makefile.in (gnucompare): Do comparison of all files using one of
the chosen methods and only afterwards decide if just warning should
be issued or comparison failure raised.

From-SVN: r106707

ifcvt.c (noce_get_alt_condition): Use prev_nonnote_insn.

* ifcvt.c (noce_get_alt_condition): Use prev_nonnote_insn.
(noce_try_abs): Negate if the comparison is reversed.
Look only one instruction backwards for a REG_EQUAL note.

From-SVN: r106706

re PR other/4372 (#pragma weak pthread* inclusion causes applications to crash without a linker error when one forgets to link with -lpthread)

gcc/ChangeLog:
PR other/4372
* gthr-dce.h, gthr-posix.h, gthr-posix95.h, gthr-solaris.h,
gthr-tpf.h: Define __gthrw.  For all identifiers that might
be weak, introduce weakrefs or non-weak aliases with __gthrw,
and prefix all uses with __ghtrw.
libobjc/ChangeLog:
PR other/4372
* thr-objc.c (_XOPEN_SOURCE): Define.

From-SVN: r106704

re PR other/4372 (#pragma weak pthread* inclusion causes applications to crash without a linker error when one forgets to link with -lpthread)

PR other/4372
* tree.h (IDENTIFIER_TRANSPARENT_ALIAS): New.
(TREE_DEPRECATED): Adjust comment.  Check for a DECL.
* c-common.c (handle_weakref_attribute): New.
(c_common_attribute_table): Add weakref.
* configure.ac (HAVE_GAS_WEAKREF): Check for weakref support
in the assembler.
* configure, config.in: Rebuilt.
* defaults.h (ASM_OUTPUT_WEAKREF): Define if HAVE_GAS_WEAKREF.
* doc/extend.texi: Document weakref attribute.
* varasm.c (ultimate_transparent_alias_target): New
(assemble_name): Use it.
(weak_finish_1): Split out of...
(weak_finish): ... and deal with weakrefs in...
(weakref_targets): ... new list.
(globalize_decl): Clean up weakref_targets.
(do_assemble_alias): Handle weakrefs.
(finish_aliases_1): Do not reject weakrefs to external symbols.
(assemble_alias): Handle weakrefs.

From-SVN: r106703

jacks.xfail (non-jls-zip-2): Now passes.

* testsuite/libjava.jacks/jacks.xfail (non-jls-zip-2): Now
passes.

From-SVN: r106702

re PR tree-optimization/24716 (Wrong code generated when optimising)

2005-11-09  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/24716
* tree-scalar-evolution.c (analyze_evolution_in_loop): Use
t_bool to track results from follow_ssa_edge.

* gcc.c-torture/execute/pr24716.c: New testcase.

From-SVN: r106700

final.c (force_source_line): New global variable.

* final.c (force_source_line): New global variable.
(final_scan_insn): Set it to true instead of clearing last_filename.
(notice_source_line): Return true if force_source_line is true,
unless source info is absent.

From-SVN: r106699

re PR c/24644 (gcc-4.1 compiled ppc64 kernels do not boot)

2005-11-09  Andrew Pinski  <pinskia@physics.uc.edu>

        PR c/24644
        * dwarf2-out.c (add_name_and_src_coords_attributes): Don't add
        a linkage name for a variable if it a register variable.
        * c-decl.c (grokdeclarator): Global register variables
        should be set as PUBLIC.

From-SVN: r106693

Added PR number.

From-SVN: r106692

pr24624.c: Disable for non S/390 targets.

2005-11-09  Andreas Krebbel  <krebbel1@de.ibm.com>

* gcc.dg/pr24624.c: Disable for non S/390 targets.

From-SVN: r106691

Moved testsuite changes from gcc/ChangeLog to gcc/testsuite/ChangeLog

From-SVN: r106690

boost_shared_ptr.h: Trivial formatting fixes.

2005-11-09  Paolo Carlini  <pcarlini@suse.de>

* include/tr1/boost_shared_ptr.h: Trivial formatting fixes.

From-SVN: r106689

s390.c (struct s390_frame_layout): New fields first_save_gpr_slot and last_save_gpr_slot.

2005-11-09  Andreas Krebbel  <krebbel1@de.ibm.com>

* config/s390/s390.c (struct s390_frame_layout): New fields
first_save_gpr_slot and last_save_gpr_slot.
(cfun_grps_save_area_size, s390_frame_info, s390_emit_prologue,
s390_emit_epilogue, s390_initial_elimination_offset): Replaced
first_save_gpr and last_save_gpr with the _slot variants.
(s390_register_info): Calculate first_save_gpr_slot and
last_save_gpr_slot using regs_ever_live.
* testsuite/gcc.dg/pr24624.c: Testcase added.

