backport: re PR rtl-optimization/46777 (ICE: in rtl_verify_flow_info, at cfgrtl.c:2164 with -O -fgcse -fno-tree-dominator-opts -funroll-loops)

Backport from mainline
2010-12-06  Jakub Jelinek  <jakub@redhat.com>

PR rtl-optimization/46777
* cfglayout.c (duplicate_insn_chain): Avoid duplicating
also barrier after tablejump.

* gcc.dg/pr46777.c: New test.

From-SVN: r167550

backport: re PR fortran/46753 (ICE: OpenMP - in extract_omp_for_data, at omp-low.c:335)

Backport from mainline
2010-12-02  Jakub Jelinek  <jakub@redhat.com>

PR fortran/46753
* trans-openmp.c (gfc_trans_omp_do): Use build2_loc instead of
fold_build2_loc for OMP_FOR conditions.

* libgomp.fortran/pr46753.f90: New test.

From-SVN: r167549

backport: re PR libgomp/45240 (parallel.c: GOMP_parallel_end locks a mutex but fails to unlock it after atomic operation complete)

Backport from mainline
2010-12-02  Jakub Jelinek  <jakub@redhat.com>

PR libgomp/45240
* parallel.c (GOMP_parallel_end): Unlock gomp_remaining_threads_lock
at the end if sync builtins aren't supported.

From-SVN: r167547

backport: re PR rtl-optimization/46614 (gcc.dg/vect/vect-strided-u8-i8-gap4.c FAILs with -fno-rename-registers -fsched2-use-superblocks)

Backport from mainline
2010-11-24  Jakub Jelinek  <jakub@redhat.com>

PR rtl-optimization/46614
* sched-deps.c (NON_FLUSH_JUMP_KIND, NON_FLUSH_JUMP_P): Define.
 (deps_analyze_insn): Mark JUMP_INSNs in
last_pending_memory_flush that weren't added through
flush_pending_lists with NON_FLUSH_JUMP_KIND.
(sched_analyze_2, sched_analyze_insn): Check NON_FLUSH_JUMP_P
on INSN_LIST instead of JUMP_P check on its operand.
* sched-rgn.c (concat_INSN_LIST): Copy over REG_NOTE_KIND.

PR rtl-optimization/46614
* gcc.dg/pr46614.c: New test.

From-SVN: r167546

backport: re PR middle-end/46629 (Failed to build 200.sixtrack in SPEC CPU 2000)

Backport from mainline
2010-11-24  Jakub Jelinek  <jakub@redhat.com>

PR middle-end/46629
* cfgexpand.c (maybe_cleanup_end_of_block): Test NEXT_INSN (insn)
instead of insn with any_condjump_p.

2010-11-23  Jakub Jelinek  <jakub@redhat.com>

PR middle-end/46499
* cfgexpand.c (maybe_cleanup_end_of_block): Remove also BARRIERs
following unconditional jumps.

* gcc.dg/pr46499-1.c: New test.
* gcc.dg/pr46499-2.c: New test.

From-SVN: r167545

backport: re PR c++/46538 (ICE: SIGSEGV in cp_make_fname_decl (decl.c:3690) on invalid code when using __PRETTY_FUNCTION__)

Backport from mainline
2010-11-20  Jakub Jelinek  <jakub@redhat.com>

PR c++/46538
* decl.c (cp_make_fname_decl): Return error_mark_node if
current_binding_level has already sk_function_parms kind.

* g++.dg/other/error34.C: New test.

From-SVN: r167543

Fix PR44676: fix invariant phi node removal.

Backport from trunk
http://gcc.gnu.org/viewcvs?view=revision&revision=163105

2010-05-07  Sebastian Pop  <sebastian.pop@amd.com>

Backport from mainline:
2010-05-07  Sebastian Pop  <sebastian.pop@amd.com>

PR tree-optimization/44676
* graphite-sese-to-poly.c (loop_entry_phi_arg): Renamed
phi_arg_in_outermost_loop.
(remove_simple_copy_phi): Call phi_arg_in_outermost_loop.
(remove_invariant_phi): Same.

From-SVN: r167542

backport: re PR target/45870 (note: non-delegitimized UNSPEC 5 found (-O1 -g))

Backport from mainline
2010-11-19  Jakub Jelinek  <jakub@redhat.com>

PR target/45870
* dwarf2out.c (const_ok_for_output_1): Don't complain about
non-delegitimized TLS UNSPECs.

