gcc.git
6 years agoPR target/82518: Fix testcase.
Christophe Lyon [Wed, 21 Mar 2018 09:26:03 +0000 (09:26 +0000)]
PR target/82518: Fix testcase.

2018-03-21  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/arm/pr82518.c: Require arm_neon_hw.

From-SVN: r258707

6 years agotest_summary: handle single quotes
Christophe Lyon [Wed, 21 Mar 2018 08:47:12 +0000 (08:47 +0000)]
test_summary: handle single quotes

2018-03-21  Christophe Lyon  <christophe.lyon@linaro.org>

* test_summary: Match possible single quotes in configure path.

From-SVN: r258706

6 years agoRemove superfluous return statement (PR ipa/84963).
Martin Liska [Wed, 21 Mar 2018 08:26:49 +0000 (09:26 +0100)]
Remove superfluous return statement (PR ipa/84963).

2018-03-21  Martin Liska  <mliska@suse.cz>

PR ipa/84963
* ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
not intended return statement.
2018-03-21  Martin Liska  <mliska@suse.cz>

PR ipa/84963
* gfortran.dg/goacc/pr84963.f90: New test.

From-SVN: r258705

6 years agoFix compile-time hog in MPX boundary checking (PR target/84988).
Martin Liska [Wed, 21 Mar 2018 08:26:05 +0000 (09:26 +0100)]
Fix compile-time hog in MPX boundary checking (PR target/84988).

2018-03-21  Martin Liska  <mliska@suse.cz>

PR target/84988
* tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
(chkp_find_bound_slots_1): Limit number of iterations.

From-SVN: r258704

6 years agoPR c++/71638, ICE with NSDMI and reference.
Marek Polacek [Wed, 21 Mar 2018 07:09:22 +0000 (07:09 +0000)]
PR c++/71638, ICE with NSDMI and reference.

* constexpr.c (cxx_eval_bare_aggregate): Update constructor's flags
even when we replace an element.

* g++.dg/cpp0x/nsdmi14.C: New test.
* g++.dg/cpp1y/nsdmi-aggr10.C: New test.

From-SVN: r258703

6 years agovect-strided-shift-1.c: Add dg-skip-if for MIPS with -mpaired-single directives.
Chenghua Xu [Wed, 21 Mar 2018 03:21:05 +0000 (03:21 +0000)]
vect-strided-shift-1.c: Add dg-skip-if for MIPS with -mpaired-single directives.

2018-03-21  Chenghua Xu  <paul.hua.gm@gmail.com>

       * gcc.dg/vect/vect-strided-shift-1.c: Add dg-skip-if for
       MIPS with -mpaired-single directives.

From-SVN: r258702

6 years agore PR fortran/85001 (ICE in gfc_build_array_type, at fortran/trans-types.c:1420)
Steven G. Kargl [Wed, 21 Mar 2018 01:07:43 +0000 (01:07 +0000)]
re PR fortran/85001 (ICE in gfc_build_array_type, at fortran/trans-types.c:1420)

2018-03-20  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/85001
* interface.c (symbol_rank): Remove bogus null pointer check that
crept in when translating a ternary operator into an if-else
constructor.

2018-03-20  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/85001
* gfortran.dg/interface_41.f90: New test.

From-SVN: r258698

6 years agoDaily bump.
GCC Administrator [Wed, 21 Mar 2018 00:16:14 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258697

6 years agore PR target/84838 (Minor grammar fixes for x86 options)
David H. Gutteridge [Tue, 20 Mar 2018 22:22:28 +0000 (22:22 +0000)]
re PR target/84838 (Minor grammar fixes for x86 options)

PR target/84838
* Minor grammar fixes for x86 options.

From-SVN: r258694

6 years agore PR libstdc++/84998 (std::hash<std::bitset<N>> fails in Debug Mode)
François Dumont [Tue, 20 Mar 2018 21:45:14 +0000 (21:45 +0000)]
re PR libstdc++/84998 (std::hash<std::bitset<N>> fails in Debug Mode)

2018-03-20  François Dumont  <fdumont@gcc.gnu.org>

PR libstdc++/84998
* include/bits/stl_bvector.h: Fix std::hash friend declaration.
* include/std/bitset: Likewise.
* include/bits/stl_map.h (std::map<>): Fix _Rb_tree_merge_helper friend
declaration.
* include/bits/stl_multimap.h (std::multimap<>): Likewise.
* include/bits/stl_multiset.h (std::multiset<>): Likewise.
* include/bits/stl_set.h (std::set<>): Likewise.
* include/bits/unordered_map.h (std::unordered_map<>): Fix
_Hash_merge_helper friend declaration.
(std::unordered_multimap<>): Likewise.
* include/bits/unordered_set.h (std::unordered_set<>): Likewise.
(std::unordered_multiset<>): Likewise.

From-SVN: r258693

6 years agore PR debug/84875 (ICE in maybe_record_trace_start, at dwarf2cfi.c:2348 on s390x)
Jakub Jelinek [Tue, 20 Mar 2018 21:07:13 +0000 (22:07 +0100)]
re PR debug/84875 (ICE in maybe_record_trace_start, at dwarf2cfi.c:2348 on s390x)

PR debug/84875
* dce.c (delete_unmarked_insns): Don't remove frame related noop moves
holding REG_CFA_RESTORE notes, instead turn them into a USE.

* gcc.dg/pr84875.c: New test.

From-SVN: r258692

6 years agore PR c++/84927 (ICE with NSDMI and reference)
Marek Polacek [Tue, 20 Mar 2018 18:13:38 +0000 (18:13 +0000)]
re PR c++/84927 (ICE with NSDMI and reference)

PR c++/84927
* constexpr.c (cxx_eval_bare_aggregate): Update constructor's flags
as we evaluate the elements.
(cxx_eval_constant_expression): Verify constructor's flags
unconditionally.

* g++.dg/cpp1y/nsdmi-aggr9.C: New test.

From-SVN: r258691

6 years agoPR c++/84978, ICE with NRVO.
Jason Merrill [Tue, 20 Mar 2018 18:05:59 +0000 (14:05 -0400)]
PR c++/84978, ICE with NRVO.

* cvt.c (cp_get_fndecl_from_callee): Add fold parameter.
(cp_get_callee_fndecl_nofold): New.
* cp-gimplify.c (cp_genericize_r): Use it instead.
* call.c (check_self_delegation): Likewise.

