rs6000: Fix rs6000_keep_leaf_when_profiled

This function is called from elsewhere, so shouldn't be static.

* config/rs6000/rs6000-internal.h (rs6000_keep_leaf_when_profiled): New
declaration.
* config/rs6000/rs6000-logue.c (rs6000_keep_leaf_when_profiled): Remove
"static".
* config/rs6000/rs6000.c (rs6000_keep_leaf_when_profiled): Delete
declaration.

---
 gcc/config/rs6000/rs6000-internal.h | 1 +
 gcc/config/rs6000/rs6000-logue.c    | 4 ++--
 gcc/config/rs6000/rs6000.c          | 1 -
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/gcc/config/rs6000/rs6000-internal.h b/gcc/config/rs6000/rs6000-internal.h
index a1acb66..22ebd37 100644
--- a/gcc/config/rs6000/rs6000-internal.h
+++ b/gcc/config/rs6000/rs6000-internal.h
@@ -99,6 +99,7 @@ extern bool save_reg_p (int reg);
 extern const char * rs6000_machine_from_flags (void);
 extern void emit_asm_machine (void);
 extern bool rs6000_global_entry_point_prologue_needed_p (void);
+extern bool rs6000_keep_leaf_when_profiled (void);

 /* Return true if the OFFSET is valid for the quad address instructions that
    use d-form (register + offset) addressing.  */
diff --git a/gcc/config/rs6000/rs6000-logue.c b/gcc/config/rs6000/rs6000-logue.c
index 9df4b5a..adc137b 100644
--- a/gcc/config/rs6000/rs6000-logue.c
+++ b/gcc/config/rs6000/rs6000-logue.c
@@ -4025,8 +4025,8 @@ rs6000_output_function_prologue (FILE *file)

 /* -mprofile-kernel code calls mcount before the function prolog,
    so a profiled leaf function should stay a leaf function.  */
-static bool
-rs6000_keep_leaf_when_profiled ()
+bool
+rs6000_keep_leaf_when_profiled (void)
 {
   return TARGET_PROFILE_KERNEL;
 }
diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index 3fc4029..bcfc881 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -1338,7 +1338,6 @@ static bool rs6000_secondary_reload_move (enum rs6000_reg_type,
    secondary_reload_info *,
    bool);
 rtl_opt_pass *make_pass_analyze_swaps (gcc::context*);
-static bool rs6000_keep_leaf_when_profiled () __attribute__ ((unused));
 static tree rs6000_fold_builtin (tree, int, tree *, bool);

 /* Hash table stuff for keeping track of TOC entries.  */
--
1.8.3.1

From-SVN: r272691

rs6000: Remove duplicated code

A large portion of the code moved from rs6000.c (to rs6000-logue.c)
was accidentally retained.  This fixes it.

* rs6000.c: Fix previous commit, it missed some changes.

From-SVN: r272690

re PR middle-end/90982 (ICE in make_decl_rtl, at varasm.c:1344)

2019-06-26  Richard Biener  <rguenther@suse.de>

PR ipa/90982
* tree-inline.c (remap_ssa_name): Copy SSA range info.

* g++.dg/torture/pr90982.C: New testcase.

From-SVN: r272688

lto-streamer.h (lto_bitmap_alloc): Remove.

2019-06-26  Richard Biener  <rguenther@suse.de>

* lto-streamer.h (lto_bitmap_alloc): Remove.
(lto_bitmap_free): Likewise.
* lto-streamer.c (lto_bitmap_alloc): Remove.
(lto_bitmap_free): Likewise.
(lto_obstack): Likewise.
(lto_obstack_initialized): Likewise.
* lto-streamer-out.c (lto_output): Use own obstack for local
bitmap, free it consistently.

From-SVN: r272687

Fix sanitizer_common/sanitizer_posix_libcdep.cc compilation on Solaris 11.5

* sanitizer_common/sanitizer_posix_libcdep.cc: Cherry-pick
compiler-rt revision 363778.

From-SVN: r272685

re PR c++/67184 (Missed optimization with C++11 final specifier)

2019-06-26  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/67184
PR c++/69445
* g++.dg/other/final3.C: New.
* g++.dg/other/final5.C: Likewise.

From-SVN: r272675

re PR target/90991 (_mm_loadu_ps instrinsic translates to vmovaps in combination with _mm512_insertf32x4)

PR target/90991
* config/i386/sse.md
(*<extract_type>_vinsert<shuffletype><extract_suf>_0): Use vmovupd,
vmovups, vmovdqu, vmovdqu32 or vmovdqu64 instead of the aligned
insns if operands[2] is misaligned_operand.

* gcc.target/i386/avx512dq-pr90991-1.c: New test.

From-SVN: r272674

[RS6000] Change maddld match_operand from DI to GPR

From PowerPC ISA3.0, the description of `maddld RT, RA.RB, RC` is as follows:
64-bit RA and RB are multiplied and then the RC is signed extend to 128 bits,
and add them together.

We only apply it to 64-bit mode (DI) when implementing maddld.  However, if we
can guarantee that the result of the maddld operation will be limited to 32-bit
mode (SI), we can still apply it to 32-bit mode (SI).

gcc/ChangeLog
2019-06-26  Li Jia He  <helijia@linux.ibm.com>

* config/rs6000/rs6000.h (TARGET_MADDLD): Remove the restriction of
TARGET_POWERPC64.
* config/rs6000/rs6000.md (maddld): Change maddld match_operand from DI
to GPR.

gcc/testsuite/ChangeLog
2019-06-26  Li Jia He  <helijia@linux.ibm.com>

* gcc.target/powerpc/maddld-1.c: New testcase.

From-SVN: r272673

doc: Fix opindex for -W options

@opindex -Wxxx is wrong; it should be @opindex Wxxx.

* doc/invoke.texi (Warning Options): Fix some @opindex syntax.

From-SVN: r272672

Fix one another thinko in tree-vect-loop.c (PR tree-optimization/90973).

2019-06-26  Martin Liska  <mliska@suse.cz>

PR tree-optimization/90973
* tree-vect-loop.c (vect_get_known_peeling_cost): Use
epilogue_cost_vec instead of prologue_cost_vec for
a epilogue cost.

From-SVN: r272671

Fix missing else keyword seen with clang-static-analyzer:

2019-06-26  Martin Liska  <mliska@suse.cz>

* bb-reorder.c (connect_better_edge_p): Add missing else
statement in the middle of if-else statements.

/home/marxin/Programming/gcc/gcc/bb-reorder.c:1031:2: warning: Value stored to 'is_better_edge' is never read
        is_better_edge = true;
        ^                ~~~~
/home/marxin/Programming/gcc/gcc/bb-reorder.c:1034:2: warning: Value stored to 'is_better_edge' is never read
        is_better_edge = false;
        ^                ~~~~~

From-SVN: r272670

PR c++/70462 - unnecessary base ctor variant with final.

As pointed out in the PR, we don't need base 'tor variants for a final
class, since it can never be a base.  I tried also dropping complete
variants for abstract classes, but that runs into ABI compatibility problems
with older releases that refer to those symbols.

