rs6000.c (rs6000_expand_vector_set): Use vnand instead of vnor to exploit possible fusion opportunity in the future.

2014-04-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* config/rs6000/rs6000.c (rs6000_expand_vector_set): Use vnand
instead of vnor to exploit possible fusion opportunity in the
future.
(altivec_expand_vec_perm_const_le): Likewise.

From-SVN: r209235

Daily bump.

From-SVN: r209233

re PR c++/59115 (ICE with invalid template parameter)

/cp
2014-04-08  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/59115
* pt.c (process_template_parm): For an invalid non-type parameter
only set TREE_TYPE to error_mark_node.
(push_inline_template_parms_recursive, comp_template_parms,
redeclare_class_template, coerce_template_template_parm,
coerce_template_template_parms, unify): Use error_operand_p.

/testsuite
2014-04-08  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/59115
* g++.dg/template/crash119.C: New.

From-SVN: r209230

re PR ada/60411 (Ada bootstrap failure on ARM)

PR ada/60411
* sigtramp.h: Add Android support.

From-SVN: r209227

sync.md (AINT mode_iterator): Move definition.

* config/rs6000/sync.md (AINT mode_iterator): Move definition.
(loadsync_<mode>): Change mode.
(load_quadpti, store_quadpti): New.
(atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
* config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.

* gcc.target/powerpc/atomic_load_store-p8.c: New.

From-SVN: r209225

gcc-dg.exp (dg-build-dso): Reset dg-do-what-default to compile.

* lib/gcc-dg.exp (dg-build-dso): Reset dg-do-what-default to
compile.

From-SVN: r209224

re PR rtl-optimization/60763 (ICE in extract_insn starting with rev 208984)

gcc/
PR target/60763
* config/rs6000/vsx.md (vsx_xscvdpspn_scalar): Change input to DImode.
* config/rs6000/rs6000.md (reload_vsx_from_gprsf): Update accordingly.
Use gen_rtx_REG rather than simplify_gen_subreg for op0_di.

From-SVN: r209223

re PR sanitizer/60745 (Many ICEs running libstdc++ testsuite with ubsan, maybe due to PCH)

PR sanitizer/60745
* c-ubsan.c: Include asan.h.
(ubsan_instrument_return): Call initialize_sanitizer_builtins.

From-SVN: r209222

re PR rtl-optimization/60776 (FAIL gcc.dg/builtin-bswap-6.c and gcc.dg/builtin-bswap-7.c)

2014-04-08  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

PR rtl-optimization/60776
* gcc.dg/builtin-bswap-6.c: Use -mbranch-cost=0 for s390.
* gcc.dg/builtin-bswap-7.c: Likewise.
* gcc.dg/builtin-bswap-6a.c: New testcase.
* gcc.dg/builtin-bswap-7a.c: New testcase.

Revert
2014-04-04  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* gcc.dg/builtin-bswap-6.c: Adjust return value to disable GCC
optimization.
* gcc.dg/builtin-bswap-7.c: Likewise.

From-SVN: r209221

re PR libfortran/60706 (FAIL: gfortran.dg/size_kind_2.f90  -O   scan-tree-dump original "var2 = 42949673 00;")

2014-04-08  Richard Biener  <rguenther@suse.de>

PR middle-end/60706
* tree-pretty-print.c (pp_double_int): For HWI32 hosts with
a 64bit widest int print double-int similar to on HWI64 hosts.

From-SVN: r209216

re PR tree-optimization/60785 (ICE in gsi_for_stmt w/ -O2 -ftree-loop-linear)

2014-04-08  Richard Biener  <rguenther@suse.de>

PR tree-optimization/60785
* graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Treat
default defs properly.

* gcc.dg/graphite/pr60785.c: New testcase.

From-SVN: r209214

XFAIL gcc.c-torture/compile/pr28865.c on Solaris 9/SPARC (PR target/60602)

PR target/60602
* gcc.c-torture/compile/pr28865.c: XFAIL on sparc*-*-solaris2.9*
with as at -O0.

From-SVN: r209213

invoke (Wnon-virtual-dtor): Update to match implementation.

* doc/invoke (Wnon-virtual-dtor): Update to match implementation.
(Weffc++): Likewise.

cp/
* class.c (check_bases_and_members): Warn about non-virtual dtors
in public bases only.  Check warn_ecpp before complaining about
non-polymorphic bases.

testsuite/
* g++.dg/warn/Wnvdtor-2.C: Add more cases.
* g++.dg/warn/Wnvdtor-3.C: Likewise.
* g++.dg/warn/Wnvdtor-4.C: Likewise.

From-SVN: r209212

re PR ada/60411 (Ada bootstrap failure on ARM)

PR ada/60411
* sigtramp-armdroid.c: New file.

