rust-demangle.c (unescape): Remove.

* rust-demangle.c (unescape): Remove.
(parse_lower_hex_nibble): New function.
(parse_legacy_escape): New function.
(is_prefixed_hash): Use parse_lower_hex_nibble.
(looks_like_rust): Use parse_legacy_escape.
(rust_demangle_sym): Use parse_legacy_escape.
* testsuite/rust-demangle-expected: Add 'llv$u6d$' test.

From-SVN: r275353

c-cppbuiltin.c (builtin_define_with_hex_fp_value): Always expand when using -fgo-dump-spec.

* c-cppbuiltin.c (builtin_define_with_hex_fp_value): Always expand
when using -fgo-dump-spec.

From-SVN: r275352

re PR target/91604 (ICE in extract_insn at recog.c:2310 since r272323)

PR target/91604
* config/i386/i386-expand.c (split_double_mode): If there is more than
one MEM operand and they are rtx_equal_p, reuse lo_half/hi_half from
already split matching MEM operand instead of calling adjust_address
again.

* gcc.target/i386/pr91604.c: New test.

From-SVN: r275344

Remove Cell Broadband Engine SPU targets

From-SVN: r275343

re PR middle-end/91603 (Unaligned access in expand_assignment)

2019-09-03  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR middle-end/91603
        PR middle-end/91612
        PR middle-end/91613
        * expr.c (expand_expr_real_1): Handle unaligned decl_rtl
        and SSA_NAME referring to CONSTANT_P correctly.

testsuite:
2019-09-03  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR middle-end/91603
        * testsuite/gcc.target/arm/pr91603.c: New test.

From-SVN: r275342

re PR other/79543 (Inappropriate "ld --version" checking)

2019-09-03  Chung-Lin Tang <cltang@codesourcery.com>

libatomic/
PR other/79543
* acinclude.m4 (LIBAT_CHECK_LINKER_FEATURES): Fix GNU ld --version
scanning to conform to the GNU Coding Standards.
* configure: Regenerate.

libffi/
PR other/79543
* acinclude.m4 (LIBAT_CHECK_LINKER_FEATURES): Fix GNU ld --version
scanning to conform to the GNU Coding Standards.
* configure: Regenerate.

libgomp/
PR other/79543
* acinclude.m4 (LIBGOMP_CHECK_LINKER_FEATURES): Fix GNU ld --version
scanning to conform to the GNU Coding Standards.
* configure: Regenerate.

libitm/
PR other/79543
* acinclude.m4 (LIBITM_CHECK_LINKER_FEATURES): Fix GNU ld --version
scanning to conform to the GNU Coding Standards.
* configure: Regenerate.

libstdc++-v3/
PR other/79543
* acinclude.m4 (GLIBCXX_CHECK_LINKER_FEATURES): Fix GNU ld --version
scanning to conform to the GNU Coding Standards.
* configure: Regenerate.

From-SVN: r275341

tree-ssa-sccvn.h (vn_nary_op_lookup): Remove.

2019-09-03  Richard Biener  <rguenther@suse.de>

* tree-ssa-sccvn.h (vn_nary_op_lookup): Remove.
(vn_nary_op_insert): Likewise.
* tree-ssa-sccvn.c (init_vn_nary_op_from_op): Remove.
(vn_nary_op_lookup): Likewise.
(vn_nary_op_insert): Likewise.

From-SVN: r275338

S/390: Commit forgotten test for r275336

gcc/testsuite/ChangeLog:

2019-09-03  Ilya Leoshkevich  <iii@linux.ibm.com>

* gcc.target/s390/sigfpe-eh.c: Forgotten test.

From-SVN: r275337

S/390: Fix failing RTL check in s390_canonicalize_comparison

The new sigfpe-eh.c fails with

    internal compiler error: RTL check: expected elt 0 type 'e' or 'u', have 'w' (rtx const_int)

This is most likely due to a typo: XEXP (*op1, 0) was used, when
XEXP (*op0, 1) was intended.  This did not cause any user-visible
problems, because reversed_comparison_code_parts ignores the
respective argument, and the release compiler is built without RTL
checks.

gcc/ChangeLog:

2019-09-03  Ilya Leoshkevich  <iii@linux.ibm.com>

* config/s390/s390.c (s390_canonicalize_comparison): Use XEXP
(*op0, 1) instead of XEXP (*op1, 0).

gcc/testsuite/ChangeLog:

2019-09-03  Ilya Leoshkevich  <iii@linux.ibm.com>

* gcc.target/s390/sigfpe-eh.c: New test.

From-SVN: r275336

[AArch64] Add support for __jcvt intrinsic

This patch implements the __jcvt ACLE intrinsic [1] that maps down to the FJCVTZS [2] instruction from Armv8.3-a.
No fancy mode iterators or nothing. Just a single builtin, UNSPEC and define_insn and the associate plumbing.
This patch also defines __ARM_FEATURE_JCVT to indicate when the intrinsic is available.

[1] https://developer.arm.com/docs/101028/latest/data-processing-intrinsics
[2] https://developer.arm.com/docs/ddi0596/latest/simd-and-floating-point-instructions-alphabetic-order/fjcvtzs-floating-point-javascript-convert-to-signed-fixed-point-rounding-toward-zero

* config/aarch64/aarch64.md (UNSPEC_FJCVTZS): Define.
(aarch64_fjcvtzs): New define_insn.
* config/aarch64/aarch64.h (TARGET_JSCVT): Define.
* config/aarch64/aarch64-builtins.c (aarch64_builtins):
Add AARCH64_JSCVT.
(aarch64_init_builtins): Initialize __builtin_aarch64_jcvtzs.
(aarch64_expand_builtin): Handle AARCH64_JSCVT.
* config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
__ARM_FEATURE_JCVT where appropriate.
* config/aarch64/arm_acle.h (__jcvt): Define.

* gcc.target/aarch64/acle/jcvt_1.c: New test.

From-SVN: r275335

[AArch64] Implement ACLE intrinsics for FRINT[32,64][Z,X]

This patch implements the ACLE intrinsics to access the FRINT[32,64][Z,X] scalar[1] and vector[2][3] instructions
from Armv8.5-a. These are enabled when the __ARM_FEATURE_FRINT macro is defined.

They're added in a fairly standard way through builtins and unspecs at the RTL level.