From-SVN: r106688

s390.c (s390_regs_ever_clobbered): Only save live eh regs for a function containing a landing pad.

2005-11-09  Andreas Krebbel  <krebbel1@de.ibm.com>

* config/s390/s390.c (s390_regs_ever_clobbered): Only save live eh regs
for a function containing a landing pad.
* testsuite/g++.dg/other/pr24623.C: Testcase added.

From-SVN: r106687

flow.c (mark_set_1): Handle CLOBBERs like SETs if the register is live afterwards.

2005-11-09  Andreas Krebbel  <krebbel1@de.ibm.com>

* flow.c (mark_set_1): Handle CLOBBERs like SETs if the register
is live afterwards.

From-SVN: r106686

t-bfin-elf (MULTILIB_OPTIONS, [...]): New.

* config/bfin/t-bfin-elf (MULTILIB_OPTIONS, MULTILIB_DEFAULTS,
MULTILIB_DIRNAMES, MULTILIB_EXCEPTIONS): New.

From-SVN: r106684

re PR fortran/22607 (Elemental character functions don't work)

fortran/
2005-11-09  Erik Edelmann  <eedelman@gcc.gnu.org>

        PR fortran/22607
        * trans-decl.c(build_function_decl): Don't set
        DECL_IS_PURE (fndecl) = 1 for return-by-reference
        functions.

testsuite/
2005-11-09  Erik Edelmann  <eedelman@gcc.gnu.org>

        PR fortran/22607
        * gfortran-dg/pure_byref_1.f90: New.
        * gfortran-dg/pure_byref_2.f90: New.

From-SVN: r106683

Add ms2 support

Add ms2 support
* config/ms1/ms1.md (UNSPEC_BLOCKAGE, UNSPEC_EI, UNSPEC_DI): New
constants.
(call,load,store): New insn types.
(mem_access, branch_access): Adjust reservation conditions.
(define_delay): Adjust condition.
(decrement_and_branch_until_zero): Allow for ms2.  Set branch
type.
(*decrement_and_rbanch_until_zero_no_clobber): Allow for ms2.
(*movqi_internal,*movsi_internal,*movsf_internal): Use load,store
insn type.
(call_internal, call_value_internal, return_internal,
return_interrupt_internal, eh_return_internal, indirect_jump,
tablejump): Set call insn type.
(blockage, ei, di): Use appropriate unspec const.
* config/ms1/ms1.c (ms1_flag_delayed_branch): New.
(ms1_get_attr_type): Adjust to give load & store types.
(ms1_final_prescan_insn): Adjust for new insn types.  Don't look
backwards past a barrier.
(ms1_override_options): Accept ms2 arch.  Copy and reset delayed
branch scheduling.
(struct branch_info, struct label_info): New.
(ms1_labels): New.
(ms1_add_branches, ms1_check_delay_slot, ms1_reorg_hazard): New.
(ms1_machine_reorg): New.
(TARGET_MACHINE_DEPENDENT_REORG): Override.
* config/ms1/crtn.asm: Add nop for ms2 JAL hazard.
* config/ms1/ms1.h (processor_type): Add PROCESSOR_MS2.
(ASM_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Add ms2.
(TARGET_MS2): New.

From-SVN: r106680

re PR c/24101 (Segfault with preprocessed source)

libcpp/

PR c/24101
* init.c (read_original_filename): Temporarily set
state.in_directive before calling _cpp_lex_direct for
CPP_HASH tokens.

gcc/

PR c/24101
* toplev.c (process_options): Initialize debug_hooks early
in case lang_hooks.post_options ends up calling a debug_hook.

gcc/testsuite/

PR c/24101
* gcc.dg/pr24101-1.i, gcc.dg/pr24101-2.i: New tests.
* gcc.dg/dg.exp: Run main loop also for *.i files.

From-SVN: r106677

Daily bump.

From-SVN: r106670

libstdc++.exp (check_v3_target_debug_mode): Use exe as the executable extension.

2005-11-08  Benjamin Kosnik  <bkoz@redhat.com>

* testsuite/lib/libstdc++.exp (check_v3_target_debug_mode): Use
exe as the executable extension.

* acinclude.m4 (GLIBCXX_ENABLE_ALLOCATOR): Change gnu defaults to
new.
(GLIBCXX_ENABLE_PCH): Adjust message order.
* configure: Regenerate.

* scripts/testsuite_flags.in (query): Correct print_usagex typo.

* include/bits/functexcept.h: Remove argument names.

* src/mt_allocator.cc: Adjust comment.

From-SVN: r106665

init.c: Use the Linux-specific section for the IA-64/Linux target.

