platform/upstream/gcc.git
6 years agoi386.c (emit_i387_cw_initialization): Always use logic instructions when changing...
Uros Bizjak [Tue, 3 Apr 2018 14:59:21 +0000 (16:59 +0200)]
i386.c (emit_i387_cw_initialization): Always use logic instructions when changing rounding bits to preserve precision...

* config/i386/i386.c (emit_i387_cw_initialization): Always use logic
instructions when changing rounding bits to preserve precision bits
in the x87 control word.

From-SVN: r259034

6 years ago[ARM,testsuite] Force hard-float on armv8_2-fp16-move-1.c
Christophe Lyon [Tue, 3 Apr 2018 14:01:52 +0000 (14:01 +0000)]
[ARM,testsuite] Force hard-float on armv8_2-fp16-move-1.c

2018-04-03  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/arm/armv8_2-fp16-move-1.c: Move -mfloat-abi=hard to
dg-additional-options.

From-SVN: r259031

6 years agoRemove UBSAN in dwarf2out.c (PR tree-optimization/82491).
Martin Liska [Tue, 3 Apr 2018 13:43:22 +0000 (15:43 +0200)]
Remove UBSAN in dwarf2out.c (PR tree-optimization/82491).

2018-04-03  Martin Liska  <mliska@suse.cz>

PR tree-optimization/82491
* rtl.h (strip_offset_and_add): Replace += suboffset with
poly_uint64 () + suboffset.

From-SVN: r259030

6 years agoBits propagation only for int and ptr types
Martin Liska [Tue, 3 Apr 2018 13:27:26 +0000 (15:27 +0200)]
Bits propagation only for int and ptr types

2018-03-29  Martin Liska  <mliska@suse.cz>
    Martin Jambor  <mjambor@suse.cz>

PR ipa/84947
* ipa-cp.c (propagate_bits_across_jump_function): Bail out if
param_type is not an integral or pointer type.

Co-Authored-By: Martin Jambor <mjambor@suse.cz>
From-SVN: r259029

6 years agors6000: Fix testcase pr82015.c
Segher Boessenkool [Tue, 3 Apr 2018 12:21:12 +0000 (14:21 +0200)]
rs6000: Fix testcase pr82015.c

It used "vector" but that is not defined.  Let's use __vector instead.

gcc/testsuite/
* gcc.target/powerpc/pr82015.c: Use __vector instead of vector.

From-SVN: r259027

6 years agosese.h (recompute_all_dominators): Remove.
Richard Biener [Tue, 3 Apr 2018 12:06:27 +0000 (12:06 +0000)]
sese.h (recompute_all_dominators): Remove.

2018-04-03  Richard Biener  <rguenther@suse.de>

* sese.h (recompute_all_dominators): Remove.

From-SVN: r259026

6 years agore PR testsuite/85154 (gcc.dg/vect/vect-95.c fail)
Richard Biener [Tue, 3 Apr 2018 09:54:22 +0000 (09:54 +0000)]
re PR testsuite/85154 (gcc.dg/vect/vect-95.c fail)

2018-04-03  Richard Biener  <rguenther@suse.de>

PR testsuite/85154
* gcc.dg/vect/vect-95.c: Remove scan for alignment peeling.

From-SVN: r259025

6 years agoDaily bump.
GCC Administrator [Tue, 3 Apr 2018 00:16:26 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r259024

6 years agoinvoke.texi (-Wrestrict): Fix typos.
Martin Sebor [Mon, 2 Apr 2018 23:36:21 +0000 (23:36 +0000)]
invoke.texi (-Wrestrict): Fix typos.

gcc/ChangeLog:

* doc/invoke.texi (-Wrestrict): Fix typos.

From-SVN: r259020

6 years agoRISC-V: Fix for combine bug with shift and AND operations.
Jim Wilson [Mon, 2 Apr 2018 22:37:21 +0000 (22:37 +0000)]
RISC-V: Fix for combine bug with shift and AND operations.

PR rtl-optimization/84660
gcc/
* config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
* config/riscv/riscv.md (<optab>si3): Use QImode shift count.
(<optab>di3, <optab>si3_extend): Likewise.
(<optab>si3_mask, <optab>si3_mask_1): New.
(<optab>di3_mask, <optab>di3_mask_1): New.
(<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
(lshrsi3_zero_extend_1): Use VOIDmode shift count.
* config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
gcc/testsuite/
* gcc.target/riscv/pr84660.c: New.
* gcc.target/riscv/shift-and-1.c: New.
* gcc.target/riscv/shift-and-2.c: New.

From-SVN: r259019

6 years agoPR c++/64095 - auto... parameter pack.
Jason Merrill [Mon, 2 Apr 2018 18:20:04 +0000 (14:20 -0400)]
PR c++/64095 - auto... parameter pack.

* parser.c (cp_parser_parameter_declaration): Handle turning autos
into packs here.
(cp_parser_parameter_declaration_list): Not here.

From-SVN: r259015

6 years agore PR fortran/85102 (ICE in gfc_conv_intrinsic_dot_product, at fortran/trans-intrinsi...
Thomas Koenig [Mon, 2 Apr 2018 16:47:48 +0000 (16:47 +0000)]
re PR fortran/85102 (ICE in gfc_conv_intrinsic_dot_product, at fortran/trans-intrinsic.c:4464)

2018-04-02  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/85102
* decl.c (variable_decl): If upper or lower bounds simplify
to a constant, use that.

2018-04-02  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/85102
* gfortran.dg/array_simplify_2.f90: New test.

From-SVN: r259014

6 years agocpp.texi (Variadic Macros): Fix line continuation in an example.
Gerald Pfeifer [Mon, 2 Apr 2018 16:04:06 +0000 (16:04 +0000)]
cpp.texi (Variadic Macros): Fix line continuation in an example.

* doc/cpp.texi (Variadic Macros): Fix line continuation in an
example.

From-SVN: r259012

6 years ago[NDS32] Implement TARGET_CANONICALIZE_COMPARISON for specific cases.
Chung-Ju Wu [Mon, 2 Apr 2018 01:55:53 +0000 (01:55 +0000)]
[NDS32] Implement TARGET_CANONICALIZE_COMPARISON for specific cases.

gcc/
* config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
(nds32_canonicalize_comparison): New function.

From-SVN: r259004

6 years agoMAINTAINERS: Remove redundant Write-After-Approval entry.
Bill Schmidt [Mon, 2 Apr 2018 01:31:19 +0000 (01:31 +0000)]
MAINTAINERS: Remove redundant Write-After-Approval entry.

2018-04-02  Bill Schmidt  <wschmidt@linux.ibm.com>

* MAINTAINERS: Remove redundant Write-After-Approval entry.