From-SVN: r258689

6 years agore PR target/83789 (__builtin_altivec_lvx fails for powerpc for altivec-4.c)
Peter Bergner [Tue, 20 Mar 2018 17:25:09 +0000 (12:25 -0500)]
re PR target/83789 (__builtin_altivec_lvx fails for powerpc for altivec-4.c)

PR target/83789
* config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
(altivec_lvx_<mode>_1op): Likewise.
(altivec_stvx_<mode>_2op): Likewise.
(altivec_stvx_<mode>_1op): Likewise.
(altivec_lvx_<VM2:mode>): New define_expand.
(altivec_stvx_<VM2:mode>): Likewise.
(altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
(altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
(altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
(altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
* config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
(rs6000_gen_lvx): Likewise.
* config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
(altivec_expand_stv_builtin): Likewise.
(altivec_expand_builtin): Likewise.
* config/rs6000/vector.md: Likewise.

From-SVN: r258688

6 years agoThis PR shows that we get the load/store_lanes logic wrong for arm big-endian.
Kyrylo Tkachov [Tue, 20 Mar 2018 17:13:16 +0000 (17:13 +0000)]
This PR shows that we get the load/store_lanes logic wrong for arm big-endian.

It is tricky to get right. Aarch64 does it by adding the appropriate lane-swapping
operations during expansion.

I'd like to do the same on arm eventually, but we'd need to port and validate the VTBL-generating
code and add it to all the right places and I'm not comfortable enough doing it for GCC 8, but I am keen
in getting the wrong-code fixed.
As I say in the PR, vectorisation on armeb is already severely restricted (we disable many patterns on BYTES_BIG_ENDIAN)
and the load/store_lanes patterns really were not working properly at all, so disabling them is not
a radical approach.

The way to do that is to return false in ARRAY_MODE_SUPPORTED_P for BYTES_BIG_ENDIAN.

Bootstrapped and tested on arm-none-linux-gnueabihf.
Also tested on armeb-none-eabi.

     PR target/82518
     * config/arm/arm.c (arm_array_mode_supported_p): Return false for
     BYTES_BIG_ENDIAN.

     * lib/target-supports.exp (check_effective_target_vect_load_lanes):
     Disable for armeb targets.
     * gcc.target/arm/pr82518.c: New test.

From-SVN: r258687

6 years ago[PR c++/84962] ICE with anon-struct member
Nathan Sidwell [Tue, 20 Mar 2018 16:01:08 +0000 (16:01 +0000)]
[PR c++/84962] ICE with anon-struct member

https://gcc.gnu.org/ml/gcc-patches/2018-03/msg00961.html
PR c++/84962
* name-lookup.c (pushdecl_class_level): Push anon-struct's
member_vec, if there is one.

PR c++/84962
* g++.dg/lookup/pr84962.C: New.

From-SVN: r258686

6 years ago[PR c++/84970] lookup marking
Nathan Sidwell [Tue, 20 Mar 2018 15:57:30 +0000 (15:57 +0000)]
[PR c++/84970] lookup marking

https://gcc.gnu.org/ml/gcc-patches/2018-03/msg00973.html
PR c++/84970
* cp-tree.h (lookup_list_keep): Declare.
* tree.c (lookup_list_keep): New, broken out of ...
(build_min): ... here.  Call it.
* decl.c (cp_finish_decl): Call lookup_list_keep.

PR c++/84970
* g++.dg/lookup/pr84970.C: New.

From-SVN: r258685

6 years agore PR target/84986 (Performance regression: loop no longer vectorized (x86-64))
Richard Biener [Tue, 20 Mar 2018 14:35:46 +0000 (14:35 +0000)]
re PR target/84986 (Performance regression: loop no longer vectorized (x86-64))

2018-03-20  Richard Biener  <rguenther@suse.de>

PR target/84986
* config/i386/i386.c (ix86_add_stmt_cost): Only cost
sign-conversions as zero, fall back to standard scalar_stmt
cost for the rest.

* gcc.dg/vect/costmodel/x86_64/costmodel-pr84986.c: New testcase.

From-SVN: r258684

6 years agoHandle -fno-guess-branch-probability properly in predict.c (PR ipa/84825).
Martin Liska [Tue, 20 Mar 2018 14:14:17 +0000 (15:14 +0100)]
Handle -fno-guess-branch-probability properly in predict.c (PR ipa/84825).

2018-03-20  Martin Liska  <mliska@suse.cz>

PR ipa/84825
* predict.c (rebuild_frequencies): Handle case when we have
PROFILE_ABSENT, but flag_guess_branch_prob is false.
2018-03-20  Martin Liska  <mliska@suse.cz>

PR ipa/84825
* g++.dg/ipa/pr84825.C: New test.

From-SVN: r258683

6 years agoRemove ICEing test-case.
Martin Liska [Tue, 20 Mar 2018 14:13:17 +0000 (15:13 +0100)]
Remove ICEing test-case.

2018-03-20  Martin Liska  <mliska@suse.cz>

* gcc.dg/lto/chkp-ctor-merge_0.c: Remove.

From-SVN: r258682

6 years agore PR target/84990 (Boostrap broken with --enable-checking=release and Ada)
Jakub Jelinek [Tue, 20 Mar 2018 13:00:48 +0000 (14:00 +0100)]
re PR target/84990 (Boostrap broken with --enable-checking=release and Ada)

PR target/84990
* dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
flag_section_anchors.
* varasm.c (use_blocks_for_decl_p): Remove hack for
dw2_force_const_mem.

From-SVN: r258681

6 years agoPR c++/84937 - ICE with class deduction and auto.
Jason Merrill [Tue, 20 Mar 2018 12:44:49 +0000 (08:44 -0400)]
PR c++/84937 - ICE with class deduction and auto.

* pt.c (rewrite_template_parm): Fix auto handling.

From-SVN: r258680

6 years agoforce-parallel-4.c: XFAIL one parallelizable loop.
Richard Biener [Tue, 20 Mar 2018 12:42:59 +0000 (12:42 +0000)]
force-parallel-4.c: XFAIL one parallelizable loop.

2018-03-20  Richard Biener  <rguenther@suse.de>

* testsuite/libgomp.graphite/force-parallel-4.c: XFAIL one
parallelizable loop.

From-SVN: r258679

6 years agore PR target/84845 (ICE: in extract_insn, at recog.c:2304: unrecognizable insn at...
Jakub Jelinek [Tue, 20 Mar 2018 10:59:26 +0000 (11:59 +0100)]
re PR target/84845 (ICE: in extract_insn, at recog.c:2304: unrecognizable insn at -O2 and above at aarch64)

PR target/84845
* config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
to ...
(*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
be created, use lowpart_subreg of operands[0] rather than operands[0]
itself.
(*aarch64_reg_<mode>3_minus_mask): Rename to ...
(*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
(*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
and n constraint instead of aarch64_shift_imm_di and Usd.
(*aarch64_reg_<optab>_minus<mode>3): Rename to ...
(*aarch64_<optab>_reg_minus<mode>3): ... this.