* init.c: Use the Linux-specific section for the IA-64/Linux target.
(__gnat_adjust_context_for_raise): Add conditional code so that the
IA-64 is also supported.

From-SVN: r106664

re PR libgcj/24587 (libgcj should set gnu.classpath.version system property)

PR libgcj/24587:
* gnu/classpath/SystemProperties.java: Set gnu.classpath.version.

From-SVN: r106659

dwarf2out.c (multiple_reg_loc_descriptor): Don't assume DBX_REGISTER_NUMBER being contiguous.

* dwarf2out.c (multiple_reg_loc_descriptor): Don't assume
DBX_REGISTER_NUMBER being contiguous.

From-SVN: r106658

re PR c++/19450 (__thread static class members)

PR c++/19450
* decl.c (redeclaration_error_message): Issue diagnostics about
olddecl and newdecl disagreement on __thread property.
(grokdeclarator): Set DECL_TLS_MODEL on class static variables.

* g++.dg/tls/diag-3.C: New test.
* g++.dg/tls/diag-4.C: New test.
* g++.dg/tls/static-1.C: New test.
* g++.dg/tls/static-1a.cc: New file.

From-SVN: r106657

re PR c++/23046 (ICE in set_value_range, at tree-vrp.c:191)

2005-11-08  James A. Morrison  <phython@gcc.gnu.org>
    Diego Novillo  <dnovillo@redhat.com>

PR 23046
* tree-vrp.c (register_edge_assert_for): Do not register
always-false predicates.

testsuite/

PR 23046
* g++.dg/tree-ssa/pr23046.C: New test.

From-SVN: r106656

PR libgcj/23763.  From aeby@graeff.com.

* java/lang/natPosixProcess.cc (nativeSpawn): Unblock SIGCHLD
before exec.

From-SVN: r106655

re PR tree-optimization/23115 (-ftree-vectorize generates wrong code)

       PR tree-optimization/23115
       * tree-if-conv.c (find_phi_replacement_condition): Check domninated_by
       relation.

       * gcc.dg/tree-ssa/pr23115.c: New.

From-SVN: r106653

dump-parse-tree.c: Fix comment typo, add a few blank lines.

* dump-parse-tree.c: Fix comment typo, add a few blank lines.
Also, reformat previous ChangeLog entry.

From-SVN: r106652

re PR java/23617 (Out of memory when classpath contains jar file with zip-style comment)

2005-11-08  Wil Mahan  <wmahan@gmail.com>

PR java/23617
* zextract.c (read_zip_archive): Fix out of memory error when
reading jar files with zip-style comments.

From-SVN: r106648

t-fprules (MULTILIB_MATCHES_FLOAT): Include -mcpu=440.

* config/rs6000/t-fprules (MULTILIB_MATCHES_FLOAT): Include
-mcpu=440.

From-SVN: r106646

re PR tree-optimization/23382 (Does not remove the old HEAP virtual variables in clobbered)

2005-11-08  Daniel Berlin  <dberlin@dberlin.org>

Fix PR tree-optimization/23382

* tree-ssa-alias.c (compute_may_aliases): Call
delete_old_heap_vars.
* tree-dfa.c (referenced_var_remove): New function.
* tree-ssa.c (delete_tree_ssa): Call delete_old_heap_vars.
* tree-flow.h (referenced_var_remove): Add prototype.
(delete_old_heap_vars): Ditto.
* tree-ssa-structalias.c (heapvars): New variable.
(oldheapvars): Ditto.
(get_constraint_for): Put heap vars on heapvars list.
(delete_old_heap_vars): New function.

From-SVN: r106643

altivec-{1-13}.C: XFAIL on AIX and SPE.

* g++.dg/ext/altivec-{1-13}.C: XFAIL on AIX and SPE.
* gcc.target/powerpc/pr16155.c: Same.
* gcc.target/powerpc/altivec-1.c: Same.
* gcc.target/powerpc/altivec-2.c: Same.
* gcc.target/powerpc/altivec-3.c: Same.
* gcc.target/powerpc/altivec-5.c: Same.
* gcc.target/powerpc/altivec-11.c: Same.
* gcc.target/powerpc/altivec-12.c: Same.
* gcc.target/powerpc/altivec-14.c: Same.
* gcc.target/powerpc/altivec-22.c: Same.
* gcc.target/powerpc/altivec-consts.c: Same.
* gcc.target/powerpc/altivec-pr22085.c: Same.
* gcc.target/powerpc/altivec-splat.c: Same.
* gcc.target/powerpc/altivec-varargs-1.c: Same.
* gcc.target/powerpc/altivec-vec-merge.c: Same.
* gcc.target/powerpc/darwin-longlong.c: Same.

