PR tree-optimization/84468 - bogus -Wstringop-truncation despite assignment after conditional strncpy

gcc/ChangeLog:

PR tree-optimization/84468
* tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
basic block when looking for nul assignment.

gcc/testsuite/ChangeLog:

PR tree-optimization/84468
* g++.dg/warn/Wstringop-truncation-2.C: New test.
* gcc.dg/Wstringop-truncation.c: New test.
* gcc.dg/Wstringop-truncation-2.c: New test.

From-SVN: r258339

re PR target/84277 (A lot of new acats testsuite failures)

PR target/84277
* except.h (output_function_exception_table): Adjust prototype.
* except.c (output_function_exception_table): Remove FNNAME parameter
and add SECTION parameter.  Ouput one part of the table at a time.
* final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
the first part of the exception table and emit unwind directives.
* config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
(i386_pe_seh_cold_init): Likewise.
* config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
(ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
* config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
(ix86_output_call_insn): Emit a nop in one more case for SEH.
* config/i386/winnt.c: Include except.h.
(struct seh_frame_state): Add reg_offset, after_prologue and
in_cold_section fields.
(i386_pe_seh_end_prologue): Set seh->after_prologue.
(i386_pe_seh_cold_init): New function.
(i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
to seh->in_cold_section.
(seh_emit_push): Record the offset of the push.
(seh_emit_save): Record the offet of the save.
(i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
Test seh->after_prologue to disregard the epilogue.
(i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
(i386_pe_end_cold_function): New function.

From-SVN: r258338

runtime: push arena on AIX higher due to clashes

    Reviewed-on: https://go-review.googlesource.com/99117

From-SVN: r258337

runtime: use a fence instruction before rdtsc

    This implements the same choices made in the gc runtime, except that
    for 32-bit x86 we only use the fence instruction if the processor
    supports SSE2.

    The code here is hacked up for speed; the gc runtime uses straight
    assembler.

    Reviewed-on: https://go-review.googlesource.com/97715

From-SVN: r258336

Ada: Fix s-oscons.ads generation

The $(GNATLIBCFLAGS) are already included in $(GNATLIBCFLAGS_FOR_C).

We must call the C compiler with the right machine flags.  So, add
$(GNATLIBCFLAGS_FOR_C) to $(OSCONS_EXTRACT).  For example, on a bi-arch
compiler supporting 32-bit and 64-bit instruction sets we pick otherwise
only one variant due to a missing -m32 or -m64 flag.

gcc/ada
* gcc-interface/Makefile.in (OSCONS_CPP): Remove redundant
$(GNATLIBCFLAGS).
(OSCONS_EXTRACT): Add $(GNATLIBCFLAGS_FOR_C).

From-SVN: r258334

re PR target/84565 (ICE in extract_insn, at recog.c:2304 on aarch64)

PR fortran/84565
* config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.

* gfortran.dg/pr84565.f90: New test.

From-SVN: r258333

re PR c++/84704 (internal compiler error: gimplification failed)

PR c++/84704
* gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
on tmp_var.
* tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
don't print names of DECL_NAMELESS DECL_IGNORED_P decls.

From-SVN: r258317

re PR middle-end/84723 (ICE in create_target_clone, at multiple_target.c:275)

PR middle-end/84723
* multiple_target.c: Include tree-inline.h and intl.h.
(expand_target_clones): Diagnose and fail if node->definition and
!tree_versionable_function_p (node->decl).

* gcc.target/i386/pr84723-1.c: New test.
* gcc.target/i386/pr84723-2.c: New test.
* gcc.target/i386/pr84723-3.c: New test.
* gcc.target/i386/pr84723-4.c: New test.
* gcc.target/i386/pr84723-5.c: New test.

From-SVN: r258316

Fix Changelog commit errro.

From-SVN: r258314

pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use sprint_ul.

* config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
sprint_ul.
(ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
(ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
* config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.

From-SVN: r258312

re PR fortran/64107 ([F95] Pure function as array size)

2018-03-06  Steven G. Kargl  <kargl@gcc.gnu.org>

        PR fortran/64107
        * gfortran.dg/pr64107.f90: New test.

From-SVN: r258311

fptr.c (_dl_read_access_allowed): New.

* config/pa/fptr.c (_dl_read_access_allowed): New.
(__canonicalize_funcptr_for_compare): Use it.

From-SVN: r258310

Daily bump.

From-SVN: r258309

re PR fortran/84697 (minloc/maxloc not simplified with zero size)

2017-03-06  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/84697
PR fortran/66128
* expr.c (simplify_parameter_variable): If p is a size zero array
and not an ARRAY_EXPR insert an empty array constructor and
return.
* gfortran.h: Add prototype for gfc_is_size_zero_array.
* simplify.c (is_size_zero_array): Make non-static and rename into
(gfc_is_size_zero_array):  Check for parameter arrays of zero
size by comparing shape and absence of constructor.
(gfc_simplify_all): Use gfc_is_size_zero_array instead of
is_size_zero_array.
(gfc_simplify_count): Likewise.
(gfc_simplify_iall): Likewise.
(gfc_simplify_iany): Likewise.
(gfc_simplify_iparity): Likewise.
(gfc_simplify_minval): Likewise.
(gfc_simplify_maxval): Likewise.
(gfc_simplify_product): Likewise.
(gfc_simplify_sum): Likewise.