From-SVN: r259003

6 years agoDaily bump.
GCC Administrator [Mon, 2 Apr 2018 00:16:23 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r259002

6 years ago[NDS32] Add relax optimization as new pass.
Chung-Ju Wu [Sun, 1 Apr 2018 10:07:40 +0000 (10:07 +0000)]
[NDS32] Add relax optimization as new pass.

gcc/
* config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
* config/nds32/constants.md (unspec_volatile_element): Add
UNSPEC_VOLATILE_RELAX_GROUP.
* config/nds32/nds32-relax-opt.c: New file.
* config/nds32/nds32-predicates.c
(nds32_symbol_load_store_p): New function.
* config/nds32/nds32-protos.h
(nds32_symbol_load_store_p): Declare function.
(make_pass_nds32_relax_opt): Declare new rtl pass function.
* config/nds32/nds32.c
(nds32_register_pass): New function to register pass.
(nds32_register_passes): New function to register passes.
* config/nds32/nds32.md (relax_group): New pattern.
* config/nds32/nds32.opt (mrelax-hint): New option.
* config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.

Co-Authored-By: Kito Cheng <kito.cheng@gmail.com>
Co-Authored-By: Kuan-Lin Chen <kuanlinchentw@gmail.com>
From-SVN: r258998

6 years ago[NDS32] Correct the dependency in t-nds32.
Kito Cheng [Sun, 1 Apr 2018 09:14:12 +0000 (09:14 +0000)]
[NDS32] Correct the dependency in t-nds32.

gcc/
* config/nds32/t-nds32: Modify files dependency.

From-SVN: r258997

6 years ago[NDS32] Implement PROFILE_HOOK.
Chung-Ju Wu [Sun, 1 Apr 2018 08:52:24 +0000 (08:52 +0000)]
[NDS32] Implement PROFILE_HOOK.

gcc/
* config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
(PROFILE_HOOK): Define its implementation.

From-SVN: r258996

6 years ago[NDS32] Change wchar type to unsinged int.
Chung-Ju Wu [Sun, 1 Apr 2018 08:29:40 +0000 (08:29 +0000)]
[NDS32] Change wchar type to unsinged int.

gcc/
* config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
type and 32-bit size.

From-SVN: r258995

6 years agore PR middle-end/85090 (wrong code with -O2 -fno-tree-dominator-opts -mavx512f -fira...
Jakub Jelinek [Sun, 1 Apr 2018 06:05:01 +0000 (08:05 +0200)]
re PR middle-end/85090 (wrong code with -O2 -fno-tree-dominator-opts -mavx512f -fira-algorithm=priority)

PR middle-end/85090
* config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
(V_128_256): New mode iterator.
(*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
(*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
(xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
of V.
* config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.

* gcc.target/i386/avx512f-pr85090-1.c: New test.
* gcc.target/i386/avx512f-pr85090-2.c: New test.
* gcc.target/i386/avx512f-pr85090-3.c: New test.
* gcc.target/i386/avx512bw-pr85090-2.c: New test.
* gcc.target/i386/avx512bw-pr85090-3.c: New test.

From-SVN: r258994

6 years agoDaily bump.
GCC Administrator [Sun, 1 Apr 2018 00:16:12 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258993

6 years ago[PR c++/85027] deal with baselink in save_expr in instantiate_type
Alexandre Oliva [Sat, 31 Mar 2018 03:44:12 +0000 (03:44 +0000)]
[PR c++/85027] deal with baselink in save_expr in instantiate_type

We use SAVE_EXPRs in conditional expressions without the middle
operand, to evaluate the first operand only once.  When the conversion
of the first operand fails, we may call instantiate_type get a better
error message.  We have code to peel off the SAVE_EXPR there, but then
we may end up with a BASELINK, and we're past the code that deals with
BASELINKs.  Reorder the tests so that we expose the saved expr first,
and then deal with BASELINKs.

for  gcc/cp/ChangeLog

PR c++/85027
* class.c (instantiate_type): Peel off SAVE_EXPR before
BASELINK.

for  gcc/testsuite/ChangeLog

PR c++/85027
* g++.dg/pr85027.C: New.

From-SVN: r258989

6 years agors6000: Fix _mm_min_ps and _mm_max_ps (PR83315)
Segher Boessenkool [Sat, 31 Mar 2018 03:05:35 +0000 (05:05 +0200)]
rs6000: Fix _mm_min_ps and _mm_max_ps (PR83315)

This makes _mm_{min,max}_ps work correctly for QNaNs.

PR target/83315
* config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
NaN inputs correctly.

gcc/testsuite/
PR target/83315
* gcc.target/powerpc/sse-maxps-2.c: New test.
* gcc.target/powerpc/sse-minps-2.c: New test.

From-SVN: r258988

6 years agore PR testsuite/80546 (FAIL: gcc.target/powerpc/bool3-p[78].c scan-assembler-not)
Peter Bergner [Sat, 31 Mar 2018 00:52:01 +0000 (19:52 -0500)]
re PR testsuite/80546 (FAIL: gcc.target/powerpc/bool3-p[78].c scan-assembler-not)

PR target/80546
* config/rs6000/vsx.md (??r): New mode attribute.
(*vsx_mov<mode>_64bit): Use it.
(*vsx_mov<mode>_32bit): Likewise.

From-SVN: r258987

6 years agoDaily bump.
GCC Administrator [Sat, 31 Mar 2018 00:16:14 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258986

6 years agoPR tree-optimization/84818 - integer_onep must not be used for i18n
Martin Sebor [Fri, 30 Mar 2018 21:04:06 +0000 (21:04 +0000)]
PR tree-optimization/84818 - integer_onep must not be used for i18n

gcc/ChangeLog:

PR tree-optimization/84818
* builtins.c (check_access): Use warning_n.

From-SVN: r258983

6 years agoFix designated initializer for anonymous union.
Jason Merrill [Fri, 30 Mar 2018 20:08:51 +0000 (16:08 -0400)]
Fix designated initializer for anonymous union.

* typeck2.c (process_init_constructor_record): Use
init_list_type_node for the CONSTRUCTOR around an anonymous union
designated initializer.

From-SVN: r258982

6 years agore PR c++/84791 (ICE with broken OpenMP reduction clause)
Jakub Jelinek [Fri, 30 Mar 2018 18:38:36 +0000 (20:38 +0200)]
re PR c++/84791 (ICE with broken OpenMP reduction clause)

PR c++/84791
* semantics.c (finish_omp_reduction_clause): If
OMP_CLAUSE_REDUCTION_PLACEHOLDER is error_mark_node, return true
even if processing_template_decl.

* g++.dg/gomp/pr84791.C: New test.

From-SVN: r258981

6 years agoMAINTAINERS: Update my email address and disambiguate myself a bit from Will Schmidt.
Bill Schmidt [Fri, 30 Mar 2018 16:17:24 +0000 (16:17 +0000)]
MAINTAINERS: Update my email address and disambiguate myself a bit from Will Schmidt.

