gcc.git
5 years agors6000: Add -mdejagnu-cpu=
Segher Boessenkool [Fri, 1 Mar 2019 17:44:15 +0000 (18:44 +0100)]
rs6000: Add -mdejagnu-cpu=

This adds an option -mdejagnu-cpu=.  This option simply overrides what
is given in -mcpu=.  The reason for this is that with older versions
of DejaGnu the value given in the RUNTESTFLAGS will override the value
a testcase wants to have.

* config/rs6000/rs6000.c (rs6000_option_override_internal): If
rs6000_dejagnu_cpu_index is set, use that to override rs6000_cpu_index.
* config/rs6000/rs6000.opt (mdejagnu-cpu=): New option.

From-SVN: r269322

5 years agohaifa-sched: handle fallthru edge to EXIT block (PR 85899)
Alexander Monakov [Fri, 1 Mar 2019 16:18:04 +0000 (19:18 +0300)]
haifa-sched: handle fallthru edge to EXIT block (PR 85899)

PR rtl-optimization/85899
* haifa-sched.c (find_fallthru_edge_from): Relax assert to account for
fallthru edges leading to the exit block.

* gcc.dg/pr85899.c: New test.

From-SVN: r269319

5 years agoPR c++/89537 - missing location for error with non-static member fn.
Marek Polacek [Fri, 1 Mar 2019 15:57:46 +0000 (15:57 +0000)]
PR c++/89537 - missing location for error with non-static member fn.

* call.c (resolve_args): Use EXPR_LOCATION.
* typeck.c (build_class_member_access_expr): Use input_location.

* g++.dg/diagnostic/member-fn-1.C: New test.

From-SVN: r269318

5 years agoPR c++/89532 - ICE with incomplete type in decltype.
Marek Polacek [Fri, 1 Mar 2019 15:55:56 +0000 (15:55 +0000)]
PR c++/89532 - ICE with incomplete type in decltype.

* semantics.c (finish_compound_literal): Return error_mark_node
if digest_init_flags returns error_mark_node.

* g++.dg/cpp2a/nontype-class14.C: New test.

From-SVN: r269317

5 years agoruntime: call execname and getpagesize on Solaris
Ian Lance Taylor [Fri, 1 Mar 2019 14:21:24 +0000 (14:21 +0000)]
runtime: call execname and getpagesize on Solaris

    Interpreting auxv as []uintptr is incorrect on 64-bit big-endian,
    as auxv alternates a 32-bit int with a 64-bit pointer.

    Patch from Rainer Orth.

    Reviewed-on: https://go-review.googlesource.com/c/164739

From-SVN: r269315

5 years agoImplement P1002R1, Try-catch blocks in constexpr functions PR c++/89513
Jakub Jelinek [Fri, 1 Mar 2019 14:20:03 +0000 (15:20 +0100)]
Implement P1002R1, Try-catch blocks in constexpr functions PR c++/89513

Implement P1002R1, Try-catch blocks in constexpr functions
PR c++/89513
* parser.c (cp_parser_ctor_initializer_opt_and_function_body):
Diagnose constexpr ctor or function with function-try-block with
pedwarn for c++17 and earlier.  Formatting fix.
(cp_parser_try_block): Use pedwarn instead of error and only for
c++17 and earlier when try block appears in constexpr function.
* constexpr.c (build_constexpr_constructor_member_initializers):
Handle TRY_BLOCK here instead of erroring on it.

* g++.dg/cpp2a/constexpr-try1.C: New test.
* g++.dg/cpp2a/constexpr-try2.C: New test.
* g++.dg/cpp2a/constexpr-try3.C: New test.
* g++.dg/cpp2a/constexpr-try4.C: New test.
* g++.dg/cpp2a/constexpr-try5.C: New test.
* g++.dg/cpp0x/constexpr-ctor10.C: Don't expect error for C++2a.

From-SVN: r269314

5 years agoFix test memory_resource to work without sized deallocation
Jonathan Wakely [Fri, 1 Mar 2019 13:50:41 +0000 (13:50 +0000)]
Fix test memory_resource to work without sized deallocation

The checking memory_resource in the testsuite assumes sized deallocation
is supported, which might not be true for other compilers.

* testsuite/util/testsuite_allocator.h (__gnu_test::memory_resource)
[!__cpp_sized_deallocation]: Do not pass size to operator delete.

From-SVN: r269312

5 years agoC++2a Utility functions to implement uses-allocator construction (P0591R4)
Jonathan Wakely [Fri, 1 Mar 2019 13:50:36 +0000 (13:50 +0000)]
C++2a Utility functions to implement uses-allocator construction (P0591R4)

* include/std/memory (uses_allocator_construction_args): New set of
overloaded functions.
(make_obj_using_allocator, uninitialized_construct_using_allocator):
New functions.
* include/std/memory_resource (polymorphic_allocator::construct)
[__cplusplus > 201703l]: Replace all overloads with a single function
using uses_allocator_construction_args.
* testsuite/20_util/polymorphic_allocator/construct_c++2a.cc: New
test.
* testsuite/20_util/uses_allocator/make_obj.cc: New test.

From-SVN: r269311

5 years agoAArch64: Make every option in options.def one line
Tamar Christina [Fri, 1 Mar 2019 13:34:14 +0000 (13:34 +0000)]
AArch64: Make every option in options.def one line

Due to config.gcc all the options need to be on one line because of the grep
lines which would select only the first line of the option.

This causes it not to select the right bits on options that are spread over
multiple lines when the --with-arch configure option is used.  The issue happens
silently and you just get a compiler with an incorrect set of default flags.

This solution just collapses everything back to one line as they were in GCC7.
Unfortunately this does make some lines quite long.

I do have an alternate patch which used the pre-processors to first flatten the
file in config.gcc.  I will send that one out for GCC 10.

gcc/ChangeLog:

PR target/89517
* config/aarch64/aarch64-option-extensions.def (fp, simd, crypto, fp16,
rdma, dotprod, sha2, sha3, sm4, fp16fml, sve): Collapse line.

From-SVN: r269309

5 years agoFix mask type choice in vectorizable_call (PR 89535)
Richard Sandiford [Fri, 1 Mar 2019 13:05:40 +0000 (13:05 +0000)]
Fix mask type choice in vectorizable_call (PR 89535)

This is another case in which we were failing to pass the expected
mask vector type to vect_get_vec_def_for_operand.

2019-02-28  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
PR tree-optimization/89535
* tree-vect-stmts.c (vectorizable_call): Record the vector types
for each operand.  Calculate the fallback choice for mask operands
and pass it to vect_get_vec_def_for_operand.

gcc/testsuite/
PR tree-optimization/89535
* gfortran.dg/vect/pr89535.f90: New test.

From-SVN: r269308

5 years agore PR tree-optimization/89541 (ICE in VN_INFO(tree_node*))
Richard Biener [Fri, 1 Mar 2019 12:52:19 +0000 (12:52 +0000)]
re PR tree-optimization/89541 (ICE in VN_INFO(tree_node*))

2019-03-01  Richard Biener  <rguenther@suse.de>

PR middle-end/89541
* tree-ssa-operands.c (add_stmt_operand): CONST_DECL may
get virtual operands.
(get_expr_operands): Handle CONST_DECL like other decls.

* gfortran.dg/pr89451.f90: New testcase.

From-SVN: r269307

5 years agod/dmd: Merge dmd upstream ed71446aa
Iain Buclaw [Fri, 1 Mar 2019 10:21:54 +0000 (10:21 +0000)]
d/dmd: Merge dmd upstream ed71446aa

Backports support for extern(C++, "namespace"), which makes the
core.stdcpp package compilable.

Added predefined condition for CppRuntime_Gcc unconditionally, as it is
unlikely that D code will be linking to anything other than libstdc++
when extern(C++) is used.

Reviewed-on: https://github.com/dlang/dmd/pull/9371

gcc/d/ChangeLog:

2019-03-01  Iain Buclaw  <ibuclaw@gdcproject.org>

* d-builtins.cc (d_init_versions): Add CppRuntime_Gcc as predefined
version condition.

From-SVN: r269304

5 years agore PR middle-end/89503 (Checking ICE in 'gcc.dg/warn-strlen-no-nul.c')
Jakub Jelinek [Fri, 1 Mar 2019 09:26:40 +0000 (10:26 +0100)]
re PR middle-end/89503 (Checking ICE in 'gcc.dg/warn-strlen-no-nul.c')

PR middle-end/89503
* fold-const.c (fold_checksum_tree): Ignore TREE_NO_WARNING bit
on DECL_P and EXPR_P.

From-SVN: r269303

5 years agore PR middle-end/89497 (ICE caused by Segmentation Fault when compiling cups 2.2...
Richard Biener [Fri, 1 Mar 2019 09:21:30 +0000 (09:21 +0000)]
re PR middle-end/89497 (ICE caused by Segmentation Fault when compiling cups 2.2.10 with LTO flags enabled)

2019-03-01  Richard Biener  <rguenther@suse.de>

PR middle-end/89497
* tree-cfgcleanup.h (cleanup_tree_cfg): Add SSA update flags
argument, defaulted to zero.
* passes.c (execute_function_todo): Pass down SSA update flags
to cleanup_tree_cfg.
* tree-cfgcleanup.c: Include tree-into-ssa.h and tree-cfgcleanup.h.
(cleanup_tree_cfg_noloop): After cleanup_control_flow_pre update SSA
form if requested.
(cleanup_tree_cfg): Get and pass down SSA update flags.