2017-03-06  Thomas Koenig  <tkoenig@gcc.gnu.org>

PR fortran/84697
PR fortran/66128
* gfortran.dg/minmaxloc_zerosize_1.f90: New test.

From-SVN: r258305

re PR libstdc++/84601 (std::optional<std::pair<int, int>> is not assignment copyable)

PR libstdc++/84601
* include/std/optional (_Optional_payload): Split into multiple
specializations that can handle different cases of trivial or
non-trivial assignment operators.
* testsuite/20_util/optional/84601.cc: New.
* testsuite/20_util/optional/cons/value_neg.cc: Adjust.

From-SVN: r258304

re PR c++/84684 (inserting random code / flags produces wrong code)

PR c++/84684
* constexpr.c (cxx_bind_parameters_in_call): Unshare evaluated
arguments.

* g++.dg/cpp1z/constexpr-84684.C: New test.

From-SVN: r258303

re PR c/84721 (ICE in c_push_function_context, at c-decl.c:9667)

PR c/84721
* c-parser.c (add_debug_begin_stmt): Don't add DEBUG_BEGIN_STMT if
!building_stmt_list_p ().

* gcc.dg/pr84721.c: New test.

From-SVN: r258302

re PR target/84710 (ICE: RTL check: expected code 'reg', have 'subreg' in rhs_regno, at rtl.h:1896 with -O -fno-forward-propagate)

PR target/84710
* combine.c (try_combine): Use reg_or_subregno instead of handling
just paradoxical SUBREGs and REGs.

* gcc.dg/pr84710.c: New test.

From-SVN: r258301

re PR fortran/56667 (Syntax error causes misleading message: "Expected PARAMETER symbol in complex constant")

2018-03-06  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/56667
* primary.c (match_sym_complex_part): Give the matcher for an implied
do-loop a chance to run.

2018-03-06  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/56667
* gfortran.dg/implied_do_2.f90: New test.
* gfortran.dg/coarray_8.f90: Update for new error message.

From-SVN: r258281

trans.c (convert_with_check): Fix typo in the condition guarding the overflow check emitted for the upper...

* gcc-interface/trans.c (convert_with_check): Fix typo in the condition
guarding the overflow check emitted for the upper bound of a floating-
point conversion.

From-SVN: r258277

* gcc-interface/utils.c (def_builtin_1): Bail out on error_mark_node.

From-SVN: r258276

[ARC] Cleanup unused functions.

Remove unused pieces of code.

gcc/
2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>

* config/arc/arc.c (arc_finalize_pic): Remove function.
(arc_must_save_register): We use single base PIC register, remove
checks to save/restore the PIC register.
(arc_expand_prologue): Likewise.
* config/arc/arc-protos.h (arc_set_default_type_attributes):
Remove.
(arc_verify_short): Likewise.
(arc_attr_type): Likewise.
* config/arc/arc.c (arc_set_default_type_attributes): Remove.
(walk_stores): Likewise.
(arc_address_cost): Make it static.
(arc_verify_short): Likewise.
(branch_dest): Likewise.
(arc_attr_type): Likewise.
* config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
(TARGET_INSN_LENGTH_PARAMETERS): Likewise.
(arc_final_prescan_insn): Remove inserting the nops due to
hardware hazards.  It is done in reorg step.
(insn_length_variant_t): Remove.
(insn_length_parameters_t): Likewise.
(arc_insn_length_parameters): Likewise.
(arc_get_insn_variants): Likewise.
* config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.

From-SVN: r258275

re PR inline-asm/84683 (internal compiler error: in move_for_stack_reg, at reg-stack.c:1173)

PR inline-asm/84683
* reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
assertion failure.

* g++.dg/ext/pr84683.C: New test.

From-SVN: r258273

re PR tree-optimization/84687 (error: invalid conversion in gimple call with -O3 and -ffast-math)

PR tree-optimization/84687
* omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
on new_node->decl.
* match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.

* gcc.dg/pr84687.c: New test.

From-SVN: r258272

[C++] [PR84231] overload on cond_expr in template

A non-type-dependent COND_EXPR within a template is reconstructed with
the original operands, after one with non-dependent proxies is built to
determine its result type.  This is problematic because the operands of
a COND_EXPR determined to be an rvalue may have been converted to denote
their rvalue nature.  The reconstructed one, however, won't have such
conversions, so lvalue_kind may not recognize it as an rvalue, which may
lead to e.g. incorrect overload resolution decisions.

If we mistake such a COND_EXPR for an lvalue, overload resolution might
regard a conversion sequence that binds it to a non-const reference as
viable, and then select that over one that binds it to a const
reference.  Only after template substitution would we rebuild the
COND_EXPR, realize it is an rvalue, and conclude the reference binding
is ill-formed, but at that point we'd have long discarded any alternate
candidates we could have used.