2018-03-30  Bill Schmidt  <wschmidt@linux.ibm.com>

* MAINTAINERS: Update my email address and disambiguate myself
a bit from Will Schmidt.

From-SVN: r258980

6 years agore PR fortran/84931 (Expansion of array constructor with constant implied-do-object...
Paul Thomas [Fri, 30 Mar 2018 12:33:49 +0000 (12:33 +0000)]
re PR fortran/84931 (Expansion of array constructor with constant implied-do-object goes sideways)

2018-03-30  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/84931
* simplify.c (gfc_convert_constant): Handle case of array
constructors within an array that has no iterator and improve
the conciseness of this section of code.

2018-03-30  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/84931
* gfortran.dg/array_constructor_53.f90: New test.

From-SVN: r258977

6 years agore PR fortran/85130 (Substrings out of range are not rejected)
Thomas Koenig [Fri, 30 Mar 2018 12:18:30 +0000 (12:18 +0000)]
re PR fortran/85130 (Substrings out of range are not rejected)

2018-03-30  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/85130
* gfortran.dg/substr_6.f90: Remove illegal test for
out-of-bounds substring.

From-SVN: r258976

6 years agore PR target/83822 (trunk/gcc/config/rs6000/rs6000-string.c:970]: (style) Redundant...
Aaron Sawdey [Fri, 30 Mar 2018 12:17:31 +0000 (12:17 +0000)]
re PR target/83822 (trunk/gcc/config/rs6000/rs6000-string.c:970]: (style) Redundant condition)

2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>

PR target/83822
* config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
condition.
* config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
condition.

From-SVN: r258975

6 years agore PR fortran/85111 (ICE in min_max_choose, at fortran/simplify.c:4884 (and others))
Thomas Koenig [Fri, 30 Mar 2018 09:58:35 +0000 (09:58 +0000)]
re PR fortran/85111 (ICE in min_max_choose, at fortran/simplify.c:4884 (and others))

2017-03-30  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/85111
* array.c (gfc_resolve_character_array_constructor): Early
exit for zero-size arrays.
* simplify.c (simplify_transformation_to_array): Exit early
if the result size is zero.
(simplify_minmaxloc_to_array): Likewise.

2017-03-30  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/85111
* gfortran.dg/zero_sized_10.f90: New test.

Add ChangeLog entries for r258973.

From-SVN: r258974

6 years agore PR fortran/85111 (ICE in min_max_choose, at fortran/simplify.c:4884 (and others))
Thomas Koenig [Fri, 30 Mar 2018 09:56:46 +0000 (09:56 +0000)]
re PR fortran/85111 (ICE in min_max_choose, at fortran/simplify.c:4884 (and others))

2017-03-30  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/85111
* array.c (gfc_resolve_character_array_constructor): Early
exit for zero-size arrays.
* simplify.c (simplify_transformation_to_array): Exit early
if the result size is zero.
(simplify_minmaxloc_to_array): Likewise.

2017-03-30  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/85111
* gfortran.dg/zero_sized_10.f90: New test.

From-SVN: r258973

6 years agoEnable tuning options for skylake-avx512.
Julia Koval [Fri, 30 Mar 2018 07:04:55 +0000 (09:04 +0200)]
Enable tuning options for skylake-avx512.

gcc/
PR target/84413
* x86-tune.def (movx, partial_reg_dependency): Enable for
m_SKYLAKE_AVX512.

From-SVN: r258972

6 years agoDaily bump.
GCC Administrator [Fri, 30 Mar 2018 00:16:24 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258970

6 years ago* gcc.pot: Regenerate.
Joseph Myers [Thu, 29 Mar 2018 23:06:47 +0000 (00:06 +0100)]
* gcc.pot: Regenerate.

From-SVN: r258967

6 years agoMore #include suggestions (PR c++/84269)
David Malcolm [Thu, 29 Mar 2018 20:43:48 +0000 (20:43 +0000)]
More #include suggestions (PR c++/84269)

PR c++/84269 reports a number of names in the C and C++ standard
libraries for which we don't yet offer #include fix-it hints.

This patch adds them (up to comment #9).

gcc/c-family/ChangeLog:
PR c++/84269
* known-headers.cc (get_stdlib_header_for_name): Add various names
from <assert.h>, <string.h>, and <memory.h>; add more names from
<stdio.h>.

gcc/cp/ChangeLog:
PR c++/84269
* name-lookup.c (get_std_name_hint): Add names from <memory>,
<tuple>, and <utility>.

gcc/testsuite/ChangeLog:
PR c++/84269
* g++.dg/lookup/missing-std-include-6.C: New test.
* g++.dg/lookup/missing-std-include.C: Add std::pair and
std::tuple tests.
* g++.dg/spellcheck-reswords.C: Expect a hint about <cstring>.
* g++.dg/spellcheck-stdlib.C: Add tests for names in <cstdio>,
<cstring>, <cassert>, and <cstdlib>.

From-SVN: r258966

6 years agoPR c++/85093 - too many template args with pack expansion.
Jason Merrill [Thu, 29 Mar 2018 19:38:41 +0000 (15:38 -0400)]
PR c++/85093 - too many template args with pack expansion.

* pt.c (coerce_template_parms): Keep pack expansion args that will
need to be empty.

From-SVN: r258964

6 years ago* pt.c (build_non_dependent_expr): Propagate expr location.
Jason Merrill [Thu, 29 Mar 2018 19:38:35 +0000 (15:38 -0400)]
* pt.c (build_non_dependent_expr): Propagate expr location.

From-SVN: r258963

6 years agoPR c++/85060 - wrong-code with call to base member in template.
Jason Merrill [Thu, 29 Mar 2018 19:38:29 +0000 (15:38 -0400)]
PR c++/85060 - wrong-code with call to base member in template.

* search.c (any_dependent_bases_p): Check uses_template_parms
rather than processing_template_decl.

From-SVN: r258962

6 years agore PR inline-asm/84985 (ICE in match_reload, at lra-constraints.c:1068)
Vladimir Makarov [Thu, 29 Mar 2018 18:29:12 +0000 (18:29 +0000)]
re PR inline-asm/84985 (ICE in match_reload, at lra-constraints.c:1068)

2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>

PR inline-asm/84985
* lra-constraints.c (process_alt_operands): Move setting
this_alternative_matches below.

2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>

PR inline-asm/84985
* gcc.target/i386/pr84985.c: New.

From-SVN: r258961

6 years agoMore underlining of bad arguments (PR c++/85110)
David Malcolm [Thu, 29 Mar 2018 14:43:01 +0000 (14:43 +0000)]
More underlining of bad arguments (PR c++/85110)

As of r256448, the C++ frontend underlines many bad arguments in its
diagnostics; those where perform_overload_resolution returns a
non-NULL candidate, but there's a failure in convert_like_real.