* config/aarch64/aarch64.md ("unspec"): Add UNSPEC_FRINT32Z,
UNSPEC_FRINT32X, UNSPEC_FRINT64Z, UNSPEC_FRINT64X.
(aarch64_<frintnzs_op><mode>): New define_insn.
* config/aarch64/aarch64.h (TARGET_FRINT): Define.
* config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
__ARM_FEATURE_FRINT when appropriate.
* config/aarch64/aarch64-simd-builtins.def: Add builtins for frint32z,
frint32x, frint64z, frint64x.
* config/aarch64/arm_acle.h (__rint32zf, __rint32z, __rint64zf,
__rint64z, __rint32xf, __rint32x, __rint64xf, __rint64x): Define.
* config/aarch64/arm_neon.h (vrnd32z_f32, vrnd32zq_f32, vrnd32z_f64,
vrnd32zq_f64, vrnd32x_f32, vrnd32xq_f32, vrnd32x_f64, vrnd32xq_f64,
vrnd64z_f32, vrnd64zq_f32, vrnd64z_f64, vrnd64zq_f64, vrnd64x_f32,
vrnd64xq_f32, vrnd64x_f64, vrnd64xq_f64): Define.
* config/aarch64/iterators.md (VSFDF): Define.
(FRINTNZX): Likewise.
(frintnzs_op): Likewise.

* gcc.target/aarch64/acle/rintnzx_1.c: New test.
* gcc.target/aarch64/simd/vrndnzx_1.c: Likewise.

From-SVN: r275334

[AArch64] Add support for missing CPUs

This patch adds '-mcpu' options for following CPUs:
Cortex-A77, Cortex-A76AE, Cortex-A65, Cortex-A65AE, and Cortex-A34.

Related specifications are as following:
https://developer.arm.com/ip-products/processors/cortex-a

Bootstraped/regtested for aarch64-none-linux-gnu.

2019-09-03  Dennis Zhang  <dennis.zhang@arm.com>

* config/aarch64/aarch64-cores.def (AARCH64_CORE): New entries
for Cortex-A77, Cortex-A76AE, Cortex-A65, Cortex-A65AE, and
Cortex-A34.
* config/aarch64/aarch64-tune.md: Regenerated.
* doc/invoke.texi: Document the new processors.

From-SVN: r275333

libphobos.exp (libphobos_init): Add multi-lib libgcc dirs to the ld_library_path var.

2019-09-03  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        * testsuite/lib/libphobos.exp (libphobos_init): Add multi-lib libgcc
        dirs to the ld_library_path var.

From-SVN: r275332

[AArch64] Add Linux hwcap strings for some extensions

This patch adds feature strings for some of the extensions. This string is what is read from /proc/cpuinfo on Linux systems
and used during -march=native detection.

The strings are taken from the kernel source tree at:
https://github.com/torvalds/linux/blob/master/arch/arm64/kernel/cpuinfo.c#L45

* config/aarch64/aarch64-option-extensions.def (sb): Add feature
string.
(ssbs): Likewise.
(sve2): Likewise.
(sve2-sm4): Likewise.
(sveaes): Likewise.
(svesha3): Likewise.
(svebitperm): Likewise.

From-SVN: r275331

re PR tree-optimization/91597 (GCC miscompiles a branch depending on a pointer tag)

PR tree-optimization/91597
* tree-vrp.c (extract_range_from_binary_expr): Remove unsafe
BIT_AND_EXPR optimization for pointers, even if both operand
ranges don't include NULL, the result can be NULL.

* gcc.c-torture/execute/pr91597.c: New test.

Co-Authored-By: Richard Biener <rguenther@suse.de>
From-SVN: r275330

[x86 testsuite] preserve full register across main

This test uses a call-saved register as a global variable.  It
attempts to preserve its value across main, but only the lower int
part is preserved, which is not good enough for x86_64, when the
runtime that calls main() happens to hold something in the chosen
register that is not a zero-extension from the 32-bit value, and
rightfully expects the full register to remain unchanged when main()
returns.

for  gcc/testsuite/ChangeLog

* gcc.target/i386/20020616-1.c: Preserve full register across
main.

From-SVN: r275329

Daily bump.

From-SVN: r275328

re PR fortran/91589 (ICE in gfc_conv_component_ref, at fortran/trans-expr.c:2447)

2019-09-02  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/91589
* primary.c (gfc_match_varspec): Return MATCH_NO on an apparent
component ref, when the primary type is intrinsic.

2019-09-02  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/91589
* gfortran.dg/pr91589.f90 : New test.

From-SVN: r275324

re PR fortran/91552 (ICE with valid array constructor)

2019-09-02  Steven G. Kargl  <kargl@gc.gnu.org>

PR fortran/91552
* array.c (walk_array_constructor): New function.
(gfc_match_array_constructor): Use it.

2019-09-02  Steven G. Kargl  <kargl@gc.gnu.org>

PR fortran/91552
* gfortran.dg/pr91552.f90: New test.

From-SVN: r275322

re PR middle-end/91605 (ICE in ix86_avx256_split_vector_move_misalign, at config/i386/i386-expand.c:489 since r274986)

2019-09-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR middle-end/91605
        * expr.c (addr_expr_of_non_mem_decl_p_1): Refactor into...
        (non_mem_decl_p): ...this.
        (mem_ref_refers_to_non_mem_p): Handle DECL_P as well ase MEM_REF.
        (expand_assignment): Call mem_ref_referes_to_non_mem_p
        unconditionally as before.

testsuite:
2019-09-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR middle-end/91605
        * g++.target/i386/pr91605.C: New test.

From-SVN: r275320

re PR tree-optimization/91632 (Probably wrong code since r275026)

PR tree-optimization/91632
* gcc.c-torture/execute/pr91632.c: New test.

From-SVN: r275318

contrib/vimrc: override formatting options for more files

* vim-gcc-dev/syntax/gcc-match.vim: Do not override 'tabstop' here.
* vimrc: Set preferred values for 'tabstop', 'softtabstop',
'shiftwidth', 'noexpandtab', 'textwidth', 'formatoptions' for all
files, not just C-like files.

From-SVN: r275316

Use __constinit keyword in libstdc++ sources

* src/c++17/memory_resource.cc: Use __constinit keyword.

From-SVN: r275315

Update libstdc++ docs for library version bumps

* doc/xml/manual/abi.xml: Document 9.x library versions.
* doc/html/*: Regenerate.