This patch modifies the logic that determines whether a
(non-type-dependent) COND_EXPR in a template is an lvalue, to rely on
its type, more specifically, on the presence of a REFERENCE_TYPE
wrapper.  In order to avoid a type bootstrapping problem, the
REFERENCE_TYPE that wraps the type of some such COND_EXPRs is
introduced earlier, so that we don't have to test for lvalueness of
the expression using the very code that we wish to change.

for  gcc/cp/ChangeLog

PR c++/84231
* tree.c (lvalue_kind): Use presence/absence of REFERENCE_TYPE
only while processing template decls.
* typeck.c (build_x_conditional_expr): Move wrapping of
reference type around type...
* call.c (build_conditional_expr_1): ... here.  Rename
is_lvalue to is_glvalue.
* parser.c (cp_parser_fold_expression): Catch REFERENCE_REF_P
INDIRECT_REF of COND_EXPR too.

for  gcc/testsuite/ChangeLog

PR c++/84231
* g++.dg/pr84231.C: New.

From-SVN: r258271

[PR c++/84593] ice on braced init with uninit ref field

If an initializer expr is to be NULL in a ctor initializer list, we
ICE in picflag_from_initializer and elsewhere.

If we're missing an initializer for a reference field, we report the
error, but then build a zero initializer to avoid the ICE.

for  gcc/cp/ChangeLog

PR c++/84593
* init.c (build_zero_init_1): Zero-initialize references.

for  gcc/testsuite/ChangeLog

PR c++/84593
* g++.dg/cpp1y/pr84593.C: New.

From-SVN: r258270

[PR c++/84492] stmt expr ending with overload

We ICEd when returning a stmt expr that ends with an overloaded
function, because instantiate_type did not know what to do with
STMT_EXPRs.  And it shouldn't have to: the expected type of a stmt
expr cannot be used to resolve its value: an unresolved overload
cannot supply the result of a stmt expr.  Catch that and report the
error in the stmt expr before we have a chance to instantiate it.

for  gcc/cp/ChangeLog

PR c++/84492
* semantics.c (finish_stmt_expr_expr): Reject unresolved
overloads used as stmt expr values.

for  gcc/testsuite/ChangeLog

PR c++/84492
* g++.dg/pr84492.C: New.

From-SVN: r258269

rs6000-builtin.def (rs6000_speculation_barrier): Rename to ppc_speculation_barrier.

[gcc]

2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
Rename to ppc_speculation_barrier.
* config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
__builtin_ppc_speculation_barrier.

[gcc/testsuite]

2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

* gcc.target/powerpc/spec-barr-1.c: Change called function name to
__builtin_ppc_speculation_barrier.

From-SVN: r258268

Daily bump.

From-SVN: r258267

PR c++/84708 - ICE with lambda in local class NSDMI.

* lambda.c (lambda_expr_this_capture): Handle local class NSDMI
context.

From-SVN: r258264

re PR target/84700 (ICE on 32-bit BE powerpc targets w/ -misel -O1)

PR target/84700
* combine.c (combine_simplify_rtx): Don't try to simplify if
if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
are equal to x.

* gcc.target/powerpc/pr84700.c: New test.

From-SVN: r258263

re PR c++/84684 (inserting random code / flags produces wrong code)

PR c++/84684
* constexpr.c (constexpr_call_hasher::equal): Return false if
lhs->hash != rhs->hash.  Change return 1 to return true and
return 0 to return false.

From-SVN: r258262

rs6000: Don't align tiny loops to 32 bytes for POWER9

For POWER4..POWER8 we align loops of 5..8 instructions to 32 bytes
(instead of to 16 bytes) because that executes faster.  This is no
longer the case on POWER9, so we can just as well only align to 16
bytes.

* config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
to 32 bytes when compiling for POWER9.

From-SVN: r258260

libgo: fix typo in mksysinfo.sh script

    Fix a small typo in the mksysinfo.sh script (incorrect input
    file for a grep command).

    Reviewed-on: https://go-review.googlesource.com/98635

From-SVN: r258259

[PR c++/84702] ICE with default tmpl arg of overload set

https://gcc.gnu.org/ml/gcc-patches/2018-03/msg00227.html
PR c++/84702
* pt.c (process_template_arg): Mark lookup_keep on a default arg.

PR c++/84702
* g++.dg/lookup/pr84702.C: New.

From-SVN: r258258

re PR target/84564 (error: unable to find a register to spill with -mforce-indirect-call)

PR target/84564
* config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
regparm >= 3 with no arg reg available also for calls with
flag_force_indirect_call.  Pass decl to ix86_function_regparm.

* gcc.target/i386/pr84564.c: New test.

From-SVN: r258256

re PR c++/84707 (internal compiler error: Segmentation fault (tree_check()/duplicate_decls()))

PR c++/84707
* decl.c (duplicate_decls): Check DECL_NAME before accessing
UDLIT_OPER_P.