* gcc.c-torture/compile/pr84845.c: New test.

From-SVN: r258678

6 years ago[ARM][PR82989] Fix unexpected use of NEON instructions for shifts
Sudakshina Das [Tue, 20 Mar 2018 10:54:42 +0000 (10:54 +0000)]
[ARM][PR82989] Fix unexpected use of NEON instructions for shifts

This patch fixes PR82989 so that we avoid NEON instructions when
-mneon-for-64bits is not enabled. This is more of a short term fix
for the real deeper problem of making an early decision of choosing
or rejecting NEON instructions. There is now a new ticket PR84467 to
deal with the longer term solution.
(Please refer to the discussion in the bug report for more details).

Sudi

*** gcc/ChangeLog ***

2018-03-20  Sudakshina Das  <sudi.das@arm.com>

PR target/82989
* config/arm/neon.md (ashldi3_neon): Update ?s for constraints
to favor GPR over NEON registers.
(<shift>di3_neon): Likewise.

*** gcc/testsuite/ChangeLog ***

2018-03-20  Sudakshina Das  <sudi.das@arm.com>

PR target/82989
* gcc.target/arm/pr82989.c: New test.

From-SVN: r258677

6 years ago[nvptx] Fix bar.sync position
Tom de Vries [Tue, 20 Mar 2018 10:31:23 +0000 (10:31 +0000)]
[nvptx] Fix bar.sync position

2018-03-20  Tom de Vries  <tom@codesourcery.com>

PR target/84952
* config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
(nvptx_process_pars): Emit bar.sync asap and alap.

From-SVN: r258676

6 years ago* c-ada-spec.c (pp_ada_tree_identifier): Deal specifically with _Bool.
Eric Botcazou [Tue, 20 Mar 2018 09:44:21 +0000 (09:44 +0000)]
* c-ada-spec.c (pp_ada_tree_identifier): Deal specifically with _Bool.

From-SVN: r258675

6 years ago[nvptx] Fix prevent_branch_around_nothing
Tom de Vries [Tue, 20 Mar 2018 09:14:07 +0000 (09:14 +0000)]
[nvptx] Fix prevent_branch_around_nothing

2018-03-20  Tom de Vries  <tom@codesourcery.com>

PR target/84954
* config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
seen_label if seen_label is already set.

From-SVN: r258674

6 years agore PR target/84945 (UBSAN: gcc/config/i386/i386.c:33312:22: runtime error: shift...
Jakub Jelinek [Tue, 20 Mar 2018 08:14:42 +0000 (09:14 +0100)]
re PR target/84945 (UBSAN: gcc/config/i386/i386.c:33312:22: runtime error: shift exponent 32 is too large for 32-bit type 'int')

PR target/84945
* config/i386/i386.c (fold_builtin_cpu): For features above 31
use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
Use 1U instead of 1.  Formatting fixes.

* gcc.target/i386/pr84945.c: New test.

* config/i386/cpuinfo.h (__cpu_features2): Declare.
* config/i386/cpuinfo.c (__cpu_features2): New variable for
ifndef SHARED only.
(set_feature): Define.
(get_available_features): Use set_feature macro.  Set __cpu_features2
to the second word of features ifndef SHARED.

From-SVN: r258673

6 years agoPR target/81647: Fix testcase.
Christophe Lyon [Tue, 20 Mar 2018 08:11:35 +0000 (08:11 +0000)]
PR target/81647: Fix testcase.

2018-03-20  Christophe Lyon  <christophe.lyon@linaro.org>

PR target/81647
* gcc.target/aarch64/pr81647.c: Require fenv_exceptions.

From-SVN: r258672

6 years agore PR c/84953 (misleading warning from strpbrk(x,""))
Jakub Jelinek [Tue, 20 Mar 2018 07:55:41 +0000 (08:55 +0100)]
re PR c/84953 (misleading warning from strpbrk(x,""))

PR c/84953
* builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
instead of TREE_TYPE (s1) for the return value.

* gcc.dg/pr84953.c: New test.

From-SVN: r258671

6 years agoDaily bump.
GCC Administrator [Tue, 20 Mar 2018 00:16:14 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258670

6 years agore PR tree-optimization/84946 (UBSAN: in mem_valid_for_store_merging ../../gcc/gimple...
Jakub Jelinek [Mon, 19 Mar 2018 20:49:57 +0000 (21:49 +0100)]
re PR tree-optimization/84946 (UBSAN: in mem_valid_for_store_merging ../../gcc/gimple-ssa-store-merging.c:3951)

PR tree-optimization/84946
* gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
bitsize + bitsize in poly_uint64 rather than poly_int64.

From-SVN: r258665

6 years agore PR sanitizer/78651 (Incorrect exception handling when catch clause uses local...
Jakub Jelinek [Mon, 19 Mar 2018 20:48:39 +0000 (21:48 +0100)]
re PR sanitizer/78651 (Incorrect exception handling when catch clause uses local class and PIC and sanitizer are active)

PR sanitizer/78651
* dwarf2asm.c: Include fold-const.c.
(dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
of decl rather than decl itself.

From-SVN: r258664

6 years agore PR sanitizer/84761 (AddressSanitizer is not compatible with glibc 2.27 on x86)
Jakub Jelinek [Mon, 19 Mar 2018 20:47:29 +0000 (21:47 +0100)]
re PR sanitizer/84761 (AddressSanitizer is not compatible with glibc 2.27 on x86)

PR sanitizer/84761
* sanitizer_common/sanitizer_linux_libcdep.cc (__GLIBC_PREREQ):
Define if not defined.
(DL_INTERNAL_FUNCTION): Don't define.
(InitTlsSize): For __i386__ if not compiled against glibc 2.27+
determine at runtime whether to use regparm(3), stdcall calling
convention for older glibcs or normal calling convention for
newer glibcs for call to _dl_get_tls_static_info.

From-SVN: r258663

6 years agore PR rtl-optimization/84643 (gcc/optabs.c:6549:26: runtime error: load of value...
Jakub Jelinek [Mon, 19 Mar 2018 20:46:23 +0000 (21:46 +0100)]
re PR rtl-optimization/84643 (gcc/optabs.c:6549:26: runtime error: load of value 131075, which is not a valid value for type 'memmodel')

PR rtl-optimization/84643
* memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.