From-SVN: r275314

Minor simplifications for std::to_chars implementation

* include/std/charconv (__detail::__to_chars_2_len): Use std::log2p1.
(__detail::__to_chars_8_len): Remove.
(__detail::__to_chars_8): Inline length calculation here.
(__detail::__from_chars_binary): Use numeric_limits instead of
CHAR_BIT.

From-SVN: r275313

re PR target/91323 (LTGT rtx produces UCOMISS instead of COMISS)

PR target/91323
* doc/generic.texi (LTGT_EXPR): Merge with other comparison operators.
* rtl.def (LTGT): Likewise.  Add note about floating-point exceptions.
* tree.def (LTGT_EXPR): Likewise.
* config/sparc/sparc.c (select_cc_mode): Return CCFPEmode for LTGT.

From-SVN: r275303

re PR go/91617 (Many go test case failures after r275026)

PR go/91617
* fold-const.c (range_check_type): For enumeral and boolean
type, pass 1 to type_for_size langhook instead of
TYPE_UNSIGNED (etype).  Return unsigned_type_for result whenever
etype isn't TYPE_UNSIGNED INTEGER_TYPE.
(build_range_check): Don't call unsigned_type_for for pointer types.
* match.pd (X / C1 op C2): Don't call unsigned_type_for on
range_check_type result.

From-SVN: r275299

Update Solaris baselines for GCC 10.0

* config/abi/post/i386-solaris/baseline_symbols.txt: Regenerate.
* config/abi/post/i386-solaris/amd64/baseline_symbols.txt: Likewise.
* config/abi/post/sparc-solaris/baseline_symbols.txt: Likewise.
* config/abi/post/sparc-solaris/sparcv9/baseline_symbols.txt:
Likewise.

From-SVN: r275298

gimple-ssa-strength-reduction.c (valid_mem_ref_cand_p): New function.

* gimple-ssa-strength-reduction.c (valid_mem_ref_cand_p): New function.
(replace_ref): Do not replace a chain of only two candidates which are
valid memory references.

From-SVN: r275297

Set tabstop=8 for gcc-match file types.

2019-09-02  Martin Liska  <mliska@suse.cz>

* vim-gcc-dev/syntax/gcc-match.vim: Set tabstop=8.

From-SVN: r275295

Fix thinko in early bail out in tree-switch-conversion.

2019-09-02  Martin Liska  <mliska@suse.cz>

* tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
Bail out when we'll end up with the same number of clusters as
at the beginning.
(bit_test_cluster::find_bit_tests): Likewise for bit tests.
(jump_table_cluster::can_be_handled): Remove the guard
as it's already handled in ::is_enabled.  Allocate output
after early bail out.

From-SVN: r275293

Use cxx_printable_name for __PRETTY_FUNCTION__ in cp_fname_init.

2019-09-02  Martin Liska  <mliska@suse.cz>

PR c++/91155
* c-common.c (fname_as_string): Use cxx_printable_name for
__PRETTY_FUNCTION__ same as was used before r265711.
2019-09-02  Martin Liska  <mliska@suse.cz>

PR c++/91155
* g++.dg/torture/pr91155.C: New test.

From-SVN: r275292

Consider also negative edges in cycle detection.

2019-09-02  Martin Liska  <mliska@suse.cz>

PR gcov-profile/91601
* gcov.c (path_contains_zero_cycle_arc): Rename to ...
(path_contains_zero_or_negative_cycle_arc): ... this and handle
also negative edges.
(circuit): Handle also negative edges as they can happen
in some situations.

From-SVN: r275291

Daily bump.

From-SVN: r275290

PR c++/91129 - wrong error with binary op in template argument.

* typeck.c (warn_for_null_address): Use fold_for_warn instead of
fold_non_dependent_expr.
(cp_build_binary_op): Likewise.

* g++.dg/cpp1y/nontype1.C: New test.

From-SVN: r275285

2019-09-01  François Dumont  <fdumont@gcc.gnu.org>

* testsuite_files/util/testsuite_performance.h
(resource_counter::start): Ignore unused malloc(0) result.

From-SVN: r275284

[testsuite] Unsupport 20190827-1.c for targets without alias support.

gcc/testsuite/

2019-09-01  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.c-torture/compile/20190827-1.c: Add dg-requires-alias.

From-SVN: r275274

target-supports.exp (check_effective_target_pthread): Add #include <pthread.h> directive to the test.

* lib/target-supports.exp (check_effective_target_pthread): Add
#include <pthread.h> directive to the test.

From-SVN: r275271

re PR target/91472 (gmp testsuite segfaults with gcc-8 and gcc-9, works fine with gcc-7)

PR target/91472
* config/sparc/sparc.c (sparc_cannot_force_const_mem): Return true
during LRA/reload in PIC mode if the PIC register hasn't been used yet.
(sparc_pic_register_p): Test reload_in_progress for consistency's sake.

From-SVN: r275270

array.c (spec_dimen_size): Check for the presence of expressions for the bounds.

2019-09-01  Paul Thomas  <pault@gcc.gnu.org>

* array.c (spec_dimen_size): Check for the presence of
expressions for the bounds.
* decl.c (gfc_match_end): Add case COMP_SELECT_RANK.
* dump-parse-tree.c(show_symbol): Show the arrayspec of class
entities.
(show_code_node): Show the code for SELECT_RANK.
* expr.c (gfc_check_vardef_context): Omit the context of
variable definition for select rank associate names since the
ASSUMED RANK throws.
* gfortran.h : Add ST_SELECT_RANK and ST_RANK to enum
gfc_statement. Add select_rank_temporary to symbol attribute
structure. Add EXEC_SELECT_RANK to enum gfc_exec_op.
* match.c (match_exit_cycle): Add COMP_SELECT_RANK.
(copy_ts_from_selector_to_associate): Add as special case for
assumed rank class variables.
(select_intrinsic_set_tmp): Clean up the code by using symbols
for references to the temporary and the selector.
(select_type_set_tmp): Ditto.
(select_rank_set_tmp): New function.
(gfc_match_select_rank): New function.
(gfc_match_rank_is): New function.
* match.h : Add prototypes for gfc_match_select_rank and
gfc_match_rank_is.
* parse.c (decode_statement): Attempt to match select_rank and
rank statements.
(next_statement, gfc_ascii_statement): Add ST_SELECT_RANK.
(parse_select_rank_block): New function.
(parse_executable): Parse select rank block for ST_SELECT_RANK.
* parse.h : Add COMP_SELECT_RANK to enum gfc_compile_state.
* resolve.c (resolve_variable): Exclude select_rank_temporaries
from the check on use of ASSUMED RANK.
(gfc_resolve_expr): Make sure that unlimited polymorphic select
rank temporaries expressions are not resolved again after being
successfully resolved.
(resolve_assoc_var): Do not do the rank check for select rank
temporaries.
(resolve_select_rank): New function.
(gfc_resolve_blocks): Deal with case EXEC_SELECT_RANK.
(resolve_symbol): Exclude select rank temporaries for check on
use of ASSUMED RANK.
* st.c (gfc_free_statement): Include EXEC_SELECT_RANK.
* trans-array.c (gfc_conv_array_ref): Select rank temporaries
may have dimen == 0.
(gfc_conv_expr_descriptor): Zero the offset of select rank
temporaries.
* trans-stmt.c (copy_descriptor): New function.
(trans_associate_var): Add code to associate select rank temps.
(gfc_trans_select_rank_cases): New function.
(gfc_trans_select_rank): New function.
* trans-stmt.h : Add prototype for gfc_trans_select_rank.
trans.c (trans_code): Add select rank case.