From-SVN: r209211

re PR ada/60411 (Ada bootstrap failure on ARM)

PR ada/60411
* gcc-interface/Makefile.in (arm% linux-android): Switch to EHABI.
* s-linux-android.ads: New file.
* s-intman-android.adb: Likewise.

From-SVN: r209210

ipa-devirt.c (maybe_record_node): When node is not recorded, set completep to false rather than true.

* ipa-devirt.c (maybe_record_node): When node is not recorded,
set completep to false rather than true.

From-SVN: r209209

Daily bump.

From-SVN: r209208

re PR target/60504 (many Ada testsuite regressions on ARM/Linux)

PR target/60504
* config/arm/arm.h (ASM_PREFERRED_EH_DATA_FORMAT): Expose from
ARM_TARGET2_DWARF_FORMAT.

From-SVN: r209201

Fix testsuite bugs.

* testsuite/20_util/exchange/1.cc: Add missing return statements.
* testsuite/20_util/pair/40925.cc: Avoid most vexing parse.
* testsuite/22_locale/codecvt_byname/50714.cc: Add missing exception
specifications.

From-SVN: r209196

* gcc-interface/Makefile.in (alpha-vms): Add missing line.

From-SVN: r209195

re PR target/60609 (Error: value of 256 too large for field of 1 bytes at 68242)

2014-04-07  Charles Baylis  <charles.baylis@linaro.org>

        PR target/60609
        * config/arm/arm.h (ASM_OUTPUT_CASE_END): Remove.
        (LABEL_ALIGN_AFTER_BARRIER): Align barriers which occur after
        ADDR_DIFF_VEC.

2014-04-07  Charles Baylis  <charles.baylis@linaro.org>

        PR target/60609
        * g++.dg/torture/pr60609.C: New test.

From-SVN: r209191

re PR tree-optimization/60766 (Wrong optimization with -O2)

2014-04-07  Richard Biener  <rguenther@suse.de>

PR tree-optimization/60766
* tree-ssa-loop-ivopts.c (cand_value_at): Compute in an
unsigned type.
(may_eliminate_iv): Convert cand_value_at result to desired
type.

* gcc.dg/torture/pr60766.c: New testcase.

From-SVN: r209190

re PR c++/60731 (dynamic library not getting reinitialized on multiple calls to dlopen())

PR c++/60731
* lib/gcc-dg.exp (dg-build-dso): New.
(gcc-dg-test-1): Handle dg-do-what "dso".
* lib/target-supports.exp (add_options_for_dlopen): New.
(check_effective_target_dlopen): Use it.
* g++.dg/dso/dlclose1.C: New.
* g++.dg/dso/dlclose1-dso.cc: New.

From-SVN: r209187

re PR c++/60731 (dynamic library not getting reinitialized on multiple calls to dlopen())

PR c++/60731
* common.opt (-fno-gnu-unique): Add.
* config/elfos.h (USE_GNU_UNIQUE_OBJECT): Check it.

From-SVN: r209186

Fix testcase for PR target/60657

From-SVN: r209185

haifa-sched.c: Fix outdated function reference and minor grammar errors in introductory comment.

* haifa-sched.c: Fix outdated function reference and minor
grammar errors in introductory comment.

From-SVN: r209184

re PR c++/60750 (double free after std::move on string inside throw when compiled with optimization)

2014-04-07  Richard Biener  <rguenther@suse.de>

PR middle-end/60750
* tree-ssa-operands.c (maybe_add_call_vops): Also add VDEFs
for noreturn calls.
* tree-cfgcleanup.c (fixup_noreturn_call): Do not remove VDEFs.

* g++.dg/torture/pr60750.C: New testcase.
* gcc.dg/tree-ssa/20040517-1.c: Adjust.

From-SVN: r209179

Daily bump.

From-SVN: r209173

invoke.texi: Add info for __AVR_ISA_RMW__ builtin macro

* doc/invoke.texi: Add info for __AVR_ISA_RMW__ builtin macro

From-SVN: r209169

* gcc.c-torture/compile/pr60655-1.c: Use __SIZE_TYPE__ for size_t.

From-SVN: r209167

re PR testsuite/60671 (FAIL: g++.dg/pr49718.C -std=gnu++98  scan-assembler-times __cyg_profile_func_enter 1)

PR testsuite/60671
g++.dg/pr49718.C: Adjust scan-assembler-times for hppa*-*-hpux*.

From-SVN: r209166

re PR testsuite/60672 (FAIL: g++.dg/cpp1y/auto-fn25.C -std=gnu++1y  (test for errors, line 7))

PR testsuite/60672
* g++.dg/cpp1y/auto-fn25.C: Require lto.