* g++.dg/cpp0x/inline-ns10.C: New test.

From-SVN: r258255

[PR c++/84694] ICE on template friend decl

https://gcc.gnu.org/ml/gcc-patches/2018-03/msg00221.html
PR c++/84694
* friend.c (do_friend): Restore check for identifier_p inside
TEMPLATE_ID_EXPR.

PR c++/84694
* g++.dg/template/pr84694.C: New.

From-SVN: r258254

re PR target/84524 (-O3 causes behavior change)

PR target/84524
* config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
orig,vex.
(*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.

* gcc.c-torture/execute/pr84524.c: New test.
* gcc.target/i386/avx512bw-pr84524.c: New test.

From-SVN: r258252

re PR target/84264 (ICE in rs6000_emit_le_vsx_store, at config/rs6000/rs6000.c:10367 starting with r256656)

gcc/
PR target/84264
* config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.

gcc/testsuite/
PR target/84264
* g++.dg/pr84264.C: New test.

From-SVN: r258251

re PR c++/84618 (ICE in build_capture_proxy, at cp/lambda.c:460)

/cp
2018-03-05  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/84618
* parser.c (cp_parser_lambda_introducer): Reject any capture not
involving a VAR_DECL or a PARM_DECL.

/testsuite
2018-03-05  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/84618
* g++.dg/cpp0x/lambda/lambda-ice29.C: New.
* g++.dg/cpp0x/lambda/lambda-ice17.C: Adjust.
* g++.dg/cpp0x/lambda/lambda-ice23.C: Likewise.

From-SVN: r258250

re PR tree-optimization/84486 (code hoisting removes alignment assumption)

2018-03-05  Richard Biener  <rguenther@suse.de>

PR tree-optimization/84486
* tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
When inserting a __builtin_assume_aligned call set the LHS
SSA name alignment info accordingly.

From-SVN: r258249

[AArch64] PR84114: Avoid reassociating FMA

As discussed in the PR, the reassociation phase runs before FMAs are formed
and so can significantly reduce FMA opportunities.  Although reassociation
could be switched off, it helps in many cases, so a better alternative is to
only avoid reassociation of floating point additions.  This fixes the testcase
and gives 1% speedup on SPECFP2017, fixing the performance regression.

    gcc/
PR tree-optimization/84114
* config/aarch64/aarch64.c (aarch64_reassociation_width)
Avoid reassociation of FLOAT_MODE addition.

From-SVN: r258248

Enable WBOINVD and PCONFIG instructions.

2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>

Enable WBOINVD and PCONFIG instructions.

gcc/
* common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
(ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
* config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
* config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
* config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
and -mwbnoinvd.
* config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
__builtin_ia32_wbinvd): New builtins.
(SPECIAL_ARGS2): New.
* config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
(SPECIAL_ARGS2): New.
* config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
(ix86_valid_target_attribute_inner_p): Ditto.
(ix86_init_mmx_sse_builtins): Add special_args2.
* config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
TARGET_WBNOINVD_P): New.
* config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
(define_insn "wbinvd", define_insn "wbnoinvd"): New.
* config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
* config/i386/immintrin.h (_wbinvd): New intrinsic.
* config/i386/pconfigintrin.h: New file.
* config/i386/wbnoinvdintrin.h: Ditto.
* config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
* doc/invoke.texi (-mpconfig, -mwbnoinvd): New.

gcc/testsuite/
* g++.dg/other/i386-2.C: Add -mpconfig and -mwbnoinvd.
* g++.dg/other/i386-3.C: Ditto.
* gcc.target/i386/sse-12.c: Ditto.
* gcc.target/i386/sse-13.c: Ditto.
* gcc.target/i386/sse-14.c: Ditto.
* gcc.target/i386/sse-23.c: Add pconfig and wbnoinvd.
* gcc.target/i386/wbinvd-1.c: New test.
* gcc.target/i386/wbnoinvd-1.c: Ditto.
* gcc.target/i386/pconfig-1.c: Ditto.

From-SVN: r258247

[PR c++/84497] ref to undefined tls init

https://gcc.gnu.org/ml/gcc-patches/2018-03/msg00209.html
PR c++/84497
* decl2.c (get_tls_init_fn): Check TYPE_HAS_TRIVIAL_DFLT too.

PR c++/84497
* g++.dg/cpp0x/pr84497.C: New.

Co-Authored-By: Jason Merrill <jason@redhat.com>
Co-Authored-By: Nathan Sidwell <nathan@acm.org>
From-SVN: r258244

re PR tree-optimization/84670 (ICE: in compute_antic_aux, at tree-ssa-pre.c:2148 with -O2 -fno-tree-dominator-opts)

2018-03-05  Richard Biener  <rguenther@suse.de>

PR tree-optimization/84670
* tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
member.
(BB_VISITED_WITH_VISITED_SUCCS): New define.
(compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
(compute_antic_aux): Only assert the number of values in ANTIC_IN
doesn't grow if all successors (recursively) were visited at least
once.