* optimize.c (populate_clone_array): Skip base variant if
CLASSTYPE_FINAL.
(maybe_clone_body): We don't need an alias if we are only defining
one clone.

From-SVN: r272669

Enable GCC support for AVX512_VP2INTERSECT which will be in tigerlaker.

There are two instructions for AVX512_VP2INTERSECT:
VP2INTERSECTD and VP2INTERSECTQ.

gcc/
2019-06-05  Hongtao Liu  <hongtao.liu@intel.com>

* common/config/i386/i386-common.c
(OPTION_MASK_ISA_AVX512VP2INTERSECT_SET,
OPTION_MASK_ISA_AVX512VP2INTERSECT_UNSET): New macros.
(OPTION_MASK_ISA2_AVX512F_UNSET): Add
OPTION_MASK_ISA_AVX512VP2INTERSECT_UNSET.
(ix86_handle_option): Handle -mavx512vp2intersect.
* config/i386/avx512vp2intersectintrin.h: New.
* config/i386/avx512vp2intersectvlintrin.h: New.
* config/i386/cpuid.h (bit_AVX512VP2INTERSECT): New.
* config/i386/driver-i386.c (host_detect_local_cpu): Detect
AVX512VP2INTERSECT.
* config/i386/i386-builtin-types.def: Add new types.
* config/i386/i386-builtin.def: Add new builtins.
* config/i386/i386-builtins.c: (enum processor_features): Add
F_AVX512VP2INTERSECT.
(static const _isa_names_table isa_names_table): Ditto.
* config/i386/i386-c.c (ix86_target_macros_internal): Define
__AVX512VP2INTERSECT__.
* config/i386/i386-expand.c (ix86_expand_builtin): Expand
IX86_BUILTIN_2INTERSECTD512, IX86_BUILTIN_2INTERSECTQ512,
IX86_BUILTIN_2INTERSECTD256, IX86_BUILTIN_2INTERSECTQ256,
IX86_BUILTIN_2INTERSECTD128, IX86_BUILTIN_2INTERSECTQ128.
* config/i386/i386-modes.def (P2QI, P2HI): New modes.
* config/i386/i386-options.c (ix86_target_string): Add
-mavx512vp2intersect.
(ix86_option_override_internal): Handle AVX512VP2INTERSECT.
* config/i386/i386.c (ix86_hard_regno_nregs): Allocate two regs for
P2HImode and P2QImode.
(ix86_hard_regno_mode_ok): Register pair only starts at even hardreg
number for P2QImode and P2HImode.
(ix86_regmode_natural_size): New function.
* config/i386/i386.h (TARGET_AVX512VP2INTERSECT,
TARGET_AVX512VP2INTERSECT_P, PTA_AVX512VP2INTERSECT
REGMODE_NATURAL_SIZE, MASK_PAIR_REGNO_P): New.
* config/i386/i386-protos.h (ix86_regmode_natural_size): Declare
* config/i386/i386.opt: Add -mavx512vp2intersect.
* config/i386/immintrin.h: Include avx512vp2intersectintrin.h and
avx512vp2intersectvlintrin.h.
* config/i386/sse.md (define_c_enum "unspec"): Add UNSPEC_VP2INTERSECT.
(define_mode_iterator VI48_AVX512VP2VL): New.
(avx512vp2intersect_2intersect<mode>,
avx512vp2intersect_2intersectv16si): New define_insn patterns.
* config.gcc: Add avx512vp2intersectvlintrin.h and
avx512vp2intersectintrin.h to extra_headers.
* doc/invoke.texi: Document -mavx512vp2intersect.

gcc/testsuite
2019-06-06  Hongtao Liu  <hongtao.liu@intel.com>
    Olga Makhotina  <olga.makhotina@intel.com>

* gcc.target/i386/avx512-check.h: Handle bit_AVX512VP2INTERSECT.
* gcc.target/i386/avx512vp2intersect-2intersect-1a.c: New test.
* gcc.target/i386/avx512vp2intersect-2intersect-1b.c: Likewise.
* gcc.target/i386/avx512vp2intersect-2intersectvl-1a.c: Likewise.
* gcc.target/i386/avx512vp2intersect-2intersectvl-1b.c: Likewise.
* gcc.target/i386/sse-12.c: Add -mavx512vp2intersect.
* gcc.target/i386/sse-13.c: Likewsie.
* gcc.target/i386/sse-14.c: Likewise.
* gcc.target/i386/sse-22.c: Likewise.
* gcc.target/i386/sse-23.c: Likewise.
* g++.dg/other/i386-2.C: Likewise.
* g++.dg/other/i386-3.C: Likewise.

Co-Authored-By: H.J. Lu <hongjiu.lu@intel.com>
Co-Authored-By: Olga Makhotina <olga.makhotina@intel.com>
From-SVN: r272668

re PR fortran/90988 (Wrong error message with variables named "PUBLIC*")

2019-06-24  Steven G. Kargl  <kargl@gcc.gnu.org>

PR Fortran/90988
* decl.c (access_attr_decl): Use temporary variable to reduce
unreadability of code.  Normalize jumping to return.
(gfc_match_protected): Fix parsing error.  Add comments to
explain code.  Remove dead code.
(gfc_match_private): Use temporary variable to reduce unreadability
of code. Fix parsing error.  Move code to test for blank PRIVATE.
Remove dead code.
(gfc_match_public): Move code to test for blank PUBLIC.  Fix
parsing error.  Remove dead code.

2019-06-24  Steven G. Kargl  <kargl@gcc.gnu.org>

PR Fortran/90988
* gfortran.dg/pr90988_1.f90: New test.
* gfortran.dg/pr90988_2.f90: Ditto.
* gfortran.dg/pr90988_3.f90: Ditto.

From-SVN: r272667

libgo/testsuite: ignore symbols with a leading dot in symtogo

    On AIX, a function has two symbols, a text symbol (with a leading dot)
    and a data one (without it).
    As the tests must be run only once, only the data symbol can be used to
    retrieve the final go symbol. Therefore, all symbols beginning with a dot
    are ignored by symtogo.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/177837

From-SVN: r272666

Daily bump.

From-SVN: r272665

cmd/go: silence ar with D flag failures

    The first call of ar must not show its output in order to avoid useless
    error messages about D flag.
    The corresponding Go toolchain patch is CL 182077.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/183817

From-SVN: r272661

Darwin, add a missed file change

From the last commit.

2019-06-25  Iain Sandoe  <iain@sandoe.co.uk>

* config.host: Add libef_ppc.a to the extra files for powerpc-darwin.

From-SVN: r272660

[Darwin, PPC] Move the out of line register save/restore to an endfile.

We have been including this in libgcc, which means that we have to append
-lgcc even when using shared libgcc. In preparation for revision of libgcc
split this into an endfile.

gcc/
2019-06-25  Iain Sandoe  <iain@sandoe.co.uk>

* config/rs6000/darwin.h (ENDFILE_SPEC): New.

libgcc/
2019-06-25  Iain Sandoe  <iain@sandoe.co.uk>

* config.host: Add libef_ppc.a to the extra files for powerpc-darwin.
* config/rs6000/t-darwin: (PPC_ENDFILE_SRC, PPC_ENDFILE_OBJS): New.
Build objects for the out of line save/restore register functions
so that they can be used for any supported Darwin version.
* config/t-darwin: Default the build Darwin version to Darwin8
(MacOS 10.4).