* gcc.dg/tree-ssa/reassoc-43.c: Avoid false match in regex.
* g++.dg/tree-prof/devirt.C: Scan tracer dump for foldings
that happen now earlier.

From-SVN: r269302

5 years agore PR bootstrap/89539 (gcc fails to build/bootstrap on MACOSX)
Jakub Jelinek [Fri, 1 Mar 2019 08:10:19 +0000 (09:10 +0100)]
re PR bootstrap/89539 (gcc fails to build/bootstrap on MACOSX)

PR bootstrap/89539
* dwarf2out.c (output_comdat_type_unit): Add ATTRIBUTE_UNUSED to
early_lto_debug argument.

From-SVN: r269301

5 years agore PR go/89406 (Go testing leaves many temporary directories in /tmp around)
Ian Lance Taylor [Fri, 1 Mar 2019 05:25:36 +0000 (05:25 +0000)]
re PR go/89406 (Go testing leaves many temporary directories in /tmp around)

PR go/89406
* Makefile.am (GOTOOLS_TEST_TIMEOUT): Double.
* Makefile.in: Regenerate.

From-SVN: r269300

5 years agocmd/go: add -O2 to invocation of gccgo
Ian Lance Taylor [Fri, 1 Mar 2019 01:23:09 +0000 (01:23 +0000)]
cmd/go: add -O2 to invocation of gccgo

    When using the go tool with gccgo, this changes the default
    compilation to use -O2.  The -gccgoflags option can be used to
    override this default.  I think this change better corresponds to what
    people expect when using the go tool.

    Reviewed-on: https://go-review.googlesource.com/c/164378

From-SVN: r269299

5 years agocommit 66ac9466852d11e968f8fd2ad6ffc7386cee49e1
Ian Lance Taylor [Fri, 1 Mar 2019 01:03:54 +0000 (01:03 +0000)]
commit 66ac9466852d11e968f8fd2ad6ffc7386cee49e1
    gotest: avoid using echo inside $()

    The handling of newlines is not portable between bash and ksh.

    Reviewed-on: https://go-review.googlesource.com/c/164597

From-SVN: r269298

5 years agoDaily bump.
GCC Administrator [Fri, 1 Mar 2019 00:16:34 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r269297

5 years ago* g++.dg/cpp1z/fold-lambda3.C: Add test for 88183.
Jason Merrill [Fri, 1 Mar 2019 00:10:28 +0000 (19:10 -0500)]
* g++.dg/cpp1z/fold-lambda3.C: Add test for 88183.

From-SVN: r269294

5 years agoPR c++/88183 - ICE with .* fold-expression.
Jason Merrill [Fri, 1 Mar 2019 00:08:58 +0000 (19:08 -0500)]
PR c++/88183 - ICE with .* fold-expression.

build_m_component_ref can't handle type-dependent operands, so let's use the
default case; tsubst_copy_and_build also uses build_x_binary_op for
substituting a DOTSTAR_EXPR.

* pt.c (fold_expression) [DOTSTAR_EXPR]: Remove special handling.

From-SVN: r269293

5 years agoPR c++/86969 - ICE with constexpr if and recursive generic lambdas.
Jason Merrill [Fri, 1 Mar 2019 00:08:21 +0000 (19:08 -0500)]
PR c++/86969 - ICE with constexpr if and recursive generic lambdas.

* class.c, lambda.c, pt.c: Revert earlier change.
* lambda.c (add_capture): Don't special-case capture of dependent
VLA.

From-SVN: r269292

5 years ago* name-lookup.c (print_binding_level): Print this_entity.
Jason Merrill [Fri, 1 Mar 2019 00:07:29 +0000 (19:07 -0500)]
* name-lookup.c (print_binding_level): Print this_entity.

From-SVN: r269291

5 years agore PR tree-optimization/89536 (wrong code at -Os and above on x86_64-linux-gnu)
Eric Botcazou [Thu, 28 Feb 2019 23:03:01 +0000 (23:03 +0000)]
re PR tree-optimization/89536 (wrong code at -Os and above on x86_64-linux-gnu)

PR tree-optimization/89536
* tree-ssa-dom.c (edge_info::derive_equivalences) <BIT_NOT_EXPR>: Test
only whether bit #0 of the value is 0 instead of the entire value.

From-SVN: r269289

5 years agoPR c++/87068 - missing diagnostic with fallthrough statement.
Marek Polacek [Thu, 28 Feb 2019 22:29:42 +0000 (22:29 +0000)]
PR c++/87068 - missing diagnostic with fallthrough statement.

* gimplify.c (expand_FALLTHROUGH_r): If IFN_FALLTHROUGH was found
at the end of a seq, save its location to walk_stmt_info.
(expand_FALLTHROUGH): Warn if IFN_FALLTHROUGH is at the end of
a switch.

* c-c++-common/Wimplicit-fallthrough-37.c: New test.

From-SVN: r269288

5 years ago[PR72741, PR89433] Repeated use of the Fortran OpenACC 'routine' directive
Thomas Schwinge [Thu, 28 Feb 2019 20:31:36 +0000 (21:31 +0100)]
[PR72741, PR89433] Repeated use of the Fortran OpenACC 'routine' directive

gcc/fortran/
PR fortran/72741
PR fortran/89433
* openmp.c (gfc_match_oacc_routine): Handle repeated use of the
Fortran OpenACC 'routine' directive.
gcc/testsuite/
PR fortran/72741
PR fortran/89433
* gfortran.dg/goacc/routine-multiple-directives-1.f90: New file.
* gfortran.dg/goacc/routine-multiple-directives-2.f90: Likewise.

Co-Authored-By: Cesar Philippidis <cesar@codesourcery.com>
From-SVN: r269287

5 years ago[PR72741] For all Fortran OpenACC 'routine' directive variants check for multiple...
Thomas Schwinge [Thu, 28 Feb 2019 20:31:23 +0000 (21:31 +0100)]
[PR72741] For all Fortran OpenACC 'routine' directive variants check for multiple clauses specifying the level of parallelism

gcc/fortran/
PR fortran/72741
* gfortran.h (enum oacc_routine_lop): Add OACC_ROUTINE_LOP_ERROR.
* openmp.c (gfc_oacc_routine_lop, gfc_match_oacc_routine): Use it.
* trans-decl.c (add_attributes_to_decl): Likewise.
gcc/testsuite/
PR fortran/72741
* gfortran.dg/goacc/routine-multiple-lop-clauses-1.f90: New file.

Co-Authored-By: Cesar Philippidis <cesar@codesourcery.com>
From-SVN: r269286

5 years ago[PR72741, PR89433] Accept intrinsic symbols in Fortran OpenACC 'routine' directives
Thomas Schwinge [Thu, 28 Feb 2019 20:31:01 +0000 (21:31 +0100)]
[PR72741, PR89433] Accept intrinsic symbols in Fortran OpenACC 'routine' directives

gcc/fortran/
PR fortran/72741
PR fortran/89433
* openmp.c (gfc_match_oacc_routine): Accept intrinsic symbols.
gcc/testsuite/
PR fortran/72741
PR fortran/89433
* gfortran.dg/goacc/routine-6.f90: Update
* gfortran.dg/goacc/routine-intrinsic-1.f: New file.
* gfortran.dg/goacc/routine-intrinsic-2.f: Likewise.

Co-Authored-By: Cesar Philippidis <cesar@codesourcery.com>
From-SVN: r269285

5 years agoPR c++/88049 - ICE with undefined destructor and anon namespace.
Jason Merrill [Thu, 28 Feb 2019 17:29:48 +0000 (12:29 -0500)]
PR c++/88049 - ICE with undefined destructor and anon namespace.

A type in an anonymous namespace can never be merged with one from
another translation unit, so a member of such a type is always its own
prevailing decl.

* lto-symtab.c (lto_symtab_prevailing_virtual_decl): Return early
for a type in an anonymous namespace.

From-SVN: r269283

5 years agore PR lto/88585 (ICE in fld_incomplete_type_of, at tree.c:5295)
Jan Hubicka [Thu, 28 Feb 2019 16:45:44 +0000 (17:45 +0100)]
re PR lto/88585 (ICE in fld_incomplete_type_of, at tree.c:5295)

PR lto/88585
* tree.c (find_atomic_core_type): Move ahead in file.
(check_base_type): Correctly compare alignments of atomic types.

From-SVN: r269282

5 years agoi386: Identify Westmere from PCLMUL
H.J. Lu [Thu, 28 Feb 2019 14:24:52 +0000 (14:24 +0000)]
i386: Identify Westmere from PCLMUL

Since AES has been removed from Westmere in GCC 9, we should identify
Westmere from PCLMUL, instead of AES.