From-SVN: r209165

stdatomic-flag.c: xfail on hppa*-*-hpux*.

* gcc.dg/atomic/stdatomic-flag.c: xfail on hppa*-*-hpux*.

From-SVN: r209164

re PR debug/55794 (FAIL: g++.dg/debug/dwarf2/non-virtual-thunk.C -std=gnu++98 and -std=gnu++11)

PR debug/55794
* config/pa/pa.c (pa_output_function_epilogue): Skip address and code
size accounting for thunks.
(pa_asm_output_mi_thunk): Use final_start_function() and
final_end_function() to output function start and end directives.

From-SVN: r209163

Daily bump.

From-SVN: r209160

warn_conversion_4.f90: Adjust test.

2014-04-05  Dominique d'Humieres <dominiq@lps.ens.fr>

* gfortran.dg/warn_conversion_4.f90: Adjust test.

From-SVN: r209151

avr-arch.h (avr_mcu_t): Add dev_attribute field to have device specific ISA/ feature information.

* config/avr/avr-arch.h (avr_mcu_t): Add dev_attribute field to have device
specific ISA/ feature information. Remove short_sp and errata_skip ds.
Add avr_device_specific_features enum to have device specific info.
* config/avr/avr-c.c (avr_cpu_cpp_builtins): use dev_attribute to check
errata_skip. Add __AVR_ISA_RMW__ builtin macro if RMW ISA available.
* config/avr/avr-devices.c (avr_mcu_types): Update AVR_MCU macro for
updated device specific info.
* config/avr/avr-mcus.def: Merge device specific details to
dev_attribute field.
* config/avr/avr.c (avr_2word_insn_p): use dev_attribute field to check
errata_skip.
* config/avr/avr.h (AVR_HAVE_8BIT_SP): same for short sp info.
* config/avr/driver-avr.c (avr_device_to_as): Pass -mrmw option to
assembler if RMW isa supported by current device.
* config/avr/genmultilib.awk: Update as device info structure changed.
* doc/invoke.texi: Add info for __AVR_ISA_RMW__ builtin macro

* gcc.target/avr/dev-specific-rmw.c: New test.

From-SVN: r209141

re PR tree-optimization/60656 (x86 vectorization produces wrong code)

2014-04-04  Cong Hou  <congh@google.com>

    PR tree-optimization/60656
    * tree-vect-stmts.c (supportable_widening_operation):
      Fix a bug that elements in a vector with vect_used_by_reduction
      property are incorrectly reordered when the operation on it is not
      consistant with the one in reduction operation.

2014-04-04  Cong Hou  <congh@google.com>

    PR tree-optimization/60656
    * gcc.dg/vect/pr60656.c: New test.

From-SVN: r209138

Daily bump.

From-SVN: r209137

re PR rtl-optimization/60155 (ICE: in get_pressure_class_and_nregs at gcse.c:3438)

PR rtl-optimization/60155
* gcse.c (record_set_data): New function.
(single_set_gcse): New function.
(gcse_emit_move_after): Use single_set_gcse instead of single_set.
(hoist_code): Likewise.
(get_pressure_class_and_nregs): Likewise.

From-SVN: r209134

check.c (gfc_check_cmplx): Fix typo.

2014-04-04  Tobias Burnus  <burnus@net-b.de>

        * check.c (gfc_check_cmplx): Fix typo.

From-SVN: r209133

builtin-bswap-6.c: Adjust return value to disable GCC optimization.

2014-04-04  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* gcc.dg/builtin-bswap-6.c: Adjust return value to disable GCC
optimization.
* gcc.dg/builtin-bswap-7.c: Likewise.

From-SVN: r209132

decl.c (duplicate_decls): Check for the return of warning_at before emitting a note.

2014-04-04  Fabien Chêne  <fabien@gcc.gnu.org>

* decl.c (duplicate_decls): Check for the return of warning_at
before emitting a note.
(warn_misplaced_attr_for_class_type): Likewise.
(check_tag_decl): Likewise.

From-SVN: r209131

* explow.c (probe_stack_range): Emit a final optimization blockage.

From-SVN: r209130

re PR bootstrap/60620 (missing gnattools dependency causes highly parallel build failure with --disable-bootstrap)

PR bootstrap/60620
* Makefile.def (dependencies): Make gnattools depend on libstdc++-v3.
* Makefile.in: Regenerate.

From-SVN: r209129

re PR c++/58207 (ICE in sort_constexpr_mem_initializers due to out of bounds vector access)

/cp
2014-04-04  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/58207
* semantics.c (sort_constexpr_mem_initializers): Robustify loop.

/testsuite
2014-04-04  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/58207
* g++.dg/cpp0x/constexpr-ice15.C: New.