From-SVN: r272659

[PATCH, rs6000] Split up rs6000.c.

The source file rs6000.c has grown to unreasonable size and is being
split up into several smaller source files.  This should improve
compilation speed for building gcc.

This is the first of several patches to do this and moves most of the
prologue/epilogue code to a new source file.

Bootstrapped and tested on powerpc64le-unknown-linux-gnu and
powerpc64-unknown-linux-gnu with no regressions.  Is this ok for trunk?

2019-06-24  Bill Seurer  <seurer@linux.vnet.ibm.com>

* config/rs6000/rs6000.c (stack_info, rs6000_pic_labelno,
savres_routine_syms, savres_routine_name, morestack_ref,
rs6000_init_machine_status, save_reg_p, first_reg_to_save,
first_fp_reg_to_save, first_altivec_reg_to_save, compute_vrsave_mask,
compute_save_world_info, is_altivec_return_reg, fixed_reg_p,
rs6000_savres_strategy, rs6000_stack_info, debug_stack_info,
rs6000_return_addr, rs6000_decl_ok_for_sibcall,
rs6000_function_ok_for_sibcall, rs6000_ra_ever_killed,
rs6000_emit_load_toc_table, rs6000_emit_eh_reg_restore, uses_TOC,
create_TOC_reference, rs6000_aix_asm_output_dwarf_table_ref,
rs6000_emit_stack_tie, rs6000_emit_allocate_stack_1,
get_stack_clash_protection_probe_interval,
get_stack_clash_protection_guard_size,
rs6000_emit_probe_stack_range_stack_clash, rs6000_emit_allocate_stack,
rs6000_emit_probe_stack_range, output_probe_stack_range_1,
interesting_frame_related_regno, output_probe_stack_range_stack_clash,
output_probe_stack_range, rs6000_frame_related, generate_set_vrsave,
gen_frame_set, gen_frame_load, gen_frame_store, emit_frame_save,
gen_frame_mem_offset, rs6000_savres_routine_name,
rs6000_savres_routine_sym, rs6000_emit_stack_reset,
ptr_regno_for_savres, rs6000_emit_savres_rtx,
rs6000_emit_prologue_move_from_cr, split_stack_arg_pointer_used_p,
rs6000_global_entry_point_prologue_needed_p,
rs6000_get_separate_components, rs6000_components_for_bb,
rs6000_disqualify_components, rs6000_emit_prologue_components,
rs6000_emit_epilogue_components, rs6000_set_handled_components,
emit_vrsave_prologue, emit_split_stack_prologue, rs6000_emit_prologue,
rs6000_output_savres_externs, rs6000_output_function_prologue,
rs6000_keep_leaf_when_profiled, load_cr_save, restore_saved_cr,
load_lr_save, restore_saved_lr, add_crlr_cfa_restore,
offset_below_red_zone_p, emit_cfa_restores, rs6000_emit_epilogue,
rs6000_output_function_epilogue, gen_add3_const,
rs6000_expand_split_stack_prologue, rs6000_live_on_entry,
rs6000_split_stack_space_check, rs6000_save_toc_in_prologue_p): Moved
to rs6000-logue.c.
(machine_function): Moved to rs6000.h.
(rs6000_stack_t, ALTIVEC_REG_BIT, quad_address_offset_p) Moved to
rs6000-internal.h.
* config/rs6000/rs6000-logue.c(stack_info, rs6000_pic_labelno,
savres_routine_syms, savres_routine_name, morestack_ref,
rs6000_init_machine_status, save_reg_p, first_reg_to_save,
first_fp_reg_to_save, first_altivec_reg_to_save, compute_vrsave_mask,
compute_save_world_info, is_altivec_return_reg, fixed_reg_p,
rs6000_savres_strategy, rs6000_stack_info, debug_stack_info,
rs6000_return_addr, rs6000_decl_ok_for_sibcall,
rs6000_function_ok_for_sibcall, rs6000_ra_ever_killed,
rs6000_emit_load_toc_table, rs6000_emit_eh_reg_restore, uses_TOC,
create_TOC_reference, rs6000_aix_asm_output_dwarf_table_ref,
rs6000_emit_stack_tie, rs6000_emit_allocate_stack_1,
get_stack_clash_protection_probe_interval,
get_stack_clash_protection_guard_size,
rs6000_emit_probe_stack_range_stack_clash, rs6000_emit_allocate_stack,
rs6000_emit_probe_stack_range, output_probe_stack_range_1,
interesting_frame_related_regno, output_probe_stack_range_stack_clash,
output_probe_stack_range, rs6000_frame_related, generate_set_vrsave,
gen_frame_set, gen_frame_load, gen_frame_store, emit_frame_save,
gen_frame_mem_offset, rs6000_savres_routine_name,
rs6000_savres_routine_sym, rs6000_emit_stack_reset,
ptr_regno_for_savres, rs6000_emit_savres_rtx,
rs6000_emit_prologue_move_from_cr, split_stack_arg_pointer_used_p,
rs6000_global_entry_point_prologue_needed_p,
rs6000_get_separate_components, rs6000_components_for_bb,
rs6000_disqualify_components, rs6000_emit_prologue_components,
rs6000_emit_epilogue_components, rs6000_set_handled_components,
emit_vrsave_prologue, emit_split_stack_prologue, rs6000_emit_prologue,
rs6000_output_savres_externs, rs6000_output_function_prologue,
rs6000_keep_leaf_when_profiled, load_cr_save, restore_saved_cr,
load_lr_save, restore_saved_lr, add_crlr_cfa_restore,
offset_below_red_zone_p, emit_cfa_restores, rs6000_emit_epilogue,
rs6000_output_function_epilogue, gen_add3_const,
rs6000_expand_split_stack_prologue, rs6000_live_on_entry,
rs6000_split_stack_space_check, rs6000_save_toc_in_prologue_p): Moved
to here from rs6000.c.
* config/rs6000/rs6000.h (machine_function): Moved to here from rs6000.c.
* config/rs6000/rs6000-internal.h: (rs6000_stack_t, ALTIVEC_REG_BIT,
quad_address_offset_p) Moved to here from rs6000.c.
* config/rs6000/t-rs6000: Add new source file rs6000-logue.c.
* config/config.gcc: Add new source file rs6000-logue.c to garbage
collector.

From-SVN: r272658

class.c (resolves_to_fixed_type_p): Check CLASSTYPE_FINAL.

* class.c (resolves_to_fixed_type_p): Check CLASSTYPE_FINAL.

If we have a pointer to final class, we know the dynamic type of the object
must be that class, because it can't have any derived classes.

From-SVN: r272656

Put hashtab_chk_error into hash-table.c.

2019-06-25  Martin Liska  <mliska@suse.cz>

* hash-table.c (hashtab_chk_error): Move here from ...
* hash-table.h (hashtab_chk_error): ... here.