PR target/89455
* config/i386/i386.c (get_builtin_code_for_version): Identify
Westmere from PCLMUL, instead of AES.

From-SVN: r269281

5 years agore PR c/89521 (ICE in expand_builtin_int_roundingfn, at builtins.c:2697)
Jakub Jelinek [Thu, 28 Feb 2019 13:49:38 +0000 (14:49 +0100)]
re PR c/89521 (ICE in expand_builtin_int_roundingfn, at builtins.c:2697)

PR c/89521
* gcc.dg/pr89521-1.c: New test.
* gcc.dg/pr89521-2.c: New test.

From-SVN: r269280

5 years agoFix test-case visibility (PR testsuite/89441).
John David Anglin [Thu, 28 Feb 2019 13:17:09 +0000 (13:17 +0000)]
Fix test-case visibility (PR testsuite/89441).

2019-02-28  John David Anglin  <danglin@gcc.gnu.org>

PR testsuite/89441
* g++.dg/ipa/pr89009.C: Update symbol visibility.

From-SVN: r269278

5 years agore PR target/89434 (wrong code with -Og and __builtin_mul_overflow())
Jakub Jelinek [Thu, 28 Feb 2019 13:13:33 +0000 (14:13 +0100)]
re PR target/89434 (wrong code with -Og and __builtin_mul_overflow())

PR target/89434
* config/arm/arm.md (*subsi3_carryin_compare_const): Use
trunc_int_for_mode (-INTVAL (...), SImode), just instead of
-UINTVAL (...).

From-SVN: r269277

5 years agoAArch64: Have empty HWCAPs string ignored during native feature detection
Tamar Christina [Thu, 28 Feb 2019 10:43:41 +0000 (10:43 +0000)]
AArch64: Have empty HWCAPs string ignored during native feature detection

This patch makes the feature detection code for AArch64 GCC not add features
automatically when the feature had no hwcaps string to match against.

This means that -mcpu=native no longer adds feature flags such as +profile.
The behavior wasn't noticed before because at the time +profile was added a bug
was preventing any feature bits from being added by native detections.

The loop has also been changed as Jakub specified in order to avoid a memory
leak that was present in the existing code and to be slightly more efficient.

gcc/ChangeLog:

PR target/88530
* config/aarch64/aarch64-option-extensions.def: Document it.
* config/aarch64/driver-aarch64.c (host_detect_local_cpu): Skip feature
if empty hwcaps.

gcc/testsuite/ChangeLog:

PR target/88530
* gcc.target/aarch64/options_set_10.c: New test.

From-SVN: r269276

5 years agore PR c++/89522 (ICE: trying to capture 'f' in instantiation of generic lambda)
Paolo Carlini [Thu, 28 Feb 2019 10:27:03 +0000 (10:27 +0000)]
re PR c++/89522 (ICE: trying to capture 'f' in instantiation of generic lambda)

2019-02-28  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/89522
* g++.dg/cpp1y/lambda-generic-ice10.C: New.

From-SVN: r269275

5 years agore PR c/89525 (inform messages from -Wbuiltin-declaration-mismatch even with -w)
Jakub Jelinek [Thu, 28 Feb 2019 07:17:49 +0000 (08:17 +0100)]
re PR c/89525 (inform messages from -Wbuiltin-declaration-mismatch even with -w)

PR c/89525
* c-typeck.c (convert_arguments): Call inform_declaration only if
the previous warning_at call returned true.

* gcc.dg/pr89525.c: New test.

From-SVN: r269274

5 years agore PR c/89520 (ICE tree check: accessed operand 4 of call_expr with 3 operands in...
Jakub Jelinek [Thu, 28 Feb 2019 07:16:38 +0000 (08:16 +0100)]
re PR c/89520 (ICE tree check: accessed operand 4 of call_expr with 3 operands in convert_to_integer_1, at convert.c:668)

PR c/89520
* convert.c (convert_to_real_1, convert_to_integer_1): Punt for
builtins if they don't have a single scalar floating point argument.
Formatting fixes.

* gcc.dg/pr89520-1.c: New test.
* gcc.dg/pr89520-2.c: New test.

From-SVN: r269273

5 years agolibgo: fix go_export extraction on Darwin
Ian Lance Taylor [Thu, 28 Feb 2019 01:01:46 +0000 (01:01 +0000)]
libgo: fix go_export extraction on Darwin

    On Darwin, the section name is prefixed with the segment name, __GNU_GO.

    Reviewed-on: https://go-review.googlesource.com/c/151097

From-SVN: r269271

5 years agoDaily bump.
GCC Administrator [Thu, 28 Feb 2019 00:16:16 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r269270

5 years agoPR c++/88857 - ICE with value-initialization of argument in template.
Marek Polacek [Thu, 28 Feb 2019 00:07:06 +0000 (00:07 +0000)]
PR c++/88857 - ICE with value-initialization of argument in template.

* call.c (convert_like_real): Don't call build_value_init in template.

* g++.dg/cpp0x/initlist-value4.C: New test.

From-SVN: r269267

5 years agore PR go/89172 (FAIL: runtime/pprof)
Ian Lance Taylor [Wed, 27 Feb 2019 22:35:10 +0000 (22:35 +0000)]
re PR go/89172 (FAIL: runtime/pprof)

PR go/89172
    internal/cpu, runtime, runtime/pprof: handle function descriptors

    When using PPC64 ELF ABI v1 a function address is not a PC, but is the
    address of a function descriptor.  The first field in the function
    descriptor is the actual PC (see
    http://refspecs.linuxfoundation.org/ELF/ppc64/PPC-elf64abi.html#FUNC-DES).
    The libbacktrace library knows about this, and libgo uses actual PC
    values consistently except for the helper function funcPC that appears
    in both runtime and runtime/pprof.

    This patch fixes funcPC by recording, in the internal/cpu package,
    whether function descriptors are being used.  We have to check for
    function descriptors using a C compiler check, because GCC can be
    configured using --with-abi to select the ELF ABI to use.

    Fixes https://gcc.gnu.org/PR89172

    Reviewed-on: https://go-review.googlesource.com/c/162978

From-SVN: r269266

5 years agoPR c++/86969 - ICE with constexpr if and recursive generic lambdas.
Jason Merrill [Wed, 27 Feb 2019 21:54:25 +0000 (16:54 -0500)]
PR c++/86969 - ICE with constexpr if and recursive generic lambdas.

Here, the problem was that extract_local_specs wasn't seeing that we use
'self' inside the lambda in the else of the inner constexpr if, because we
don't walk into lambda bodies and we didn't capture it in the lambda because
'self' is still dependent.  Marek recently changed process_outer_var_ref to
do more implicit capture in templates; this example shows that we should
always capture non-packs, so that we can continue to not walk into lambda
bodies.  We do walk into lambda bodies for pack expansions, so we can delay
deciding whether we're capturing a single element or the entire pack.

Immediately capturing a VLA means we need to create a dependent VLA capture
type, and not in the context of the lambda op(), since trying to look up the
instantiation of the op() while we're substituting into the capture list
would crash.  So I force TYPE_CONTEXT and the binding level out to the
enclosing function before pushtag, avoid adding a TAG_DEFN, and instead
force the type to be complete in tsubst_lambda_expr.

* semantics.c (process_outer_var_ref): Do capture dependent vars.
* class.c (finish_struct): Only add TAG_DEFN if T is in
current_function_decl.
* lambda.c (vla_capture_type): Force the capture type out into the
lambda's enclosing function.
(add_capture): Pass in the lambda.
* pt.c (tsubst_lambda_expr): complete_type a VLA capture type.

From-SVN: r269265

5 years agore PR rtl-optimization/89490 (char array constant put in string merge section)
Bernd Edlinger [Wed, 27 Feb 2019 20:14:55 +0000 (20:14 +0000)]
re PR rtl-optimization/89490 (char array constant put in string merge section)

2019-02-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR rtl-optimization/89490
        * varasm.c (get_block_for_section): Bail out for mergeable sections.
        (default_use_anchors_for_symbol_p, output_object_block): Assert the
        block section is not mergeable.

From-SVN: r269264

5 years agoPR c++/89511 - ICE with using-declaration and unscoped enumerator.
Marek Polacek [Wed, 27 Feb 2019 19:07:18 +0000 (19:07 +0000)]
PR c++/89511 - ICE with using-declaration and unscoped enumerator.

* parser.c (cp_parser_using_declaration): For an unscoped enum
only use its context if it's not a function declaration.

* g++.dg/cpp0x/using-enum-3.C: New test.

From-SVN: r269262

5 years agore PR target/70341 (cost model for addresses is incorrect, slsr is using reg + reg...
Jakub Jelinek [Wed, 27 Feb 2019 18:11:26 +0000 (19:11 +0100)]
re PR target/70341 (cost model for addresses is incorrect, slsr is using reg + reg + CST for arm)

PR target/70341
* config/aarch64/aarch64.md (casesi): Create the casesi_dispatch
MEM manually here, set MEM_READONLY_P and MEM_NOTRAP_P on it.