From-SVN: r106638

* scripts/makemake.tcl: Set package gnu/javax/sound/midi to bc.

From-SVN: r106636

re PR c++/21123 (ICE in cp_expr_size, at cp/cp-objcp-common.c:101)

        PR c++/21123
        * cp/method.c (use_thunk): Use build_cplus_new instead of
        force_target_expr.

        * tree.h (CALL_FROM_THUNK_P): Add CALL_EXPR_CHECK.

From-SVN: r106634

re PR target/19340 (Compilation SEGFAULTs with -O1 -fschedule-insns2 -fsched2-use-traces on an x86 architecture.)

PR target/19340
* reg-stack.c (reg_to_stack): Update register liveness also
for flag_sched2_use_traces.

testsuite/

PR target/19340
* gcc.dg/pr19340.c: New test.

From-SVN: r106632

re PR target/23704 (gcc.dg/rs6000-fpint.c fails)

PR target/23704
* config/rs6000/rs6000.c (rs6000_handle_option <OPT_m64>): Don't
override prior explicit -mno-powerpc-gfxopt.

From-SVN: r106631

Removed decompilation for non-ELF targets.

From-SVN: r106628

Daily bump.

From-SVN: r106625

Added Terry Laurenzo

From-SVN: r106620

mauve-libgcj: Disable javax.rmi.* and java.io.InputStreamReader.getEncoding.

* mauve-libgcj: Disable javax.rmi.* and
java.io.InputStreamReader.getEncoding.

From-SVN: r106614

*  Update ChangeLog with the missing entry from my previous commit.

From-SVN: r106611

expmed.c (extract_bit_field): Do not use insv/extv/extzv patterns if the bitsize is zero.

* expmed.c (extract_bit_field): Do not use insv/extv/extzv patterns
if the bitsize is zero.
* doc/md.texi (Standard Pattern Names): Document it.

* config/ia64/ia64.c (ia64_pass_by_reference): Delete.
(TARGET_PASS_BY_REFERENCE): Likewise.

From-SVN: r106605

re PR rtl-optimization/24683 (ICE in in extract_insn, at recog.c:2084)

./: PR rtl-optimization/24683
* config/i386/i386.c (legitimize_pic_address): If constant operand
to PLUS is too large, put it in a register.
testsuite/:
PR rtl-optimization/24683
* gcc.dg/pr24683.c: New test.

From-SVN: r106601

configure.ac: Enable checking assembler dwarf2 support for bfin target.

* configure.ac: Enable checking assembler dwarf2 support for bfin
target.
* configure: Regenerate.

From-SVN: r106598

configure.ac: Enable checking assembler dwarf2 support for bfin target.

* configure.ac: Enable checking assembler dwarf2 support for bfin
target.
* configure: Regenerate.

From-SVN: r106597

error.c: Use flag_fatal_error.

    *error.c: Use flag_fatal_error.
    *invoke.texi:  Remove -Werror from list of options.

From-SVN: r106595

pr24225.c: Add cleanup-coverage-files.

* gcc.dg/vect/pr24225.c: Add cleanup-coverage-files.
* gcc.dg/vect/pr24300.c: Add cleanup-tree-dump.

From-SVN: r106592

pr23376: Compile with -mmmx.

* gcc.target/i386/pr23376: Compile with -mmmx.

From-SVN: r106591

* g++.dg/abi/empty6.C (B): Align to twice pointer size.

From-SVN: r106589

re PR target/24230 (ICE in extract_insn with altivec)

2005-11-07  Paolo Bonzini  <bonzini@gnu.org>

PR target/24230

* config/rs6000/rs6000.c (easy_vector_splat_const, easy_vector_same,
gen_easy_vector_constant_add_self): Delete.
(vspltis_constant, easy_altivec_constant, gen_easy_altivec_constant):
New.
(output_vec_const_move): Use gen_easy_altivec_constant.
(rs6000_expand_vector_init): Do not emit a set of a VEC_DUPLICATE.
* config/rs6000/predicates.md (easy_vector_constant): Reorganize tests.
(easy_vector_constant_add_self): Rewritten.
* config/rs6000/rs6000-protos.h (easy_vector_splat_const,
easy_vector_same, gen_easy_vector_constant_add_self): Remove prototype.
(easy_altivec_constant, gen_easy_altivec_constant): Add prototype.

testsuite:
2005-11-07  Paolo Bonzini  <bonzini@gnu.org>

PR target/24230

        * gcc.target/powerpc/altivec-consts.c,
        gcc.target/powerpc/altivec-splat.c: New testcase.

From-SVN: r106588

c-typeck.c (build_c_cast): Try using a shared constant...