From-SVN: r272655

re PR c++/90969 (ICE: tree check: expected array_type, have vector_type in array_ref_low_bound, at tree.c:13570)

PR c++/90969
* constexpr.c (cxx_eval_array_reference): Don't look through VCE from
vector type if lval.

* g++.dg/ext/vector38.C: New test.

From-SVN: r272654

Fix non-portable use of std::abs(double) in constexpr function

Although libstdc++ adds 'constexpr' to its std::abs(floating-point)
overloads (as a non-conforming extension), those overloads are not used
if the target libc provides them, which is the case on Solaris.

The fix is to avoid std::abs and simply apply the negation when needed.

* include/std/numeric (midpoint(T, T)): Avoid std::abs in constexpr
function.

From-SVN: r272653

Transform filter-rtags-warnings to filter-clang-warnings.

2019-06-25  Martin Liska  <mliska@suse.cz>

contrib/filter-clang-warnings.py: Transform from
filter-rtags-warnings.py.

From-SVN: r272652

Remove dead code in df-scan.c (PR tree-optimization/90978).

2019-06-25  Martin Liska  <mliska@suse.cz>

PR tree-optimization/90978
* df-scan.c (df_update_entry_block_defs): Remove dead else
branch.
(df_update_exit_block_uses): Likewise.

From-SVN: r272651

Do not call strlen with NULL argument in libgcov.

2019-06-25  Martin Liska  <mliska@suse.cz>

* libgcov-driver-system.c (replace_filename_variables): Do not
call strlen with NULL argument.

From-SVN: r272650

Enable full GFortran library for AMD GCN

2019-06-25  Kwok Cheung Yeung  <kcy@codesourcery.com>
            Andrew Stubbs  <ams@codesourcery.com>

libgfortran/
* configure: Regenerate.
* configure.ac (LIBGFOR_MINIMAL): Do not use on AMD GCN.

Co-Authored-By: Andrew Stubbs <ams@codesourcery.com>
From-SVN: r272649

Stub implementation of unwinding for AMD GCN.

2019-06-25  Andrew Stubbs  <ams@codesourcery.com>

libgcc/
* config/gcn/t-amdgcn (LIB2ADD): Add unwind-gcn.c.
* config/gcn/unwind-gcn.c: New file.

From-SVN: r272648

Create GCN-specific gthreads

2019-06-25  Kwok Cheung Yeung  <kcy@codesourcery.com>
            Andrew Stubbs  <ams@codesourcery.com>

        gcc/
* config.gcc (thread_file): Set to gcn for AMD GCN.
* config/gcn/gcn.c (gcn_emutls_var_init): New function.
(TARGET_EMUTLS_VAR_INIT): New hook.

config/
* gthr.m4 (GCC_AC_THREAD_HEADER): Add case for gcn.

libgcc/
* configure: Regenerate.
* config/gcn/gthr-gcn.h: New.

Co-Authored-By: Andrew Stubbs <ams@codesourcery.com>
From-SVN: r272647

[PR 90939] Remove outdated assert in ipcp_bits_lattice::meet_with

2019-06-25  Martin Jambor  <mjambor@suse.cz>

PR ipa/90939
* ipa-cp.c (ipcp_bits_lattice::meet_with): Remove assert.

testsuite/
* g++.dg/lto/pr90939_[01].C: New test.

From-SVN: r272646

re PR tree-optimization/90930 (Excessive memory consumption)

2019-06-25  Richard Biener  <rguenther@suse.de>

PR tree-optimization/90930
* tree-ssa-reassoc.c (reassociate_bb): Only rewrite expression
into parallel form in the last pass instance.

* gcc.dg/tree-ssa/reassoc-24.c: Adjust.
* gcc.dg/tree-ssa/reassoc-25.c: Likewise.

From-SVN: r272644

[ARC] Fix for PR89838

gcc/
xxxx-xx-xx  Claudiu Zissulescu  <claziss@synopsys.com>

* config/arc/arc.c (arc_symbol_binds_local_p): New function.
(arc_legitimize_pic_address): Simplify and cleanup the function.
(SYMBOLIC_CONST): Remove.
(prepare_pic_move): Likewise.
(prepare_move_operands): Handle complex mov cases here.
(arc_legitimize_address_0): Remove call to
arc_legitimize_pic_address.
(arc_legitimize_address): Remove call to
arc_legitimize_tls_address.
* config/arc/arc.md (movqi_insn): Allow Cm3 match.
(movhi_insn): Likewise.

/gcc/testsuite
xxxx-xx-xx  Claudiu Zissulescu  <claziss@synopsys.com>

* gcc.target/arc/pr89838.c: New file.

From-SVN: r272643

Implement alternate "__intN__" form of "__intN" type

gcc/ChangeLog:

* gcc/c-family/c-common.c (c_common_nodes_and_builtins): Define
alternate "__intN__" name for "__intN" types.
* gcc/c/c-parser.c (c_parse_init): Create keyword for "__intN__" type.
* gcc/cp/lex.c (init_reswords): Likewise.
* gcc/config/msp430/msp430.h: Use __int20__ for SIZE_TYPE and
PTRDIFF_TYPE.
* gcc/cp/cp-tree.h (cp_decl_specifier_seq): New bitfield "int_n_alt".
* gcc/c/c-decl.c (declspecs_add_type): Don't pedwarn about "__intN" ISO
C incompatibility if alternate "__intN__" form is used.
* gcc/cp/decl.c (grokdeclarator): Likewise.
* gcc/cp/parser.c (cp_parser_simple_type_specifier): Set
decl_specs->int_n_alt if "__intN__" form is used.
* gcc/gimple-ssa-sprintf.c (build_intmax_type_nodes): Accept "__intN__"
format of "__intN" types for UINTMAX_TYPE.
* gcc/brig/brig-lang.c (brig_build_c_type_nodes): Accept "__intN__"
format of "__intN" types for SIZE_TYPE.
* gcc/lto/lto-lang.c (lto_build_c_type_nodes): Likewise.
* gcc/stor-layout.c (initialize_sizetypes): Accept "__intN__"
format of "__intN" types for SIZETYPE.
* gcc/tree.c (build_common_tree_nodes): Accept "__intN__"
format of "__intN" types for SIZE_TYPE and PTRDIFF_TYPE.
* gcc/doc/invoke.texi: Document that __intN__ disables pedantic
warnings.

gcc/testsuite/ChangeLog:

* gcc.target/msp430/mlarge-pedwarns.c: New test.

From-SVN: r272640

tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Check that base2_alias_set is non-zero before doing TBAA based disambiguation.

* tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Check that
base2_alias_set is non-zero before doing TBAA based disambiguation.

From-SVN: r272639

Properly sum costs in tree-vect-loop.c (PR tree-optimization/90973).

2019-06-25  Martin Liska  <mliska@suse.cz>

PR tree-optimization/90973
* tree-vect-loop.c (vect_get_known_peeling_cost): Sum retval
of prologue and epilogue.

From-SVN: r272638

decl.c (gnat_to_gnu_entity): Remove superfluous test in previous change.