From-SVN: r269260

5 years agoruntime: align first persistentalloc chunk as requested
Ian Lance Taylor [Wed, 27 Feb 2019 15:51:49 +0000 (15:51 +0000)]
runtime: align first persistentalloc chunk as requested

    Backport of upstream https://golang.org/cl/163859.

    This fixes various failures on 32-bit SPARC.

    Patch from Eric Boctazou.

    Reviewed-on: https://go-review.googlesource.com/c/163860

From-SVN: r269258

5 years agore PR tree-optimization/89280 (ICE: Segmentation fault (in is_gimple_reg_type))
Jakub Jelinek [Wed, 27 Feb 2019 15:22:16 +0000 (16:22 +0100)]
re PR tree-optimization/89280 (ICE: Segmentation fault (in is_gimple_reg_type))

PR tree-optimization/89280
* gcc.dg/torture/pr57147-2.c (SetNaClSwitchExpectations): Add static
keyword.

From-SVN: r269256

5 years agore PR target/70341 (cost model for addresses is incorrect, slsr is using reg + reg...
Jakub Jelinek [Wed, 27 Feb 2019 14:50:35 +0000 (15:50 +0100)]
re PR target/70341 (cost model for addresses is incorrect, slsr is using reg + reg + CST for arm)

PR target/70341
* config/arm/arm.md (arm_casesi_internal): New define_expand.  Rename
old define_insn to ...
(*arm_casesi_internal): ... this.  Add mode to LABEL_REFs.
* config/arm/thumb2.md (thumb2_casesi_internal): New define_expand.
Rename old define_insn to ...
(*thumb2_casesi_internal): ... this.  Add mode to LABEL_REFs.
(thumb2_casesi_internal_pic): New define_expand.  Rename old
define_insn to ...
(*thumb2_casesi_internal_pic): ... this.  Add mode to LABEL_REFs.

From-SVN: r269255

5 years agore PR debug/88878 (.debug_pubnames/types empty with -flto)
Richard Biener [Wed, 27 Feb 2019 14:04:18 +0000 (14:04 +0000)]
re PR debug/88878 (.debug_pubnames/types empty with -flto)

2019-02-27  Richard Biener  <rguenther@suse.de>

PR debug/88878
* dwarf2out.c (use_debug_types): Disable when in_lto_p.

From-SVN: r269253

5 years agopasses.c (should_skip_pass_p): Do not skip cgraph-edge building.
Richard Biener [Wed, 27 Feb 2019 13:24:40 +0000 (13:24 +0000)]
passes.c (should_skip_pass_p): Do not skip cgraph-edge building.

2019-02-27  Richard Biener  <rguenther@suse.de>

* passes.c (should_skip_pass_p): Do not skip cgraph-edge
building.

* gcc.dg/gimplefe-36.c: New testcase.

From-SVN: r269252

5 years agore PR debug/88878 (.debug_pubnames/types empty with -flto)
Richard Biener [Wed, 27 Feb 2019 12:51:43 +0000 (12:51 +0000)]
re PR debug/88878 (.debug_pubnames/types empty with -flto)

2019-02-27  Richard Biener  <rguenther@suse.de>

PR debug/88878
* dwarf2out.c (output_comdat_type_unit): Add early_lto_debug
parameter, prefix section name with .gnu.debuglto_ if true.
(dwarf2out_finish): Pass false to output_comdat_type_unit.
(dwarf2out_early_finish): Pass true to output_comdat_type_unit.

From-SVN: r269251

5 years agore PR lto/89514 (-g -fdebug-types-section -flto gives 'Dwarf Error: bad length' in...
Richard Biener [Wed, 27 Feb 2019 11:59:07 +0000 (11:59 +0000)]
re PR lto/89514 (-g -fdebug-types-section -flto gives 'Dwarf Error: bad length' in gdb)

2019-02-27  Richard Biener  <rguenther@suse.de>

PR debug/89514
* dwarf2out.c (size_of_die): Key on AT_ref (a)->comdat_type_p
rather than on use_debug_types, doing what output_die does.
(value_format): Likewise.

From-SVN: r269250

5 years agoPR libstdc++/89466 avoid slow xsltproc command in configure
Jonathan Wakely [Wed, 27 Feb 2019 11:25:44 +0000 (11:25 +0000)]
PR libstdc++/89466 avoid slow xsltproc command in configure

Certain broken versions of xsltproc ignore the --nonet option and will
attempt to fetch the docbook stylesheet from the WWW when it isn't in
the local XML catalog.

This patch checks for the local stylesheet directory first, and doesn't
use xsltproc if no local stylesheets are found. Checking for the local
directory is done using xmlcatalog if available, only checking the
hardcoded list of directories if xmlcatalog fails. The right directory
for Suse is added to the hardcoded list.

This should avoid doing an xsltproc check that would need to download
the stylesheet, so no network connection is made even if a broken
xsltproc is present.

PR libstdc++/89466
* acinclude.m4 (GLIBCXX_CONFIGURE_DOCBOOK): Reorder check for local
stylesheet directories before check for xsltproc. Try to use
xmlcatalog to find local stylesheet directory before trying hardcoded
paths. Add path used by suse to hardcoded paths. Adjust xsltproc
check to look for the same stylesheet as doc/Makefile.am uses. Don't
use xsltproc if xmlcatalog fails to find a local stylesheet.
* configure.ac: Check for xmlcatalog.
* Makefile.in: Regenerate.
* configure: Likewise.
* doc/Makefile.in: Likewise.
* include/Makefile.in: Likewise.
* libsupc++/Makefile.in: Likewise.
* po/Makefile.in: Likewise.
* python/Makefile.in: Likewise.
* src/Makefile.in: Likewise.
* src/c++11/Makefile.in: Likewise.
* src/c++17/Makefile.in: Likewise.
* src/c++98/Makefile.in: Likewise.
* src/filesystem/Makefile.in: Likewise.
* testsuite/Makefile.in: Likewise.

From-SVN: r269249

5 years agoReplaced spaces with tab in Changelog, committed as obvious
Alejandro Martinez [Wed, 27 Feb 2019 10:38:25 +0000 (10:38 +0000)]
Replaced spaces with tab in Changelog, committed as obvious

From-SVN: r269248

5 years agoAdd myself to write after approval.
Alejandro Martinez [Wed, 27 Feb 2019 10:23:54 +0000 (10:23 +0000)]
Add myself to write after approval.

From-SVN: r269247

5 years ago[doc] Reword description of -Wno-absolute-value
Martin Jambor [Wed, 27 Feb 2019 10:07:01 +0000 (11:07 +0100)]
[doc] Reword description of -Wno-absolute-value

2019-02-27  Martin Jambor  <mjambor@suse.cz>
    Martin Sebor  <msebor@redhat.com>

* doc/invoke.texi (Warning Options): Reword description of
-Wno-absolute-value.

Co-Authored-By: Martin Sebor <msebor@redhat.com>
From-SVN: r269246

5 years agore PR c++/89488 (ICE in merge_exception_specifiers, at cp/typeck2.c:2395)
Paolo Carlini [Wed, 27 Feb 2019 10:02:23 +0000 (10:02 +0000)]
re PR c++/89488 (ICE in merge_exception_specifiers, at cp/typeck2.c:2395)

/cp
2019-02-27  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/89488
* method.c (process_subob_fn): When maybe_instantiate_noexcept
returns false don't call merge_exception_specifiers.

/testsuite
2019-02-27  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/89488
* g++.dg/cpp0x/nsdmi15.C: New.

/cp
2019-02-27  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/88987
* parser.c (cp_parser_noexcept_specification_opt): Return NULL_TREE
for a non-constant parsed expression.

/testsuite
2019-02-27  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/88987
* g++.dg/cpp0x/pr88987.C: New.

From-SVN: r269245

5 years agore PR tree-optimization/89280 (ICE: Segmentation fault (in is_gimple_reg_type))
Jakub Jelinek [Wed, 27 Feb 2019 08:41:01 +0000 (09:41 +0100)]
re PR tree-optimization/89280 (ICE: Segmentation fault (in is_gimple_reg_type))

PR tree-optimization/89280
* tree-cfgcleanup.c (maybe_dead_abnormal_edge_p,
builtin_setjmp_setup_bb): New functions.
(cleanup_control_flow_pre): Ignore maybe_dead_abnormal_edge_p edges.
When visiting __builtin_setjmp_setup block, queue in special
setjmp_vec vector edges from .ABNORMAL_DISPATCHER to corresponding
__builtin_setjmp_receiver.  Remove .ABNORMAL_DISPATCHER basic blocks
from visited after the loop if they don't have any visited successor
blocks.

* gcc.c-torture/compile/pr89280.c: New test.
* gcc.dg/torture/pr57147-2.c: Don't expect a setjmp after noreturn
function.  Skip the test for -O0.

From-SVN: r269243

5 years agocompiler: check duplicate numeric keys in map literals
Ian Lance Taylor [Wed, 27 Feb 2019 04:56:12 +0000 (04:56 +0000)]
compiler: check duplicate numeric keys in map literals

    Updates golang/go#28104

    Reviewed-on: https://go-review.googlesource.com/c/162882

From-SVN: r269242

5 years agocompiler: check recursive inherited interface aliases
Ian Lance Taylor [Wed, 27 Feb 2019 01:41:25 +0000 (01:41 +0000)]
compiler: check recursive inherited interface aliases

    Fixes golang/go#25302.