2019-09-01  Paul Thomas  <pault@gcc.gnu.org>

* gfortran.dg/select_rank_1.f90 : New test.
* gfortran.dg/select_rank_2.f90 : New test.

From-SVN: r275269

policy_data_structures_biblio.xml (COM: Component Model Object Technologies): Adjust name and link.

* doc/xml/manual/policy_data_structures_biblio.xml (COM: Component
Model Object Technologies): Adjust name and link.

From-SVN: r275268

re PR middle-end/91623 (-msse4.1 -O3 segfault in /usr/lib/gcc/x86_64-pc-linux-gnu/8.3.0/include/smmintrin.h:270:10)

PR middle-end/91623
* optabs.c (expand_vec_cond_expr): If op0 is a VECTOR_CST and only
EQ_EXPR/NE_EXPR is supported, verify that op0 only contains
zeros or negative elements and use NE_EXPR instead of LT_EXPR against
zero vector.

* gcc.target/i386/pr91623.c: New test.

From-SVN: r275267

re PR lto/91572 (lto1: error: type variant has different ‘TREE_TYPE’ since r269862)

PR lto/91572
* tree.c (find_decls_types_in_node): Also walk TREE_PURPOSE of
GIMPLE_ASM TREE_LIST operands.

* g++.dg/lto/pr91572_0.C: New test.

From-SVN: r275266

Fix wrong dates in ChangeLog

From-SVN: r275265

Daily bump.

From-SVN: r275264

Add source location to TRAIT_EXPR.

Since TRAIT_EXPR is exceptional, maybe_wrap_with_location won't wrap it, so
we need to put its location in the TRAIT_EXPR node itself.

* cp-tree.h (TRAIT_EXPR_LOCATION): New.
(struct tree_trait_expr): Add locus field.
* parser.c (cp_parser_trait_expr): Pass trait_loc down.
* pt.c (tsubst_copy_and_build) [TRAIT_EXPR]: Likewise.
* semantics.c (finish_trait_expr): Add location parm.
* tree.c (cp_expr_location): Handle TRAIT_EXPR.

From-SVN: r275260

* es.po: Update.

From-SVN: r275258

generic.texi (Unary and Binary Expressions): Mark up an instance of TYPE_MIN.

* doc/generic.texi (Unary and Binary Expressions): Mark up
an instance of TYPE_MIN.

From-SVN: r275243

or1k: Fix issue with set_got clobbering LR (r9)

When compiling glibc we found that the GOT register was being allocated
r9 when the instruction was still set_got_tmp.  That is a problem
because r9 is the Link Register (LR) in OpenRISC which is used/clobbered
in set_got.  We cannot use r9 as the GOT register.  Also, we cannot
simply say set_got_tmp clobbers r9 as this is the reason for having the
temporary set_got_tmp.

Fix by using a register class constraint that does not allow r9 during
register allocation.

gcc/ChangeLog:

        * config/or1k/constraints.md (t): New constraint.
        * config/or1k/or1k.h (GOT_REGS): New register class.
        * config/or1k/or1k.md (set_got_tmp, set_got): Use t contraint.

From-SVN: r275242

compiler: check for notinheap struct at each struct field

    When generating write barriers, we were only checking for a notinheap
    struct at the outermost struct.  That mishandled the case of setting a
    pointer to a notinheap struct as a field of another struct that is not
    notinheap.  This caused an invalid write barrier error when building
    the 1.13 version of the runtime.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/192279

From-SVN: r275240

compiler, runtime: support and use single argument go:linkname

    The gc compiler has started permitting go:linkname comments with a
    single argument to mean that a function should be externally visible
    outside the package.  Implement this in the Go frontend.

    Change the libgo runtime package to use it, rather than repeating the
    name just to export a function.

    Remove a couple of unnecessary go:linkname comments on declarations.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/192197

From-SVN: r275239

compiler: don't report runtime escapes if we've seen errors

    If we get errors during compilation, we skip the escape analysis pass.
    If we are compiling the runtime package, we report an error if a bound
    method expression escapes.  The effect is that if we get an error
    while compiling the runtime package, we would report confusing and
    meaningless errors about bound method expressions escaping.

    This CL stops doing that.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/192720

From-SVN: r275238

runtime: always build panic32.go

    Avoids problems with arm64 ILP32 mode.  We might want to handle that
    mode better in general, but always building panic32.go is simple and
    fixes the build.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/192723

From-SVN: r275237

re PR fortran/91587 (ICE in gfc_resolve_filepos, at fortran/io.c:2913)

2019-08-30  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/91587
* io.c (match_filepos): MATCH_ERROR should branch to a syntax error.

2019-08-30  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/91587
* gfortran.dg/pr91587.f90: New test.

From-SVN: r275236

Daily bump.

From-SVN: r275235

RISC-V: Disable -msave-restore for shared libraries.

This was noticed while trying to test -msave-restore support.  The
save/restore routines use the alternate return register t0/x5 which is
clobbered by the PLT header, so we can't use them in shared libraries.
This patch disables -msave-restore when -fpic (and -mplt), and emits a
warning if the user explicitly turned on -msave-restore.

gcc/
* config/riscv/riscv.c (riscv_option_override): If -msave-restore
and -fpic and -mplt then disable -msave-restore and warn.