From-SVN: r258662

6 years agore PR c++/84927 (ICE with NSDMI and reference)
Marek Polacek [Mon, 19 Mar 2018 20:46:16 +0000 (20:46 +0000)]
re PR c++/84927 (ICE with NSDMI and reference)

PR c++/84927
* constexpr.c (cxx_eval_bare_aggregate): Update constructor's flags
as we evaluate the elements.
(cxx_eval_constant_expression): Verify constructor's flags
unconditionally.

* g++.dg/cpp1y/nsdmi-aggr9.C: New test.

From-SVN: r258661

6 years agore PR c++/84925 (ICE with segfault in __PRETTY_FUNCTION__)
Marek Polacek [Mon, 19 Mar 2018 20:34:45 +0000 (20:34 +0000)]
re PR c++/84925 (ICE with segfault in __PRETTY_FUNCTION__)

PR c++/84925
* pt.c (enclosing_instantiation_of): Check if fn is null.

* g++.dg/cpp1z/lambda-__func__.C: New test.

From-SVN: r258660

6 years agoPR c++/71834 - template-id with too few arguments.
Jason Merrill [Mon, 19 Mar 2018 20:32:57 +0000 (16:32 -0400)]
PR c++/71834 - template-id with too few arguments.

* pt.c (coerce_template_parms): Check fixed_parameter_pack_p.

From-SVN: r258659

6 years agore PR sanitizer/78651 (Incorrect exception handling when catch clause uses local...
Maxim Ostapenko [Mon, 19 Mar 2018 19:59:56 +0000 (19:59 +0000)]
re PR sanitizer/78651 (Incorrect exception handling when catch clause uses local class and PIC and sanitizer are active)

2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>

gcc/

PR sanitizer/78651
* dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
calling assemble_variable.

gcc/testsuite/

PR sanitizer/78651
* g++.dg/asan/pr78651.C: New test.

From-SVN: r258658

6 years ago[testsuite] Add nvptx xfail to pr84512.c
Tom de Vries [Mon, 19 Mar 2018 19:12:53 +0000 (19:12 +0000)]
[testsuite] Add nvptx xfail to pr84512.c

2018-03-19  Tom de Vries  <tom@codesourcery.com>

* gcc.dg/tree-ssa/pr84512.c: Don't require effective target
vect_int_mult.  Add nvptx xfail for PR84958.

From-SVN: r258656

6 years ago[PR c++/84835] ICE with generic lambda in extern "C"
Nathan Sidwell [Mon, 19 Mar 2018 18:56:22 +0000 (18:56 +0000)]
[PR c++/84835] ICE with generic lambda in extern "C"

https://gcc.gnu.org/ml/gcc-patches/2018-03/msg00890.html
PR c++/84835
* lambda.c (maybe_add_lambda_conv_op): Force C++ linkage.
* pt.c (build_template_decl): Propagate language linkage.

PR c++/84835
* g++.dg/cpp1y/pr84835.C: New.

From-SVN: r258655

6 years ago[PR81647][AARCH64] Fix handling of Unordered Comparisons in aarch64-simd.md
Sudakshina Das [Mon, 19 Mar 2018 18:50:32 +0000 (18:50 +0000)]
[PR81647][AARCH64] Fix handling of Unordered Comparisons in aarch64-simd.md

This patch fixes the inconsistent behavior observed at -O3 for the unordered
comparisons. According to the online docs (https://gcc.gnu.org/onlinedocs
/gcc-7.2.0/gccint/Unary-and-Binary-Expressions.html), all of the following
should not raise an FP exception:
- UNGE_EXPR
- UNGT_EXPR
- UNLE_EXPR
- UNLT_EXPR
- UNEQ_EXPR
Also ORDERED_EXPR and UNORDERED_EXPR should only return zero or one.

The aarch64-simd.md handling of these were generating exception raising
instructions such as fcmgt. This patch changes the instructions that are
emitted in order to not give out the exceptions. We first check each
operand for NaNs and force any elements containing NaN to zero before using
them in the compare.

Example: UN<cc> (a, b) -> UNORDERED (a, b)
  | (cm<cc> (isnan (a) ? 0.0 : a, isnan (b) ? 0.0 : b))

The ORDERED_EXPR is now handled as (cmeq (a, a) & cmeq (b, b)) and
UNORDERED_EXPR as ~ORDERED_EXPR and UNEQ as (~ORDERED_EXPR | cmeq (a,b)).

ChangeLog Entries:

*** gcc/ChangeLog ***

2018-03-19  Sudakshina Das  <sudi.das@arm.com>

PR target/81647
* config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.

*** gcc/testsuite/ChangeLog ***

2018-03-19  Sudakshina Das  <sudi.das@arm.com>

PR target/81647
* gcc.target/aarch64/pr81647.c: New.

From-SVN: r258653

6 years ago* es.po, sv.po: Update.
Joseph Myers [Mon, 19 Mar 2018 18:29:34 +0000 (18:29 +0000)]
* es.po, sv.po: Update.

From-SVN: r258651

6 years agoRISC-V: Fix bootstrap failure.
Jim Wilson [Mon, 19 Mar 2018 18:08:24 +0000 (18:08 +0000)]
RISC-V: Fix bootstrap failure.

gcc/
PR bootstrap/84856
* config/riscv/riscv.c (riscv_function_arg_boundary): Use
PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
(riscv_first_stack_step): Likewise.
(riscv_option_override): Use STACK_BOUNDARY instead of
MIN_STACK_BOUNDARY.
* config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
MIN_STACK_BOUNDARY.
(BIGGEST_ALIGNMENT): Set to 128.
(PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
(RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
STACK_BOUNDARY.

From-SVN: r258650

6 years agore PR tree-optimization/84933 (ICE in set_value_range, at tree-vrp.c:288 since r257852)
Richard Biener [Mon, 19 Mar 2018 14:11:05 +0000 (14:11 +0000)]
re PR tree-optimization/84933 (ICE in set_value_range, at tree-vrp.c:288 since r257852)

2018-03-19  Richard Biener  <rguenther@suse.de>

PR tree-optimization/84933
* tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.

* g++.dg/pr84933.C: New testcase.

From-SVN: r258646

6 years agore PR tree-optimization/84859 (bogus -Warray-bounds on a memcpy in a loop)
Richard Biener [Mon, 19 Mar 2018 14:08:58 +0000 (14:08 +0000)]
re PR tree-optimization/84859 (bogus -Warray-bounds on a memcpy in a loop)

2018-03-19  Richard Biener  <rguenther@suse.de>

PR tree-optimization/84859
* tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
(cond_if_else_store_replacement): Perform sinking operation on
single-store BBs regardless of MAX_STORES_TO_SINK setting.
Generalize what a BB with a single eligible store is.