    Reviewed-on: https://go-review.googlesource.com/c/163298

From-SVN: r269241

5 years agocmd/go: preserve CC for TestScript child processes
Ian Lance Taylor [Wed, 27 Feb 2019 01:11:20 +0000 (01:11 +0000)]
cmd/go: preserve CC for TestScript child processes

    Reviewed-on: https://go-review.googlesource.com/c/158838

From-SVN: r269240

5 years agoDaily bump.
GCC Administrator [Wed, 27 Feb 2019 00:16:42 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r269239

5 years agosimd-builtins-1.f90: Update for aarch64*-*-*.
Steve Ellcey [Wed, 27 Feb 2019 00:02:53 +0000 (00:02 +0000)]
simd-builtins-1.f90: Update for aarch64*-*-*.

2018-02-26  Steve Ellcey  <sellcey@marvell.com>

* gfortran.dg/simd-builtins-1.f90: Update for aarch64*-*-*.
* gfortran.dg/simd-builtins-2.f90: Ditto.
* gfortran.dg/simd-builtins-6.f90: Ditto.
* gfortran.dg/simd-builtins-8.f90: New test.
* gfortran.dg/simd-builtins-8.h: New header file.

From-SVN: r269236

5 years agoaarch64.c (aarch64_get_multilib_abi_name): New function.
Steve Ellcey [Wed, 27 Feb 2019 00:00:56 +0000 (00:00 +0000)]
aarch64.c (aarch64_get_multilib_abi_name): New function.

2018-02-26  Steve Ellcey  <sellcey@marvell.com>

* config/aarch64/aarch64.c (aarch64_get_multilib_abi_name):
New function.
(TARGET_GET_MULTILIB_ABI_NAME): New macro.

From-SVN: r269235

5 years agoPR libstdc++/89477 constrain deduction guides for maps and sets
Jonathan Wakely [Tue, 26 Feb 2019 23:12:44 +0000 (23:12 +0000)]
PR libstdc++/89477 constrain deduction guides for maps and sets

The Compare, Hash, and Pred template parameters should be constrained in
the C++17 deduction guides for associative and unordered containers.

The deduction guides for stack, queue and priority_queue are already
constrained, but this patch makes them use the _RequireNotAllocator
helper instead of reproducing the logic each time.

PR libstdc++/89477
* include/bits/alloc_traits.h (_RequireNotAllocator): New helper for
container deduction guides.
* include/bits/hashtable.h (_RequireNotAllocatorOrIntegral): Likewise.
* include/bits/stl_map.h (map): Use _RequireNotAllocator to constrain
parameters in deduction guides.
* include/bits/stl_multimap.h (multimap): Likewise.
* include/bits/stl_multiset.h (multiset): Likewise.
* include/bits/stl_queue.h (queue, priority_queue): Likewise.
* include/bits/stl_set.h (set): Likewise.
* include/bits/stl_stack.h (stack): Likewise.
* include/bits/unordered_map.h (unordered_map, unordered_multimap):
use _RequireNotAllocator and _RequireNotAllocatorOrIntegral to
constrain parameters in deduction guides.
* include/bits/unordered_set.h (unordered_set, unordered_multiset):
Likewise.
* testsuite/23_containers/map/cons/deduction.cc: Test additional
deduction cases.
* testsuite/23_containers/multiset/cons/deduction.cc: Likewise.
* testsuite/23_containers/set/cons/deduction.cc: Likewise.
* testsuite/23_containers/unordered_map/cons/deduction.cc: Likewise.
* testsuite/23_containers/unordered_multimap/cons/deduction.cc:
Likewise.
* testsuite/23_containers/unordered_multiset/cons/deduction.cc:
Likewise.
* testsuite/23_containers/unordered_set/cons/deduction.cc: Likewise.

From-SVN: r269234

5 years agore PR c++/89507 (bogus "size of array exceeds maximum object size")
Jakub Jelinek [Tue, 26 Feb 2019 21:27:33 +0000 (22:27 +0100)]
re PR c++/89507 (bogus "size of array exceeds maximum object size")

PR c++/89507
* tree.c (valid_constant_size_p): Deal with size INTEGER_CSTs
with types other than sizetype/ssizetype.

* g++.dg/other/new2.C: New test.

From-SVN: r269233

5 years agosparc-opts.h (enum processor_type): Rename to...
Eric Botcazou [Tue, 26 Feb 2019 21:08:22 +0000 (21:08 +0000)]
sparc-opts.h (enum processor_type): Rename to...

* config/sparc/sparc-opts.h (enum processor_type): Rename to...
(enum sparc_processor_type): ...this.
(enum sparc_code_model_type): New enumeration type.
(enum sparc_memory_model_type): Tweak comments.
* config/sparc/sparc.opt (mcpu): Adjust to above renaming.
(mtune): Likewise.
(mcmodel): Use sparc_code_model enumeration and variable.
(sparc_code_model): New enumeration.
(mdebug): Add Undocumented marker.
* config/sparc/sparc.h (enum cmodel): Delete.
(sparc_cmodel): Likewise.
(TARGET_CM_MEDLOW): Adjust to above renaming.
(TARGET_CM_MEDMID): Likewise.
(TARGET_CM_MEDANY): Likewise.
(TARGET_CM_EMBMEDANY): Likewise.
* config/sparc/sparc.c (sparc_cmodel): Delete.
(sparc_option_override): Remove string/value mapping support for the
code model.  Move code and memory model support to after the handling
of target flags.  Do private machine setup last.
(sparc_emit_set_symbolic_const64): Use sparc_code_model.
(sparc_legitimize_reload_address): Likewise.
(sparc_output_mi_thunk): Likewise.
* config/sparc/sparc.md (cpu): Adjust comment to above renaming.

From-SVN: r269232

5 years agore PR tree-optimization/89500 (ICE: tree check: expected integer_cst, have ssa_name...
Jakub Jelinek [Tue, 26 Feb 2019 20:36:29 +0000 (21:36 +0100)]
re PR tree-optimization/89500 (ICE: tree check: expected integer_cst, have ssa_name in get_len, at tree.h:5653)

PR tree-optimization/89500
* tree-ssa-strlen.c (stridx_strlenloc): Adjust comment.
(handle_builtin_strlen): Remove noncst_bound variable.  Always
optimize strnlen (x, 0) to 0.  Optimize strnlen (x, cst) to
cst if the first cst bytes starting at x are known to be non-zero,
even if the string is not zero terminated.  Don't try to modify
*si for strnlen.  Update strlen_to_stridx only for strlen or if
we can prove strnlen returns the same value as strlen would.

* gcc.dg/pr89500.c: New test.
* gcc.dg/Wstringop-overflow-10.c: New test.
* gcc.dg/strlenopt-60.c: New test.

From-SVN: r269230

5 years agoPR libstdc++/89416 fix alloc insertable trait for clang (again)
Jonathan Wakely [Tue, 26 Feb 2019 20:34:46 +0000 (20:34 +0000)]
PR libstdc++/89416 fix alloc insertable trait for clang (again)

PR libstdc++/89416
* include/bits/alloc_traits.h (__is_alloc_insertable_impl): Change
to class template and partial specialization using void_t.
(__is_copy_insertable, __is_move_insertable): Adjust base class.

From-SVN: r269229

5 years agore PR fortran/89492 (Endless compilation of an invalid TRANSFER after r269177)
Harald Anlauf [Tue, 26 Feb 2019 20:03:08 +0000 (20:03 +0000)]
re PR fortran/89492 (Endless compilation of an invalid TRANSFER after r269177)

2019-02-26  Harald Anlauf  <anlauf@gmx.de>

PR fortran/89492
* check.c (gfc_calculate_transfer_sizes): Handle cases where
storage size of elements of MOLD is 0.

PR fortran/89492
* gfortran.dg/pr89492.f90: New test.

From-SVN: r269227

5 years agore PR fortran/89496 (gcc/fortran/trans-types.c:3015:9: runtime error: member access...
Thomas Koenig [Tue, 26 Feb 2019 19:10:00 +0000 (19:10 +0000)]
re PR fortran/89496 (gcc/fortran/trans-types.c:3015:9: runtime error: member access within null pointer of type 'struct gfc_formal_arglist')

2019-02-26  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/89496
* trans-types.c (get_formal_from_actual_arglist): If
the actual arglist has no expression, the corresponding
formal arglist is an alternate return.

2019-02-26  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/89496
* gfortran.dg/altreturn_9_0.f90: New file.
* gfortran.dg/altreturn_9_1.f90: New file.

From-SVN: r269226

5 years agotransfer.c (transfer_array_inner): Do not cast charlen to index_type.
Uros Bizjak [Tue, 26 Feb 2019 18:41:53 +0000 (19:41 +0100)]
transfer.c (transfer_array_inner): Do not cast charlen to index_type.