2005-11-07  Paolo Bonzini  <bonzini@gnu.org>

        * c-typeck.c (build_c_cast): Try using a shared constant, and see
        if TREE_OVERFLOW or TREE_CONSTANT_OVERFLOW really changed.

testsuite:
2005-11-07  Paolo Bonzini  <bonzini@gnu.org>

        * gcc.dg/overflow-2.c: New testcase.

From-SVN: r106587

re PR rtl-optimization/23567 (if-conversion causes wrong code)

PR rtl-optimization/23567
* ifcvt.c (noce_mem_write_may_trap_or_fault_p): New function.
(noce_process_if_block): Don't do any optimizations except
if (cond) x = x; if !set_b and write into orig_x may trap
or fault.  Remove the MEM_READONLY_P check.

* gcc.c-torture/execute/20051104-1.c: New test.

From-SVN: r106585

splay-tree.c (rotate_left): New function.

2005-11-07  Richard Guenther  <rguenther@suse.de>

* splay-tree.c (rotate_left): New function.
(rotate_right): Likewise.
(splay_tree_splay_helper): Remove.
(splay_tree_splay): Re-implement.

From-SVN: r106584

Fix PR number.

From-SVN: r106583

re PR ada/21952 (Annoying "attribute directive ignored" warnings)

2005-11-07  James A. Morrison  <phython@gcc.gnu.org>

        PR treelang/21952
        * treetree.c (LANG_HOOKS_ATTRIBUTE_TABLE): Set to
        treelang_attribute_table.
        (handle_attribute): New function.
        (treelang_attribute_table): New attribute table.

From-SVN: r106582

re PR c++/17256 (undefined but used static or inline functions should be diagnosed)

        PR c++/17256
        * decl2.c (cp_finish_file): Fix conditions for undefined warning.
        Set TREE_NO_WARNING instead of TREE_PUBLIC.
        * pt.c (instantiate_pending_templates): Set DECL_INITIAL to avoid
        a warning on a function we didn't instantiate because of excessive
        recursion.

Co-Authored-By: James A. Morrison <phython@gcc.gnu.org>
From-SVN: r106581

class.c (record_subobject_offsets): Don't record offsets past biggest empty class for non-empty base classes.

* class.c (record_subobject_offsets): Don't record offsets past
biggest empty class for non-empty base classes.
(layout_class_type): Use TYPE_SIZE_UNIT, not TYPE_SIZE, when
keeping track of the size of emptyclasses.

From-SVN: r106580

Daily bump.

From-SVN: r106577

MAINTAINERS (Write After Approval): Add myself.

2005-11-06  Janne Blomqvist  <jb@gcc.gnu.org>

* MAINTAINERS (Write After Approval): Add myself.

From-SVN: r106569

Reset svn:executable property for those testcases that had it set.

From-SVN: r106568

re PR fortran/24534 (PUBLIC derived types with private components)

2005-11-06  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/24534
* resolve.c (resolve_symbol): Exclude case of PRIVATE declared
within derived type from error associated with PRIVATE type
components within derived type.

PR fortran/20838
PR fortran/20840
* gfortran.h: Add prototype for gfc_has_vector_index.
* io.c (gfc_resolve_dt): Error if internal unit has a vector index.
* expr.c (gfc_has_vector_index): New function to check if any of
the array references of an expression have vector inidices.
(gfc_check_pointer_assign): Error if internal unit has a vector index.

PR fortran/17737
* data.c (gfc_assign_data_value): Remove gcc_assert that caused the ICE
and replace by a standard dependent warning/error if overwriting an
existing initialization.
* decl.c (gfc_data_variable): Remove old error for already initialized
variable and the unused error check for common block variables.  Add
error for hots associated variable and standard dependent error for
common block variables, outside of blockdata.
* symbol.c (check_conflict): Add constraints for DATA statement.

2005-11-06  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/24534
gfortran.dg/private_type_2.f90: Modified to check that case with
PRIVATE declaration within derived type is accepted.

PR fortran/20838
gfortran.dg/pointer_assign_1.f90: New test.

PR fortran/20840
* gfortran.dg/arrayio_0.f90: New test.

PR fortran/17737
gfortran.dg/data_initialized.f90: New test.
gfortran.dg/data_constraints_1.f90: New test.
gfortran.dg/data_constraints_2.f90: New test.

From-SVN: r106567

re PR c++/21308 (Very high compile time)

PR c++/21308
* class.c (sizeof_biggest_empty_class): New variable.
(record_subobject_offsets): Don't record offsets past biggest
empty class for data members.  Replace vbases_p parameter with
is_data_member parameter.
(build_base_field): Adjust call.
(layout_class_type): Likewise.  Maintain
sizeof_biggest_empty_class.