* gcc.dg/pr84670-1.c: New testcase.
* gcc.dg/pr84670-2.c: Likewise.
* gcc.dg/pr84670-3.c: Likewise.
* gcc.dg/pr84670-4.c: Likewise.

From-SVN: r258243

re PR tree-optimization/84650 ([graphite] ICE: Segmentation fault (in create_new_iv))

2018-03-05  Richard Biener  <rguenther@suse.de>

PR tree-optimization/84650
* tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
if executed in the loop pipeline.

* gcc.dg/graphite/pr84650.c: New testcase.

From-SVN: r258242

re PR c++/82022 (constexpr lambda in template context: expression ‘<lambda>’ is not a constant expression)

2018-03-05  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/82022
* g++.dg/cpp1z/constexpr-lambda20.C: New.

From-SVN: r258241

configfiles.texi (Configuration Files): Move info about conditionalizing $target-protos.h to...

2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>

gcc/
* doc/configfiles.texi (Configuration Files): Move info about
conditionalizing $target-protos.h to...
* doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
differs from $target-protos.h.

From-SVN: r258240

[NDS32] Implment setmem pattern.

gcc/
* config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
* config/nds32/nds32-multiple.md (setmemsi): Define.
* config/nds32/nds32-memory-manipulation.c
(nds32_gen_dup_4_byte_to_word_value): New.
(emit_setmem_word_loop): New.
(emit_setmem_byte_loop): New.
(nds32_expand_setmem_loop): New.
(nds32_expand_setmem_loop_v3m): New.
(nds32_expand_setmem_unroll): New.
(nds32_expand_setmem): New.

Co-Authored-By: Chung-Ju Wu <jasonwucj@gmail.com>
From-SVN: r258239

Daily bump.

From-SVN: r258238

[NDS32] Rename nds32_expand_movmemqi to nds32_expand_movmemsi and rewrite its implementation.

gcc/
*config/nds32/nds32-memory-manipulation.c
(nds32_emit_load_store): New.
(nds32_emit_post_inc_load_store): New.
(nds32_emit_mem_move): New.
(nds32_emit_mem_move_block): New.
(nds32_expand_movmemsi_loop_unknown_size): New.
(nds32_expand_movmemsi_loop_known_size): New.
(nds32_expand_movmemsi_loop): New.
(nds32_expand_movmemsi_unroll): New.
(nds32_expand_movmemqi): Rename ...
(nds32_expand_movmemsi): ... to this.
*config/nds32/nds32-multiple.md (movmemqi): Rename ...
(movmemsi): ... to this.
*config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
(nds32_expand_movmemsi): ... to this.

Co-Authored-By: Chung-Ju Wu <jasonwucj@gmail.com>
From-SVN: r258235

[NDS32] Refine load_multiple and store_multiple.

gcc/
* config/nds32/nds32-protos.h
(nds32_expand_load_multiple): New arguments.
(nds32_expand_store_multiple): Ditto.
(nds32_valid_multiple_load_store): Rename ...
(nds32_valid_multiple_load_store_p): ... to this.
* config/nds32/nds32-memory-manipulation.c
(nds32_expand_load_multiple): Refine implementation.
(nds32_expand_store_multiple): Ditto.
* config/nds32/nds32-multiple.md
(load_multiple): Update nds32_expand_load_multiple interface.
(store_multiple): Update nds32_expand_store_multiple interface.
* config/nds32/nds32-predicates.c
(nds32_valid_multiple_load_store): Rename ...
(nds32_valid_multiple_load_store_p): ... to this and refine
implementation.
* config/nds32/predicates.md
(nds32_load_multiple_and_update_address_operation): New predicate.
(nds32_store_multiple_and_update_address_operation): New predicate.

Co-Authored-By: Chung-Ju Wu <jasonwucj@gmail.com>
Co-Authored-By: Monk Chiang <sh.chiang04@gmail.com>
From-SVN: r258234

[NDS32] Add load_multiple,store_multiple and new attribute combo.

gcc/
* config/nds32/nds32.md (type): Add load_multiple and store_multiple.
(combo): New attribute.
* config/nds32/nds32-multiple.md: Refine patterns with new attributes.

Co-Authored-By: Chung-Ju Wu <jasonwucj@gmail.com>
From-SVN: r258232

PR c++/84686 - missing volatile loads.

* cvt.c (convert_to_void): Call maybe_undo_parenthesized_ref.

From-SVN: r258231

re PR fortran/71085 (ICE with some intrinsic functions specifying array function result dimension)

2018-03-03  Harald Anlauf  <anlauf@gmx.de>

PR fortran/71085
* trans-expr.c (gfc_apply_interface_mapping_to_expr): Do not
dereference NULL pointer.

2018-03-03  Harald Anlauf  <anlauf@gmx.de>

PR fortran/71085
* gfortran.dg/pr71085.f90: New test.

From-SVN: r258228

Daily bump.

From-SVN: r258227

re PR fortran/66128 (ICE for some intrinsics with zero sized array parameter)

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

PR fortran/66128
* gfortran.dg/zero_sized_8.f90: Really commit new test.