* io/transfer.c (transfer_array_inner): Do not
cast charlen to index_type.

From-SVN: r269224

5 years agoImprove memory statistics report readability.
Martin Liska [Tue, 26 Feb 2019 17:27:52 +0000 (18:27 +0100)]
Improve memory statistics report readability.

2019-02-26  Martin Liska  <mliska@suse.cz>

* alloc-pool.h (struct pool_usage): Remove extra
print_dash_line.
* bitmap.h (struct bitmap_usage): Likewise.
* ggc-common.c (struct ggc_usage): Likewise.
* mem-stats.h (struct mem_usage): Likewise.
(mem_alloc_description::dump): Print dash lines
here and repeat header at the end of a table report.
It's then more readable.
* tree-phinodes.c (phinodes_print_statistics): Make
horizontal alignment.
* tree-ssanames.c (ssanames_print_statistics): Likewise.
* vec.c (struct vec_usage): Remove extra print_dash_line.
* vec.h (vec_safe_grow_cleared): Pass PASS_MEM_STAT.
2019-02-26  Martin Liska  <mliska@suse.cz>

* symtab.c (ht_dump_statistics): Make
horizontal alignment for statistics.

From-SVN: r269221

5 years agoRemove a legacy lto-symtab.c file.
Martin Liska [Tue, 26 Feb 2019 17:27:36 +0000 (18:27 +0100)]
Remove a legacy lto-symtab.c file.

2019-02-26  Martin Liska  <mliska@suse.cz>

* lto-symtab.c: Remove.

From-SVN: r269220

5 years agoextend.texi (__builtin_object_size): Use @pxref instead of @xref inside parenthesis.
Uros Bizjak [Tue, 26 Feb 2019 17:21:43 +0000 (18:21 +0100)]
extend.texi (__builtin_object_size): Use @pxref instead of @xref inside parenthesis.

* doc/extend.texi (__builtin_object_size):
Use @pxref instead of @xref inside parenthesis.
(__builtin_has_attribute): Add missing comma after @xref.
(__builtin_object_size): Ditto.
* doc/md.texi (cond_*{mode}): Use @samp instead of @var around op1[i].
* fortran/invoke.texi (-ffpe-trap): Use @var for every item
in the list.

From-SVN: r269219

5 years agore PR rtl-optimization/87761 ([MIPS] New FAIL: gcc.target/mips/fix-r4000-10.c ...
Jeff Law [Tue, 26 Feb 2019 17:08:06 +0000 (10:08 -0700)]
re PR rtl-optimization/87761 ([MIPS] New FAIL: gcc.target/mips/fix-r4000-10.c   -O1 start with r265398)

PR rtl-optimization/87761
* regcprop.c (copyprop_hardreg_forward_1): Use REG_UNUSED notes to
detect obviously dead insns and delete them.

From-SVN: r269218

5 years ago[libbacktrace] Fix btest-lto for older gcc
Tom de Vries [Tue, 26 Feb 2019 16:09:40 +0000 (16:09 +0000)]
[libbacktrace] Fix btest-lto for older gcc

With gcc 4.8.5, btest_lto ends up with a privatized name for the variable with
name 'global':
...
$ nm btest_lto | grep ' d ' | grep global
0000000000617150 d global.2530
...
which makes test5 fail:
...
    test5: unexpected syminfo name got global.2530 expected global
...

Fix this failure by accepting this type of name as a valid name in btest_lto.

2019-02-26  Tom de Vries  <tdevries@suse.de>

* btest.c (test5): Allow global.* as minimal symbol name for global.

From-SVN: r269217

5 years agolibgo: update to Go 1.12 release
Ian Lance Taylor [Tue, 26 Feb 2019 15:38:12 +0000 (15:38 +0000)]
libgo: update to Go 1.12 release

    Reviewed-on: https://go-review.googlesource.com/c/163742

From-SVN: r269216

5 years agore PR go/86535 (FreeBSD/PowerPC64 - Building Go Frontend support for gcc 7.3.0 fails)
Ian Lance Taylor [Tue, 26 Feb 2019 14:46:56 +0000 (14:46 +0000)]
re PR go/86535 (FreeBSD/PowerPC64 - Building Go Frontend support for gcc 7.3.0 fails)

PR go/86535
    runtime: always declare nanotime in Go

    For libgo it's always defined in C.

    Updates https://gcc.gnu.org/PR86535

    Reviewed-on: https://go-review.googlesource.com/c/163743

From-SVN: r269214

5 years agore PR c++/89481 (constexpr function allows writing one active union member and readin...
Jakub Jelinek [Tue, 26 Feb 2019 14:37:39 +0000 (15:37 +0100)]
re PR c++/89481 (constexpr function allows writing one active union member and reading another)

PR c++/89481
* constexpr.c (cxx_eval_store_expression): When changing active union
member, set no_zero_init.

* g++.dg/cpp1y/constexpr-89481.C: New test.

From-SVN: r269213

5 years agore PR tree-optimization/89505 (LibreOffice miscompilation starting with r260383)
Richard Biener [Tue, 26 Feb 2019 14:09:19 +0000 (14:09 +0000)]
re PR tree-optimization/89505 (LibreOffice miscompilation starting with r260383)

2019-02-26  Richard Biener  <rguenther@suse.de>

PR tree-optimization/89505
* tree-ssa-structalias.c (compute_dependence_clique): Make sure
to handle restrict pointed-to vars with multiple subvars
correctly.

* gcc.dg/torture/pr89505.c: New testcase.

From-SVN: r269212

5 years agore PR tree-optimization/89489 (ICE in gimple_duplicate_bb, at tree-cfg.c:6257)
Richard Biener [Tue, 26 Feb 2019 11:03:45 +0000 (11:03 +0000)]
re PR tree-optimization/89489 (ICE in gimple_duplicate_bb, at tree-cfg.c:6257)

2019-02-26  Richard Biener  <rguenther@suse.de>

PR tree-optimization/89489
* tree-parloops.c (create_loop_fn): Copy over last_clique.

From-SVN: r269210

5 years agoopt77.adb: New test.
Eric Botcazou [Tue, 26 Feb 2019 10:55:16 +0000 (10:55 +0000)]
opt77.adb: New test.

* gnat.dg/opt77.adb: New test.
* gnat.dg/opt77_pkg.ad[sb]: New helper.

From-SVN: r269208

5 years agore PR fortran/43210 (Initializer of huge static arrays should be improved)
Jakub Jelinek [Tue, 26 Feb 2019 10:36:05 +0000 (11:36 +0100)]
re PR fortran/43210 (Initializer of huge static arrays should be improved)

PR fortran/43210
* trans-array.c (gfc_conv_array_initializer): Use RANGE_EXPR instead
of duplicating the initializer possibly many times.

From-SVN: r269207

5 years agore PR target/89474 (ice in df_reg_chain_verify_unmarked, at df-scan.c:3995)
Jakub Jelinek [Tue, 26 Feb 2019 10:08:29 +0000 (11:08 +0100)]
re PR target/89474 (ice in df_reg_chain_verify_unmarked, at df-scan.c:3995)

PR target/89474
* config/i386/i386.c (remove_partial_avx_dependency): Call
df_analyze etc. before creation of the v4sf_const0 pseudo, rather than
after changing possibly many instructions to use that pseudo.  Fix up
insertion of v4sf_const0 setter at the start of bb.

* gcc.target/i386/pr89474.c: New test.

From-SVN: r269206

5 years ago[libbacktrace] Require dwz for b3test_dwz_buildid
Tom de Vries [Tue, 26 Feb 2019 07:54:57 +0000 (07:54 +0000)]
[libbacktrace] Require dwz for b3test_dwz_buildid

If dwz is not available, then the libbacktrace test b3test_dwz_buildid fails
like this:
...
gmake[4]: *** No rule to make target 'b3test_dwz_buildid'
...

Fix this by guarding the test with HAVE_DWZ.

Tested on x86_64 with and without dwz installed.

2019-02-26  Tom de Vries  <tdevries@suse.de>

* Makefile.am (TESTS): Only add b3test_dwz_buildid if HAVE_DWZ.
* Makefile.in: Regenerate.

From-SVN: r269204

5 years agore PR c/80409 (Document that va_arg(ap, void*) can be used to consume any pointer...
Sandra Loosemore [Tue, 26 Feb 2019 02:33:26 +0000 (21:33 -0500)]
re PR c/80409 (Document that va_arg(ap, void*) can be used to consume any pointer argument)

2019-02-25  Sandra Loosemore  <sandra@codesourcery.com>

PR c/80409

gcc/
* doc/extend.texi (Variadic Pointer Args): New section.