However, for the case where perform_overload_resolution fails, but
there's a single non-viable candidate, the error is diagnosed by
cp_build_function_call_vec, and that currently doesn't underline
the bad argument:

$ cat test.cc
void callee (int one, const char **two, int three);

void
caller (const char *fmt)
{
  callee (1, fmt, 3);
}

We emit:

$ g++ test.cc
test.cc: In function 'void caller(const char*)':
test.cc:6:20: error: cannot convert 'const char*' to 'const char**' for argument '2' to 'void callee(int, const char**, int)'
   callee (1, fmt, 3);
                    ^

It's going through convert_for_assignment, and
implicitly using input_location.

This patch updates convert_for_assignment for this case, using
an EXPR_LOCATION if there is one, or falling back to input_location
otherwise, underlining the argument in question:

test.cc: In function 'void caller(const char*)':
test.cc:6:14: error: cannot convert 'const char*' to 'const char**' for argument '2' to 'void callee(int, const char**, int)'
   callee (1, fmt, 3);
              ^~~

gcc/cp/ChangeLog:
PR c++/85110
* typeck.c (convert_for_assignment): When complaining due to
conversions for an argument, attempt to use the location of the
argument.

gcc/testsuite/ChangeLog:
PR c++/85110
* g++.dg/diagnostic/param-type-mismatch-2.C: New test.

From-SVN: r258957

6 years agoi386: Enable AVX/AVX512 features only if supported by OSXSAVE
H.J. Lu [Thu, 29 Mar 2018 13:14:06 +0000 (13:14 +0000)]
i386: Enable AVX/AVX512 features only if supported by OSXSAVE

Enable AVX and AVX512 features only if their states are supported by
OSXSAVE.

PR target/85100
* config/i386/cpuinfo.c (XCR_XFEATURE_ENABLED_MASK): New.
(XSTATE_FP): Likewise.
(XSTATE_SSE): Likewise.
(XSTATE_YMM): Likewise.
(XSTATE_OPMASK): Likewise.
(XSTATE_ZMM): Likewise.
(XSTATE_HI_ZMM): Likewise.
(XCR_AVX_ENABLED_MASK): Likewise.
(XCR_AVX512F_ENABLED_MASK): Likewise.
(get_available_features): Enable AVX and AVX512 features only
if their states are supported by OSXSAVE.

From-SVN: r258954

6 years agoDocumentation tweaks.
Martin Liska [Thu, 29 Mar 2018 13:02:23 +0000 (15:02 +0200)]
Documentation tweaks.

2018-03-29  Martin Liska  <mliska@suse.cz>

        PR lto/84995.
        * doc/invoke.texi: Document how LTO works with debug info.
        Describe auto-load support of binutils.  Mention 'x86-64'
        as valid option value of -march option.

From-SVN: r258953

6 years agore PR c++/85108 (Incorrect warning for constexpr lambda with -Wunused-but-set-variable)
Jakub Jelinek [Thu, 29 Mar 2018 11:18:04 +0000 (13:18 +0200)]
re PR c++/85108 (Incorrect warning for constexpr lambda with -Wunused-but-set-variable)

PR c++/85108
* g++.dg/warn/Wunused-var-31.C: New test.

From-SVN: r258952

6 years ago* config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
Jakub Jelinek [Thu, 29 Mar 2018 10:38:51 +0000 (12:38 +0200)]
* config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.

From-SVN: r258951

6 years agore PR c/85094 (-g with any optimization suppresses -Wduplicated-branches)
Jakub Jelinek [Thu, 29 Mar 2018 10:37:58 +0000 (12:37 +0200)]
re PR c/85094 (-g with any optimization suppresses -Wduplicated-branches)

PR c/85094
* fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
OEP_NO_HASH_CHECK for recursive call, to avoid exponential
checking.

* c-c++-common/Wduplicated-branches-14.c: New test.

From-SVN: r258950

6 years agore PR c++/84606 (internal compiler error: Segmentation fault (enclosing_instantiation...
Paolo Carlini [Thu, 29 Mar 2018 08:43:43 +0000 (08:43 +0000)]
re PR c++/84606 (internal compiler error: Segmentation fault (enclosing_instantiation_of()))

2018-03-29  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/84606
* g++.dg/cpp0x/pr84606.C: New.

From-SVN: r258947

6 years agoDaily bump.
GCC Administrator [Thu, 29 Mar 2018 00:16:22 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258946

6 years agore PR target/84912 (__builtin_divde* produce Internal Compiler Error when compiled...
Peter Bergner [Thu, 29 Mar 2018 00:06:31 +0000 (19:06 -0500)]
re PR target/84912 (__builtin_divde* produce Internal Compiler Error when compiled -m32)

PR target/84912
* config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
(RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
* config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
for RS6000_BTM_POWERPC64.
(rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
(rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
* config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
definition.
(DIVDE): Use it.
(DIVDEU): Likewise.

From-SVN: r258943

6 years agorevert: rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
Carl Love [Wed, 28 Mar 2018 23:08:16 +0000 (23:08 +0000)]
revert: rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.

gcc/ChangeLog:

2018-03-20 Carl Love  <cel@us.ibm.com>

Revert
2017-09-27  Carl Love  <cel@us.ibm.com>

* config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
(FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
* config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
fctiw instruction.

gcc/testsuite/ChangeLog:

2018-03-20  Carl Love  <cel@us.ibm.com>

Revert
2017-09-27  Carl Love  <cel@us.ibm.com>

* gcc.target/powerpc/builtin-fctid-fctiw-runnable.c: New test file
for the __builtin_fctid and __builtin_fctiw.

From-SVN: r258942

6 years agoxmmintrin.h (_mm_max_pi16): Use __vector __bool instead of __vector bool.
Bill Schmidt [Wed, 28 Mar 2018 20:40:36 +0000 (20:40 +0000)]
xmmintrin.h (_mm_max_pi16): Use __vector __bool instead of __vector bool.

2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
instead of __vector bool.
(_mm_max_pu8): Likewise.
(_mm_min_pi16): Likewise.

From-SVN: r258936

6 years agore PR fortran/69497 (ICE in gfc_free_namespace, at fortran/symbol.c:3701)
Mikael Morin [Wed, 28 Mar 2018 20:12:55 +0000 (20:12 +0000)]
re PR fortran/69497 (ICE in gfc_free_namespace, at fortran/symbol.c:3701)

PR fortran/69497
* symbol.c (gfc_symbol_done_2): Start freeing namespaces
from the root.
(gfc_free_namespace): Restore assert (revert r258839).