From-SVN: r106566

MAINTAINERS (Write After Approval): Add myself.

2005-11-06  Erik Edelmann  <eedelman@gcc.gnu.org>

        * MAINTAINERS (Write After Approval): Add myself.

From-SVN: r106565

gcc_release: Further update for SVN.

* gcc_release: Further update for SVN.  Don't set EXPORTTAG or
EXPORTDATE; always determine SVNREV the same way.  Correct message
about what sources are being retrived.  Include branch information
in LAST_UPDATED.
* crontab: Don't include initial / in branch name.

From-SVN: r106564

gfortran ChangeLog

2005-11-06  Janne Blomqvist <jb@gcc.gnu.org>

PR fortran/24174
PR fortran/24305
* fortran/trans-io.c (gfc_build_io_library_fndecls): Add kind
argument to transfer_array.
(transfer_array_desc): Add kind argument.

testsuite ChangeLog:

2005-11-06  Janne Blomqvist <jb@gcc.gnu.org>

PR fortran/24174
PR fortran/24305
* testsuite/gfortran.dg/large_real_kind_form_io_1.f90: New file.

libgfortran Changelog:

2005-11-06  Janne Blomqvist <jb@gcc.gnu.org>

PR fortran/24174
PR fortran/24305
* io/io.h: Add argument to prototypes, add prototypes for
size_from_*_kind functions.
* io/list_read.c (read_complex): Add size argument, use
it.
(list_formatted_read): Add size argument, cleanup.
(list_formatted_read_scalar): Add size argument.
(nml_read_obj): Fix for padding.
* io/transfer.c: Add argument to transfer function pointer.
(unformatted_read): Add size argument.
(unformatted_write): Likewise.
(formatted_transfer_scalar): Fix for padding with complex(10).
(formatted_transfer): Add size argument, cleanup.
(transfer_integer): Add size argument to transfer call.
(transfer_real): Likewise.
(transfer_logical): Likewise.
(transfer_character): Likewise.
(transfer_complex): Likewise.
(transfer_array): New kind argument, use it.
(data_transfer_init): Add size argument to formatted_transfer
call.
(iolength_transfer): Add size argument, cleanup.
* io/write.c (write_complex): Add size argument, fix for padding
with complex(10).
(list_formatted_write): Add size argument, cleanup.
(list_formatted_write_scalar): Add size argument, use it.
(nml_write_obj): Fix for size vs. kind issue.
* io/size_from_kind.c: New file.
* Makefile.am: Add io/size_from_kind.c.
* configure: Regenerate.
* Makefile.in: Regenerate.

From-SVN: r106563

re PR tree-optimization/24670 (VRP ICE in compare_name_with_value)

PR 24670
* tree-vrp.c (fix_equivalence_set): New.
(extract_range_from_assert): Call it.

testsuite/

PR 24670
* gcc.dg/tree-ssa/pr24670.c: New test.

From-SVN: r106562

re PR libstdc++/18174 (documentation example for std::priority_queue usage)

2005-11-06  Paolo Carlini  <pcarlini@suse.de>

PR libstdc++/18174
* include/bits/stl_queue.h (priority_queue): Tweak a bit the
comment describing the container.

From-SVN: r106560

intrinsic.c (add_functions): Add ctime and fdate intrinsics.

* intrinsic.c (add_functions): Add ctime and fdate intrinsics.
(add_subroutines): Likewise.
* intrinsic.h: Prototypes for gfc_check_ctime,
gfc_check_ctime_sub, gfc_check_fdate_sub, gfc_resolve_ctime,
gfc_resolve_fdate, gfc_resolve_ctime_sub, gfc_resolve_fdate_sub.
* gfortran.h: Add GFC_ISYM_CTIME and GFC_ISYM_FDATE.
* iresolve.c (gfc_resolve_ctime, gfc_resolve_fdate,
gfc_resolve_ctime_sub, gfc_resolve_fdate_sub): New functions.
* trans-decl.c (gfc_build_intrinsic_function_decls): Add
gfor_fndecl_fdate and gfor_fndecl_ctime.
* check.c (gfc_check_ctime, gfc_check_ctime_sub,
gfc_check_fdate_sub): New functions.
* trans-intrinsic.c (gfc_conv_intrinsic_ctime,
gfc_conv_intrinsic_fdate): New functions.
(gfc_conv_intrinsic_function): Add cases for GFC_ISYM_CTIME
and GFC_ISYM_FDATE.
* intrinsic.texi: Documentation for the new CTIME and FDATE
intrinsics.
* trans.h: Declarations for gfor_fndecl_ctime and gfor_fndecl_fdate.