From-SVN: r269203

5 years agolibgo: update to Go1.12rc1
Ian Lance Taylor [Tue, 26 Feb 2019 01:00:39 +0000 (01:00 +0000)]
libgo: update to Go1.12rc1

    Reviewed-on: https://go-review.googlesource.com/c/162881

From-SVN: r269202

5 years agoDaily bump.
GCC Administrator [Tue, 26 Feb 2019 00:16:21 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r269201

5 years agore PR c/89495 (gcc/c-family/c-format.c:1272:20: runtime error: signed integer overflo...
Jakub Jelinek [Mon, 25 Feb 2019 23:43:51 +0000 (00:43 +0100)]
re PR c/89495 (gcc/c-family/c-format.c:1272:20: runtime error: signed integer overflow: 214748365 * 10 cannot be represented in type 'int')

PR c/89495
* c-format.c (maybe_read_dollar_number): Compute nargnum in
HOST_WIDE_INT type to avoid overflows and change overflow_flag
checking.

From-SVN: r269198

5 years agore PR c/77754 (internal compiler error : tree code 'call_expr' is not supported in...
Jakub Jelinek [Mon, 25 Feb 2019 22:58:45 +0000 (23:58 +0100)]
re PR c/77754 (internal compiler error : tree code 'call_expr' is not supported in LTO streams)

PR c/77754
* gcc.c-torture/compile/pr77754-1.c: New test.
* gcc.c-torture/compile/pr77754-2.c: New test.
* gcc.c-torture/compile/pr77754-3.c: New test.
* gcc.c-torture/compile/pr77754-4.c: New test.
* gcc.c-torture/compile/pr77754-5.c: New test.
* gcc.c-torture/compile/pr77754-6.c: New test.

From-SVN: r269197

5 years agoc.opt (Wmissing-attributes): Clean up doc string.
Sandra Loosemore [Mon, 25 Feb 2019 20:38:42 +0000 (15:38 -0500)]
c.opt (Wmissing-attributes): Clean up doc string.

2019-02-25  Sandra Loosemore  <sandra@codesourcery.com>
    Martin Sebor  <msebor@gmail.com>

gcc/
* c-family/c.opt (Wmissing-attributes): Clean up doc string.
* common.opt (Wattribute-alias): Likewise.
* doc/invoke.texi (Option Summary): List general form of
-Wattribute-alias=.  List positive form of -Wmissing-attributes.
(-Wmissing-attributes): Invert entry, rewrite and correct default.
Add cross-references.
(-Wattribute-alias): Rewrite and correct default.  Mention
considered attributes (same as for -Wmissing-attributes).

Co-Authored-By: Martin Sebor <msebor@gmail.com>
From-SVN: r269196

5 years ago[rs6000] PR89338, PR89339: Fix compat vector intrinsics for BE and 32-bit
Paul A. Clarke [Mon, 25 Feb 2019 19:36:05 +0000 (19:36 +0000)]
[rs6000] PR89338, PR89339: Fix compat vector intrinsics for BE and 32-bit

Test FAILS: sse2-cvtpd2dq-1, sse2-cvtpd2ps, sse2-cvttpd2dq on powerpc64
(big-endian).

_mm_cvtpd_epi32, _mm_cvtpd_ps, _mm_cvttpd_epi32: Type conversion from
vector doubleword type to vector word type leaves the results in even
lanes in big endian mode.

Test FAILS: sse-cvtss2si-1, sse-cvtss2si-2, sse-movmskb-1 on powerpc
(32-bit big-endian).

Incorrect type for interpreting the result from mfvsrd instruction leads
to incorrect results.  Also, mfvsrd instruction only works as expected in
64-bit mode or for 32-bit quantities in 32-bit mode.  A more general,
if slower, solution is needed for 32-bit mode.

2019-02-25  Paul A. Clarke  <pc@us.ibm.com>

[gcc]

* config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Fix big endian.
(_mm_cvtpd_ps): Likewise.
(_mm_cvttpd_epi32): Likewise.

PR target/89338
* config/rs6000/xmmintrin.h (_mm_cvtss_f32):  Fix type mismatch.
(_mm_cvt_ss2si): Fix type mismatch and 32-bit.

PR target/89339
* config/rs6000/xmmintrin.h (_mm_movemask_pi8): Fix 32-bit.

From-SVN: r269195

5 years ago* fr.po: Update.
Joseph Myers [Mon, 25 Feb 2019 18:48:48 +0000 (18:48 +0000)]
* fr.po: Update.

From-SVN: r269194

5 years agoAArch64: Fix command line options canonicalization version #2. (PR target/88530)
Tamar Christina [Mon, 25 Feb 2019 17:57:01 +0000 (17:57 +0000)]
AArch64: Fix command line options canonicalization version #2. (PR target/88530)

Commandline options on AArch64 don't get canonicalized into the smallest
possible set before output to the assembler. This means that overlapping feature
sets are emitted with superfluous parts.

Normally this isn't an issue, but in the case of crypto we have retro-actively
split it into aes and sha2. We need to emit only +crypto to the assembler
so old assemblers continue to work.

Because of how -mcpu=native and -march=native work they end up enabling all
feature bits. Instead we need to get the smallest possible set, which would also
fix the problem with older the assemblers and the retro-active split.

The function that handles this is called quite often.  It is called for any
push/pop options or attribute that changes arch, cpu etc.  In order to not make
this search for the smallest set too expensive we sort the options based on the
number of features (bits) they enable.  This allows us to process the list
linearly instead of quadratically (Once we have enabled a feature, we know that
anything else that enables it can be ignored.  By sorting we'll get the biggest
groups first and thus the smallest combination of commandline flags).

The Option handling structures have been extended to have a boolean to indicate
whether the option is synthetic, with that I mean if the option flag itself
enables a new feature.

e.g. +crypto isn't an actual feature, it just enables other features, but others
like +rdma enable multiple dependent features but is itself also a feature.

There are two ways to solve this.

1) Either have the options that are feature bits also turn themselves on, e.g.
   change rdma to turn on FP, SIMD and RDMA as dependency bits.

2) Make a distinction between these two different type of features and have the
   framework handle it correctly.

Even though it's more code I went for the second approach, as it's the one
that'll be less fragile (people can't forget it) and gives the least surprises.

Effectively this patch changes the following:

The values before the => are the old compiler and after the => the new code.

-march=armv8.2-a+crypto+sha2 => -march=armv8.2-a+crypto
-march=armv8.2-a+sha2+aes => -march=armv8.2-a+crypto

The remaining behaviors stay the same.

gcc/ChangeLog:

PR target/88530
* common/config/aarch64/aarch64-common.c
(struct aarch64_option_extension): Add is_synthetic.
(all_extensions): Use it.
(TARGET_OPTION_INIT_STRUCT): Define hook.
(struct gcc_targetm_common): Moved to end.
(all_extensions_by_on): New.
(opt_ext_cmp, typedef opt_ext): New.
(aarch64_option_init_struct): New.
(aarch64_contains_opt): New.
(aarch64_get_extension_string_for_isa_flags): Output smallest set.
* config/aarch64/aarch64-option-extensions.def
(AARCH64_OPT_EXTENSION): Explicitly include AES and SHA2 in crypto.
(fp, simd, crc, lse, fp16, rcpc, rdma, dotprod, aes, sha2, sha3,
sm4, fp16fml, sve, profile, rng, memtag, sb, ssbs, predres):
Set is_synthetic to false.
(crypto): Set is_synthetic to true.
* config/aarch64/driver-aarch64.c (AARCH64_OPT_EXTENSION): Add
SYNTHETIC.

gcc/testsuite/ChangeLog:

PR target/88530
* gcc.target/aarch64/options_set_1.c: New test.
* gcc.target/aarch64/options_set_2.c: New test.
* gcc.target/aarch64/options_set_3.c: New test.
* gcc.target/aarch64/options_set_4.c: New test.
* gcc.target/aarch64/options_set_5.c: New test.
* gcc.target/aarch64/options_set_6.c: New test.
* gcc.target/aarch64/options_set_7.c: New test.
* gcc.target/aarch64/options_set_8.c: New test.
* gcc.target/aarch64/options_set_9.c: New test.

From-SVN: r269193

5 years agoArm: Update Armv8.4-a's FP16 FML intrinsics
Tamar Christina [Mon, 25 Feb 2019 17:48:55 +0000 (17:48 +0000)]
Arm: Update Armv8.4-a's FP16 FML intrinsics

This patch updates the Armv8.4-a FP16 FML intrinsics's suffixes from u32 to f16
to be more consistent with the naming convention for intrinsics.

The specifications for these intrinsics have not been published yet so we do
not need to maintain the old names.

