Make max_align_t respect _Float128.

The _FloatN, _FloatNx, _DecimalN and _DecimalNx types are specified in
such a way that they are basic types, meaning that max_align_t must be
at least as aligned as those types.

On 32-bit x86, max_align_t is currently 8-byte aligned, but
_Decimal128 and _Float128 are 16-byte aligned, so the alignment of
max_align_t needs to increase to meet the standard requirements for
these types.

This patch implements such an increase.  Because max_align_t needs to
be usable for C++ as well as for C, <stddef.h> can't actually refer to
_Float128, but needs to use __float128 (or some other notation for the
type) instead.  And since __float128 is architecture-specific, there
isn't a preprocessor conditional that means "__float128 is available"
(whereas one could test __FLT128_MANT_DIG__ to see if _Float128 is
available; __SIZEOF_FLOAT128__ is available on x86 only).  But I
believe the only case that actually has an alignment problem here is
32-bit x86, and <stddef.h> already has lots of conditional specific to
particular architectures or OSes, so this patch uses a conditional on
__i386__; that also is the minimal change that ensures neither size
nor alignment of max_align_t is changed in any case other than where
it is necessary.  If any other architectures turn out to have such an
issue, it will show up as failures of one of the testcases added by
this patch.

Such an increase is of course an ABI change, but a reasonably safe
one, in that max_align_t doesn't tend to appear in library interfaces
(rather, it's something to use when writing allocators and similar
code; most matches found on codesearch.debian.net look like copies of
the gnulib stddef.h module rather than anything actually using
max_align_t at all).

max_align_t_align has a corresponding change (adding _Float128 to the
types considered).

(I think glibc malloc alignment should also increase to 16-byte on
32-bit x86 so it works for allocating objects of these types, which is
now straightforward given the fix made for 32-bit powerpc.)

Bootstrapped with no regressions on x86_64-pc-linux-gnu, and
spot-tested with -m32 that the new float128-align.c test now compiles
where previously it didn't.

gcc:
* ginclude/stddef.h (max_align_t) [__i386__]: Add __float128
element.

gcc/c-family:
* c-common.c (max_align_t_align): Also consider alignment of
float128_type_node.

gcc/testsuite:
* gcc.dg/float128-align.c, gcc.dg/float128x-align.c,
gcc.dg/float16-align.c, gcc.dg/float32-align.c,
gcc.dg/float32x-align.c, gcc.dg/float64-align.c,
gcc.dg/float64x-align.c, gcc.dg/floatn-align.h: New tests.

From-SVN: r240248

re PR rtl-optimization/77416 (LRA rematerializing use of CA reg across function call)

2016-09-19  Vladimir Makarov  <vmakarov@redhat.com>

PR rtl-optimization/77416
* lra-remat.c (operand_to_remat): Process hard coded insn
registers.

2016-09-19  Vladimir Makarov  <vmakarov@redhat.com>

PR rtl-optimization/77416
* gcc.target/powerpc/pr77416.c: New.

From-SVN: r240247

Fix PR c++/77639 (ICE during error recovery)

gcc/cp/ChangeLog:

PR c++/77639
* parser.c (cp_parser_class_head): When
processing_template_parmlist, don't assume that the
class-head may start an explicit specialization.

gcc/testsuite/ChangeLog:

PR c++/77639
* g++.dg/template/error-recovery4.C: New test.

From-SVN: r240245

libstdc++/77645 fix deque and vector xmethods for Python 3

PR libstdc++/77645
* python/libstdcxx/v6/xmethods.py (DequeWorkerBase.__init__)
(DequeWorkerBase.index, VectorWorkerBase.get): Cast results of
division to int to work with Python 3.

From-SVN: r240241

pr57558-1.c: Use unsigned int instead of unsigned long.

gcc/testsuite
* gcc.dg/vect/pr57558-1.c: Use unsigned int instead of unsigned long.

From-SVN: r240239

[simplify-rtx] (GTU (PLUS a C) (C - 1)) --> (LTU a -C)

* simplify-rtx.c (simplify_relational_operation_1): Add transformation
(GTU (PLUS a C) (C - 1)) --> (LTU a -C).