From-SVN: r275231

compile, runtime: permit anonymous and empty fields in C header

    Permit putting structs with anonymous and empty fields in the C header
    file runtime.inc that is used to build the C runtime code.  This is
    required for upcoming 1.13 support, as the m struct has picked up an
    anonymous field.

    Doing this lets the C header contain all the type descriptor structs,
    so start using those in the C code.  This cuts the number of copies of
    type descriptor definitions from 3 to 2.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/192343

From-SVN: r275227

Fix ChangeLog

From-SVN: r275223

PR middle-end/91599 - GCC does not say where warning is happening

gcc/ChangeLog:

PR middle-end/91599
* tree-ssa-strlen.c (handle_store): Use a fallback location if
the statement doesn't have one.
* gimple-pretty-print.c (percent_G_format): Same.

gcc/testsuite/ChangeLog:

PR middle-end/91599
* gcc.dg/Wstringop-overflow-16.c: New test.

From-SVN: r275211

PR middle-end/91584 - Bogus warning from -Warray-bounds during string assignment

gcc/ChangeLog:

PR middle-end/91584
* tree-vrp.c (vrp_prop::check_mem_ref): Normalize type domain bounds
before using them to validate MEM_REF offset.

gcc/testsuite/ChangeLog:
* gfortran.dg/char_array_constructor_4.f90: New test.

From-SVN: r275210

Optimize std::to_chars

Bunch of micro optimizations for std::to_chars:
* For base == 8 replacing the lookup in __digits table with arithmetic
computations leads to a same CPU cycles for a loop (exchanges two movzx
with 3 bit ops). However this saves 129 bytes of data and totally avoids
a chance of cache misses on __digits.
* For base == 16 replacing the lookup in __digits table with
arithmetic computations leads to a few additional instructions, but
totally avoids a chance of cache misses on __digits (- ~9 cache misses
for worst case) and saves 513 bytes of const data.
* Replacing __first[pos] and __first[pos - 1] with __first[1] and
__first[0] on final iterations saves ~2% of code size.
* Removing trailing '\0' from arrays of digits allows the linker to
merge the symbols (so that "0123456789abcdefghijklmnopqrstuvwxyz" and
"0123456789abcdef" could share the same address). This improves data
locality and reduces binary sizes.
* Using __detail::__to_chars_len_2 instead of a generic
__detail::__to_chars_len makes the operation O(1) instead of O(N). It
also makes the code two times shorter.

In sum: this significantly reduces the size of a binary (for about 4KBs
only for base-8 conversion), deals with latency (CPU cache misses)
without changing the iterations count and without adding costly
instructions into the loops.

2019-08-30  Antony Polukhin  <antoshkka@gmail.com>

* include/std/charconv (__detail::__to_chars_8)
__detail::__to_chars_16): Replace array of precomputed digits with
arithmetic operations to avoid CPU cache misses. Remove zero
termination from array of digits to allow symbol merge with generic
implementation of __detail::__to_chars. Replace final offsets with
constants. Use __detail::__to_chars_len_2 instead of a generic
__detail::__to_chars_len.
(__detail::__to_chars): Remove zero termination from array of digits.
(__detail::__to_chars_2): Leading digit is always '1'.

From-SVN: r275205

Fix errors in new test

* testsuite/23_containers/vector/cons/89164_c++17.cc: Fix errors.

From-SVN: r275204

decl.c (maybe_saturate_size): New function.

* gcc-interface/decl.c (maybe_saturate_size): New function.
(gnat_to_gnu_entity): Invoke it on the Esize of types before sending
it for back-annotations.
* gcc-interface/trans.c: Fix typo.

From-SVN: r275200

baseline_symbols.txt: Update.

* config/abi/post/alpha-linux-gnu/baseline_symbols.txt: Update.

From-SVN: r275199

ada-tree.h (DECL_FORCED_BY_REF_P): New macro.

* gcc-interface/ada-tree.h (DECL_FORCED_BY_REF_P): New macro.
* gcc-interface/decl.c (gnat_to_gnu_param): Set it on parameters
whose mechanism was forced to by-reference.
* gcc-interface/trans.c (Call_to_gnu): Do not issue a warning about a
misaligned actual parameter if it is based on a CONSTRUCTOR.  Remove
  obsolete warning for users of Starlet.  Issue a warning if a temporary
is make around the call for a parameter with DECL_FORCED_BY_REF_P set.
(addressable_p): Return true for REAL_CST and ADDR_EXPR.

From-SVN: r275198

trans.c (gnat_to_gnu): Do not set the location on an expression used for a tag.

* gcc-interface/trans.c (gnat_to_gnu): Do not set the location on an
expression used for a tag.

From-SVN: r275197

gigi.h (aggregate_type_contains_array_p): Declare.

* gcc-interface/gigi.h (aggregate_type_contains_array_p): Declare.
* gcc-interface/decl.c (gnat_to_gnu_entity) <E_Record_Type>: For an
extension, test Has_Record_Rep_Clause instead of Has_Specified_Layout.
(adjust_packed): Return 0 if the type of the field is an aggregate
type that contains (or is) a self-referential array.
(type_has_variable_size): Delete.
* gcc-interface/utils.c (inish_record_type): Constify a variable.
(aggregate_type_contains_array_p): Add parameter SELF_REFERENTIAL.
<RECORD_TYPE>: Pass it in the recursive call.
<ARRAY_TYPE>: If it is true, return true only if the array type is
self-referential.
(create_field_decl): Streamline the setting of the alignment on the
field.  Pass false to aggregate_type_contains_array_p.

From-SVN: r275196

* gcc.target/mips/r10k-cache-barrier-9.c: Suppress warnings.

From-SVN: r275195

trans.c (lvalue_required_p): Adjust GNU_TYPE in the recursive call.

* gcc-interface/trans.c (lvalue_required_p) <N_Slice>: Adjust GNU_TYPE
in the recursive call.
<N_Selected_Component>: Likewise.

From-SVN: r275191

utils.c (build_template): Deal with parameters passed by pointer to component of multi-dimensional arrays.

* gcc-interface/utils.c (build_template): Deal with parameters
passed by pointer to component of multi-dimensional arrays.

From-SVN: r275190

decl.c (annotate_value): Inline the call also if List_Representation_Info is greater than 3.

* gcc-interface/decl.c (annotate_value) <CALL_EXPR>: Inline the call
also if List_Representation_Info is greater than 3.

From-SVN: r275188

* gcc-interface/trans.c (Attribute_to_gnu) <Attr_Size>: Add assertion.