* gcc.dg/tree-ssa/pr84859.c: New testcase.
* gcc.dg/tree-ssa/pr35286.c: Disable cselim.
* gcc.dg/tree-ssa/split-path-6.c: Likewise.
* gcc.dg/tree-ssa/split-path-7.c: Likewise.

From-SVN: r258645

6 years ago[C++/84812] ICE with local fn decl
Nathan Sidwell [Mon, 19 Mar 2018 14:07:07 +0000 (14:07 +0000)]
[C++/84812] ICE with local fn decl

https://gcc.gnu.org/ml/gcc-patches/2018-03/msg00872.html
PR c++/84812
* name-lookup.c (set_local_extern_decl_linkage): Defend against
ambiguous lookups.

PR c++/84812
* g++.dg/lookup/pr84812.C: New.

From-SVN: r258644

6 years agore PR tree-optimization/84929 (ICE at -O3 on valid code on x86_64-linux-gnu: tree...
Richard Biener [Mon, 19 Mar 2018 12:49:30 +0000 (12:49 +0000)]
re PR tree-optimization/84929 (ICE at -O3 on valid code on x86_64-linux-gnu: tree check: expected polynomial_chrec, have nop_expr in analyze_siv_subscript_cst_affine, at tree-data-ref.c:3018)

2018-03-19  Richard Biener  <rguenther@suse.de>

PR tree-optimization/84929
* tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
chrec_is_positive against non-chrec arg.

* gcc.dg/torture/pr84929.c: New testcase.

From-SVN: r258643

6 years agore PR target/84711 (AArch32 big-endian fails when taking subreg of a vector mode...
Tamar Christina [Mon, 19 Mar 2018 09:14:25 +0000 (09:14 +0000)]
re PR target/84711 (AArch32 big-endian fails when taking subreg of a vector mode to a scalar mode.)

gcc/
2018-03-19  Tamar Christina  <tamar.christina@arm.com>

PR target/84711
* config/arm/arm.c (arm_can_change_mode_class): revert r258554.

gcc/testsuite/
2018-03-19  Tamar Christina  <tamar.christina@arm.com>

PR target/84711
* gcc.target/arm/big-endian-subreg.c: Delete.

From-SVN: r258642

6 years agore PR fortran/84931 (Expansion of array constructor with constant implied-do-object...
Thomas Koenig [Mon, 19 Mar 2018 07:04:35 +0000 (07:04 +0000)]
re PR fortran/84931 (Expansion of array constructor with constant implied-do-object goes sideways)

2018-03-19  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/84931
* simplify.c (gfc_convert_constant): Correctly handle iterators
for type conversion.

2018-03-19  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/84931
* gfortran.dg/array_constructor_52.f90: New test.

From-SVN: r258641

6 years agoDaily bump.
GCC Administrator [Mon, 19 Mar 2018 00:16:19 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258640

6 years agoapi.xml: www.fsf.org has moved to https.
Gerald Pfeifer [Sun, 18 Mar 2018 23:52:00 +0000 (23:52 +0000)]
api.xml: www.fsf.org has moved to https.

* doc/xml/api.xml: www.fsf.org has moved to https. Also omit
trailing slash for domain level link.
* doc/xml/faq.xml: Ditto.
* doc/xml/manual/appendix_free.xml (software): Ditto.
* doc/xml/manual/intro.xml: Ditto.
* doc/xml/manual/spine.xml: Ditto.
* doc/xml/spine.xml: Ditto.

From-SVN: r258637

6 years agodocumentation_hacking.xml: Adjust link to docbook.org.
Gerald Pfeifer [Sun, 18 Mar 2018 23:38:23 +0000 (23:38 +0000)]
documentation_hacking.xml: Adjust link to docbook.org.

* doc/xml/manual/documentation_hacking.xml: Adjust link to
docbook.org.

From-SVN: r258636

6 years agoFix UBSAN in regrename.c (PR rtl-optimization/84635).
Martin Liska [Sun, 18 Mar 2018 20:17:10 +0000 (21:17 +0100)]
Fix UBSAN in regrename.c (PR rtl-optimization/84635).

2018-03-18  Martin Liska  <mliska@suse.cz>

PR rtl-optimization/84635
* regrename.c (build_def_use): Use matches_mode only when
matches >= 0.

From-SVN: r258634

6 years agore PR fortran/77414 (ICE in create_function_arglist, at fortran/trans-decl.c:2410)
Steven G. Kargl [Sun, 18 Mar 2018 17:51:57 +0000 (17:51 +0000)]
re PR fortran/77414 (ICE in create_function_arglist, at fortran/trans-decl.c:2410)

2018-03-18  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/77414
* decl.c (get_proc_name):  Check for a subroutine re-defined in
the contain portion of a subroutine.  Change language of existing
error message to better describe the issue. While here fix whitespace
issues.

2018-03-18  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/77414
* gfortran.dg/pr77414.f90: New test.
* gfortran.dg/internal_references_1.f90: Adjust error message.

From-SVN: r258633

6 years agore PR fortran/65453 (ICE in build_function_decl, at fortran/trans-decl.c:2001)
Steven G. Kargl [Sun, 18 Mar 2018 16:33:55 +0000 (16:33 +0000)]
re PR fortran/65453 (ICE in build_function_decl, at fortran/trans-decl.c:2001)

2018-03-18  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/65453
* decl.c (get_proc_name): Catch clash between a procedure statement
and a contained subprogram

2018-03-18  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/65453
* gfortran.dg/pr65453.f90: New test.

From-SVN: r258632

6 years agoDon't try to vectorise COND_EXPR reduction chains (PR 84913)
Richard Sandiford [Sun, 18 Mar 2018 10:25:29 +0000 (10:25 +0000)]
Don't try to vectorise COND_EXPR reduction chains (PR 84913)

The testcase ICEd for both SVE and AVX512 because we were trying
to vectorise a chain of COND_EXPRs as a reduction and getting
confused by reduc_index == -1.

2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>

gcc/
PR tree-optimization/84913
* tree-vect-loop.c (vectorizable_reduction): Don't try to
vectorize chains of COND_EXPRs.

gcc/testsuite/
PR tree-optimization/84913
* gfortran.dg/vect/pr84913.f90: New test.