From-SVN: r167541

backport: re PR middle-end/46534 (ICE optimizing printf ("...>10MBstring\n") into puts)

Backport from mainline
2010-11-18  Jakub Jelinek  <jakub@redhat.com>

PR middle-end/46534
* builtins.c (fold_builtin_printf): Don't copy and modify string
before build_string_literal, instead modify what
build_string_literal returned.

* gcc.c-torture/compile/pr46534.c: New test.

From-SVN: r167540

backport: re PR rtl-optimization/46440 (ICE: in rtl_verify_flow_info, at cfgrtl.c:2165 with -fstack-protector-all -fno-tree-dominator-opts -fno-tree-fre)

Backport from mainline
2010-11-17  Jakub Jelinek  <jakub@redhat.com>

PR rtl-optimization/46440
* combine.c (update_cfg_for_uncondjump): When changing
an indirect jump into unconditional jump, remove BARRIERs
from bb's footer.

* gcc.dg/pr46440.c: New test.

From-SVN: r167539

backport: re PR c++/46401 (very slow compile time with -Wsequence-point)

Backport from mainline
2010-11-16  Jakub Jelinek  <jakub@redhat.com>

PR c++/46401
* c-common.c (warning_candidate_p): Don't track non-const calls
or STRING_CSTs.

* g++.dg/warn/Wsequence-point-3.C: New test.

From-SVN: r167538

Daily bump.

From-SVN: r167524

gcc:

Backport from mainline:
2010-07-23  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* doc/sourcebuild.texi (Add Options): Document tls.

gcc/testsuite:
Backport from mainline:
2010-07-23  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* lib/target-supports.exp (add_options_for_tls): New proc.
* g++.dg/tls/static-1.C: Use dg-add-options tls.
* g++.dg/tls/static-1a.cc: Likewise.
* gcc.dg/tls/emutls-1.c: Likewise.
* gcc.dg/tls/opt-11.c: Likewise.
* gcc.dg/tls/opt-12.c: Likewise.
* gcc.dg/tls/pr24428-2.c: Likewise.
* gcc.dg/tls/pr24428.c: Likewise.

2010-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

PR target/46131
* lib/target-supports.exp (add_options_for_tls): Pass -pthread for
*-*-osf*.

From-SVN: r167507

backport: re PR testsuite/43925 (Plugin tests unresolved on IRIX 6.5: libintl.h: No such file or directory)

Backport from mainline:
2010-07-16  Iain Sandoe  <iains@gcc.gnu.org>

PR testsuite/43925
* lib/plugin-support.exp (plugin-test-execute):
Add ${gcc_objdir}/intl to includes.

From-SVN: r167506

configure.ac (libffi_cv_as_ascii_pseudo_op): Use double backslashes.

* configure.ac (libffi_cv_as_ascii_pseudo_op): Use double
backslashes.
(libffi_cv_as_string_pseudo_op): Likewise.
* configure: Regenerate.

From-SVN: r167503

run_acats (which): Assign output to temporary variable, only use if successful.

* ada/acats/run_acats (which): Assign output to temporary
variable, only use if successful.
Use last field of type output.

From-SVN: r167500

re PR tree-optimization/46806 (-O3 implies 0 <= 8 <= 7)

2010-12-06  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/46806
* tree-ssa-copyrename.c (copy_rename_partition_coalesce): Do not
coalesce different types.

* g++.dg/torture/20100825.C: New testcase.

From-SVN: r167492

Daily bump.

From-SVN: r167477

re PR tree-optimization/46663 (ICE: SIGSEGV in vect_recog_pow_pattern (gimple.h:2055) with -O -fexceptions -ftree-vectorize)

PR tree-optimization/46663
* tree-vect-patterns.c (vect_recog_pow_pattern): Check that
FUNCTION_DECL exists and that it's a builtin.

Co-Authored-By: Ira Rosen <irar@il.ibm.com>
From-SVN: r167469

Daily bump.

From-SVN: r167463

re PR fortran/46794 (ICE on valid code involving power of small integer kinds)

2010-12-04  Daniel Kraft  <d@domob.eu>

PR fortran/46794
* trans-expr.c (gfc_conv_power_op): Handle kind of result expression
correctly for integer kind 1 and 2 operands.