From-SVN: r275187

* doc/invoke.texi (-Wvolatile): Use @code for volatile.

From-SVN: r275178

PR libstdc++/89164 enforce constraints for uninitialized algos

The memmove optimizations for std::uninitialized_copy/fill/_n will
compile even if the type is not copy constructible, because std::copy
doesn't require copy construction to work. But the uninitialized
algorithms do require it.

This adds explicit static assertions to ensure we don't allow ill-formed
initializations.

PR libstdc++/89164
* include/bits/stl_algobase.h (__copy_move): Give descriptive names
to template parameters.
* include/bits/stl_uninitialized.h (uninitialized_copy)
(uninitialized_fill, uninitialized_fill_n): Add static assertions to
diagnose invalid uses.
* testsuite/20_util/specialized_algorithms/uninitialized_copy/1.cc:
Adjust expected error.
* testsuite/20_util/specialized_algorithms/uninitialized_copy/89164.cc:
New test.
* testsuite/20_util/specialized_algorithms/uninitialized_copy_n/
89164.cc: New test.
* testsuite/20_util/specialized_algorithms/uninitialized_fill/89164.cc:
New test.
* testsuite/20_util/specialized_algorithms/uninitialized_fill_n/
89164.cc: New test.
* testsuite/23_containers/vector/cons/89164.cc: New test.
* testsuite/23_containers/vector/cons/89164_c++17.cc: New test.

From-SVN: r275177

gigi.h (gigi_checking_assert): New macro.

* gcc-interface/gigi.h (gigi_checking_assert): New macro.
* gcc-interface/decl.c (gnat_to_gnu_entity) <E_Modular_Integer_Type>:
Remove redundant test and adjust comments.  Minor tweaks.
* gcc-interface/trans.c (Call_to_gnu): Do not generate range checks,
instead assert that the Do_Range_Check flag is not set.  Adjust call
to convert_with_check.
(gnat_to_gnu): Likewise.
(assoc_to_constructor): Likewise.
(pos_to_constructor): Likewise.  Remove GNAT_COMPONENT_TYPE parameter.
(emit_range_check): Delete.
(convert_with_check): Remove RANGE_P parameter and adjust.  Do a single
overflow check for modular types.

From-SVN: r275174

arm.md (unaligned_loaddi, [...]): New unspec insn patterns.

2019-08-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        * config/arm/arm.md (unaligned_loaddi,
        unaligned_storedi): New unspec insn patterns.
        * config/arm/neon.md (unaligned_storev8qi): Likewise.
        * config/arm/arm.c (gen_cpymem_ldrd_strd): Use unaligned_loaddi
        and unaligned_storedi for 4-byte aligned memory.
        (arm_block_set_aligned_vect): Use unaligned_storev8qi for
        4-byte aligned memory.

From-SVN: r275063

[PR 91579] Avoid creating redundant PHI nodes in tail-call pass

2019-08-30  Martin Jambor  <mjambor@suse.cz>

tree-optimization/91579
* tree-tailcall.c (tailr_arg_needs_copy): New variable.
(find_tail_calls): Allocate tailr_arg_needs_copy and set its bits as
appropriate.
(arg_needs_copy_p): Removed.
(eliminate_tail_call): Test tailr_arg_needs_copy instead of calling
arg_needs_copy_p.
(tree_optimize_tail_calls_1): Likewise.  Free tailr_arg_needs_copy.

testsuite/
* gcc.dg/tree-ssa/pr91579.c: New test.

From-SVN: r275062

Daily bump.

From-SVN: r275059

* config/i386/i386-features.c
(general_scalar_chain::compute_convert_gain):
Correct cost for double-word shifts.
(general_scalar_to_vector_candidate_p): Reject count operands
greater or equal to mode bitsize.

From-SVN: r275055

i386.c (inline_secondary_memory_needed): Return true for moves between SSE and non-general registers and between mask...

* config/i386/i386.c (inline_secondary_memory_needed): Return true
for moves between SSE and non-general registers and between
mask and non-general registers.
(ix86_register_move_cost): Remove stalled comment.

From-SVN: r275050

* ChangeLog: Fix wrong ChangeLog of my last entry.

From-SVN: r275049

[preprocessor] Include stacking

https://gcc.gnu.org/ml/gcc-patches/2019-08/msg01971.html
* internal.h (enum include_type): Add IT_MAIN, IT_DIRECTIVE_HWM,
IT_HEADER_HWM.
(_cpp_stack_file): Take include_type, not a bool.
* files.c (_cpp_find_file): Refactor to not hide an if inside a
for conditional.
(should_stack_file): Break apart to ...
(is_known_idempotent_file, has_unique_contents): ... these.
(_cpp_stack_file): Replace IMPORT boolean with include_type enum.
Refactor to use new predicates.  Do linemap compensation here ...
(_cpp_stack_include): ... not here.
* init.c (cpp_read_main_file): Pass IT_MAIN to _cpp_stack_file.

From-SVN: r275034

PR libstdc++/91067 add more missing exports for directory iterators

PR libstdc++/91067
* acinclude.m4 (libtool_VERSION): Bump to 6:28:0.
* configure: Regenerate.
* config/abi/pre/gnu.ver (GLIBCXX_3.4.28): Add new version. Export
missing symbols.
* testsuite/27_io/filesystem/iterators/91067.cc: Test move
constructors.
* testsuite/util/testsuite_abi.cc: Add new symbol version.

From-SVN: r275032

i386-features.c (general_scalar_chain::convert_insn): Guard debug work with MAY_HAVE_DEBUG_BIND_INSNS.

2019-08-29  Richard Biener  <rguenther@suse.de>

* config/i386/i386-features.c (general_scalar_chain::convert_insn):
Guard debug work with MAY_HAVE_DEBUG_BIND_INSNS.

From-SVN: r275031

re PR bootstrap/91580 (i686-{darwin, linux} bootstrap fails after r274926)

2019-08-29  Richard Biener  <rguenther@suse.de>

PR bootstrap/91580
* config/i386/i386-features.c (general_scalar_chain::convert_insn):
Do not emit scalar copies for debug-insns, instead replace
their uses with the reg copy used in the chain or reset them
if there is a reaching definition outside of the chain as well.

From-SVN: r275030

baseline_symbols.txt: Update.

* config/abi/post/aarch64-linux-gnu/baseline_symbols.txt: Update.
* config/abi/post/s390x-linux-gnu/baseline_symbols.txt: Update.