* gcc-interface/decl.c (gnat_to_gnu_entity): Remove superfluous test
in previous change.
* gcc-interface/gigi.h (maybe_character_type): Fix formatting.
(maybe_character_value): Likewise.

From-SVN: r272637

re PR sanitizer/90954 (ICE: combining undefined behavior sanitizer with openmp)

PR sanitizer/90954
* c-omp.c (c_finish_omp_atomic): Allow tree_invariant_p in addition
to SAVE_EXPR in first operand of a COMPOUND_EXPR.

* c-c++-common/gomp/pr90954.c: New test.

From-SVN: r272635

pstl_config.h (_PSTL_PRAGMA_SIMD_SCAN, [...]): Define to OpenMP 5.0 pragmas even for GCC 10.0+.

* include/pstl/pstl_config.h (_PSTL_PRAGMA_SIMD_SCAN,
_PSTL_PRAGMA_SIMD_INCLUSIVE_SCAN, _PSTL_PRAGMA_SIMD_EXCLUSIVE_SCAN):
Define to OpenMP 5.0 pragmas even for GCC 10.0+.
(_PSTL_UDS_PRESENT): Define to 1 for GCC 10.0+.

From-SVN: r272634

runtime: mark memequal and memclrNoHeapPointers nosplit

    They are wrappers of libc functions that use no stack. Mark them
    nosplit so the linker won't patch it to call __morestack_non_split.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/183629

From-SVN: r272633

Daily bump.

From-SVN: r272632

ipa-utils.h (type_with_linkage_p): Verify that type is CXX_ODR_P.

* ipa-utils.h (type_with_linkage_p): Verify that type is
CXX_ODR_P.
(odr_type_p): Remove extra return.
* lto-streamer-out.c (hash_tree): Hash TYPE_CXX_ODR_P;
hash STRING_FLAG only for arrays and integers.
* tree-stremaer-in.c (unpack_ts_type_common_value_fields):
Update analogously.
* tree-streamer-out.c (pack_ts_type_common_value_fields):
Likewise.
* print-tree.c (print_node): Print cxx-odr-p
and string-flag.
* tree.c (need_assembler_name_p): Also check that type
is CXX_ODR_TYPE_P
(verify_type_variant): Update verification of SRING_FLAG;
also check CXX_ODR_P.
* tree.h (ARRAY_OR_INTEGER_TYPE_CHECK): New macro.
(TYPE_STRING_FLAG): Use it.
(TYPE_CXX_ODR_P): New macro.

* lto-common.c (compare_tree_sccs_1): Compare CXX_ODR_P;
compare STRING_FLAG only for arrays and integers.

* gcc-interface/decl.c (gnat_to_gnu_entity): Check that
type is array or integer prior checking string flag.
* gcc-interface/gigi.h (gnat_signed_type_for,
maybe_character_value): Likewise.

* c-common.c (braced_lists_to_strings): Check that
type is array or integer prior checking string flag.

* lex.c (cxx_make_type): Set TYPE_CXX_ODR_P.

* dwarf2out.c (gen_array_type_die): First check that type
is an array and then test string flag.

* trans-expr.c (gfc_conv_substring): Check that
type is array or integer prior checking string flag.
(gfc_conv_string_parameter): Likewise.
* trans-openmp.c (gfc_omp_scalar_p): Likewise.
* trans.c (gfc_build_array_ref): Likewise.

From-SVN: r272628

[Darwin, PPC, testsuite] Skip tests for unimplemented functionality.

The -mno-speculate-indirect-jumps functionality is not implemented for
Darwin and, given that it's deprecated, is unlikely to be.

2019-06-24  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.target/powerpc/safe-indirect-jump-1.c: Skip for Darwin.
* gcc.target/powerpc/safe-indirect-jump-7.c: Likewise.

From-SVN: r272627

[Darwin, PPC, testsuite] Fix spec-barr-1.c for Darwin.

We just needed to adjust the regex to accept Darwin's register names.

2019-06-24  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.target/powerpc/spec-barr-1.c: Adjust scan assembler regex
to recognise Darwin's register names.

From-SVN: r272626

[Darwin, testsuite] Fix isystem-2.c.

For the test to succeed there needs to be some header that is to be found in
the 'expected' place i.e. <sysroot>/usr/include/.  It's important that it is
not the name of a header for which fixincludes have been applied, since such
headers will be found in the gcc include-fixed dir and, in general, reference
additional headers.  The dummy sysroot will prevent the additional headers
from being found, resulting in a failed test.  The fix is to use a header name
that isn't expected to be present in a real sysroot.

2019-06-24  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.dg/cpp/isysroot-1.c (main): Use <example.h> as the test header.
* gcc.dg/cpp/usr/include/stdio.h: Rename...
* gcc.dg/cpp/usr/include/example.h: ... to this.

From-SVN: r272625

compiler: open code string equality

    Open code string equality with builtin memcmp. This allows
    further optimizations in the backend.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/183538

From-SVN: r272624

re PR target/90972 (ICE in fold_convert_loc, at fold-const.c:2429)

2019-06-24  Richard Biener  <rguenther@suse.de>

PR tree-optimization/90972
* tree-vect-stmts.c (vect_init_vector): Handle CONSTANT_CLASS_P
in common code, dealing with STRING_CST properly.

* gcc.dg/torture/pr90972.c: New testcase.

From-SVN: r272623

re PR tree-optimization/90930 (Excessive memory consumption)

2019-06-24  Richard Biener  <rguenther@suse.de>

PR tree-optimization/90930
PR tree-optimization/90316
* tree-ssa-alias.c (walk_non_aliased_vuses): Add missing
decrement of limit.

From-SVN: r272621

compiler: use builtin memcmp directly

    Instead of going through a C function __go_memcmp, we can just
    use __builtin_memcmp directly. This allows more optimizations in
    the compiler backend.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/183537

From-SVN: r272620

utils.c (handle_nonnull_attribute): Quote attribute name.

gcc/ada/ChangeLog:

* gcc-interface/utils.c (handle_nonnull_attribute): Quote attribute
name.

gcc/c/ChangeLog:

* c-typeck.c (build_binary_op): Hyphenate floating-point.

gcc/testsuite/ChangeLog:

* gcc.dg/Wfloat-equal-1.c: Adjust text of expected diagnostic.
* gcc.dg/misc-column.c: Ditto.

gcc/ChangeLog:

* tree-pretty-print.h: Remove unnecessary punctuation characters
from a diagnostic.
* tree-ssa.c (release_defs_bitset): Correct preprocessor conditional.

From-SVN: r272619

Don't use a separate CLASSTYPE_AS_BASE if it's the same size.

* class.c (layout_class_type): Don't use a separate
CLASSTYPE_AS_BASE if it's the same size.

From-SVN: r272618

Fix std::midpoint for denormal values

* include/std/numeric (midpoint(T, T)): Change implementation for
floating-point types to avoid incorrect rounding of denormals.
* testsuite/26_numerics/midpoint/floating.cc: Add check for correct
rounding with denormals.
* testsuite/26_numerics/gcd/gcd_neg.cc: Adjust dg-error line numbers.
* testsuite/26_numerics/lcm/lcm_neg.cc: Likewise.