From-SVN: r258224

re PR fortran/66128 (ICE for some intrinsics with zero sized array parameter)

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

PR fortran/66128
* simplify.c (is_size_zero_array): New function to check for size
zero array.
(gfc_simplify_all, gfc_simplify_any, gfc_simplify_count,
 gfc_simplify_iall, gfc_simplify_iany, gfc_simplify_iparity,
 gfc_simplify_minval, gfc_simplify_maxval, gfc_simplify_norm2,
 gfc_simplify_product, gfc_simplify_sum): Use it, and implement
requirements from F2018.

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

PR fortran/66128
* gfortran.dg/zero_sized_8.f90: New test.

From-SVN: r258223

* da.po: Update.

From-SVN: r258221

re PR fortran/51434 (ICE with scalar init of an array parameter, used in DT default init with transfer)

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

PR fortran/51434
* simplify.c (gfc_simplify_transfer): Resolve mold.

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

PR fortran/51434
* gfortran.dg/pr51434.f90: New test.

From-SVN: r258220

re PR c++/61358 (Bogus "duplicate label" error after label used within C++11 lambda)

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

PR c++/61358
* g++.dg/cpp0x/lambda/lambda-61358.C: New.

From-SVN: r258219

re PR c++/67370 (Invalid "parameter packs not expanded" error in lambda capture)

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

PR c++/67370
* g++.dg/cpp0x/lambda/lambda-67370.C: New.

From-SVN: r258218

re PR c++/71464 (ICE on invalid code (with redeclared constructor) at -Os: Segmentation fault)

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

PR c++/71464
* optimize.c (maybe_thunk_body): Bail out immediately if either
fns[0] or fns[1] is null.

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

PR c++/71464
* g++.dg/torture/pr71464.C: New.

From-SVN: r258216

[NDS32] Change to large model by default.

gcc/
* config/nds32/nds32.opt: Change -mcmodel= default value.

From-SVN: r258215

[NDS32] Add intrinsic functions for unalignment memory access.

gcc/
* config/nds32/constants.md (unspec_element): New enum.
* config/nds32/constraints.md (Umw): New constraint.
* config/nds32/nds32-intrinsic.c: Add more builtin functions.
* config/nds32/nds32-intrinsic.md: Likewise.
* config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
(nds32_valid_smw_lwm_base_p): New.
(nds32_output_smw_single_word): New.
(nds32_output_lmw_single_word): New.
(nds32_expand_unaligned_load): New.
(nds32_expand_unaligned_store): New.
* config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
(nds32_output_smw_single_word): Declare.
(nds32_output_lmw_single_word): Declare.
(nds32_expand_unaligned_load): Declare.
(nds32_expand_unaligned_store): Declare.
* config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
NDS32_BUILTIN_UASTORE_DW.
* config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
predicate.

Co-Authored-By: Chung-Ju Wu <jasonwucj@gmail.com>
Co-Authored-By: Monk Chiang <sh.chiang04@gmail.com>
From-SVN: r258214

[NDS32] Rewrite infrastructure for intrinsic.

gcc/
* config/nds32/nds32-intrinsic.c
(nds32_expand_builtin_null_ftype_reg): Delete.
(nds32_expand_builtin_reg_ftype_imm): Ditto.
(nds32_expand_builtin_null_ftype_reg_imm): Ditto.
(nds32_read_argument): New.
(nds32_legitimize_target): Ditto.
(nds32_legitimize_argument): Ditto.
(nds32_check_constant_argument): Ditto.
(nds32_expand_unop_builtin): Ditto.
(nds32_expand_unopimm_builtin): Ditto.
(nds32_expand_binop_builtin): Ditto.
(nds32_builtin_decl_impl): Ditto.
(builtin_description): Ditto.
(nds32_expand_builtin_impl): Rewrite with new infrastructure.
(nds32_init_builtins_impl): Ditto.
* config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
(nds32_builtin_decl): New.
* config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
* config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.

Co-Authored-By: Chung-Ju Wu <jasonwucj@gmail.com>
Co-Authored-By: Kito Cheng <kito.cheng@gmail.com>
From-SVN: r258211

re PR fortran/80965 (ICE with class argument and -O2 optimization)

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

PR fortran/80965
* resolve.c (build_loc_call): Change symtree name from 'loc' to
'_loc'.

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

PR fortran/80965
* gfortran.dg/select_type_41.f90: New test.

From-SVN: r258195

re PR c++/61135 (It seems to be not able to call virtual method of literal object in lambda expression)

2018-03-02  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/61135
* g++.dg/cpp0x/lambda/lambda-61135.C: New.

From-SVN: r258165

Daily bump.

From-SVN: r258163

* es.po: Update.

From-SVN: r258160

reorg.c (stop_search_p): Handle DEBUG_INSN.

* reorg.c (stop_search_p): Handle DEBUG_INSN.
(redundant_insn, fill_simple_delay_slots): Likewise.
(fill_slots_from_thread): Likewise.
* resource.c (mark_referenced_resources): Likewise.
(mark_set_resources, find_dead_or_set_registers): Likewise.