From-SVN: r258935

6 years agoFix typo in last ChangeLog entry
Peter Bergner [Wed, 28 Mar 2018 20:10:18 +0000 (15:10 -0500)]
Fix typo in last ChangeLog entry

From-SVN: r258934

6 years agore PR target/84912 (__builtin_divde* produce Internal Compiler Error when compiled...
Peter Bergner [Wed, 28 Mar 2018 20:09:02 +0000 (15:09 -0500)]
re PR target/84912 (__builtin_divde* produce Internal Compiler Error when compiled -m32)

gcc/
PR target/84912
* config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
(DIVWEUO): Likewise.
(DIVDEO): Likewise.
(DIVDEUO): Likewise.
* config/rs6000/rs6000.c (builtin_function_type): Remove support for
DIVWEUO and DIVDEUO.
* config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
(UNSPEC_DIV_EXTEND): Remove deleted unspecs.
(div_extend): Likewise.
* doc/extend.texi (__builtin_divweo): Remove documention for deleted
builtin function.
(__builtin_divweuo): Likewise.
(__builtin_divdeo): Likewise.
(__builtin_divdeuo): Likewise.

gcc/testsuite/
PR target/84912
* gcc.target/powerpc/extend-divide-1.c (div_weo): Remove test for
deleted builtin function.
(div_weuo): Likewise.
* gcc.target/powerpc/extend-divide-2.c (div_deo): Likewise.
(div_deuo): Likewise.

From-SVN: r258933

6 years agore PR c++/85028 (ICE on invalid C++ code: in tsubst_default_argument, at cp/pt.c...
Paolo Carlini [Wed, 28 Mar 2018 19:21:41 +0000 (19:21 +0000)]
re PR c++/85028 (ICE on invalid C++ code: in tsubst_default_argument, at cp/pt.c:12340)

/cp
2018-03-28  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/85028
* pt.c (tsubst_default_argument): Early return if the type of the
parameter is erroneous.

/testsuite
2018-03-28  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/85028
* g++.dg/other/default13.C: New.

From-SVN: r258932

6 years agore PR target/85095 (worse code generated)
Jakub Jelinek [Wed, 28 Mar 2018 19:16:50 +0000 (21:16 +0200)]
re PR target/85095 (worse code generated)

PR target/85095
* config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
*sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.

* gcc.target/i386/pr85095-1.c: New test.
* gcc.target/i386/pr85095-2.c: New test.
* gcc.c-torture/execute/pr85095.c: New test.

From-SVN: r258931

6 years agore PR middle-end/82004 (SPEC CPU2017 628.pop2_s miscompare)
Jakub Jelinek [Wed, 28 Mar 2018 19:15:39 +0000 (21:15 +0200)]
re PR middle-end/82004 (SPEC CPU2017 628.pop2_s miscompare)

PR tree-optimization/82004
* gimple-match-head.c (optimize_pow_to_exp): New function.
* match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
Don't fold to exp if optimize_pow_to_exp is false.

* gcc.dg/pr82004.c: New test.

From-SVN: r258930

6 years agoFix typos (PR other/84819).
Martin Liska [Wed, 28 Mar 2018 14:51:09 +0000 (16:51 +0200)]
Fix typos (PR other/84819).

2018-03-28  Martin Liska  <mliska@suse.cz>

PR other/84819
* calls.c (initialize_argument_information): Fix trailing space.
* common.opt: Fix typo and provide better explanation for
-fsanitize-coverage option.
* config/i386/i386.opt: Fix typo.

From-SVN: r258925

6 years agoFix wrong use-after-scope sanitization for omp variable (PR sanitizer/85081).
Jakub Jelinek [Wed, 28 Mar 2018 14:45:21 +0000 (16:45 +0200)]
Fix wrong use-after-scope sanitization for omp variable (PR sanitizer/85081).

2018-03-28  Jakub Jelinek  <jakub@redhat.com>
    Martin Liska  <mliska@suse.cz>

PR sanitizer/85081
* gimplify.c (asan_poison_variable): Don't do the check for
gimplify_omp_ctxp here.
(gimplify_decl_expr): Do it here.
(gimplify_target_expr): Likewise.
2018-03-28  Jakub Jelinek  <jakub@redhat.com>
    Martin Liska  <mliska@suse.cz>

PR sanitizer/85081
* g++.dg/asan/pr85081.C: New test.

Co-Authored-By: Martin Liska <mliska@suse.cz>
From-SVN: r258924

6 years agoDo not call chkp_type_bounds_count if MPX is not enabled (PR target/84988).
Martin Liska [Wed, 28 Mar 2018 14:22:27 +0000 (16:22 +0200)]
Do not call chkp_type_bounds_count if MPX is not enabled (PR target/84988).

2018-03-28  Martin Liska  <mliska@suse.cz>

PR target/84988
* config/i386/i386.c (ix86_function_arg_advance): Do not call
chkp_type_bounds_count if MPX is not enabled.

From-SVN: r258923

6 years ago[NDS32] Adjust BRANCH_COST to prevent too much jump when compile with -Os.
Chung-Ju Wu [Wed, 28 Mar 2018 13:26:42 +0000 (13:26 +0000)]
[NDS32] Adjust BRANCH_COST to prevent too much jump when compile with -Os.

gcc/
* config/nds32/nds32.h (BRANCH_COST): Adjust cost.

From-SVN: r258921

6 years agogfortran.h (gfc_dt): Rename default_exp field to dec_ext.
Jakub Jelinek [Wed, 28 Mar 2018 11:59:06 +0000 (13:59 +0200)]
gfortran.h (gfc_dt): Rename default_exp field to dec_ext.

* gfortran.h (gfc_dt): Rename default_exp field to dec_ext.
* ioparm.def (IOPARM_dt_default_exp): Rename to ...
(IOPARM_dt_dec_ext): ... this.
* trans-io.c (build_dt): Adjust for default_exp renaming to
dec_ext and IOPARM_dt_default_exp renaming to IOPARM_dt_dec_ext.
* io.c (match_io): Likewise.

* io/io.h (IOPARM_DT_DEFAULT_EXP): Rename to ...
(IOPARM_DT_DEC_EXT): ... this.
* io/list_read.c (parse_real): Adjust for IOPARM_DT_DEFAULT_EXP
renaming to IOPARM_DT_DEC_EXT.
(read_real): Likewise.
* io/read.c (read_f): Likewise.

From-SVN: r258920

6 years ago[PR c++/84789] adjust testcase for -fconcepts
Alexandre Oliva [Wed, 28 Mar 2018 05:05:30 +0000 (05:05 +0000)]
[PR c++/84789] adjust testcase for -fconcepts

When compiling with -fconcepts,
cp_parser_template_declaration_after_export calls
cp_parser_template_introduction and that preparses qualified-ids not
preceded by typename in such a way that, when we get to
cp_parser_parse_and_diagnose_invalid_type_name and then
cp_parser_diagnose_invalid_type_name, the nested name specifier no
longer carries the previous template-dependent context, so we don't
stand a chance to suggest the use of 'typename' any more.  Thus,
tolerate in the testcase the poorer error messages we get.

for  gcc/testsuite/ChangeLog

PR c++/84789
* g++.dg/template/pr84789.C: Adjust for testing with
-fconcepts too.