From-SVN: r275028

re PR target/91560 (Try harder for AVX non-AVX2 cross-lane permutations)

PR target/91560
* config/i386/i386-expand.c (expand_vec_perm_movs,
expand_vec_perm_blend, expand_vec_perm_vpermil,
expand_vec_perm_pshufb, expand_vec_perm_1,
expand_vec_perm_pshuflw_pshufhw, expand_vec_perm_palignr,
expand_vec_perm_interleave2, expand_vec_perm_vpermq_perm_1,
expand_vec_perm_vperm2f128, expand_vec_perm_interleave3,
expand_vec_perm_vperm2f128_vblend, expand_vec_perm_2vperm2f128_vshuf,
expand_vec_perm_even_odd, expand_vec_perm_broadcast): Adjust function
comments - replace ix86_expand_vec_perm_builtin_1 with
ix86_expand_vec_perm_const_1.
(expand_vec_perm2_vperm2f128_vblend): New function.
(ix86_expand_vec_perm_const_1): New forward declaration.  Call
expand_vec_perm2_vperm2f128_vblend as last resort.
(canonicalize_perm): Formatting fix.

* gcc.dg/torture/vshuf-8.inc: Add two further permutations.

From-SVN: r275027

re PR tree-optimization/91351 (-fstrict-enums generates incorrect code)

PR tree-optimization/91351
* tree-cfg.c (generate_range_test): Use range_check_type instead of
unsigned_type_for.
* tree-cfgcleanup.c (convert_single_case_switch): Punt if
range_check_type returns NULL.
* tree-switch-conversion.c (switch_conversion::build_one_array):
Use range_check_type instead of unsigned_type_for, don't perform
linear opt if it returns NULL.
(bit_test_cluster::find_bit_tests): Formatting fix.
(bit_test_cluster::emit): Use range_check_type instead of
unsigned_type_for.
(switch_decision_tree::try_switch_expansion): Punt if range_check_type
returns NULL.

* g++.dg/opt/pr91351.C: New test.

From-SVN: r275026

decl.c (check_var_type): Add location_t parameter and use it.

/cp
2019-08-29  Paolo Carlini  <paolo.carlini@oracle.com>

* decl.c (check_var_type): Add location_t parameter and use it.
(grokdeclarator): Adjust call.
* pt.c (tsubst_decl): Likewise.
* cp-tree.h: Adjust declaration.

/testsuite
2019-08-29  Paolo Carlini  <paolo.carlini@oracle.com>

* g++.dg/spellcheck-typenames.C: Adjust expected locations.
* g++.dg/cpp0x/pr84676.C: Check locations.
* g++.dg/other/pr88187.C: Likewise.
* g++.dg/parse/crash13.C: Likewise.
* g++.dg/parse/crash46.C: Likewise.
* g++.dg/parse/template28.C: Likewise.
* g++.dg/parse/typename4.C: Likewise.

From-SVN: r275025

re PR tree-optimization/91568 (internal compiler error: in vect_schedule_slp_instance, at tree-vect-slp.c:3922)

2019-08-29  Richard Biener  <rguenther@suse.de>

PR tree-optimization/91568
* tree-vectorizer.h (_slp_tree::max_nunits): Add.
(vect_update_max_nunits): Add overload for poly_uint64.
* tree-vect-slp.c (vect_create_new_slp_node): Initialize it.
(vect_build_slp_tree): Record max_nunits into the subtree
and merge it upwards.
(vect_print_slp_tree): Print max_nunits.

* gfortran.dg/pr91568.f: New testcase.

From-SVN: r275023

Implement P1152R4: Deprecating some uses of volatile.

PR c++/91361
* c-opts.c (c_common_post_options): Enable -Wvolatile by
default for C++2a, unless -Wno-deprecated.
* c.opt (Wvolatile): New warning.

* cp-gimplify.c (cp_fold): Set TREE_THIS_VOLATILE.
* decl.c (grokdeclarator): Warn about a volatile-qualified structured
binding and return type.
(grokparms): Warn about a volatile-qualified function parameter.
* expr.c (mark_use) <case MODIFY_EXPR>: Emit a -Wvolatile warning.
* typeck.c (cp_build_unary_op): Emit a -Wvolatile warning for pre and
post ++/-- on a volatile operand.
(genericize_compound_lvalue): Use a better location.  Don't lose
TREE_THIS_VOLATILE.
(cp_build_modify_expr): Emit a -Wvolatile warning for a compound
assignment whose LHS is volatile-qualified.  Build the assignment with
a more precise location.

* doc/invoke.texi: Document -Wvolatile.

* c-c++-common/Wbool-operation-1.c: Use -Wno-volatile in C++.
* c-c++-common/gomp/atomic-1.c: Likewise.
* c-c++-common/gomp/atomic-9.c: Likewise.
* c-c++-common/gomp/depend-iterator-1.c: Likewise.
* c-c++-common/gomp/loop-1.c: Adjust warning location for C++.
* c-c++-common/gomp/order-3.c: Likewise.
* c-c++-common/pr69733.c: Use -Wno-volatile in C++.
* c-c++-common/spec-barrier-2.c: Likewise.
* c-c++-common/tm/pr54893.c: Likewise.
* g++.dg/cpp0x/pr65327.C: Add dg-warning.
* g++.dg/cpp0x/rv-conv2.C: Likewise.
* g++.dg/cpp0x/rv1n.C: Likewise.
* g++.dg/cpp0x/rv1p.C: Likewise.
* g++.dg/cpp0x/rv2n.C: Likewise.
* g++.dg/cpp0x/rv2p.C: Likewise.
* g++.dg/cpp0x/rv3n.C: Likewise.
* g++.dg/cpp0x/rv3p.C: Likewise.
* g++.dg/cpp0x/rv4n.C: Likewise.
* g++.dg/cpp0x/rv4p.C: Likewise.
* g++.dg/cpp0x/rv5n.C: Likewise.
* g++.dg/cpp0x/rv5p.C: Likewise.
* g++.dg/cpp0x/rv6n.C: Likewise.
* g++.dg/cpp0x/rv6p.C: Likewise.
* g++.dg/cpp0x/rv7n.C: Likewise.
* g++.dg/cpp0x/rv7p.C: Likewise.
* g++.dg/cpp0x/rv8p.C: Likewise.
* g++.dg/cpp0x/trailing14.C: Use -Wno-volatile.
* g++.dg/cpp1y/new1.C: Add dg-warning.
* g++.dg/cpp2a/volatile1.C: New test.
* g++.dg/cpp2a/volatile2.C: New test.
* g++.dg/cpp2a/volatile3.C: New test.
* g++.dg/cpp2a/volatile4.C: New test.
* g++.dg/expr/bool3.C: Add dg-warning.
* g++.dg/expr/bool4.C: Likewise.
* g++.dg/expr/cond9.C: Likewise.
* g++.dg/ext/vector25.C: Likewise.
* g++.dg/gomp/depend-iterator-1.C: Use -Wno-volatile.
* g++.dg/inherit/covariant21.C: Add dg-warning.
* g++.dg/init/ref18.C: Likewise.
* g++.dg/ipa/pr63838.C: Likewise.
* g++.dg/overload/rvalue2.C: Likewise.
* g++.dg/parse/semicolon4.C: Likewise.
* g++.dg/warn/Wreturn-type-4.C: Likewise.
* g++.dg/warn/pr36069.C: Likewise.
* g++.old-deja/g++.mike/p9506.C: Likewise.
* g++.old-deja/g++.other/volatile1.C: Likewise.