From-SVN: r258158

PR libstdc++/84671 handle digit separators in duration literals

PR libstdc++/84671
* include/bits/parse_numbers.h (_Number_help): Add partial
specialization to handle digit separators. Adjust partial
specialization for recursion temrination to require _Pow == 1ULL.
* testsuite/20_util/duration/literals/84671.cc: New

From-SVN: r258157

re PR c++/84578 (ICE with flexible array member and constexpr)

PR c++/84578
* constexpr.c (get_array_or_vector_nelts): New.
(cxx_eval_array_reference): Use it.
(cxx_eval_vec_init_1): Likewise.
(cxx_eval_store_expression): Likewise.

* g++.dg/ext/flexary29.C: New test.

From-SVN: r258156

libiberty: fix URL for demangler ABI

libiberty/ChangeLog:
* cp-demangle.c: Update URL for g++ V3 ABI.

From-SVN: r258155

substring-locations.h (format_warning_va): Formatting fix for ATTRIBUTE_GCC_DIAG.

* substring-locations.h (format_warning_va): Formatting fix for
ATTRIBUTE_GCC_DIAG.
(format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
argument.
(format_warning_n_va, format_warning_at_substring_n): New prototypes.
* substring-locations.c: Include intl.h.
(format_warning_va): Turned into small wrapper around
format_warning_n_va, renamed to ...
(format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
use ngettext.
(format_warning_at_substring_n): New function.
* gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
(fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
format_warning_at_substring with just a shorter name instead of
const function pointer.
(fmtwarn_n): New function.
(maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
appropriate, get rid of all the fmtstr temporaries, move conditionals
with G_() wrapped string literals directly into fmtwarn arguments,
cast dir.len to (int), formatting fixes.

From-SVN: r258154

* c-c++-common/Warray-bounds-2.c: Fix a comment typo.

From-SVN: r258153

* semantics.c (force_paren_expr): Remove redundant test.

From-SVN: r258151

re PR middle-end/81812 (Empty virtual function fails to compile)

2018-03-02  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/81812
* g++.dg/torture/pr81812.C: New.

From-SVN: r258150

re PR c++/84664 (internal compiler error: in cp_perform_integral_promotions, at cp/typeck.c:2172)

PR c++/84664
* typeck.c (cp_perform_integral_promotions): Check the result of
mark_rvalue_use.

* g++.dg/cpp0x/lambda/lambda-ice28.C: New test.

From-SVN: r258149

re PR c++/84663 (internal compiler error: tree check: expected array_type, have error_mark in cp_complete_array_type, at cp/decl.c:8334)

PR c++/84663
* decl.c (cp_complete_array_type): Check error_mark_node.

* g++.dg/parse/array-size3.C: New test.

From-SVN: r258148

re PR c++/84171 (ICE with -Wsign-compare)

PR c++/84171
* c-warn.c (warn_for_sign_compare): Bail out if any of the operands
is erroneous.

* g++.dg/warn/Wsign-compare-8.C: New test.

From-SVN: r258147

re PR c++/84662 (internal compiler error: tree check: expected class 'type', have 'exceptional' (error_mark) in is_bitfield_expr_with_lowered_type, at cp/typeck.c:1944)

PR c++/84662
* pt.c (tsubst_copy_and_build) <case TEMPLATE_ID_EXPR>: Use
RETURN instead of return.
<case POINTER_PLUS_EXPR>: Likewise.
<case CONVERT_EXPR>: If op0 is error_mark_node, just return
it instead of wrapping it into CONVERT_EXPR.

* g++.dg/cpp1y/pr84662.C: New test.

From-SVN: r258146

Fix MIPS16 ICE.

* pt.c (type_dependent_expression_p): Check DECL_LANG_SPECIFIC.

From-SVN: r258143

Fix "dg-lto-options" misuse

gcc/testsuite/
* g++.dg/lto/20080915_0.C: Don't use "dg-lto-options".
* g++.dg/lto/20080907_0.C: Use "#pragma GCC" instead of
"dg-lto-options".
* g++.dg/lto/20101010-1_0.C: Likewise.
* g++.dg/lto/20101010-2_0.C: Likewise.

Reviewed-by: Jason Merrill <jason@redhat.com>
From-SVN: r258142

More Cilk Plus removal

contrib/
* update-copyright.py: Remove "libcilkrts" reference.
gcc/c-family/
* c-attribs.c (c_common_attribute_table): Remove "cilk simd
function".
gcc/
* doc/invoke.texi: Remove "Cilk Plus" references.

From-SVN: r258141

re PR ipa/84628 (attribute(warning/error) functions should not be merged together)

PR ipa/84628
* expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
for error or warning attributes if CALL_FROM_THUNK_P is set.
Formatting fixes.

* gcc.dg/pr84628.c: New test.