From-SVN: r258631

6 years agore PR fortran/79929 (Bogus Warning: '__builtin_memset': specified size 4294967291...
Thomas Koenig [Sun, 18 Mar 2018 09:20:37 +0000 (09:20 +0000)]
re PR fortran/79929 (Bogus Warning: '__builtin_memset': specified size 4294967291 exceeds maximum object size 2147483647)

2018-03-18  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/79929
* gfortran.dg/warn_concat.f90: New test.

From-SVN: r258630

6 years ago[NDS32] Fix wrong MAX_REGS_PER_ADDRESS value.
Chung-Ju Wu [Sun, 18 Mar 2018 06:44:31 +0000 (06:44 +0000)]
[NDS32] Fix wrong MAX_REGS_PER_ADDRESS value.

gcc/
* config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.

From-SVN: r258629

6 years ago[NDS32] Define LOGICAL_OP_NON_SHORT_CIRCUIT.
Chung-Ju Wu [Sun, 18 Mar 2018 06:18:18 +0000 (06:18 +0000)]
[NDS32] Define LOGICAL_OP_NON_SHORT_CIRCUIT.

gcc/
* config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.

From-SVN: r258628

6 years ago[NDS32] Define CLZ_DEFINED_VALUE_AT_ZERO.
Chung-Ju Wu [Sun, 18 Mar 2018 04:47:10 +0000 (04:47 +0000)]
[NDS32] Define CLZ_DEFINED_VALUE_AT_ZERO.

gcc/
* config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.

From-SVN: r258627

6 years agoDaily bump.
GCC Administrator [Sun, 18 Mar 2018 00:16:25 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258626

6 years agoFix new test that fails in C++98 mode
Jonathan Wakely [Sat, 17 Mar 2018 13:44:16 +0000 (13:44 +0000)]
Fix new test that fails in C++98 mode

* testsuite/20_util/function_objects/comparisons_pointer.cc: Adjust
to compile as C++98.

From-SVN: r258623

6 years ago[NDS32] Implment ADJUST_REG_ALLOC_ORDER for performance requirement.
Chung-Ju Wu [Sat, 17 Mar 2018 13:06:12 +0000 (13:06 +0000)]
[NDS32] Implment ADJUST_REG_ALLOC_ORDER for performance requirement.

gcc/
* config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
* config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
(nds32_adjust_reg_alloc_order): New function.
* config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.

Co-Authored-By: Kito Cheng <kito.cheng@gmail.com>
From-SVN: r258621

6 years ago[NDS32] Use HOST_WIDE_INT_PRINT_DEC instead of %ld or %d.
Kito Cheng [Sat, 17 Mar 2018 12:11:22 +0000 (12:11 +0000)]
[NDS32] Use HOST_WIDE_INT_PRINT_DEC instead of %ld or %d.

gcc/
* config/nds32/nds32.c (nds32_asm_output_mi_thunk,
nds32_print_operand, nds32_print_operand_address): Use
HOST_WIDE_INT_PRINT_DEC instead.

From-SVN: r258620

6 years ago[NDS32] Optimize nds32_register_priority.
Chung-Ju Wu [Sat, 17 Mar 2018 11:51:46 +0000 (11:51 +0000)]
[NDS32] Optimize nds32_register_priority.

gcc/
* config/nds32/nds32.c (nds32_register_priority): Modify cost.

From-SVN: r258619

6 years agore PR target/84902 (549.fotonik3d_r from SPEC2017 fails verification with -Ofast...
Jakub Jelinek [Sat, 17 Mar 2018 11:12:00 +0000 (12:12 +0100)]
re PR target/84902 (549.fotonik3d_r from SPEC2017 fails verification with -Ofast -march=native on Zen since r258518)

PR target/84902
* config/i386/i386.c (initial_ix86_tune_features,
initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
unsigned long long.
(set_ix86_tune_features): Change ix86_tune_mask from unsigned int
to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
rather than 1u << ix86_tune.  Formatting fix.
(ix86_option_override_internal): Change ix86_arch_mask from
unsigned int to unsigned HOST_WIDE_INT, initialize to
HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
(ix86_function_specific_restore): Likewise.

From-SVN: r258618

6 years agoDaily bump.
GCC Administrator [Sat, 17 Mar 2018 00:16:13 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258617

6 years ago* de.po: Update.
Joseph Myers [Fri, 16 Mar 2018 23:50:29 +0000 (23:50 +0000)]
* de.po: Update.

From-SVN: r258614

6 years agore PR target/84899 (ICE: in final_scan_insn_1, at final.c:3139 (error: could not...
Jakub Jelinek [Fri, 16 Mar 2018 21:01:16 +0000 (22:01 +0100)]
re PR target/84899 (ICE: in final_scan_insn_1, at final.c:3139 (error: could not split insn))

PR target/84899
* postreload.c (reload_combine_recognize_pattern): Perform
INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
truncate_int_for_mode the result for the destination's mode.

* gcc.dg/pr84899.c: New test.

From-SVN: r258610

6 years agore PR c/84909 (typo: conversion from %qT to to %qT)
Jakub Jelinek [Fri, 16 Mar 2018 20:12:27 +0000 (21:12 +0100)]
re PR c/84909 (typo: conversion from %qT to to %qT)

PR c/84909
* c-warn.c (conversion_warning): Replace "to to" with "to" in
diagnostics.

* hsa-gen.c (mem_type_for_type): Fix comment typo.
* tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
Likewise.
* gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
Likewise.

From-SVN: r258609

6 years agore PR c/84910 (typo: "%qs follows inline declaration ")
Jakub Jelinek [Fri, 16 Mar 2018 20:11:40 +0000 (21:11 +0100)]
re PR c/84910 (typo: "%qs follows inline declaration ")

PR c/84910
* c-warn.c (diagnose_mismatched_attributes): Remove trailing space from
diagnostics.

* parser.c (cp_parser_lambda_introducer): Remove trailing space from
diagnostics.
* method.c (synthesize_method): Likewise.
* pt.c (convert_nontype_argument): Likewise.

From-SVN: r258608

6 years agore PR fortran/69395 (ICE on declaring array with more than 7 dimensions+codimensions)
Steven G. Kargl [Fri, 16 Mar 2018 19:20:15 +0000 (19:20 +0000)]
re PR fortran/69395 (ICE on declaring array with more than 7 dimensions+codimensions)

2018-03-16  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/69395
* decl.c (merge_array_spec): Correct the error condition.