From-SVN: r209128

Fix typo in recent patch.

From-SVN: r209127

re PR c++/44613 (Declaring an array with non-constant length inside a switch corrupts stack pointer.)

PR c++/44613
* semantics.c (add_stmt): Set STATEMENT_LIST_HAS_LABEL.
* decl.c (cp_finish_decl): Create a new BIND_EXPR before
instantiating a variable-sized type.

From-SVN: r209125

re PR c++/21113 (Jumps into VLA or VM scope not rejected for C++)

PR c++/21113
* decl.c (decl_jump_unsafe): Consider variably-modified decls.

From-SVN: r209124

re PR lto/59626 (/usr/include/bits/unistd.h:173:1: error: inlining failed in call to always_inline 'readlinkat': recursive inlining)

PR ipa/59626
* lto-cgraph.c (input_overwrite_node): Check that partitioning
flags are set only during streaming.
* ipa.c (process_references, walk_polymorphic_call_targets,
symtab_remove_unreachable_nodes): Drop bodies of always inline
after early inlining.
(symtab_remove_unreachable_nodes): Remove always_inline attribute.
* gcc.dg/lto/pr59626_0.c: New testcase.
* gcc.dg/lto/pr59626_1.c: New testcase.

From-SVN: r209123

Fix PR debug/60655 - part 1

This is a partial fix for PR60655 where dwarf2out.c rejects NOT of a
value in const_ok_for_output_1. There is still a problem with the
testcase on armhf where we get operations of the form, const (minus
(const_int) (symref)) without the -fdata-sections option which is just
weird. I'm not yet sure where this is produced from and will not have
the time to dig further today.

As Jakub said on IRC, const_ok_for_output_1 is called only with
partial rtx's and therefore disabling minus (const_int) (symref) might
not be the best thing to do especially if this were part of plus
(symref) (minus (const int) (symref)) and both symrefs were in the
same section.

Bootstrapped and regtested on armhf

Bootstrap and regression test running on x86_64.

Ok to commit ?

regards
Ramana

gcc/

<DATE>   Jakub Jelinek <jakub@redhat.com>
         Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>

* dwarf2out.c (const_ok_for_output_1): Reject expressions
containing a NOT.

gcc/testsuite

<DATE>  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>

* gcc.c-torture/compile/pr60655-1.c: New test.

From-SVN: r209121

[ARM/AArch64] Reduce divider reservation duration in Cortex-A53 pipeline decription

PR bootstrap/60743
* config/arm/cortex-a53.md (cortex_a53_fdivs): Reduce reservation
duration.
(cortex_a53_fdivd): Likewise.

From-SVN: r209119

class.c (find_abi_tags_r): Check for the return of warning before emitting a note.

2014-04-04  Fabien Chêne  <fabien@gcc.gnu.org>

* class.c (find_abi_tags_r): Check for the return of warning
before emitting a note.
(one_inherited_ctor): Likewise.

From-SVN: r209118

re PR ipa/60640 (ICE edge points to wrong declaration / verify_cgraph_node failed)

2014-04-04  Martin Jambor  <mjambor@suse.cz>

PR ipa/60640
        * cgraph.h (cgraph_clone_node): New parameter added to declaration.
        Adjust all callers.
* cgraph.c (clone_of_p): Also return true if thunks match.
(verify_edge_corresponds_to_fndecl): Removed extraneous call to
cgraph_function_or_thunk_node and an obsolete comment.
        * cgraphclones.c (build_function_type_skip_args): Moved upwards in the
        file.
        (build_function_decl_skip_args): Likewise.
(set_new_clone_decl_and_node_flags): New function.
        (duplicate_thunk_for_node): Likewise.
        (redirect_edge_duplicating_thunks): Likewise.
        (cgraph_clone_node): New parameter args_to_skip, pass it to
        redirect_edge_duplicating_thunks which is called instead of
        cgraph_redirect_edge_callee.
        (cgraph_create_virtual_clone): Pass args_to_skip to cgraph_clone_node,
moved setting of a lot of flags to set_new_clone_decl_and_node_flags.

testsuite/
        * g++.dg/ipa/pr60640-1.C: New test.
        * g++.dg/ipa/pr60640-2.C: Likewise.
        * g++.dg/ipa/pr60640-3.C: Likewise.
        * g++.dg/ipa/pr60640-4.C: Likewise.

From-SVN: r209097

re PR fortran/60191 (test case gfortran.dg/dynamic_dispatch_1/3.f03 fail on ARMv7)

2014-04-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR fortran/60191
        * fortran/trans-types.c (gfc_get_function_type): In case of recursion
        build a variadic function type with empty argument list instead of a
        stdarg-like function type with incomplete argument list.