From-SVN: r272616

Define C11 macros such as FLT_DECIMAL_DIG for C++17

* testsuite/18_support/headers/cfloat/values_c++17.cc: New test.

From-SVN: r272615

Daily bump.

From-SVN: r272613

compiler: edit error messages to avoid -Wformat-diag warnings

    GCC recently introduced -Wformat-diag to scrutinize GCC error messages.
    It reports a number of warnings about gofrontend code, such as:

    ../../trunk/gcc/go/gofrontend/import.cc: In member function ‘Type* Import::type_for_index(int, const string&, size_t, bool*)’:
    ../../trunk/gcc/go/gofrontend/import.cc:1129:48: warning: unquoted operator ‘>=’ in format [-Wformat-diag]
     1129 |         "error in %s at %lu: bad type index %d >= %d",
          |                                                ^~

    ../../trunk/gcc/go/gofrontend/ast-dump.cc: In member function ‘void Ast_dump_context::dump(Gogo*, const char*)’:
    ../../trunk/gcc/go/gofrontend/ast-dump.cc:203:25: warning: unquoted option name ‘-fgo-dump-ast’ in format [-Wformat-diag]
      203 |     "cannot open %s:%m, -fgo-dump-ast ignored", dumpname.c_str());
          |                         ^~~~~~~~~~~~~

    ../../trunk/gcc/go/gofrontend/expressions.cc: In static member function ‘static Bexpression* Func_expression::get_code_pointer(Gogo*, Named_object*, Location)’:
    ../../trunk/gcc/go/gofrontend/expressions.cc:1350:29: warning: misspelled term ‘builtin function’ in format; use ‘built-in function’ instead [-Wformat-diag]
     1350 |     "invalid use of special builtin function %qs; must be called",
          |                             ^~~~~~~~~~~~~~~~

    ../../trunk/gcc/go/gofrontend/gogo.cc: In member function ‘void Gogo::add_linkname(const string&, bool, const string&, Location)’:
    ../../trunk/gcc/go/gofrontend/gogo.cc:2527:4: warning: unquoted sequence of 2 consecutive punctuation characters ‘//’ in format [-Wformat-diag]
     2527 |   ("%s is not a function; "
          |   ~^~~~~~~~~~~~~~~~~~~~~~~~
     2528 |    "//go:linkname is only supported for functions"),
          |    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

    This CL edits error messages to avoid these warnings.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/183497

* go.test/test/blank1.go: Update for diagnostic message changes.

From-SVN: r272608

compiler: add go_debug and use it for debug messages

    GCC recently added a new warning -Wformat-diag which does a lot of
    rigorous checks on GCC diagnostic messages.  This produces a number of
    unnecessary diagnostics on gofrontend diagnostic output, such as

    ../../trunk/gcc/go/gofrontend/escape.cc: In member function ‘virtual int Escape_analysis_assign::statement(Block*, size_t*, Statement*)’:
    ../../trunk/gcc/go/gofrontend/escape.cc:1336:33: warning: spurious leading punctuation sequence ‘[’ in format [-Wformat-diag]
     1336 |       go_inform(s->location(), "[%d] %s esc: %s",
          |                                 ^

    ../../trunk/gcc/go/gofrontend/escape.cc: In member function ‘void Escape_analysis_assign::call(Call_expression*)’:
    ../../trunk/gcc/go/gofrontend/escape.cc:1964:17: warning: unquoted operator ‘::’ in format [-Wformat-diag]
     1964 |         "esccall:: indirect call <- %s, untracked",
          |                 ^~

    ../../trunk/gcc/go/gofrontend/escape.cc:1964:34: warning: unbalanced punctuation character ‘<’ in format [-Wformat-diag]
     1964 |         "esccall:: indirect call <- %s, untracked",
          |                                  ^

    Avoid these messages by adding a new function go_debug that uses only
    printf formatting, not GCC diagnostic formatting, and change all the
    optimization debugging messages to use it.  None of the debugging
    messages used the GCC diagnostic formatting specifiers anyhow.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/183437

From-SVN: r272607

[Darwin, PPC, testsuite] Fix builtins-2 for Darwin.

This cannot pass for current Darwin, since it requires VSX and we
don't have any hardware supporting that.  Add a dg-requires clause
for this.

2019-06-23  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.target/powerpc/builtins-2.c: Require VSX hardware support.

From-SVN: r272606

[Darwin, PPC, testsuite] Fix pr80125 testcase for Darwin.

Darwin (unlike most of the members of the PowerPC port family)
defaults to signed chars, so the test was failing to compile with
a "mismatched parameters" error.

2019-06-23  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.target/powerpc/pr80125.c (foo): Use an unsigned char
vector explicitly for the vec_perm.

From-SVN: r272605

[Darwin, PPC, testsuite] Fix builtins-1 testcase for Darwin.

This needs to account for Darwin's __USER_LABEL_PREFIX__.

2019-06-23  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.target/powerpc/builtins-1.c: Account for Darwin's use of
__USER_LABEL_PREFIX__.

From-SVN: r272604

[Darwin, PPC] Handle GCC target pragma.

For compatibility with other members of the port.
Note, that we do not handle the longcall attribute, since longcall
is not required/used on current Darwin.

2019-06-23  Iain Sandoe  <iain@sandoe.co.uk>

* config/rs6000/darwin.h: Handle GCC target pragma.

From-SVN: r272603

[Darwin, PPC] Emit uppercase versions of ppc defines.

Emit __PPC__ (32b) and __PPC64__ (64bit) as per the other members
of the PowerPC port.

2019-06-23  Iain Sandoe  <iain@sandoe.co.uk>

* config/rs6000/darwin.h: (__PPC__, __PPC64__): New.

From-SVN: r272602

[Darwin, PPC, testsuite] Fix pr71785 testcase for Darwin.

Firstly, we adjust the test conditions to use non-PIC code for Darwin.
Secondly, we have to account for out-of-line GPR restores which gives
a false positive on one of the scan-assembler-not.  Lastly, we make the
test a bit more specific for Darwin - that it looks for absence of
branches to local labels.

2019-06-23  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.target/powerpc/pr71785.c: For Darwin, make test non-PIC,
expect the out-of-line GPR restore, and test specifically for
absence of branches to local labels.

From-SVN: r272601

Rename DEFAULT_ARG to DEFERRED_PARSE.

From-SVN: r272600

Daily bump.

From-SVN: r272599

Fix typo in ChangeLog

From-SVN: r272595

re PR fortran/89782 (Can do an internal READ of a character array when it is a parameter, but not a scalar character parameter)

2019-06-22  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR fortran/89782
* io.c (gfc_resolve_dt): Check that internal units are not
character PARAMETER.

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

From-SVN: r272594

[Darwin, PPC, testsuite] Add requires for DFP to two tests.

The two tests use decimal floating point, add the relevant dg-requires
so that they are unsupported on platforms without DFP.

2019-06-22  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.target/powerpc/pr64205.c: Require effective target dfp.
* gcc.target/powerpc/pr79909.c: Likewise.

From-SVN: r272592

[Darwin, PPC, testsuite] Fix darwin-bool-1.c.