From-SVN: r258915

6 years ago[PR c++/84973] don't defer output of uninstantiated templates
Alexandre Oliva [Wed, 28 Mar 2018 05:05:14 +0000 (05:05 +0000)]
[PR c++/84973] don't defer output of uninstantiated templates

When an anon struct gets a name through a typedef, we reset its
linkage and that of its members.  Member functions may get vague
linkage, which schedules them for deferred output, but we don't want
to add them to the queue if they're uninstantiated templates,
e.g. because the enclosing function is a template.  They will be added
as needed when the enclosing template is instantiated.

for  gcc/cp/ChangeLog

PR c++/84973
* decl2.c (note_vague_linkage_fn): Don't defer uninstantiated
templates.

for  gcc/testsuite/ChangeLog

PR c++/84973
* g++.dg/template/pr84973.C: New.
* g++.dg/template/pr84973-2.C: New.
* g++.dg/template/pr84973-3.C: New.

From-SVN: r258914

6 years ago[PR c++/84968] reject stmt-exprs in noexcept constexprs
Alexandre Oliva [Wed, 28 Mar 2018 05:04:59 +0000 (05:04 +0000)]
[PR c++/84968] reject stmt-exprs in noexcept constexprs

We reject extended statement-expressions in template parameters, so we
might as well reject them in constant expressions used in noexcept
specifications.

for  gcc/cp/ChangeLog

PR c++/84968
* tree.c (strip_typedefs_expr): Reject STATEMENT_LISTs.

for  gcc/testsuite/ChangeLog

PR c++/84968
* g++.dg/eh/pr84968.C: New.

From-SVN: r258913

6 years agoDaily bump.
GCC Administrator [Wed, 28 Mar 2018 00:16:22 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258912

6 years agore PR target/84914 (PowerPC complex multiply/divide calls the wrong function when...
Michael Meissner [Tue, 27 Mar 2018 23:14:22 +0000 (23:14 +0000)]
re PR target/84914 (PowerPC complex multiply/divide calls the wrong function when -mabi=ieeelongdouble)

[gcc]
2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/84914
* config/rs6000/rs6000.c (create_complex_muldiv): New helper
function to create the function decl for complex long double
multiply and divide for -mabi=ieeelongdouble.
(init_float128_ieee): Call it.

[gcc/testsuite]
2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>

PR target/84914
* gcc.target/powerpc/mulkc-2.c: New tests to make sure complex
long double multiply/divide uses the correct function.
* gcc.target/powerpc/mulkc-3.c: Likewise.
* gcc.target/powerpc/divkc-2.c: Likewise.
* gcc.target/powerpc/divkc-3.c: Likewise.

From-SVN: r258908

6 years ago* es.po: Update.
Joseph Myers [Tue, 27 Mar 2018 22:31:08 +0000 (23:31 +0100)]
* es.po: Update.

From-SVN: r258906

6 years agore PR c++/85067 (ICE with volatile parameter in defaulted copy-constructor)
Paolo Carlini [Tue, 27 Mar 2018 21:19:25 +0000 (21:19 +0000)]
re PR c++/85067 (ICE with volatile parameter in defaulted copy-constructor)

/cp
2018-03-27  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/85067
* method.c (defaulted_late_check): Partially revert r253321 changes,
do not early return upon error.

/testsuite
2018-03-27  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/85067
* g++.dg/cpp0x/defaulted51.C: New.
* g++.dg/cpp0x/constexpr-68754.C: Adjust.

From-SVN: r258904

6 years agore PR c++/85077 (V[248][SD]F abs not optimized to)
Jakub Jelinek [Tue, 27 Mar 2018 20:00:56 +0000 (22:00 +0200)]
re PR c++/85077 (V[248][SD]F abs not optimized to)

PR c++/85077
* cp-gimplify.c (cp_fold) <case CONSTRUCTOR>: For ctors with vector
type call fold to generate VECTOR_CSTs when possible.

* g++.dg/ext/vector35.C: New test.

From-SVN: r258903

6 years agore PR c++/85061 (ICE with __builtin_offsetof applied to static member)
Jakub Jelinek [Tue, 27 Mar 2018 19:59:30 +0000 (21:59 +0200)]
re PR c++/85061 (ICE with __builtin_offsetof applied to static member)

PR c++/85061
* c-common.c (fold_offsetof_1) <case COMPOUND_EXPR>: Assert that
get_base_address of the second operand is a VAR_P, rather than the
operand itself, and use gcc_checking_assert instead of gcc_assert.

* g++.dg/ext/builtin-offsetof3.C: New test.

From-SVN: r258902

6 years agore PR c++/85076 (ICE with invalid template used as lambda argument)
Jakub Jelinek [Tue, 27 Mar 2018 19:58:30 +0000 (21:58 +0200)]
re PR c++/85076 (ICE with invalid template used as lambda argument)

PR c++/85076
* tree.c (cp_build_reference_type): If to_type is error_mark_node,
return it right away.

* g++.dg/cpp1y/pr85076.C: New test.

From-SVN: r258901

6 years agore PR fortran/85084 (ICE: out of memory allocating 18446744073709551600 bytes ...)
Thomas Koenig [Tue, 27 Mar 2018 18:42:02 +0000 (18:42 +0000)]
re PR fortran/85084 (ICE: out of memory allocating 18446744073709551600 bytes ...)

2018-03-27  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/85084
* frontend-passes.c (gfc_run_passes): Do not run front-end
optimizations if a previous error occurred.

2018-03-27  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/85084
* gfortran.dg/matmul_rank_1.f90: New test.

From-SVN: r258900

6 years agore PR fortran/85083 (ICE in gfc_convert_to_structure_constructor, at fortran/primary...
Thomas Koenig [Tue, 27 Mar 2018 17:28:35 +0000 (17:28 +0000)]
re PR fortran/85083 (ICE in gfc_convert_to_structure_constructor, at fortran/primary.c:2915)

2018-03-27  Thomas Koenig  <tkoenig@gcc.gnu.org>
Harald Anlauf  <anlauf@gmx.de>

PR fortran/85083
* primary.c (gfc_convert_to_structure_constructor): Check
conformance of argument types in structure constructor.

2018-03-27  Thomas Koenig  <tkoenig@gcc.gnu.org>
Harald Anlauf  <anlauf@gmx.de>

* gfortran.dg/structure_constructor_15.f90: New test.

Co-Authored-By: Harald Anlauf <anlauf@gmx.de>
From-SVN: r258899

6 years agosearch.c (check_final_overrider): Use inform instead of error for the diagnostics...
Volker Reichelt [Tue, 27 Mar 2018 17:21:20 +0000 (17:21 +0000)]
search.c (check_final_overrider): Use inform instead of error for the diagnostics of the overridden functions.

2018-03-27  Volker Reichelt  <v.reichelt@netcologne.de>

        * search.c (check_final_overrider): Use inform instead of error
        for the diagnostics of the overridden functions.  Tweak wording.