From-SVN: r209091

re PR target/60657 (ICE: error: insn does not satisfy its constraints)

        PR target/60657
* config/arm/predicates.md (const_int_I_operand): New predicate.
(const_int_M_operand): Similarly.
* config/arm/arm.md (insv_zero): Use const_int_M_operand instead of
const_int_operand.
(insv_t2, extv_reg, extzv_t2): Likewise.
(load_multiple_with_writeback): Similarly for const_int_I_operand.
(pop_multiple_with_writeback_and_return): Likewise.
(vfp_pop_multiple_with_writeback): Likewise

PR target/60657
* gcc.target/arm/pr60657.c: New test.

From-SVN: r209085

re PR middle-end/60746 (ICE segfault in gimple-expr.c:314)

2014-04-04  Richard Biener  <rguenther@suse.de>

PR ipa/60746
* tree-ssanames.c (make_ssa_name_fn): Fix assert.
* gimple.c (gimple_set_bb): Avoid ICEing for NULL cfun for
non-GIMPLE_LABELs.
* gimplify.h (gimple_add_tmp_var_fn): Declare.
* gimplify.c (gimple_add_tmp_var_fn): New function.
* gimple-expr.h (create_tmp_reg_fn): Declare.
* gimple-expr.c (create_tmp_reg_fn): New function.
* gimple-low.c (record_vars_into): Don't change cfun.
* cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Fix
code generation without cfun.

* g++.dg/torture/pr60746.C: New testcase.

From-SVN: r209079

Fix shell scripting.

PR bootstrap/60719
* Makefile.in (install-driver): Fix shell scripting.

From-SVN: r209072

decl.c (duplicate_decls): Check for the return of permerror before emitting a note.

2014-04-04  Fabien Chêne  <fabien@gcc.gnu.org>

* decl.c (duplicate_decls): Check for the return of permerror
before emitting a note.

2014-04-04  Fabien Chêne  <fabien@gcc.gnu.org>

* g++.old-deja/g++.robertl/eb121.C: Adjust.
* g++.old-deja/g++.jason/overload21.C: Likewise.
* g++.old-deja/g++.law/init5.C: Likewise.

From-SVN: r209071

Daily bump.

From-SVN: r209068

re PR tree-optimization/60505 (Warning caused by GCC vectorizer.)

2014-04-03  Cong Hou  <congh@google.com>

PR tree-optimization/60505
    * tree-vectorizer.h (struct _stmt_vec_info): Add th field as the
    threshold of number of iterations below which no vectorization will be
    done.
    * tree-vect-loop.c (new_loop_vec_info):
    Initialize LOOP_VINFO_COST_MODEL_THRESHOLD.
    * tree-vect-loop.c (vect_analyze_loop_operations):
    Set LOOP_VINFO_COST_MODEL_THRESHOLD.
    * tree-vect-loop.c (vect_transform_loop):
    Use LOOP_VINFO_COST_MODEL_THRESHOLD.
    * tree-vect-loop.c (vect_analyze_loop_2): Check the maximum number
    of iterations of the loop and see if we should build the epilogue.

2014-04-03  Cong Hou  <congh@google.com>

    PR tree-optimization/60505
    * gcc.dg/vect/pr60505.c: New test.

From-SVN: r209065

tree-streamer.h (struct streamer_tree_cache_d): Add next_idx member.

2014-04-03  Richard Biener  <rguenther@suse.de>

* tree-streamer.h (struct streamer_tree_cache_d): Add next_idx
member.
(streamer_tree_cache_create): Adjust.
* tree-streamer.c (streamer_tree_cache_add_to_node_array): Adjust
to allow optional nodes array.
(streamer_tree_cache_insert_1): Use next_idx to assign idx.
(streamer_tree_cache_append): Likewise.
(streamer_tree_cache_create): Create nodes array optionally
as specified by parameter.
* lto-streamer-out.c (create_output_block): Avoid maintaining
the node array in the writer cache.
(DFS_write_tree): Remove assertion.
(produce_asm_for_decls): Free the out decl state hash table
early.
* lto-streamer-in.c (lto_data_in_create): Adjust for
streamer_tree_cache_create prototype change.

From-SVN: r209059

tree-streamer-out.c (streamer_write_chain): Do not temporarily set TREE_CHAIN to NULL_TREE.

2014-04-03  Richard Biener  <rguenther@suse.de>

* tree-streamer-out.c (streamer_write_chain): Do not temporarily
set TREE_CHAIN to NULL_TREE.

From-SVN: r209058

re PR tree-optimization/60740 (ICE in extract_affine w/ -O2 -ftree-loop-linear)

2014-04-03  Richard Biener  <rguenther@suse.de>

PR tree-optimization/60740
* graphite-scop-detection.c (stmt_simple_for_scop_p): Iterate
over all GIMPLE_COND operands.