This test is failing because of a pedantic warning that is unrelated to the
purpose of the test.  Fixed by suppressing that warning.

2019-06-22  Iain Sandoe  <iain@sandoe.co.uk>

* gcc.target/powerpc/darwin-bool-1.c: Suppress the pedantic
warning about _Bool.

From-SVN: r272591

* config/avr/avr.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Remove.

From-SVN: r272590

re PR c++/65707 (internal compiler error: in unify, at cp/pt.c:18577)

PR c++/65707
PR c++/89480
PR c++/58836
* g++.dg/cpp0x/nondeduced5.C: New test.
* g++.dg/cpp0x/nondeduced6.C: New test.
* g++.dg/cpp0x/nondeduced7.C: New test.

From-SVN: r272589

re PR c++/66256 (noexcept evaluation done before end of class)

PR c++/66256
* g++.dg/cpp0x/noexcept54.C: New test.

From-SVN: r272588

alias-access-path-6.c: New testcase.

* gcc.dg/tree-ssa/alias-access-path-6.c: New testcase.
* tree-ssa-alias.c (nonoverlapping_component_refs_p): Do not
give up on bitfields; continue searching for different refs
appearing later.

From-SVN: r272587

PR c++/86476 - noexcept-specifier is a complete-class context.

PR c++/52869
* cp-tree.def (DEFAULT_ARG): Update commentary.
* cp-tree.h (UNPARSED_NOEXCEPT_SPEC_P): New macro.
(tree_default_arg): Use tree_base instead of tree_common.
(do_push_parm_decls, maybe_check_overriding_exception_spec): Declare.
* decl.c (do_push_parm_decls): New function, broken out of...
(store_parm_decls): ...here.  Call it.
* except.c (nothrow_spec_p): Accept DEFAULT_ARG in the assert.
* parser.c (cp_parser_noexcept_specification_opt,
cp_parser_late_noexcept_specifier, noexcept_override_late_checks):
Forward-declare.
(unparsed_noexcepts): New macro.
(push_unparsed_function_queues): Update initializer.
(cp_parser_direct_declarator): Pass FRIEND_P to
cp_parser_exception_specification_opt.
(inject_parm_decls): New.
(pop_injected_parms): New.
(cp_parser_class_specifier_1): Implement delayed parsing of
noexcept-specifiers.
(cp_parser_save_noexcept): New.
(cp_parser_late_noexcept_specifier): New.
(noexcept_override_late_checks): New.
(cp_parser_noexcept_specification_opt): Add FRIEND_P parameter.  Call
cp_parser_save_noexcept instead of the normal processing if needed.
(cp_parser_exception_specification_opt): Add FRIEND_P parameter and
pass it to cp_parser_noexcept_specification_opt.
(cp_parser_save_member_function_body): Fix comment.
(cp_parser_save_default_args): Maybe save the noexcept-specifier to
post process.
(cp_parser_transaction): Update call to
cp_parser_noexcept_specification_opt.
(cp_parser_transaction_expression): Likewise.
* parser.h (cp_unparsed_functions_entry): Add new field to carry
a noexcept-specifier.
* pt.c (dependent_type_p_r): Handle unparsed noexcept expression.
* search.c (maybe_check_overriding_exception_spec): New function, broken
out of...
(check_final_overrider): ...here.  Call
maybe_check_overriding_exception_spec.
* tree.c (canonical_eh_spec): Handle UNPARSED_NOEXCEPT_SPEC_P.
(cp_tree_equal): Handle DEFAULT_ARG.

* g++.dg/cpp0x/noexcept45.C: New test.
* g++.dg/cpp0x/noexcept46.C: New test.
* g++.dg/cpp0x/noexcept47.C: New test.
* g++.dg/cpp0x/noexcept48.C: New test.
* g++.dg/cpp0x/noexcept49.C: New test.
* g++.dg/cpp0x/noexcept50.C: New test.
* g++.dg/cpp0x/noexcept51.C: New test.
* g++.dg/cpp0x/noexcept52.C: New test.
* g++.dg/cpp0x/noexcept53.C: New test.
* g++.dg/eh/shadow1.C: Adjust dg-error.

From-SVN: r272586

PR c++/90881 - bogus -Wunused-value in unevaluated context.

* cvt.c (convert_to_void): Don't emit unused warnings in
an unevaluated context.

* g++.dg/cpp0x/Wunused-value1.C: New test.

From-SVN: r272585

decl.c (grokdeclarator): Use id_loc...

/cp
2019-06-22  Paolo Carlini  <paolo.carlini@oracle.com>

* decl.c (grokdeclarator): Use id_loc, typespec_loc, and
locations[ds_storage_class] in a few additional places.

/testsuite
2019-06-22  Paolo Carlini  <paolo.carlini@oracle.com>

* g++.dg/diagnostic/auto-storage-1.C: New.
* g++.dg/diagnostic/no-type-1.C: Likewise.
* g++.dg/diagnostic/no-type-2.C: Likewise.
* g++.dg/diagnostic/top-level-auto-1.C: Likewise.
* g++.dg/cpp0x/auto9.C: Test some locations too.
* g++.dg/cpp1z/register1.C: Likewise.
* g++.dg/cpp1z/register2.C: Likewise.
* g++.dg/cpp1z/register3.C: Likewise.
* g++.dg/other/error34.C: Likewise.

From-SVN: r272584

Daily bump.

From-SVN: r272583

compiler: intrinsify some math/bits functions

    Let the Go frontend recognize some math/bits functions and turn
    them into intrinsics.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/183266

* go-gcc.cc (Gcc_backend::Gcc_backend): Define math/bits
builtins.

From-SVN: r272579

runtime: inline and remove eqtype

    Now that type equality is just a pointer equality, write it
    inlined and remove the eqtype function.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/182978

From-SVN: r272578

compiler: open code some type assertions

    Now that type equality is just simple pointer equality, we can
    open code some type assertions instead of making runtime calls.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/182977

From-SVN: r272577

re PR c++/90909 (call devirtualized to pure virtual)

2019-06-21  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/90909
* g++.dg/other/final7.C: New.

From-SVN: r272576

tree-vect-data-refs.c (vect_find_stmt_data_reference): Handle even zero DR_OFFSET...

* tree-vect-data-refs.c (vect_find_stmt_data_reference): Handle
even zero DR_OFFSET, but DR_BASE_ADDRESS of POINTER_PLUS_EXPR
containing the offset as possible simd lane access.  Look through
widening conversion.  Move the
TREE_CODE (DR_INIT (newdr)) == INTEGER_CST test earlier and reindent.

* g++.dg/vect/simd-2.cc: Don't xfail, instead expect vectorization on
x86.
* g++.dg/vect/simd-5.cc: Likewise.

From-SVN: r272575

re PR c++/90909 (call devirtualized to pure virtual)

/cp
2019-06-21  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/90909
Revert:
2019-05-21  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/67184
PR c++/69445
* call.c (build_over_call): Devirtualize when the final overrider
comes from the base.