2010-12-04  Daniel Kraft  <d@domob.eu>

PR fortran/46794
* gfortran.dg/power2.f90: New test.

From-SVN: r167454

Daily bump.

From-SVN: r167449

re PR debug/46123 (ICE: in output_aranges, at dwarf2out.c:11531 with -feliminate-dwarf2-dups -g)

PR debug/46123
* dwarf2out.c (gen_type_die_with_usage): Don't put local types in
a declaration DIE.

From-SVN: r167441

re PR c++/46058 (gcc crashes with lvalue error on the following Code)

PR c++/46058
* tree.c (lvalue_kind) [SCOPE_REF]: Handle non-dependent case.

From-SVN: r167438

backport: osf.h (MD_UNWIND_SUPPORT): Define.

2010-12-03  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

Backport from mainline:
2010-09-15  Olivier Hainque  <hainque@adacore.com>
             Jose Ruiz  <ruiz@adacore.com>

* config/alpha/osf.h (MD_UNWIND_SUPPORT): Define.
* config/alpha/osf-unwind.h: New file.

From-SVN: r167424

backport: rtti3.C: Scan for .weakext on alpha*-dec-osf*.

Backport from mainline:
2010-11-10  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* g++.dg/abi/rtti3.C: Scan for .weakext on alpha*-dec-osf*.
* g++.dg/abi/thunk4.C: Likewise.
* g++.dg/other/anon5.C: Skip on alpha*-dec-osf*.
* g++.dg/warn/pr31246.C: XFAIL on alpha*-dec-osf*.
* g++.dg/warn/weak1.C: Skip on alpha*-dec-osf*.
* g++.old-deja/g++.eh/badalloc1.C [!STACK_SIZE && __osf__]: Use
large arena_size.
* gcc.dg/attr-weakref-1.c: Skip on alpha*-dec-osf*.
* gcc.dg/intmax_t-1.c: Likewise.

From-SVN: r167423

backport: re PR target/45693 (All Tru64 UNIX C++ EH tests fail)

Backport from mainline:
2010-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

PR target/45693
* configure.host (osf*): Set os_include_dir to os/generic.
Add -lpthread to OPT_LDFLAGS.

From-SVN: r167422

backport: jvmti.exp (gcj_jvmti_compile_cxx_to_o): Add -pthread on alpha*-dec-osf*.

Backport from mainline:
2010-11-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* testsuite/libjava.jvmti/jvmti.exp (gcj_jvmti_compile_cxx_to_o):
Add -pthread on alpha*-dec-osf*.
* testsuite/libjava.jvmti/jvmti-interp.exp
(gcj_jni_compile_c_to_so): Likewise.

From-SVN: r167420

gcc:

Backport from mainline:
2010-11-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* config/mips/iris6.h [!IRIX_USING_GNU_LD]
(SUPPORTS_INIT_PRIORITY): Define.

gcc/testsuite:
Backport from mainline:
2010-11-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* lib/prune.exp (prune_gcc_output): Ignore IRIX 6 linker multiline
warning.
* g++.dg/abi/packed1.C (struct INNER): Ignore warning on
mips-sgi-irix*.
* g++.dg/cpp/_Pragma1.C: Skip on mips-sgi-irix*.
* gcc.target/mips/call-3.c: Add dg-require-visibility.

From-SVN: r167419

backport: inclhack.def (irix_pthread_init): New fix.

Backport from mainline:
2010-11-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* inclhack.def (irix_pthread_init): New fix.
* fixincl.x: Regenerate.
* tests/base/pthread.h [IRIX_PTHREAD_INIT_CHECK]: New test.

From-SVN: r167418

backport: configure.ac (*-*-irix6*): Disable set_have_as_tls.

Backport from mainline:
2010-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* configure.ac (*-*-irix6*): Disable set_have_as_tls.
(*-*-osf*): Likewise.
* configure: Regenerate.

From-SVN: r167417

backport: vla8.f90: Use dg-timeout-factor 2.0.

Backport from mainline:
2010-12-01  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* testsuite/libgomp.fortran/vla8.f90: Use dg-timeout-factor 2.0.

From-SVN: r167410

Daily bump.

From-SVN: r167404

Daily bump.

From-SVN: r167362

Daily bump.