* intrinsics/ctime.c: New file.
* configure.ac: Add check for ctime.
* Makefile.am: Add ctime.c
* configure: Regenerate.
* config.h.in: Regenerate.
* Makefile.in: Regenerate.

From-SVN: r106558

re PR target/22432 (Wrong code generation using MMX intrinsics on amd64)

./: PR target/22432
* combine.c (apply_distributive_law): Don't distribute across a
vector mode subreg.
testsuite/:
PR target/22432
* gcc.target/i386/pr22432.c: New test.

From-SVN: r106555

gcc_release: Always determine revision number from date or tag...

2005-11-05  Daniel Berlin  <dberlin@dberlin.org>

* gcc_release:  Always determine revision number from date or tag,
and use that, instead of fragile date based export.
Put back setting of non-local CVSROOT.

From-SVN: r106554

Daily bump.

From-SVN: r106547

asm-1.c: Move from gcc.dg/asm-5.c.

        * gcc.target/alpha/asm-1.c: Move from gcc.dg/asm-5.c.
        * gcc.target/alpha/cix-1.c, gcc.target/alpha/cix-2.c,
        gcc.target/alpha/max-1.c, gcc.target/alpha/max-2.c,
        gcc.target/alpha/base-1.c, gcc.target/alpha/base-2.c: Move
        from gcc.dg/alpha-*.
        * gcc.target/alpha/980217-1.c, gcc.target/alpha/20011018-1.c,
        gcc.target/alpha/20000715-1.c: Move from gcc.dg/.

From-SVN: r106538

configure.ac: Use AM_FCFLAGS for extra flags, not FCFLAGS.

2005-11-05  Richard Guenther  <rguenther@suse.de>

* configure.ac: Use AM_FCFLAGS for extra flags, not FCFLAGS.
* configure: Regenerate.

From-SVN: r106535

decl.c, [...]: Fix comment typos.

* decl.c, trans-decl.c: Fix comment typos.
* gfortran.texi: Fix a typo.

From-SVN: r106534

* decl2.c, init.c, typeck.c: Fix comment typos.

From-SVN: r106533

c-typeck.c, [...]: Fix comment typos.

* c-typeck.c, config/i386/netware.h, config/m32c/cond.md,
config/ms1/ms1.h, config/rs6000/predicates.md,
config/s390/s390.c, params.def, postreload-gcse.c,
tree-flow-inline.h, tree-ssa-operands.c, tree-vectorizer.c,
tree-vrp.c, tree.c: Fix comment typos.
* doc/invoke.texi: Fix typos.

From-SVN: r106532

gcc_release: Make tags go in /tags.

2005-11-05  Daniel Berlin  <dberlin@dberlin.org>

* gcc_release: Make tags go in /tags.
Don't tag for snapshots.
Determine revision number of export and put into snapshot
announcement.

From-SVN: r106531

lambda-code.c (lambda_transform_legal_p): Use DDR_NUM_DIST_VECTS for testing whether the data_dependence_relation...

* lambda-code.c (lambda_transform_legal_p): Use DDR_NUM_DIST_VECTS
for testing whether the data_dependence_relation contains distance
vectors.  Iterate over all distance vectors of the ddr.
* lambda.h: Define a vec of lambda_vector pointers.
* tree-data-ref.c (dump_data_dependence_relation,
dump_data_dependence_direction): Iterate over all distance and
direction vectors of the ddr.
(initialize_data_dependence_relation): Initialize DDR_DIR_VECTS and
DDR_DIST_VECTS.
(build_classic_dist_vector, build_classic_dir_vector): Push a set
of distance/direction vectors instead of a single one.
* tree-data-ref.h (dir_vects, dist_vects): Replace dir/dist
lambda_vectors with a vec of lambda_vectors.
(DDR_DIR_VECT, DDR_DIST_VECT): Redefined as operations on vec.
(DDR_DIR_VECTS, DDR_DIST_VECTS, DDR_NUM_DIR_VECTS,
DDR_NUM_DIST_VECTS): New.
* tree-loop-linear.c (gather_interchange_stats): Test for the
existence of distance vectors only after having checked that there
is a dependence.  Iterate over all distance vectors of the ddr.
(linear_transform_loops): Use dump_data_dependence_relation.
* tree-vect-analyze.c (vect_analyze_data_ref_dependence): Test for
distance vectors using DDR_NUM_DIST_VECTS.  Iterate over all the
distance vectors of the ddr.

From-SVN: r106530

check.tpl: Adopt for move to subversion.

        * check.tpl: Adopt for move to subversion.

From-SVN: r106527

bfin.c (n_dregs_to_save, [...]): New argument IS_INTHANDLER; all callers changed.