Co-Authored-By: Richard Biener <rguenther@suse.de>
From-SVN: r258140

re PR target/56540 (No __SIZEOF__XXX__ macro for __float128)

PR target/56540
* config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
__SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.

From-SVN: r258139

re PR target/56540 (No __SIZEOF__XXX__ macro for __float128)

PR target/56540
* config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
__SIZEOF_{FPREG,FLOAT{80,128}}__ macros.

* gcc.target/ia64/pr56540.c: New test.

From-SVN: r258138

altivec-7-be.c: Remove VSX content, allow 32-bit target.

[testsuite]

2018-03-02  Will Schmidt  <will_schmidt@vnet.ibm.com>

* gcc.target/powerpc/altivec-7-be.c: Remove VSX content, allow
32-bit target.
* gcc.target/powerpc/altivec-7.h: Remove VSX content.
* gcc.target/powerpc/vsx-7-be.c: New test (VSX content).
* gcc.target/powerpc/vsx-7.h: New include (VSX content).
* gcc.target/powerpc/altivec-7-le.c: Add vsx-7.h include.

From-SVN: r258137

predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED instead of -1U in last predictors element's probability...

* predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
instead of -1U in last predictors element's probability member.

From-SVN: r258136

re PR ipa/83983 (FAIL: g++.dg/lto/pr83121  (test for LTO warnings, pr83121_0.C line 8))

PR ipa/83983
* ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
arguments if they are comparable.

From-SVN: r258133

re PR c++/84590 (-fsanitize=undefined internal compiler error: tree check: expected constructor, have target_expr in split_nonconstant_init_1, at cp/typeck2.c:629)

PR c++/84590
* cp-gimplify.c (cp_fully_fold): Unwrap TARGET_EXPR or a CONSTRUCTOR
wrapped in VIEW_CONVERT_EXPR.

* c-c++-common/ubsan/shift-11.c: New test.

From-SVN: r258132

Avoid &LOOP_VINFO_MASKS for bb vectorisation (PR 84634)

We were computing &LOOP_VINFO_MASKS even for bb vectorisation,
which is UB.

2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>

gcc/
PR tree-optimization/84634
* tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
masks and masked_loop_p with a single loop_masks, making sure it's
null for bb vectorization.

From-SVN: r258131

Use loop->safelen rather than loop->force_vectorize

...since the latter doesn't guarantee independence by itself.

2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>

gcc/
* tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
(vect_analyze_data_ref_access): Use loop->safe_len rather than
loop->force_vectorize to check whether there is no alias.

gcc/testsuite/
* gcc.dg/vect/vect-alias-check-13.c: New test.

From-SVN: r258130

re PR rtl-optimization/84614 (wrong code with u16->u128 extension at aarch64 -fno-split-wide-types -g3 --param=max-combine-insns=3)

PR target/84614
* rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
prototypes.
* emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
comments.
(next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
* cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
instead of a loop around prev_real_insn.
* combine.c (move_deaths): Use prev_real_nondebug_insn instead of
prev_real_insn.

* gcc.dg/pr84614.c: New test.

From-SVN: r258129

re PR fortran/84219 (Failure to generate error for IO of transfer intrinsic, when MOLD has derived type components.)

2018-03-02  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/84219
* gfortran.dg/coarray_47.f90: Use the correct test.

From-SVN: r258128

re PR inline-asm/84625 (ICE with empty constraint and vector constant)

PR inline-asm/84625
* config/i386/i386.c (ix86_print_operand): Use conditional
output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
zero vector.

* gcc.target/i386/pr84625.c: New test.

From-SVN: r258125

re PR tree-optimization/84427 (gcc ICE at -O3 on x86_64-linux-gnu in compute_antic, at tree-ssa-pre.c:2356)

2018-03-02  Richard Biener  <rguenther@suse.de>

PR tree-optimization/84427
* tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
(bitmap_set_subtract_values): Rewrite to handle multiple
exprs per value.
(clean): Likewise.
(prune_clobbered_mems): Likewise.
(phi_translate): Take edge instead of pred/phiblock.
(phi_translate_1): Likewise.
(phi_translate_set): Likewise.  Insert all translated
exprs for a value into the set, keeping possibly multiple
expressions per value.
(compute_antic_aux): Adjust for phi_translate changes.
When intersecting union the expressions and prune those
not in the final value set, keeping possibly multiple
expressions per value.  Do not use value-insertion
for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
all expressions.  Add verification that the value-sets
only shrink during iteration.
(compute_partial_antic_aux): Adjust for the phi_translate changes.
(do_pre_regular_insertion): Likewise.
(do_pre_partial_partial_insertion): Likewise.

* gcc.dg/torture/pr84427.c: New testcase.

From-SVN: r258124

re PR target/82005 (Early lto debug not implemented on Darwin)

2018-03-02  Richard Biener  <rguenther@suse.de>

PR target/82005
* config/darwin.c (saved_debug_info_level): New static global.
(darwin_asm_lto_start): Disable debug info generation for LTO out.
(darwin_asm_lto_end): Restore debug info generation settings.

From-SVN: r258123