From-SVN: r167321

backport: re PR libgcj/42986 (natVMSecureRandom.cc  error: expected type-specifier)

2010-11-30  Kai Tietz  <kai.tietz@onevision.com>

        Backport from mainline
        2010-08-12  Tom Tromey  <tromey@redhat.com>

        PR libgcj/42986
        * gnu/java/security/jce/prng/natVMSecureRandomWin32.cc
        (natGenerateSeed): Add missing "::".

From-SVN: r167314

Daily bump.

From-SVN: r167281

Use SSA_NAME_DEF_STMT only on SSA_NAMEs.

gcc/

2010-11-29  H.J. Lu  <hongjiu.lu@intel.com>

Backport from mainline
PR middle-end/46651
2010-07-22  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Use
SSA_NAME_DEF_STMT only on SSA_NAMEs.

gcc/testsuite/

2010-11-29  H.J. Lu  <hongjiu.lu@intel.com>

Backport from mainline
2010-07-22  Sebastian Pop  <sebastian.pop@amd.com>

PR middle-end/46651
* gcc.dg/graphite/id-24.c: New.

From-SVN: r167274

[multiple changes]

2010-11-29  Zdenek Dvorak  <rakdver@kam.uniff.cz>

PR tree-optimization/46675
* tree-ssa-loop-niter.c (split_to_var_and_offset): Avoid overflow
in offset calculation.

testsuite/ChangeLog:

2010-11-29  Richard Guenther  <rguenther@suse.de>
    Zdenek Dvorak  <rakdver@kam.uniff.cz>

PR tree-optimization/46675
* gcc.dg/pr46675.c: New test.

From-SVN: r167267

re PR debug/46258 (gcc.dg/tree-ssa/pr23820.c: -fcompare-debug failure with -O3 -ffloat-store)

PR debug/46258
* tree-cfg.c (replace_uses_by): Don't mark BBs as altered on
debug stmts.

From-SVN: r167264

Daily bump.

From-SVN: r167228

aliasing2.adb (dg-final): Robustify pattern matching.

* gnat.dg/aliasing2.adb (dg-final): Robustify pattern matching.
* gnat.dg/pack9.adb (dg-final): Likewise.

From-SVN: r167223

Daily bump.

From-SVN: r167215

re PR fortran/46638 (Wrong result with TRANSFER from string to DT)

2010-11-27  Tobias Burnus  <burnus@net-b.de>

        PR fortran/46638
        PR fortran/46668
        * gfortran.dg/transfer_simplify_10.f90: Fix endian issue.

From-SVN: r167210

re PR ada/40777 (compile error on gcc-interface/targtyps.c)

PR ada/40777
* gcc-interface/targtyps.c (get_target_double_scalar_alignment): Guard
use of TARGET_64BIT macro.

From-SVN: r167206

Daily bump.

From-SVN: r167193

gnat.exp: Load gcc.exp.

* lib/gnat.exp: Load gcc.exp.
(gnat_target_compile): Use gcc_target_compile for *.c sources.
* gnat.dg/sse_nolib.adb: Use dg-require-effective-target sse_runtime.

Backport from mainline:
2010-11-25  Eric Botcazou  <ebotcazou@adacore.com>

* gnat.dg/loop_optimization7.adb: Remove dg-require-effective-target.
* gnat.dg/sse_nolib.adb: Likewise.

From-SVN: r167187

Daily bump.

From-SVN: r167157

re PR fortran/46638 (Wrong result with TRANSFER from string to DT)

2010-11-25  Tobias Burnus  <burnus@net-b.de>

        PR fortran/46638
        * target-memory.c (gfc_interpret_derived): Correctly handle
        component offset.

2010-11-25  Tobias Burnus  <burnus@net-b.de>

        PR fortran/46638
        * gfortran.dg/transfer_simplify_10.f90: New.

From-SVN: r167139

Daily bump.

From-SVN: r167134

Daily bump.

From-SVN: r167102

Daily bump.

From-SVN: r167067

re PR middle-end/43057 ([LTO] fold check: original tree changed by fold)

PR middle-end/43057
* fold-const.c (pedantic_non_lvalue_loc): Unshare x before
setting location.

From-SVN: r167042

Daily bump.

From-SVN: r167017

re PR rtl-optimization/46571 (bootstrap comparison failure in fortran/trans-openmp.c)

PR rtl-optimization/46571
* gcse.c (hash_scan_set): Use next_nonnote_nondebug_insn.
(compute_hash_table_work): Use NONDEBUG_INSN_P.