* gcc.dg/graphite/pr60740.c: New testcase.

From-SVN: r209057

invoke.texi (Wnon-virtual-dtor): Adjust documentation.

* doc/invoke.texi (Wnon-virtual-dtor): Adjust documentation.
(Weffc++): Remove Scott's numbering, merge lists and reference
Wnon-virtual-dtor.

c-family/
* c.opt (Wnon-virtual-dtor): Auto set when Weffc++.

cp/
* class.c (accessible_nvdtor_p): New.
(check_bases): Don't check base destructor here ...
(check_bases_and_members): ... check them here.  Trigger on
Wnon-virtual-dtor flag.
(finish_struct_1): Use accessible_nvdtor_p.

testsuite/
* g++.dg/warn/Wnvdtor.C: Add non-polymorphic case.
* g++.dg/warn/Wnvdtor-2.C: New.
* g++.dg/warn/Wnvdtor-3.C: New.
* g++.dg/warn/Wnvdtor-4.C: New.
* g++.dg/warn/Weff1.C: Delete.
* g++.old-deja/g++.benjamin/15309-1.C: Delete.
* g++.old-deja/g++.benjamin/15309-2.C: Delete.

From-SVN: r209056

rl78-expand.md (movqi): Handle (SUBREG (SYMBOL_REF)) properly.

* config/rl78/rl78-expand.md (movqi): Handle (SUBREG (SYMBOL_REF))
properly.

From-SVN: r209055

ipa-cp.c (ipcp_verify_propagated_values): Also dump symtab and mention gcc_unreachable before failing.

2014-04-03  Martin Jambor  <mjambor@suse.cz>

* ipa-cp.c (ipcp_verify_propagated_values): Also dump symtab and
mention gcc_unreachable before failing.
* ipa.c (symtab_remove_unreachable_nodes): Also print order of
removed symbols.

From-SVN: r209052

re PR ipa/60659 (ICE in get_polymorphic_call_info, at ipa-devirt.c:1292)

PR ipa/60659
* ipa-devirt.c (get_polymorphic_call_info): Do not ICE on type inconsistent
code and instead mark the context inconsistent.
(possible_polymorphic_call_targets): For inconsistent contexts
return empty complete list.
* testsuite/g++.dg/torture/pr60659.C: New testcase.

From-SVN: r209048

This patch does three related things for the moxie port...

1. Changes char to be unsigned by default
2. Changes WCHAR_TYPE from long int to unsigned int
3. Zero- and sign-extends values properly, sometimes using the new
    sign-extension instructions.

From-SVN: r209047

Daily bump.

From-SVN: r209045

re PR tree-optimization/60733 (ICE due to SLSR on valid code at -O3 on x86_64-linux-gnu)

[gcc]

2014-04-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

PR tree-optimization/60733
* gimple-ssa-strength-reduction.c (ncd_with_phi): Change required
insertion point for PHI candidates to be the end of the feeding
block for the PHI argument.

[gcc/testsuite]

2014-04-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

PR tree-optimization/60733
* gcc.dg/torture/pr60733.c:  New test.

From-SVN: r209040

re PR rtl-optimization/60650 ([ARM] LRA ICE in assign_by_spills)

2014-04-02  Vladimir Makarov  <vmakarov@redhat.com>

PR rtl-optimization/60650
* lra-constraints.c (process_alt_operands): Decrease reject for
earlyclobber matching.

2014-04-02  Vladimir Makarov  <vmakarov@redhat.com>

PR rtl-optimization/60650
* gcc.target/arm/pr60650-2.c: New.

From-SVN: r209038

libgomp: Fix default futex vs errno

* config/linux/futex.h (futex_wait): Get error value from errno.
(futex_wake): Likewise.

From-SVN: r209035

avx2-vpand-3.c, [...]: Use -mno-prefer-avx128.

* gcc.target/i386/avx2-vpand-3.c,
gcc.target/i386/avx256-unaligned-load-2.c: Use -mno-prefer-avx128.

From-SVN: r209034

funcspec-2.c, [...]: Skip for -march= options different from those in dg-options.

* gcc.target/i386/funcspec-2.c, gcc.target/i386/funcspec-3.c,
gcc.target/i386/funcspec-9.c, gcc.target/i386/isa-1.c,
gcc.target/i386/memcpy-strategy-1.c,
gcc.target/i386/memcpy-strategy-2.c,
gcc.target/i386/memcpy-vector_loop-1.c,
gcc.target/i386/memcpy-vector_loop-2.c,
gcc.target/i386/memset-vector_loop-1.c,
gcc.target/i386/memset-vector_loop-2.c,
gcc.target/i386/sse2-init-v2di-2.c, gcc.target/i386/ssetype-1.c,
gcc.target/i386/ssetype-2.c, gcc.target/i386/ssetype-5.c: Skip for
-march= options different from those in dg-options.