2018-03-16  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/69395
* gfortran.dg/pr69395.f90: Add test for max dimensions

From-SVN: r258607

6 years agolibgo: add runtime/pprof/internal/profile.gox to noinst_DATA
Ian Lance Taylor [Fri, 16 Mar 2018 19:01:40 +0000 (19:01 +0000)]
libgo: add runtime/pprof/internal/profile.gox to noinst_DATA

    Also add noinst_DATA to CHECK_DEPS; it's not needed in practice since
    `make` will build noinst_DATA, but it's logically required and will
    make a difference if any of the noinst_DATA sources change between
    `make` and `make check`.

    Tony Reix figured out why omitting packages from noinst_DATA didn't
    seem to matter: because if gccgo can't find foo.gox, it will fall back
    to reading the export data in foo.o, and foo.o will exist because
    these packages go into libgo.a.

    Reviewed-on: https://go-review.googlesource.com/101077

From-SVN: r258606

6 years agoPR c++/84720 - ICE with rvalue ref non-type argument.
Jason Merrill [Fri, 16 Mar 2018 18:56:06 +0000 (14:56 -0400)]
PR c++/84720 - ICE with rvalue ref non-type argument.

* pt.c (invalid_nontype_parm_type_p): Prohibit rvalue reference.
(convert_nontype_argument): Revert earlier change.

From-SVN: r258605

6 years agoPR c++/80227 - SFINAE and negative array size.
Jason Merrill [Fri, 16 Mar 2018 18:56:01 +0000 (14:56 -0400)]
PR c++/80227 - SFINAE and negative array size.

* decl.c (compute_array_index_type): Use
build_converted_constant_expr and valid_constant_size_p.

From-SVN: r258604

6 years agoPR c++/84906 - silent wrong code with ambiguous conversion.
Jason Merrill [Fri, 16 Mar 2018 18:55:54 +0000 (14:55 -0400)]
PR c++/84906 - silent wrong code with ambiguous conversion.

* call.c (build_user_type_conversion_1): Set need_temporary_p on
ambiguous conversion.
(convert_like_real): Check it.

From-SVN: r258603

6 years agore PR target/84876 (ICE on invalid code in lra_assign at gcc/lra-assigns.c:1601 since...
Vladimir Makarov [Fri, 16 Mar 2018 18:48:26 +0000 (18:48 +0000)]
re PR target/84876 (ICE on invalid code in lra_assign at gcc/lra-assigns.c:1601 since r258504)

2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>

PR target/84876
* lra-assigns.c (lra_split_hard_reg_for): Don't use
regno_allocno_class_array and sorted_pseudos.
* lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
insns where regno is used.

2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>

PR target/84876
* gcc.target/i386/pr84876.c: New test.

From-SVN: r258602

6 years ago* uk.po: Update.
Joseph Myers [Fri, 16 Mar 2018 17:34:56 +0000 (17:34 +0000)]
* uk.po: Update.

From-SVN: r258601

6 years ago* de.po: Update.
Joseph Myers [Fri, 16 Mar 2018 16:58:25 +0000 (16:58 +0000)]
* de.po: Update.

From-SVN: r258599

6 years agoRedirect reference in the symbol table (PR ipa/84833).
Martin Liska [Fri, 16 Mar 2018 14:53:45 +0000 (15:53 +0100)]
Redirect reference in the symbol table (PR ipa/84833).

2018-03-16  Martin Liska  <mliska@suse.cz>

PR ipa/84833
* multiple_target.c (create_dispatcher_calls): Redirect
reference in the symbol table.
2018-03-16  Martin Liska  <mliska@suse.cz>

PR ipa/84833
* gcc.target/i386/mvc11.c: New test.

From-SVN: r258596

6 years agoProperly redirect alias for MVC (PR ipa/84722).
Martin Liska [Fri, 16 Mar 2018 14:53:24 +0000 (15:53 +0100)]
Properly redirect alias for MVC (PR ipa/84722).

2018-03-16  Martin Liska  <mliska@suse.cz>

PR ipa/84722
* multiple_target.c (create_dispatcher_calls): Redirect also
an alias.
2018-03-16  Martin Liska  <mliska@suse.cz>

PR ipa/84722
* gcc.target/i386/mvc10.c: New test.

From-SVN: r258595

6 years agoPR c++/83937 - wrong C++17 handling of init-list ctor argument.
Jason Merrill [Fri, 16 Mar 2018 14:35:47 +0000 (10:35 -0400)]
PR c++/83937 - wrong C++17 handling of init-list ctor argument.

* call.c (build_special_member_call): Don't convert an init-list
argument directly to the class type.

From-SVN: r258594

6 years agore PR c++/79937 (ICE in replace_placeholders_r)
Jakub Jelinek [Fri, 16 Mar 2018 12:46:12 +0000 (13:46 +0100)]
re PR c++/79937 (ICE in replace_placeholders_r)

PR c++/79937
PR c++/82410
* tree.h (TARGET_EXPR_NO_ELIDE): Define.
* gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.

* cp-tree.h (CONSTRUCTOR_PLACEHOLDER_BOUNDARY): Define.
(find_placeholder): Declare.
* tree.c (struct replace_placeholders_t): Add exp member.
(replace_placeholders_r): Don't walk into ctors with
CONSTRUCTOR_PLACEHOLDER_BOUNDARY flag set, unless they are equal to
d->exp.  Replace PLACEHOLDER_EXPR with unshare_expr (x) rather than x.
(replace_placeholders): Initialize data.exp.
(find_placeholders_r, find_placeholders): New functions.
* typeck2.c (process_init_constructor_record,
process_init_constructor_union): Set CONSTRUCTOR_PLACEHOLDER_BOUNDARY
if adding NSDMI on which find_placeholder returns true.
* call.c (build_over_call): Don't call replace_placeholders here.
* cp-gimplify.c (cp_genericize_r): Set TARGET_EXPR_NO_ELIDE on
TARGET_EXPRs with CONSTRUCTOR_PLACEHOLDER_BOUNDARY set on
TARGET_EXPR_INITIAL.
(cp_fold): Copy over CONSTRUCTOR_PLACEHOLDER_BOUNDARY bit to new
ctor.

* g++.dg/cpp1y/pr79937-1.C: New test.
* g++.dg/cpp1y/pr79937-2.C: New test.
* g++.dg/cpp1y/pr79937-3.C: New test.
* g++.dg/cpp1y/pr79937-4.C: New test.
* g++.dg/cpp1y/pr82410.C: New test.

From-SVN: r258593

6 years agoPR c++/83911 - ICE with multiversioned constructor.
Jason Merrill [Fri, 16 Mar 2018 12:38:42 +0000 (08:38 -0400)]
PR c++/83911 - ICE with multiversioned constructor.

* cp-gimplify.c (cp_genericize_r): Replace versioned function with
dispatchere here.
* call.c (build_over_call): Not here.
PR c++/83911 - ICE with multiversioned constructor.