From-SVN: r167008

mips.c (machine_function): Rename load_label_length to load_label_num_insns.

* config/mips/mips.c (machine_function): Rename load_label_length to
load_label_num_insns.
(mips_load_label_length): Rename to...
(mips_load_label_num_insns): ...this.  Adjust to first renaming.
(mips_adjust_insn_length): Adjust to second renaming.  Fix thinko.

From-SVN: r167002

Daily bump.

From-SVN: r166990

re PR target/31100 (ASM_OUTPUT_ALIGN_WITH_NOP wrong with GNU as)

PR target/31100
* config/sparc/sparc.h (ASM_OUTPUT_ALIGN_WITH_NOP): Move to...
* config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): ...here.
* config/sparc/sol2-gas.h (ASM_OUTPUT_ALIGN_WITH_NOP): Undefine.

From-SVN: r166987

Daily bump.

From-SVN: r166971

re PR c/46547 (internal compiler error when converting a complex to a bool)

PR c/46547
* objc-act.c (in_late_binary_op): Remove.

From-SVN: r166965

re PR c/46547 (internal compiler error when converting a complex to a bool)

PR c/46547
* c-common.c (in_late_binary_op): Define.
(c_common_truthvalue_conversion): Check in_late_binary_op before
calling c_save_expr.
* c-common.h (in_late_binary_op): Declare.
* c-tree.h (in_late_binary_op): Move to c-common.h.
* c-typeck.c (in_late_binary_op): Move to c-common.c.

testsuite:
* gcc.c-torture/compile/pr46547-1.c: New test.

From-SVN: r166957

Daily bump.

From-SVN: r166925

Daily bump.

From-SVN: r166893

re PR tree-optimization/46498 (ICE: in eliminate_unnecessary_stmts, at tree-ssa-dce.c:1112 with -O -funsafe-math-optimizations -fno-tree-dominator-opts -fno-tree-reassoc)

2010-11-17  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/46498
* tree-ssa-dce.c (perform_tree_ssa_dce): Compute dominator info.

From-SVN: r166862

re PR target/45807 (Lying eh_frame r2 save info causes crashes with static libgcc_eh and libstdc++)

PR target/45807
Backport r164685,r164825,r165101
* config/rs6000/aix.h (SETUP_FRAME_ADDRESSES): Delete.
* config/rs6000/linux64.h (SETUP_FRAME_ADDRESSES): Delete.
* config/rs6000/rs6000-protos.h (rs6000_aix_emit_builtin_unwind_init):
Delete.
* config/rs6000/rs6000.c (rs6000_aix_emit_builtin_unwind_init): Delete.
(rs6000_emit_prologue): Don't just create frame save info for r2,
actually save r2.

From-SVN: r166857

re PR target/44266 (stack frame lacks parameter save area)

PR target/44266
Backport r159963
* config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Use
emit_library_call machinery to set up __tls_get_addr calls.

From-SVN: r166856

Daily bump.

From-SVN: r166843

re PR rtl-optimization/46315 (-O2 -fno-strict-overflow causes wrong code generation)

PR rtl-optimization/46315
* rtl.h (remove_reg_equal_equiv_notes_for_regno): Declare.
* rtlanal.c (remove_reg_equal_equiv_notes_for_regno): New function
extracted from...
* dce.c (delete_corresponding_reg_eq_notes): ...here.  Rename into...
(remove_reg_equal_equiv_notes_for_defs): ...this.
(delete_unmarked_insns): Adjust to above renaming.
* ifcvt.c (dead_or_predicable): Remove REG_EQUAL and REG_EQUIV notes
referring to registers set in the insns being moved, if any.

* df-core.c (df_ref_dump): New function extracted from...
(df_refs_chain_dump): ...here.  Call it.
(df_regs_chain_dump): Likewise.
* df-problems.c (df_chain_dump): Print 'e' for uses in notes.
* df-scan.c (df_scan_start_dump): Likewise.  Fix long line.

From-SVN: r166828

re PR tree-optimization/44545 (internal compiler error: in remove_unreachable_handlers, at tree-eh)

2010-11-16  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/44545
* tree-ssa-reassoc.c (linearize_expr_tree): Possibly throwing
statements are not reassociatable.
(reassociate_bb): Likewise.

* gcc.dg/pr44545.c: New testcase.

From-SVN: r166796

Daily bump.