From-SVN: r209033

s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.

2014-04-02  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* config/s390/s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.

From-SVN: r209032

spu.c (pad_bb): Do not crash when the last insn is CODE_FOR_blockage.

* config/spu/spu.c (pad_bb): Do not crash when the last
insn is CODE_FOR_blockage.

From-SVN: r209029

spu.md ("insv"): Fail if bitoffset+bitsize lies outside the target mode.

* config/spu/spu.md ("insv"): Fail if bitoffset+bitsize
lies outside the target mode.

From-SVN: r209028

libitm: Remove unused PAGE_SIZE macros

        * config/alpha/target.h (PAGE_SIZE, FIXED_PAGE_SIZE): Remove.
        * config/arm/target.h, config/sh/target.h: Likewise.
        * config/sparc/target.h, config/x86/target.h: Likewise.

From-SVN: r209027

re PR target/60735 (GCC targeting E500 with SPE has errors with the _Decimal64 type)

2014-04-02  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/60735
* config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
software floating point or no floating point registers, do not
allow any type in the FPRs.  Eliminate a test for SPE SIMD types
in GPRs that occurs after we tested for GPRs that would never be
true.

* config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
since the FMOVE64 type is DFmode/DDmode.  If TARGET_E500_DOUBLE,
specifically allow DDmode, since that does not use the SPE SIMD
instructions.

From-SVN: r209025

c-common.h (c_expand_expr): Remove declaration.

* c-common.h (c_expand_expr): Remove declaration.

From-SVN: r209020

387-3.c, [...]: Use -mtune=generic.

* gcc.target/i386/387-3.c, gcc.target/i386/387-4.c,
gcc.target/i386/pr30970.c: Use -mtune=generic.
* gcc.target/i386/avx2-vpaddb-3.c,
gcc.target/i386/avx2-vpaddd-3.c, gcc.target/i386/avx2-vpaddq-3.c,
gcc.target/i386/avx2-vpaddw-3.c, gcc.target/i386/avx2-vpmulld-3.c,
gcc.target/i386/avx2-vpmullw-3.c, gcc.target/i386/avx2-vpsrad-3.c,
gcc.target/i386/avx2-vpsraw-3.c, gcc.target/i386/avx2-vpsrld-3.c,
gcc.target/i386/avx2-vpsrlw-3.c, gcc.target/i386/avx2-vpsubb-3.c,
gcc.target/i386/avx2-vpsubd-3.c, gcc.target/i386/avx2-vpsubq-3.c,
gcc.target/i386/avx2-vpsubw-3.c,
gcc.target/i386/avx256-unaligned-load-1.c,
gcc.target/i386/avx256-unaligned-load-4.c,
gcc.target/i386/avx256-unaligned-store-1.c,
gcc.target/i386/avx256-unaligned-store-2.c,
gcc.target/i386/avx256-unaligned-store-4.c: Use
-mno-prefer-avx128.

From-SVN: r209019

re PR target/54083 (FAIL: gcc.dg/torture/pr53922.c on *-apple-darwin*)

2014-03-26  Dominique d'Humieres  <dominiq@lps.ens.fr>
    Iain Sandoe <iain@codesourcery.com>

PR target/54083
* gcc.dg/attr-weakref-1.c: Allow the test on darwin
with the additional options -Wl,-undefined,dynamic_lookup
and -Wl,-flat_namespace.
* gcc.dg/torture/pr53922.c: Additional option
-Wl,-flat_namespace for darwin[89].
* gcc.dg/torture/pr60092.c: Additional options
-Wl,-undefined,dynamic_lookup and -Wl,-flat_namespace
for darwin[89].

Co-Authored-By: Iain Sandoe <iain@codesourcery.com>
From-SVN: r209018

re PR target/54407 (FAIL: 30_threads/condition_variable/54185.cc execution test  program timed out on powerpc-apple-darwin9 and x86_64-apple-darwin10)

2014-04-02  Dominique d'Humieres  <dominiq@lps.ens.fr>
    Jack Howarth <howarth@bromo.med.uc.edu>

PR target/54407
* testsuite/30_threads/condition_variable/54185.cc: Skip for
darwin < 11.

Co-Authored-By: Jack Howarth <howarth@bromo.med.uc.edu>
From-SVN: r209017

re PR target/43751 (dsymutil is not called for fortran and, under some circumstances not for other FEs.)