        * g++.dg/cpp0x/defaulted2.C: Use dg-message instead of dg-error
        for the diagnostics of overridden functions.  Adjust for new wording.
        * g++.dg/cpp0x/implicit1.C: Likewise.
        * g++.dg/cpp0x/override1.C: Likewise.
        * g++.dg/cpp1y/auto-fn18.C: Likewise.
        * g++.dg/eh/shadow1.C: Likewise.
        * g++.dg/inherit/covariant12.C: Likewise.
        * g++.dg/inherit/covariant14.C: Likewise.
        * g++.dg/inherit/covariant15.C: Likewise.
        * g++.dg/inherit/covariant16.C: Likewise.
        * g++.dg/inherit/covariant22.C: Likewise.
        * g++.dg/inherit/crash3.C: Likewise.
        * g++.dg/inherit/error2.C: Likewise.
        * g++.dg/template/crash100.C: Likewise.
        * g++.old-deja/g++.eh/spec6.C: Likewise.
        * g++.old-deja/g++.mike/p811.C: Likewise.
        * g++.old-deja/g++.other/virtual11.C: Likewise.
        * g++.old-deja/g++.other/virtual4.C: Likewise.

From-SVN: r258898

6 years agoi386: Insert ENDBR to trampoline for -fcf-protection=branch -mibt
H.J. Lu [Tue, 27 Mar 2018 17:18:51 +0000 (17:18 +0000)]
i386: Insert ENDBR to trampoline for -fcf-protection=branch -mibt

When -fcf-protection=branch -mibt are used, we need to insert ENDBR
to trampoline.  TRAMPOLINE_SIZE is creased by 4 bytes to accommodate
4-byte ENDBR instruction.

gcc/

PR target/85044
* config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
-fcf-protection=branch -mibt.
* config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.

gcc/testsuite/

PR target/85044
* gcc.target/i386/pr85044.c: New test.

From-SVN: r258897

6 years agoPR testsuite/83462 - c-c++-common/Warray-bounds-3.c fails on arm-none-eabi
Martin Sebor [Tue, 27 Mar 2018 17:08:41 +0000 (17:08 +0000)]
PR testsuite/83462 - c-c++-common/Warray-bounds-3.c fails on arm-none-eabi

gcc/testsuite/ChangeLog:
* c-c++-common/Warray-bounds-4.c: Disable assertion for targets
other than x86.

From-SVN: r258896

6 years ago[AArch64] XFAIL gcc.target/aarch64/store_v2vec_lanes.c for ILP32
Kyrylo Tkachov [Tue, 27 Mar 2018 16:52:10 +0000 (16:52 +0000)]
[AArch64] XFAIL gcc.target/aarch64/store_v2vec_lanes.c for ILP32

The test in question fails for ilp32. The initial analysis I did in the PR for it
is that for ILP32 we generate somewhat different address forms that we'd need to adjust aarch64_classify_address to catch.
Given the optimisation this test checks for was added for GCC 8 it is not a regression, and improving the codegen on ILP32
would be an enhancement rather than a fix. So Richi has asked for it to be marked as XFAIL on ILP32, which is what this
patch does.
Checked that the test still passes on LP64 and appears as XFAIL on -mabi=ilp32.

PR target/83009
* gcc.target/aarch64/store_v2vec_lanes.c: XFAIL for ilp32.

From-SVN: r258894

6 years ago[Patch ARM] Fix PR target/81863
Ramana Radhakrishnan [Tue, 27 Mar 2018 14:06:20 +0000 (14:06 +0000)]
[Patch ARM] Fix PR target/81863

This has been in my patch stack for quite some time. The problem here
was that we weren't handling arm_word_relocations in
arm_valid_symbolic_address and is the surest fix for this
for GCC8 and GCC7.

Regression tested on arm-none-linux-gnueabihf . Applying to
trunk and backporting to GCC-7 in a day or so.

regards
Ramana

2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>

PR target/81863
* config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations

2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>

PR target/81863
* gcc.target/arm/pr81863.c: New test.

From-SVN: r258886

6 years agore PR target/85056 ([nvptx] wrong declaration of external arrays)
Cesar Philippidis [Tue, 27 Mar 2018 13:54:03 +0000 (06:54 -0700)]
re PR target/85056 ([nvptx] wrong declaration of external arrays)

PR target/85056

gcc/
* config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
extern array declarations.

gcc/testsuite/
* testsuite/gcc.target/nvptx/pr85056.c: New test.
* testsuite/gcc.target/nvptx/pr85056a.c: New test.

From-SVN: r258885

6 years agore PR testsuite/82847 (gcc.dg/vect/slp-perm-9.c fail)
Richard Biener [Tue, 27 Mar 2018 13:26:04 +0000 (13:26 +0000)]
re PR testsuite/82847 (gcc.dg/vect/slp-perm-9.c fail)

2018-03-27  Richard Biener  <rguenther@suse.de>

PR testsuite/82847
* lib/target-supports.exp (check_ssse3_available): New function.
(check_effective_target_vect_perm_short): Enable for x86 if
check_ssse3_available.

From-SVN: r258882

6 years agore PR middle-end/84067 (gcc.dg/wmul-1.c regression on aarch64 after r257077)
Richard Biener [Tue, 27 Mar 2018 13:23:15 +0000 (13:23 +0000)]
re PR middle-end/84067 (gcc.dg/wmul-1.c regression on aarch64 after r257077)

2018-03-27  Richard Biener  <rguenther@suse.de>

PR middle-ed/84067
* match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
explicit single_use checks.

From-SVN: r258881

6 years agore PR tree-optimization/85082 (ICE with -Ofast in vn_reference_insert_pieces, at...
Richard Biener [Tue, 27 Mar 2018 12:13:55 +0000 (12:13 +0000)]
re PR tree-optimization/85082 (ICE with -Ofast in vn_reference_insert_pieces, at tree-ssa-sccvn.c:2624)

2018-03-27  Richard Biener  <rguenther@suse.de>

PR tree-optimization/85082
* tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
Valueize the VUSE.

* gfortran.dg/pr85082.f90: New testcase.

From-SVN: r258877

6 years agoumips-stroe16-2.c: Change "length = 2" to "l=2" in dg-final.
Chenghua Xu [Tue, 27 Mar 2018 11:43:43 +0000 (11:43 +0000)]
umips-stroe16-2.c: Change "length = 2" to "l=2" in dg-final.

2018-03-27  Chenghua Xu  <paul.hua.gm@gmail.com>

    * gcc.target/mips/umips-stroe16-2.c: Change "length = 2"
    to "l=2" in dg-final.

From-SVN: r258876

6 years agofix-r4000-1.c: Delete "[^\n]+" in dg-final.
Chenghua Xu [Tue, 27 Mar 2018 11:35:07 +0000 (11:35 +0000)]
fix-r4000-1.c: Delete "[^\n]+" in dg-final.