* config/bfin/bfin.c (n_dregs_to_save, n_pregs_to_save,
expand_prologue_reg_save, expand_epilogue_reg_restore): New argument
IS_INTHANDLER; all callers changed.
(n_regs_saved_by_prologue): Take interrupt handler attributes into
account.
(do_link, do_unlink): New argument ALL; all callers changed.
(expand_interrupt_handler_prologue, expand_interrupt_handler_epilogue):
If function isn't leaf, save and restore all registers.
(bfin_function_ok_for_sibcall): Only true if not an interrupt or
exception handler.

From-SVN: r106526

configure.host: Add | rs6000 to the cpu_defines_dir switch, consistently with the try_cpu one.

2005-11-05  Paolo Carlini  <pcarlini@suse.de>

* configure.host: Add | rs6000 to the cpu_defines_dir switch,
consistently with the try_cpu one.

From-SVN: r106525

re PR libstdc++/22203 (std::numeric_limits<int>::traps is wrong on PPC)

2005-11-05  Paolo Carlini  <pcarlini@suse.de>

PR libstdc++/22203
* include/bits/c++config: Include cpu_defines.h.
* include/Makefile.am: Add cpu_defines.h to host_headers.
* configure.host: Add cpu_defines_dir.
* configure.ac: Use it.
* config/cpu/powerpc/cpu_defines.h: New.
* config/cpu/generic/cpu_defines.h: Likewise.
* configure: Regenerate.
* Makefile.in: Likewise.
* include/Makefile.in: Likewise.
* libmath/Makefile.in: Likewise.
* libsupc++/Makefile.in: Likewise.
* po/Makefile.in: Likewise.
* src/Makefile.in: Likewise.
* testsuite/Makefile.in: Likewise.

From-SVN: r106524

intrinsic.c (add_functions): Add function version of TTYNAM.

* intrinsic.c (add_functions): Add function version of TTYNAM.
* intrinsic.h: Add prototypes for gfc_check_ttynam and
gfc_resolve_ttynam.
* gfortran.h: Add case for GFC_ISYM_TTYNAM.
* iresolve.c (gfc_resolve_ttynam): New function.
* trans-decl.c (gfc_build_intrinsic_function_decls): Add a tree
for function call to library ttynam.
* check.c (gfc_check_ttynam): New function.
* trans-intrinsic.c (gfc_conv_intrinsic_ttynam): New function.
(): Call gfc_conv_intrinsic_ttynam.
* trans.h: Add prototype for gfor_fndecl_ttynam.

* intrinsics/tty.c (ttynam): New function.

From-SVN: r106522

20020523-2.c: Fix path for i386-cpuid.h.

2005-11-05  Andreas Jaeger  <aj@suse.de>

        * gcc.target/i386/20020523-2.c: Fix path for i386-cpuid.h.

From-SVN: r106521

re PR rtl-optimization/23490 (Long compile time for array initializer with inlined constructor)

PR rtl-optimization/23490
* doc/invoke.texi (max-predicted-iterations, max-cse-insns,
max-flow-memory-location): Document.
* flow.c: Include params.h
(MAX_MEM_SET_LIST_LEN): Kill.
(add_to_mem_set_list): Use new param.
* cse.c (cse_basic_block): Replace 1000 by new param.
* params.def (PARAM_MAX_PREDICTED_ITERATIONS, PARAM_MAX_CSE_INSNS,
PARAM_MAX_FLOW_MEMORY_LOCATIONS): New.
* predict.c (predict_loops): Use new param.
* predict.def (MAX_PRED_LOOP_ITERATIONS): Remove.

From-SVN: r106520

ipa-inline.c (cgraph_decide_inlining_of_small_function, [...]): Do not hold memory returned by cgraph_node_name across other call.

* ipa-inline.c (cgraph_decide_inlining_of_small_function,
cgraph_decide_inlining, cgraph_decide_inlining_incrementally):
Do not hold memory returned by cgraph_node_name across other call.

From-SVN: r106519

Daily bump.

From-SVN: r106516

re PR fortran/24636 (gfortran: STOP without stop-code too noisy, regression w.r.t. g77)

PR fortran/24636
* match.c (gfc_match_stopcode):  Set stop_code = -1.
* runtime/stop.c (stop_numeric): Use stop_code = -1.

From-SVN: r106509

re PR target/23424 (cris.md rtl canonicalization bug)

PR target/23424
* gcc.dg/torture/pr23424-1.c: New test.

From-SVN: r106507

re PR target/23424 (cris.md rtl canonicalization bug)

PR target/23424
* config/cris/predicates.md ("cris_bdap_sign_extend_operand"):
Disable.

From-SVN: r106506