From-SVN: r166785

re PR tree-optimization/46491 (ipa-pure-const.c miscompilation)

PR tree-optimization/46491
Backport from mainline

2010-05-14  Jan Hubicka  <jh@suse.cz>
* ipa-pure-const.c (check_stmt): Do not use memory_identifier_string.

* gcc.target/i386/pr46491.c: New test.

From-SVN: r166774

Daily bump.

From-SVN: r166741

Daily bump.

From-SVN: r166727

re PR fortran/45742 (VOLATILE has no effect)

2010-11-13  Tobias Burnus  <burnus@net-b.de>

        PR fortran/45742
        * trans-common.c (build_field): Add TREE_SIDE_EFFECTS for
        * volatile.
        * trans-decl.c (gfc_finish_var_decl): Ditto.
        (create_function_arglist): Handle volatile dummy arguments.

2010-11-13  Tobias Burnus  <burnus@net-b.de>

        PR fortran/45742
        * gfortran.dg/volatile12.f90: New.

From-SVN: r166714

Daily bump.

From-SVN: r166693

re PR tree-optimization/46177 (ICE: in prop_phis, at tree-loop-distribution.c:327 with -fno-tree-copy-prop -ftree-loop-distribution)

PR tree-optimization/46177
* gcc.dg/pr46177.c: Fix dg-options.

From-SVN: r166679

rs6000.c (rs6000_expand_ternop_builtin): Rewrite switch on insn codes as sequence of ifs.

        * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Rewrite
        switch on insn codes as sequence of ifs.

From-SVN: r166659

backport: re PR rtl-optimization/44691 (ICE: RTL check: expected code 'reg', have 'plus' in rhs_regno, at rtl.h:1050)

Backport from mainline:
2010-08-19  Andrey Belevantsev  <abel@ispras.ru>

PR rtl-optimization/44691
* sel-sched.c (count_occurrences_1): Also punt when SUBREG_REG
is not a register.

* gfortran.dg/pr44691.f: New test.

From-SVN: r166653

backport: [multiple changes]

2010-11-12  Richard Guenther  <rguenther@suse.de>

Backport from mainline
2010-11-09  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/46355
* tree-loop-distribution.c (tree_loop_distribution): Do not
distribute loops without a single exit.

* gcc.dg/pr46355.c: New testcase.

2010-11-02  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/46177
* tree-loop-distribution.c (prop_phis): Use
mark_virtual_phi_result_for_renaming.

* gcc.dg/pr46177.c: New testcase.

2010-11-09  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/46177
* tree-loop-distribution.c (prop_phis): Remove.
(generate_builtin): Re-use the old loop exit edge to avoid
needing to update PHI nodes.

* gcc.dg/pr46177-2.c: New testcase.

From-SVN: r166652

Daily bump.

From-SVN: r166640

backport: re PR c++/45894 ([C++0x] ICE: segmentation fault with -Wall)

Backport from mainline
2010-11-09  Jakub Jelinek  <jakub@redhat.com>

PR c++/45894
* tree.c (lvalue_p_1): Don't crash if ref has NULL type.

* g++.dg/warn/Wsequence-point-2.C: New test.

From-SVN: r166621

backport: re PR c/44772 (-Wc++-compat warns incorrectly for anonymous unions)

Backport from mainline
2010-11-05  Jakub Jelinek  <jakub@redhat.com>

PR c/44772
* c-decl.c (warn_cxx_compat_finish_struct): Don't call
pointer_set_contains if DECL_NAME is NULL.

* gcc.dg/Wcxx-compat-21.c: New test.

From-SVN: r166620

backport: re PR c++/46160 (ICE with volatile structure and enum)

2010-11-11  Jakub Jelinek  <jakub@redhat.com>

Backport from mainline
2010-11-05  Jakub Jelinek  <jakub@redhat.com>

PR c++/46160
* cp-gimplify.c (cp_gimplify_expr): Drop volatile INDIRECT_REFs
on the RHS to avoid infinite recursion with gimplify_expr.

* g++.dg/opt/empty2.C: New test.