2014-04-02  Dominique d'Humieres  <dominiq@lps.ens.fr>

PR target/43751
* lib/prune.exp: Modify the regular express to prune
the new warnings introduced by r205679 on darwin9.

From-SVN: r209016

re PR middle-end/60729 (Compiler failure for combination of -ftrapv and -O3: compiler error: in prepare_cmp_insn)

2014-04-02  Richard Biener  <rguenther@suse.de>

PR middle-end/60729
* optabs.c (expand_abs_nojump): Honor flag_trapv only for
MODE_INTs.  Properly use negv_optab.
(expand_abs): Likewise.

* g++.dg/vect/pr60729.cc: New testcase.

From-SVN: r209012

lto-plugin.c (onload): Fail to load if -fno-use-linker-plugin is set in COLLECT_GCC_OPTIONS.

2014-04-02  Richard Biener  <rguenther@suse.de>

* lto-plugin.c (onload): Fail to load if -fno-use-linker-plugin
is set in COLLECT_GCC_OPTIONS.

From-SVN: r209011

re PR bootstrap/60719 (With --program-prefix=$target_alias --program-suffix=-$version install-driver breaks)

2014-04-02  Richard Biener  <rguenther@suse.de>

PR bootstrap/60719
* Makefile.in (install-driver): Guard extra installs with special
names properly.

From-SVN: r209010

Daily bump.

From-SVN: r209007

backwards_compatibility.xml (backwards.third.headers): Update link.

* doc/xml/manual/backwards_compatibility.xml (backwards.third.headers):
Update link.
* doc/xml/manual/policy_data_structures_biblio.xml (bibliography):
Fix broken links.
* doc/xml/manual/shared_ptr.xml (shared_ptr.impl): Likewise.
* doc/xml/manual/using_exceptions.xml (bibliography): Likewise.
* doc/xml/manual/concurrency_extensions.xml
(manual.ext.concurrency.impl.atomic_fallbacks): Likewise.
* doc/html/*: Regenerate.

From-SVN: r209002

fix gcc/cp/ChangeLog formatting

From-SVN: r209001

pt.c (process_partial_specialization): Say "not deducible" rather than "not used".

* pt.c (process_partial_specialization): Say "not deducible"
rather than "not used".  Use inform.

From-SVN: r209000

re PR c++/60374 ([c++11] ICE with invalid template specialization)

PR c++/60374
* pt.c (coerce_template_parms): Check that the pack expansion
pattern works with the first matching parameter.

From-SVN: r208999

init.c (perform_member_init): Homogenize uninitialized diagnostics.

2014-04-01  Fabien Chêne  <fabien@gcc.gnu.org>

* cp/init.c (perform_member_init): Homogenize uninitialized
diagnostics.

2014-04-01  Fabien Chêne  <fabien@gcc.gnu.org>

        * g++.dg/init/ctor4.C: Adjust.
* g++.dg/init/ctor4-1.C: New.
* g++.dg/cpp0x/defaulted2.C: Adjust.

From-SVN: r208998

re PR c++/60708 (An array temporary causes an ICE in gimplify)

PR c++/60708
* call.c (build_array_conv): Call complete_type.

From-SVN: r208996

re PR c++/60713 (ICE in iterative_hash_expr)

PR c++/60713
* typeck2.c (PICFLAG_SIDE_EFFECTS): New.
(picflag_from_initializer): Return it.
(process_init_constructor): Handle it.

From-SVN: r208995

extend.texi (PowerPC AltiVec/VSX Built-in Functions): Document vec_vgbbd.

2014-04-01  Michael Meissner  <meissner@linux.vnet.ibm.com>

* doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
Document vec_vgbbd.

From-SVN: r208993

re PR c++/60642 (Unclear diagnostic with invalid use of abi_tag attribute on explicit instantiation)

PR c++/60642
* decl2.c (is_late_template_attribute): Don't defer abi_tag.
* mangle.c (write_unqualified_name): Fix abi_tag on templates.
* pt.c (get_template_info): Handle NAMESPACE_DECL.
(most_general_template): Handle more kinds of template.
* tree.c (handle_abi_tag_attribute): Ignore abi_tag on template
instantiations and specializations.

From-SVN: r208992

re PR target/60704 (ICE: in extract_constrain_insn_cached, at recog.c:2156 with -flive-range-shrinkage -march=amdfam10)

PR target/60704

        * gcc.dg/pr60704.c: New file.

From-SVN: r208990

re PR target/60704 (ICE: in extract_constrain_insn_cached, at recog.c:2156 with -flive-range-shrinkage -march=amdfam10)

PR target/60704

        * config/i386/i386.md (*float<SWI48><MODEF>2_sse): Leave the second
        alternative enabled before register allocation.

From-SVN: r208989