From-SVN: r275022

Daily bump.

From-SVN: r275021

* es.po: Update.

From-SVN: r275012

Update ABI baselines for x86 and powerpc GNU targets

* config/abi/post/i386-linux-gnu/baseline_symbols.txt: Update.
* config/abi/post/i486-linux-gnu/baseline_symbols.txt: Update.
* config/abi/post/powerpc-linux-gnu/baseline_symbols.txt: Update.
* config/abi/post/powerpc64-linux-gnu/32/baseline_symbols.txt: Update.
* config/abi/post/powerpc64-linux-gnu/baseline_symbols.txt: Update.
* config/abi/post/x86_64-linux-gnu/32/baseline_symbols.txt: Update.
* config/abi/post/x86_64-linux-gnu/baseline_symbols.txt: Update.

From-SVN: r275011

runtime: move osinit to Go

    This is a step toward updating libgo to 1.13.  This adds the 1.13
    version of the osinit function to Go code, and removes the
    corresponding code from the C runtime.  This should simplify future updates.
    Some additional 1.13 code was brought in to simplify this change.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/191717

From-SVN: r275010

re PR fortran/91551 (ICE in sort_actual, at fortran/intrinsic.c:4193)

2019-08-28  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/91551
* intrinsic.c (sort_actual): ALLOCATED has one argument. Check for
no argument case.

2019-08-28  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/91551
* gfortran.dg/allocated_3.f90

From-SVN: r275009

PR c++/91360 - Implement C++20 P1143R2: constinit.

* c-common.c (c_common_reswords): Add constinit and __constinit.
(keyword_is_decl_specifier): Handle RID_CONSTINIT.
* c-common.h (enum rid): Add RID_CONSTINIT, RID_FIRST_CXX20, and
RID_LAST_CXX20.
(D_CXX20): Define.
* c-cppbuiltin.c (c_cpp_builtins): Define __cpp_constinit.
* c-format.c (cxx_keywords): Add "constinit".
* c.opt (Wc++2a-compat, Wc++20-compat): New options.

* cp-tree.h (TINFO_VAR_DECLARED_CONSTINIT): Define.
(LOOKUP_CONSTINIT): Define.
(enum cp_decl_spec): Add ds_constinit.
* decl.c (check_tag_decl): Give an error for constinit in type
declarations.
(check_initializer): Also check LOOKUP_CONSTINIT.
(cp_finish_decl): Add checking for a constinit declaration.  Set
TINFO_VAR_DECLARED_CONSTINIT.
(grokdeclarator): Add checking for a declaration with the constinit
specifier.
* lex.c (init_reswords): Handle D_CXX20.
* parser.c (cp_lexer_get_preprocessor_token): Pass a better location
to warning_at.  Warn about C++20 keywords.
(cp_keyword_starts_decl_specifier_p): Handle RID_CONSTINIT.
(cp_parser_diagnose_invalid_type_name): Add an inform about constinit.
(cp_parser_decomposition_declaration): Maybe pass LOOKUP_CONSTINIT to
cp_finish_decl.
(cp_parser_decl_specifier_seq): Handle RID_CONSTINIT.
(cp_parser_init_declarator): Maybe pass LOOKUP_CONSTINIT to
cp_finish_decl.
(set_and_check_decl_spec_loc): Add "constinit".
* pt.c (tsubst_decl): Set TINFO_VAR_DECLARED_CONSTINIT.
(instantiate_decl): Maybe pass LOOKUP_CONSTINIT to cp_finish_decl.
* typeck2.c (store_init_value): If a constinit variable wasn't
initialized using a constant initializer, give an error.

* doc/invoke.texi: Document -Wc++20-compat.

* g++.dg/cpp2a/constinit1.C: New test.
* g++.dg/cpp2a/constinit2.C: New test.
* g++.dg/cpp2a/constinit3.C: New test.
* g++.dg/cpp2a/constinit4.C: New test.
* g++.dg/cpp2a/constinit5.C: New test.
* g++.dg/cpp2a/constinit6.C: New test.
* g++.dg/cpp2a/constinit7.C: New test.
* g++.dg/cpp2a/constinit8.C: New test.
* g++.dg/cpp2a/constinit9.C: New test.
* g++.dg/cpp2a/constinit10.C: New test.
* g++.dg/cpp2a/constinit11.C: New test.
* g++.dg/cpp2a/constinit12.C: New test.

From-SVN: r275008

re PR fortran/91565 (ICE in gfc_simplify_reshape, at fortran/simplify.c:6707 etc.)

2019-08-28  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/91565
* simplify.c (gfc_simplify_reshape): Add additional checks of the
ORDER dummy argument.

2019-08-28  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/91565
* gfortran.dg/pr91565.f90: New test.

From-SVN: r275007

re PR fortran/91564 (ICE in gimplify_expr, at gimplify.c:14147)

2019-08-28  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/91564
* check.c (gfc_check_kill_sub): Additional checks on status dummy
argument.

2019-08-28  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/91564
* gfortran.dg/pr91564.f90: New test.

From-SVN: r275005

Add .clangd and compile_commands.json to .gitignore.

2019-08-28  Martin Liska  <mliska@suse.cz>

* .gitignore: Add .clangd and compile_commands.json
to .gitignore.

From-SVN: r275004