/testsuite
2019-06-21  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/90909
Revert:
2019-05-21  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/67184
PR c++/69445
* g++.dg/other/final3.C: New.
* g++.dg/other/final4.C: Likewise.
* g++.dg/other/final5.C: Likewise.

* g++.dg/other/final6.C: New.

From-SVN: r272574

re PR c++/90909 (call devirtualized to pure virtual)

/cp
2019-06-21  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/90909
Revert:
2019-05-21  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/67184
PR c++/69445
* call.c (build_over_call): Devirtualize when the final overrider
comes from the base.

/testsuite
2019-06-21  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/90909
Revert:
2019-05-21  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/67184
PR c++/69445
* g++.dg/other/final3.C: New.
* g++.dg/other/final4.C: Likewise.
* g++.dg/other/final5.C: Likewise.

* g++.dg/other/final6.C: New.

From-SVN: r272573

PR c++/61490 - qualified-id in friend function definition.

* decl.c (grokdeclarator): Diagnose qualified-id in friend function
definition.  Improve location for diagnostics of friend functions.

* g++.dg/diagnostic/friend2.C: New test.
* g++.dg/diagnostic/friend3.C: New test.

From-SVN: r272572

PR c++/60223 - ICE with T{} in non-deduced context.

* pt.c (unify): Allow COMPOUND_LITERAL_P in a non-deduced context.

* g++.dg/cpp0x/nondeduced1.C: New test.
* g++.dg/cpp0x/nondeduced2.C: New test.
* g++.dg/cpp0x/nondeduced3.C: New test.
* g++.dg/cpp0x/nondeduced4.C: New test.

From-SVN: r272571

PR c++/64235 - missing syntax error with invalid alignas.

* parser.c (cp_parser_std_attribute_spec): Commit to tentative parse
if there's a missing close paren.

* g++.dg/parse/alignas1.C: New test.

From-SVN: r272570

re PR fortran/67884 (Missing error message on required allocatable attribute)

2019-06-21  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/67884
* resolve.c (deferred_requirements) : Check only the result variable.
(resolve_fl_procedure): Check deferred requirements on functions.

2019-06-21  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/67884
* gfortran.dg/dummy_procedure_8.f90: Remove a test that is ...
* gfortran.dg/pr67884.f90: ... covered here.  New test.

From-SVN: r272569

PR c++/90490 - fix decltype issues in noexcept-specifier.

* except.c (build_noexcept_spec): Call
instantiate_non_dependent_expr_sfinae before
build_converted_constant_expr instead of calling
instantiate_non_dependent_expr after it.  Add
processing_template_decl_sentinel.

* g++.dg/cpp0x/noexcept43.C: New test.
* g++.dg/cpp0x/noexcept44.C: New test.

From-SVN: r272568

PR c++/90875 - added -Wswitch-outside-range option

From-SVN: r272567

[Darwin, fixincludes] Fix Darwin9/10 math.h issues.

Darwin has had long long functions for some considerable time and these are
exposed in Darwin8 and Darwin11+ headers.  However, for some reason it was
elected to hide them behind __STRICT_ANSI__ and __STDC_VERSION__ on Darwin9
and Darwin10.  This is a problem for G++/libstdc++ that expects the functions
to be available for strict ansi (-std=c++14, for example) and without
defining __STDC_VERSION__.  The fix here follows the pattern used in
Darwin11+ headers where the functions may be explicitly hidden by defining
__DARWIN_NO_LONG_LONG.

This fixes the tr1 testsuite fails seen on Darwin9 and 10.

2019-06-21  Iain Sandoe  <iain@sandoe.co.uk>

* inclhack.def: Replace the complex test using __STRICT_ANSI__ and
__STDC_VERSION__ with a test using __DARWIN_NO_LONG_LONG.
Ensure that the top level math.h uses <> to wrap included headers
rather than "".
* fixincl.x: Regenerated.
* tests/base/architecture/ppc/math.h: Update test to include the
__DARWIN_NO_LONG_LONG case.

From-SVN: r272563

[Darwin, fixincludes] Fix Darwin14 header issues.

There are two issues with the Darwin14 (SDK) headers in which unguarded
advanced syntax elements causes any code including these headers to fail.

2019-06-21  Iain Sandoe  <iain@sandoe.co.uk>

* inclhack.def: Guard __has_attribute and __has_extension in
os/base.h.
Guard Apple blocks syntax in dispatch/object.h.
* fixincl.x: Regenerate.
* tests/base/dispatch/object.h: New file.
* tests/base/os/base.h: New file.

From-SVN: r272561

re PR tree-optimization/90930 (Excessive memory consumption)

2019-06-21  Richard Biener  <rguenther@suse.de>

PR tree-optimization/90930
* tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Set visited
flag on new stmts to avoid re-processing them.

From-SVN: r272560

PR c++/90875 - added -Wswitch-outside-range option

* doc/invoke.texi (Wswitch-outside-range): Document.

* c.opt (Wswitch-outside-range): Added new option.
* c-warn.c (c_do_switch_warnings): Use OPT_Wswitch-outside-range.

* c-c++-common/Wswitch-outside-range-1.c: New test.
* c-c++-common/Wswitch-outside-range-2.c: New test.
* c-c++-common/Wswitch-outside-range-3.c: New test.
* c-c++-common/Wswitch-outside-range-4.c: New test.

From-SVN: r272559

re PR fortran/51991 (Wrong error message with variables named "SAVE*")

2019-06-21  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/51991
* decl.c (gfc_match_save): If SAVE was not seen, return MATCH_NO
instead issuing an error message and returning MATCH_ERROR.

2019-06-21  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/51991
gfortran.dg/pr51991.f90

From-SVN: r272556

re PR tree-optimization/90949 (null pointer check removed)

PR tree-optimization/90949
* tree-ssa-copy.c (fini_copy_prop): Use reset_flow_sensitive_info.
* tree-ssanames.c (reset_flow_sensitive_info): Reset non-null state.

* gcc.c-torture/execute/pr90949.c: New test.

From-SVN: r272555

Fix missing name in ChangeLog

Fix another case in which a name in the commit message wasn't copied to
the ChangeLog -- sorry!

From-SVN: r272553

PR c++/90953 - ICE with -Wmissing-format-attribute.

* c-common.c (check_function_arguments_recurse): Use
get_attribute_name.
(check_missing_format_attribute): Likewise.

* g++.dg/warn/miss-format-7.C: New test.

From-SVN: r272552

Fix missing name in ChangeLog

It was in the commit message but got dropped from the ChangeLog entry
due to a scripting bug.

From-SVN: r272551

compiler: omit write barrier for assignment to *(convert(&local))

    Assignments to local variables don't need a write barrier. But
    currently the compiler inserts a write barrier if the LHS is a
    local variable with type converted, as *(convert(&local)). Let
    the compiler recognize this pattern and omit the write barrier.

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/182541

From-SVN: r272550

compiler: open code string slice expressions

    Currently a string slice expression is implemented with a runtime
    call __go_string_slice. Change it to open code it, which is more
    efficient, and allows the backend to further optimize it.

    Also omit the write barrier for length-only update (i.e.
    s = s[:n]).

    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/182540

From-SVN: r272549