From-SVN: r166619

backport: re PR tree-optimization/46099 (ICE: in replace_ssa_name, at tree-cfg.c:5643 with -ftree-parallelize-loops -g)

Backport from mainline
2010-11-05  Jakub Jelinek  <jakub@redhat.com>

PR tree-optimization/46099
* tree-parloops.c (take_address_of): Add GSI argument.  Return NULL
if it is NULL and uid wasn't found in the hash table.  Just fold the
result if it is NULL otherwise.  Insert other potentially needed
stmts right before current stmt instead of on the entry edge.
(struct elv_data): Add gsi and reset fields.
(eliminate_local_variables_1): Adjust caller.  If take_address_of
failed for debug stmt, set dta->reset and return.
(eliminate_local_variables_stmt): Change STMT argument for GSI,
pass GSI through to the callback, handle resetting of debug stmts.
(eliminate_local_variables): Adjust caller.  Process debug stmts
in second phase.

* gcc.dg/autopar/pr46099.c: New test.

From-SVN: r166618

backport: re PR middle-end/43690 (Internal compiler error detected by avr-gcc.)

Backport from mainline
2010-11-05  Jakub Jelinek  <jakub@redhat.com>

PR middle-end/43690
* gimplify.c (gimplify_asm_expr): If a "m" input is a
{pre,post}{in,de}crement, fail.

* c-c++-common/pr43690.c: New test.

From-SVN: r166617

backport: re PR tree-optimization/46165 (ICE: verify_flow_info failed when casting-out attribute noreturn with -fno-tree-ccp -fno-tree-copy-prop -fno-tree-dce)

Backport from mainline
2010-11-03  Jakub Jelinek  <jakub@redhat.com>

PR tree-optimization/46165
* tree-ssa-pre.c (eliminate): Return TODO_cleanup_cfg if changing
a normal call into noreturn call.

* gcc.dg/pr46165.c: New test.

From-SVN: r166616

backport: re PR tree-optimization/46107 (-ftree-loop-distribute-patterns caused verify_loop_structure problem)

Backport from mainline
2010-11-03  Jakub Jelinek  <jakub@redhat.com>

PR tree-optimization/46107
* cfgloopmanip.c (loop_version): Set irred_flag back into entry->flags
if cfg_hook_duplicate_loop_to_header_edge failed.

* gcc.c-torture/compile/pr46107.c: New test.

From-SVN: r166615

re PR libfortran/46373 (fflush called when reading from a string)

2010-11-10  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libgfortran/46373
Backport from mainline:
* io/transfer.c (data_transfer_init): Do not call flush_if_preconnected
if this is an internal unit.

From-SVN: r166588

Daily bump.

From-SVN: r166581

re PR target/46419 (_mm_cvtpu16_ps (and hence _mm_cvtpu8_ps) returns false result)

PR middle-end/46419
* config/i386/xmmintrin.h (_mm_cvtpi16_ps): Swap __hisi and __losi.
(_mm_cvtpu16_ps): Ditto.

testsuite/ChangeLog:

PR target/46419
* gcc-target/i386/pr46419.c: New test.

From-SVN: r166571

Daily bump.

From-SVN: r166513

Fix Changelog entry.

2010-11-09  Sebastian Pop  <sebastian.pop@amd.com>

   PR tree-optimization/45314
   Backport of revision 163132.
   * graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Always insert out
   of SSA copies on edges except for loop->latch.

From-SVN: r166507

Fix PR45314: backport revision 163123.

2010-11-09  Sebastian Pop  <sebastian.pop@amd.com>

PR tree-optimization/45902
Backport of revision 163132.
* graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Always insert out
of SSA copies on edges except for loop->latch.

From-SVN: r166506

re PR rtl-optimization/46237 (SPECint2006 403.gcc fails with -mzarch -m31)

2010-11-09  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

PR rtl-optimization/46237
* ira-conflicts.c (ira_build_conflicts): Record conflicts for call
saved hard regs if they might get partially clobbered.

From-SVN: r166496

Daily bump.

From-SVN: r166464

backport: i386.exp (FLAGS): Remove.

Backport from mainline:
2010-11-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

* gcc.target/i386/i386.exp (FLAGS): Remove.
Wrap gcc_target_compile with $clearcap_ldflags added to options.

From-SVN: r166436

Daily bump.

From-SVN: r166425

Daily bump.

From-SVN: r166411

re PR debug/45939 (Wrong debug info: Negative location range generated)

2010-11-06  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

PR debug/45939
* var-tracking.c (emit_note_insn_var_location): Make sure that
call related var location notes come before the normal ones.

From-SVN: r166395

Daily bump.

From-SVN: r166389