2018-03-27  Chenghua Xu  <paul.hua.gm@gmail.com>

    * gcc.target/mips/fix-r4000-1.c: Delete "[^\n]+" in dg-final.
    * gcc.target/mips/fix-r4000-2.c: Likewise.
    * gcc.target/mips/fix-r4000-3.c: Likewise.
    * gcc.target/mips/fix-r4000-4.c: Likewise.
    * gcc.target/mips/fix-r4000-5.c: Likewise.
    * gcc.target/mips/fix-r4000-6.c: Likewise.
    * gcc.target/mips/fix-r4000-7.c: Likewise.
    * gcc.target/mips/fix-r4000-8.c: Likewise.
    * gcc.target/mips/fix-r4000-9.c: Likewise.
    * gcc.target/mips/fix-r4000-10.c: Likewise.
    * gcc.target/mips/fix-r4000-7.c: Change dg-final
    "mulditi3_r4000" instead of "mulditi3".
    * gcc.target/mips/fix-r4000-8.c: Change dg-final
    "umulditi3_r4000" instead of "umulditi3".

From-SVN: r258875

6 years agore PR c++/85068 (ICE with invalid covariant return type hierarchy)
Jakub Jelinek [Tue, 27 Mar 2018 11:02:08 +0000 (13:02 +0200)]
re PR c++/85068 (ICE with invalid covariant return type hierarchy)

PR c++/85068
* class.c (update_vtable_entry_for_fn): Don't ICE if base_binfo
is NULL.  Assert if thunk_binfo is NULL then errorcount is non-zero.

* g++.dg/inherit/covariant22.C: New test.

From-SVN: r258873

6 years agore PR testsuite/84004 (gcc.dg/vect/vect-95.c XPASSes)
Richard Biener [Tue, 27 Mar 2018 10:01:46 +0000 (10:01 +0000)]
re PR testsuite/84004 (gcc.dg/vect/vect-95.c XPASSes)

2018-03-27  Richard Biener  <rguenther@suse.de>

PR testsuite/84004
* gcc.dg/vect/vect-95.c: Never expect the loop to be peeled for
alignment.

From-SVN: r258872

6 years ago[Patch AArch64] Turn on -fasynchronous-unwind-tables and -funwind-tables by default
Ramana Radhakrishnan [Tue, 27 Mar 2018 09:26:16 +0000 (09:26 +0000)]
[Patch AArch64] Turn on -fasynchronous-unwind-tables and -funwind-tables by default

2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>

* config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
* common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
Turn on fasynchronous-unwind-tables and funwind-tables.

From-SVN: r258871

6 years agore PR c++/84632 (internal compiler error: tree check: expected record_type or union_t...
Paolo Carlini [Tue, 27 Mar 2018 06:52:04 +0000 (06:52 +0000)]
re PR c++/84632 (internal compiler error: tree check: expected record_type or union_type or qual_union_type, have array_type in reduced_constant_expression_p, at cp/constexpr.c:1778)

/cp
2018-03-27  Paolo Carlini  <paolo.carlini@oracle.com>
    Jason Merrill  <jason@redhat.com>

PR c++/84632
* init.c (build_aggr_init): When initializing from array,
reject anything but CONSTRUCTORs and TARGET_EXPRs.
(build_vec_init): Handle separately ARRAY_TYPEs.

/testsuite
2018-03-27  Paolo Carlini  <paolo.carlini@oracle.com>
    Jason Merrill  <jason@redhat.com>

PR c++/84632
* g++.dg/init/array49.C: New.
* g++.dg/torture/pr70499.C: Adjust.

Co-Authored-By: Jason Merrill <jason@redhat.com>
From-SVN: r258870

6 years agoDaily bump.
GCC Administrator [Tue, 27 Mar 2018 00:16:18 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r258869

6 years ago* es.po, sv.po: Update.
Joseph Myers [Mon, 26 Mar 2018 22:02:59 +0000 (23:02 +0100)]
* es.po, sv.po: Update.

From-SVN: r258866

6 years agoruntime: don't check for stale runtime
Ian Lance Taylor [Mon, 26 Mar 2018 19:29:27 +0000 (19:29 +0000)]
runtime: don't check for stale runtime

    The gccgo runtime is never stale, and on a system with gc sources in
    ~/go the test may wind up checking whether the gc runtime is stale.

    Reviewed-on: https://go-review.googlesource.com/102282

From-SVN: r258865

6 years agore PR target/85073 ([x86] extra check after BLSR)
Uros Bizjak [Mon, 26 Mar 2018 19:26:19 +0000 (21:26 +0200)]
re PR target/85073 ([x86] extra check after BLSR)

PR target/85073
* config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
(*bmi_blsr_<mode>_ccz): Ditto.

testsuite/ChangeLog:

PR target/85073
* gcc.target/i386/pr85073.c: New test.

From-SVN: r258864

6 years agoPR c++/85062 - ICE with alignas in wrong place.
Jason Merrill [Mon, 26 Mar 2018 16:30:41 +0000 (12:30 -0400)]
PR c++/85062 - ICE with alignas in wrong place.

* decl.c (grokdeclarator): Ignore attributes on type-specifiers
here.

From-SVN: r258859

6 years agoPR c++/85049 - ICE with __integer_pack.
Jason Merrill [Mon, 26 Mar 2018 14:37:50 +0000 (10:37 -0400)]
PR c++/85049 - ICE with __integer_pack.

* pt.c (unify_pack_expansion): Don't try to deduce generated packs.
* cp-tree.h (TEMPLATE_PARM_P): New.

From-SVN: r258856

6 years agoAdd workaround to std::variant for Clang bug 31852
Jonathan Wakely [Mon, 26 Mar 2018 13:09:10 +0000 (14:09 +0100)]
Add workaround to std::variant for Clang bug 31852

* include/std/variant (__get): Qualify calls to avoid ADL.
(__select_index): Adjust whitespace.
(variant): Add using-declaration to workaround Clang bug.

From-SVN: r258854

6 years agoFix switch conversion in offloading functions
Tom de Vries [Mon, 26 Mar 2018 09:45:49 +0000 (09:45 +0000)]
Fix switch conversion in offloading functions

2018-03-26  Tom de Vries  <tom@codesourcery.com>

PR tree-optimization/85063
* omp-general.c (offloading_function_p): New function.  Factor out
of ...
* omp-offload.c (pass_omp_target_link::gate): ... here.
* omp-general.h (offloading_function_p): Declare.
* tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
with attribute omp declare target for offloading functions.

* testsuite/libgomp.c/switch-conversion-2.c: New test.
* testsuite/libgomp.c/switch-conversion.c: New test.
* testsuite/libgomp.oacc-c-c++-common/switch-conversion-2.c: New test.
* testsuite/libgomp.oacc-c-c++-common/switch-conversion.c: New test.

From-SVN: r258852