The patch was created with the following script:

grep -lIE "(vfml[as].+)_u32" -r gcc/ | grep -iEv ".+Changelog.*" \
  | xargs sed -i -E -e "s/(vfml[as].+)_u32/\1_f16/g"

gcc/ChangeLog:

* config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32, vfmlslq_low_u32,
vfmlalq_high_u32, vfmlslq_high_u32, vfmlal_lane_low_u32,
vfmlal_lane_high_u32, vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32,
vfmlal_laneq_low_u32, vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32,
vfmlal_laneq_high_u32, vfmlsl_lane_low_u32, vfmlsl_lane_high_u32,
vfmlslq_laneq_low_u32, vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32,
vfmlslq_laneq_high_u32, vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32):
Rename ...
(vfmlal_low_f16, vfmlsl_low_f16, vfmlal_high_f16, vfmlsl_high_f16,
vfmlalq_low_f16, vfmlslq_low_f16, vfmlalq_high_f16, vfmlslq_high_f16,
vfmlal_lane_low_f16, vfmlal_lane_high_f16, vfmlalq_laneq_low_f16,
vfmlalq_lane_low_f16, vfmlal_laneq_low_f16, vfmlalq_laneq_high_f16,
vfmlalq_lane_high_f16, vfmlal_laneq_high_f16, vfmlsl_lane_low_f16,
vfmlsl_lane_high_f16, vfmlslq_laneq_low_f16, vfmlslq_lane_low_f16,
vfmlsl_laneq_low_f16, vfmlslq_laneq_high_f16, vfmlslq_lane_high_f16,
vfmlsl_laneq_high_f16): ... To this.
* config/arm/neon.md: Update comments.

gcc/testsuite/ChangeLog:

* gcc.target/arm/simd/fp16fml_high.c (test_vfmlal_high_u32,
test_vfmlalq_high_u32, test_vfmlsl_high_u32, test_vfmlslq_high_u32):
Rename ....
(test_vfmlal_high_f16, test_vfmlalq_high_f16, test_vfmlsl_high_f16,
test_vfmlslq_high_f16): ... To this.
* gcc.target/arm/simd/fp16fml_lane_high.c (test_vfmlal_lane_high_u32,
tets_vfmlsl_lane_high_u32, test_vfmlal_laneq_high_u32,
test_vfmlsl_laneq_high_u32, test_vfmlalq_lane_high_u32,
test_vfmlslq_lane_high_u32, test_vfmlalq_laneq_high_u32,
test_vfmlslq_laneq_high_u32): Rename ...
(test_vfmlal_lane_high_f16, tets_vfmlsl_lane_high_f16,
test_vfmlal_laneq_high_f16, test_vfmlsl_laneq_high_f16,
test_vfmlalq_lane_high_f16, test_vfmlslq_lane_high_f16,
test_vfmlalq_laneq_high_f16, test_vfmlslq_laneq_high_f16): ... To this.
* gcc.target/arm/simd/fp16fml_lane_low.c (test_vfmlal_lane_low_u32,
test_vfmlsl_lane_low_u32, test_vfmlal_laneq_low_u32,
test_vfmlsl_laneq_low_u32, test_vfmlalq_lane_low_u32,
test_vfmlslq_lane_low_u32, test_vfmlalq_laneq_low_u32,
test_vfmlslq_laneq_low_u32): Rename ...
(test_vfmlal_lane_low_f16, test_vfmlsl_lane_low_f16,
test_vfmlal_laneq_low_f16, test_vfmlsl_laneq_low_f16,
test_vfmlalq_lane_low_f16, test_vfmlslq_lane_low_f16,
test_vfmlalq_laneq_low_f16, test_vfmlslq_laneq_low_f16): ... To this.
* gcc.target/arm/simd/fp16fml_low.c (test_vfmlal_low_u32,
test_vfmlalq_low_u32, test_vfmlsl_low_u32, test_vfmlslq_low_u32):
Rename ...
(test_vfmlal_low_f16, test_vfmlalq_low_f16, test_vfmlsl_low_f16,
test_vfmlslq_low_f16): ... To this.

From-SVN: r269192

5 years agoAArch64: Update Armv8.4-a's FP16 FML intrinsics
Tamar Christina [Mon, 25 Feb 2019 17:46:16 +0000 (17:46 +0000)]
AArch64: Update Armv8.4-a's FP16 FML intrinsics

This patch updates the Armv8.4-a FP16 FML intrinsics's suffixes from u32 to f16
to be more consistent with the naming convention for intrinsics.

The specifications for these intrinsics have not been published yet so we do
not need to maintain the old names.

The patch was created with the following script:

grep -lIE "(vfml[as].+)_u32" -r gcc/ | grep -iEv ".+Changelog.*" \
  | xargs sed -i -E -e "s/(vfml[as].+)_u32/\1_f16/g"

gcc/ChangeLog:

* config/aarch64/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
vfmlalq_low_u32, vfmlslq_low_u32, vfmlal_high_u32, vfmlsl_high_u32,
vfmlalq_high_u32, vfmlslq_high_u32, vfmlal_lane_low_u32,
vfmlsl_lane_low_u32, vfmlal_laneq_low_u32, vfmlsl_laneq_low_u32,
vfmlalq_lane_low_u32, vfmlslq_lane_low_u32, vfmlalq_laneq_low_u32,
vfmlslq_laneq_low_u32, vfmlal_lane_high_u32, vfmlsl_lane_high_u32,
vfmlal_laneq_high_u32, vfmlsl_laneq_high_u32, vfmlalq_lane_high_u32,
vfmlslq_lane_high_u32, vfmlalq_laneq_high_u32, vfmlslq_laneq_high_u32):
Rename ...
(vfmlal_low_f16, vfmlsl_low_f16, vfmlalq_low_f16, vfmlslq_low_f16,
vfmlal_high_f16, vfmlsl_high_f16, vfmlalq_high_f16, vfmlslq_high_f16,
vfmlal_lane_low_f16, vfmlsl_lane_low_f16, vfmlal_laneq_low_f16,
vfmlsl_laneq_low_f16, vfmlalq_lane_low_f16, vfmlslq_lane_low_f16,
vfmlalq_laneq_low_f16, vfmlslq_laneq_low_f16, vfmlal_lane_high_f16,
vfmlsl_lane_high_f16, vfmlal_laneq_high_f16, vfmlsl_laneq_high_f16,
vfmlalq_lane_high_f16, vfmlslq_lane_high_f16, vfmlalq_laneq_high_f16,
vfmlslq_laneq_high_f16): ... To this.

gcc/testsuite/ChangeLog:

* gcc.target/aarch64/fp16_fmul_high.h (test_vfmlal_high_u32,
test_vfmlalq_high_u32, test_vfmlsl_high_u32, test_vfmlslq_high_u32):
Rename ...
(test_vfmlal_high_f16, test_vfmlalq_high_f16, test_vfmlsl_high_f16,
test_vfmlslq_high_f16): ... To this.
* gcc.target/aarch64/fp16_fmul_lane_high.h (test_vfmlal_lane_high_u32,
tets_vfmlsl_lane_high_u32, test_vfmlal_laneq_high_u32,
test_vfmlsl_laneq_high_u32, test_vfmlalq_lane_high_u32,
test_vfmlslq_lane_high_u32, test_vfmlalq_laneq_high_u32,
test_vfmlslq_laneq_high_u32): Rename ...
(test_vfmlal_lane_high_f16, tets_vfmlsl_lane_high_f16,
test_vfmlal_laneq_high_f16, test_vfmlsl_laneq_high_f16,
test_vfmlalq_lane_high_f16, test_vfmlslq_lane_high_f16,
test_vfmlalq_laneq_high_f16, test_vfmlslq_laneq_high_f16): ... To this.
* gcc.target/aarch64/fp16_fmul_lane_low.h (test_vfmlal_lane_low_u32,
test_vfmlsl_lane_low_u32, test_vfmlal_laneq_low_u32,
test_vfmlsl_laneq_low_u32, test_vfmlalq_lane_low_u32,
test_vfmlslq_lane_low_u32, test_vfmlalq_laneq_low_u32,
test_vfmlslq_laneq_low_u32): Rename ...
(test_vfmlal_lane_low_f16, test_vfmlsl_lane_low_f16,
test_vfmlal_laneq_low_f16, test_vfmlsl_laneq_low_f16,
test_vfmlalq_lane_low_f16, test_vfmlslq_lane_low_f16,
test_vfmlalq_laneq_low_f16, test_vfmlslq_laneq_low_f16): ... To this.
* gcc.target/aarch64/fp16_fmul_low.h (test_vfmlal_low_u32,
test_vfmlalq_low_u32, test_vfmlsl_low_u32, test_vfmlslq_low_u32):
Rename ...
(test_vfmlal_low_f16, test_vfmlalq_low_f16, test_vfmlsl_low_f16,
test_vfmlslq_low_f16): ... To This.
* lib/target-supports.exp
(check_effective_target_arm_fp16fml_neon_ok_nocache): Update test.

From-SVN: r269191

5 years agore PR fortran/89282 (Garbage arithmetics results in fortran with -O3 and overloaded...
Dominique d'Humieres [Mon, 25 Feb 2019 15:19:45 +0000 (16:19 +0100)]
re PR fortran/89282 (Garbage arithmetics results in fortran with -O3 and overloaded operators)

2019-02-25  Dominique d'Humieres  <dominiq@gcc.gnu.org>

PR fortran/89282
* gfortran.dg/overload_3.f90: New test.

From-SVN: r269190

5 years agodf-scan: fix use of mw_order in df_mw_compare (PR 86096)
Alexander Monakov [Mon, 25 Feb 2019 15:14:39 +0000 (18:14 +0300)]
df-scan: fix use of mw_order in df_mw_compare (PR 86096)

PR rtl-optimization/86096
* df-scan.c (df_mw_compare): Do not check mw_reg fields when
comparing mw_order values.

From-SVN: r269189