From-SVN: r258592

6 years agore PR c/84873 (ICE: verify_ssa failed (error: definition in block 3 does not dominate...
Richard Biener [Fri, 16 Mar 2018 12:04:29 +0000 (12:04 +0000)]
re PR c/84873 (ICE: verify_ssa failed (error: definition in block 3 does not dominate use in block 4))

2018-03-16  Richard Biener  <rguenther@suse.de>

PR c/84873
* c-gimplify.c (c_gimplify_expr): Revert previous change.  Instead
unshare the possibly folded expression.

From-SVN: r258591

6 years ago[testsuite] Require vect_int_mult in pr84512.c
Tom de Vries [Fri, 16 Mar 2018 11:38:57 +0000 (11:38 +0000)]
[testsuite] Require vect_int_mult in pr84512.c

2018-03-16  Tom de Vries  <tom@codesourcery.com>

* gcc.dg/tree-ssa/pr84512.c: Require effective target vect_int_mult.

From-SVN: r258590

6 years ago[testsuite] Require label_values in builtin-unreachable-6.c
Tom de Vries [Fri, 16 Mar 2018 11:38:45 +0000 (11:38 +0000)]
[testsuite] Require label_values in builtin-unreachable-6.c

2018-03-16  Tom de Vries  <tom@codesourcery.com>

* gcc.dg/builtin-unreachable-6.c: Require effective target label_values.

From-SVN: r258589

6 years agore PR c++/84874 (internal compiler error: in reshape_init_class, at cp/decl.c:5800)
Jakub Jelinek [Fri, 16 Mar 2018 08:44:56 +0000 (09:44 +0100)]
re PR c++/84874 (internal compiler error: in reshape_init_class, at cp/decl.c:5800)

PR c++/84874
* g++.dg/cpp2a/desig8.C: New test.

From-SVN: r258588

6 years agoFix documentation for CLWB ISA.
Julia Koval [Fri, 16 Mar 2018 08:11:27 +0000 (09:11 +0100)]
Fix documentation for CLWB ISA.

gcc/
* doc/invoke.texi (Skylake Server): Add CLWB.
Cannonlake): Remove CLWB.

From-SVN: r258587

6 years agore PR tree-optimization/84841 (ICE: tree check: expected ssa_name, have real_cst...
Jakub Jelinek [Fri, 16 Mar 2018 08:06:41 +0000 (09:06 +0100)]
re PR tree-optimization/84841 (ICE: tree check: expected ssa_name, have real_cst in rewrite_expr_tree_parallel, at tree-ssa-reassoc.c:4624)

PR tree-optimization/84841
* tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1 << 3.
(FLOAT_ONE_CONST_TYPE): Define.
(constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
(sort_by_operand_rank): Put entries with higher constant_type last
rather than first to match comments.

* gcc.dg/pr84841.c: New test.

From-SVN: r258586

6 years agore PR c++/84874 (internal compiler error: in reshape_init_class, at cp/decl.c:5800)
Jakub Jelinek [Fri, 16 Mar 2018 08:05:06 +0000 (09:05 +0100)]
re PR c++/84874 (internal compiler error: in reshape_init_class, at cp/decl.c:5800)

PR c++/84874
* decl.c (reshape_init_class): Don't assert d->cur->index == field
if d->cur->index is a FIELD_DECL, instead set field to d->cur->index.

* g++.dg/cpp2a/desig7.C: New test.

From-SVN: r258585

6 years ago03-16-2018 Mark Doffman <mark.doffman@codethink.co.uk>
Mark Doffman [Fri, 16 Mar 2018 05:07:39 +0000 (23:07 -0600)]
03-16-2018  Mark Doffman  <mark.doffman@codethink.co.uk>
            Jim MacArthur  <jim.macarthur@codethink.co.uk>

* gfortran.dg/automatic_1.f90: New test.
* gfortran.dg/automatic_repeat.f90: New test
* gfortran.dg/automatic_save.f90: New test.
* gfortran.dg/vax_structure.f90: New test.

From-SVN: r258584

6 years agore PR fortran/78741 (ICE in gfc_get_symbol_decl, at fortran/trans-decl.c:1534)
Steven G. Kargl [Fri, 16 Mar 2018 02:56:34 +0000 (02:56 +0000)]
re PR fortran/78741 (ICE in gfc_get_symbol_decl, at fortran/trans-decl.c:1534)

2018-03-15  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/78741
* decl.c (get_proc_name):  Check for clash of entry name with
subroutine name.

2018-03-15  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/78741
* gfortran.dg/pr78741.f90: New test.

From-SVN: r258581

6 years agore PR fortran/69395 (ICE on declaring array with more than 7 dimensions+codimensions)
Steven G. Kargl [Fri, 16 Mar 2018 02:43:02 +0000 (02:43 +0000)]
re PR fortran/69395 (ICE on declaring array with more than 7 dimensions+codimensions)

2018-03-15  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/69395
* decl.c (merge_array_spec): Limit the merging to maximum allowed
dimensions, and issue error message if limit is exceeded.

2018-03-15  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/69395
* gfortran.dg/pr69395.f90: New test.

From-SVN: r258580

6 years agoDaily bump.
GCC Administrator [Fri, 16 Mar 2018 00:16:13 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258579

6 years agonios2.md (movsi_internal): Fix thinko in split predicate.
Sandra Loosemore [Thu, 15 Mar 2018 22:33:34 +0000 (18:33 -0400)]
nios2.md (movsi_internal): Fix thinko in split predicate.

2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>

gcc/
* config/nios2/nios2.md (movsi_internal): Fix thinko in
split predicate.

From-SVN: r258575

6 years agore PR c++/79085 (ICE with placement new to unaligned location)
Jakub Jelinek [Thu, 15 Mar 2018 21:09:24 +0000 (22:09 +0100)]
re PR c++/79085 (ICE with placement new to unaligned location)

PR c++/79085
* calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
check and use address of target always.

* g++.dg/opt/pr79085.C: New test.

From-SVN: r258574

6 years agore PR libfortran/84880 (libgfortran fail to build on aarch64/arm 32bit cross toolchain)
Jakub Jelinek [Thu, 15 Mar 2018 18:04:59 +0000 (19:04 +0100)]
re PR libfortran/84880 (libgfortran fail to build on aarch64/arm 32bit cross toolchain)

PR libgfortran/84880
* intrinsics/kill.c (kill): Rename to...
(PREFIX (kill)): ... this.  Use export_proto_np instead of export_proto.

From-SVN: r258571