* gcc.target/aarch64/gtu_to_ltu_cmp_1.c: New test.
* gcc.target/aarch64/gtu_to_ltu_cmp_2.c: New test.

From-SVN: r240238

Make libstdc++ tests FAIL on Python errors

* testsuite/lib/gdb-test.exp (gdb-test): Fail if Python error occurs.

From-SVN: r240237

Fix lra_p text

* target.def (lra_p): Wordsmithing.
* doc/tm.texi: Regenerate.

From-SVN: r240236

Improve pretty printer for std::variant

* python/libstdcxx/v6/printers.py (SingleObjContainerPrinter): Allow
display_hint to be set by subclasses.
(StdVariantPrinter): Use array for display_hint. Adjust output to be
more similar to std::any and std::optional output.
(register_type_printers): Add type printers for basic_string_view
typedefs and experimental::any. Adjust type printers for
fundamentals_v1 templates to match fundamentals_v2 and later.
* testsuite/libstdc++-prettyprinters/cxx17.cc: New.

From-SVN: r240234

libstdc++/77645 Fix xmethods for std::list

PR libstdc++/77645
* python/libstdcxx/v6/xmethods.py (DequeWorkerBase.index): Rename
argument.
(ListWorkerBase.get_value_from_node): Define new method.
(ListFrontWorker.__call__, ListBackWorker.__call__): Use it.

From-SVN: r240233

re PR ipa/77587 (C compiler produces incorrect stack alignment with __attribute__((weak)))

PR target/77587
* cgraph.c (cgraph_node::rtl_info): Pass &avail to
ultimate_alias_target call, return NULL if avail < AVAIL_AVAILABLE.
Call ultimate_alias_target just once, not up to 4 times.

* gcc.dg/pr77587.c: New test.
* gcc.dg/pr77587a.c: New file.

Co-Authored-By: Jan Hubicka <jh@suse.cz>
From-SVN: r240232

libcaf.h: Add caf_reference_type.

libgfortran/ChangeLog:

2016-09-19  Andre Vehreschild  <vehre@gcc.gnu.org>

* caf/libcaf.h: Add caf_reference_type.
* caf/mpi.c: Adapted signature of caf_register().
* caf/single.c (struct caf_single_token): Added to keep the pointer
to the memory registered and array descriptor.
(caf_internal_error): Added convenience interface.
(_gfortran_caf_register): Adapted to work with caf_single_token and
return memory in the array descriptor.
(_gfortran_caf_deregister): Same.
(assign_char1_from_char4): Fixed style.
(convert_type): Fixed incorrect conversion.
(_gfortran_caf_get): Adapted to work with caf_single_token.
(_gfortran_caf_send): Same.
(_gfortran_caf_sendget): Same.
(copy_data): Added to stop repeating it in all _by_ref functions.
(get_for_ref): Recursive getting of coarray data using a chain of
references.
(_gfortran_caf_get_by_ref): Driver for computing the memory needed for
the get and checking properties of the operation.
(send_by_ref): Same as get_for_ref but for sending data.
(_gfortran_caf_send_by_ref): Same like caf_get_by_ref but for sending.
(_gfortran_caf_sendget_by_ref): Uses get_by_ref and send_by_ref to
implement sendget for reference chains.
(_gfortran_caf_atomic_define): Adapted to work with caf_single_token.
(_gfortran_caf_atomic_ref): Likewise.
(_gfortran_caf_atomic_cas): Likewise.
(_gfortran_caf_atomic_op): Likewise.
(_gfortran_caf_event_post): Likewise.
(_gfortran_caf_event_wait): Likewise.
(_gfortran_caf_event_query): Likewise.
(_gfortran_caf_lock): Likewise.
(_gfortran_caf_unlock): Likewise.

gcc/testsuite/ChangeLog:

2016-09-19  Andre Vehreschild  <vehre@gcc.gnu.org>

* gfortran.dg/coarray/alloc_comp_4.f90: New test.
* gfortran.dg/coarray_38.f90:
* gfortran.dg/coarray_alloc_comp_1.f08: New test.
* gfortran.dg/coarray_alloc_comp_2.f08: New test.
* gfortran.dg/coarray_allocate_7.f08: New test.
* gfortran.dg/coarray_allocate_8.f08: New test.
* gfortran.dg/coarray_allocate_9.f08: New test.
* gfortran.dg/coarray_lib_alloc_1.f90: Adapted scan-tree-dumps to expect
new caf_register.
* gfortran.dg/coarray_lib_alloc_2.f90: Same.
* gfortran.dg/coarray_lib_alloc_3.f90: Same.
* gfortran.dg/coarray_lib_comm_1.f90: Adapted scan-tree-dumps to expect
get_by_refs.
* gfortran.dg/coarray_lib_token_3.f90: Same as for coarray_lib_alloc2.
* gfortran.dg/coarray_lock_7.f90: Same.
* gfortran.dg/coarray_poly_5.f90: Same.
* gfortran.dg/coarray_poly_6.f90: Same.
* gfortran.dg/coarray_poly_7.f90: Same.
* gfortran.dg/coarray_poly_8.f90: Same.
* gfortran.dg/coindexed_1.f90: Changed errors expected.

gcc/fortran/ChangeLog:

2016-09-19  Andre Vehreschild  <vehre@gcc.gnu.org>

* expr.c (gfc_check_assign): Added flag to control whether datatype
conversion is allowed.
* gfortran.h: Added caf-token-tree to gfc_component.  Changed
prototypes mostly to add whether datatype conversion is allowed.
* gfortran.texi: Added documentation for the caf_reference_t and the
caf_*_by_ref function.
* primary.c (caf_variable_attr): Similar to gfc_variable_attr but
focused on the needs of coarrays.
(gfc_caf_attr): Same.
* resolve.c (resolve_ordinary_assign): Set the conversion allowed
flag when not in a coarray.
* trans-array.c (gfc_array_init_size): Moved setting of array
descriptor's datatype before the alloc, because caf_register needs it.
(gfc_array_allocate): Changed notion of whether an array is a coarray.
(gfc_array_deallocate): Same.
(gfc_alloc_allocatable_for_assignment): Added setting of coarray's
array descriptor datatype before the register.  And using deregister/
register to mimmick a realloc for coarrays.
* trans-decl.c (gfc_build_builtin_function_decls): Corrected signatures
of old caf-functions and added signature definitions of the _by_ref
ones.
(generate_coarray_sym_init): Adapted to new caf_register signature.
* trans-expr.c (gfc_conv_scalar_to_descriptor): Make sure a constant
is translated to an lvalue expression before use in an array
descriptor.
(gfc_get_ultimate_alloc_ptr_comps_caf_token): New function.  Get the
last allocatable component's coarray token.
(gfc_get_tree_for_caf_expr): For top-level object get the coarray
token and check for unsupported features.
(gfc_get_caf_token_offset): Getting the offset might procude new
statements, which now are stored in the pre and post of the current se.
(gfc_caf_get_image_index): For this image return a call to
caf_this_image.
(expr_may_alias_variables): Check that the result is set for testing
its properties.
(alloc_scalar_allocatable_for_assignment): Added auto allocation of
coarray components.
(gfc_trans_assignment_1): Rewrite an assign to a coarray object to
be a sendget.
* trans-intrinsic.c (conv_caf_vector_subscript_elem): Corrected
wrong comment.
(compute_component_offset): Compute the correct offset a structure
member.
(conv_expr_ref_to_caf_ref): Convert to a chain of refs into
caf_references.
(gfc_conv_intrinsic_caf_get): Call caf_get_by_ref instead of caf_get.
(conv_caf_send): Call caf_*_by_ref for coarrays that need
reallocation.
(gfc_conv_intrinsic_function): Adapted to new signuature of the caf
drivers.
(conv_intrinsic_atomic_op): Add pre and post statements correctly.
(conv_intrinsic_atomic_ref): Same.
(conv_intrinsic_atomic_cas): Same.
(conv_intrinsic_event_query): Same.
* trans-stmt.c (gfc_trans_lock_unlock): Same.
(gfc_trans_event_post_wait): Same.
(gfc_trans_allocate): Support allocation of allocatable coarrays.
(gfc_trans_deallocate): And there deallocation.
* trans-types.c (gfc_typenode_for_spec): Added flag to control whether
a component is part of coarray.  When so, then add space to store a
coarray token.
(gfc_build_array_type): Same.
(gfc_get_array_descriptor_base): Same.
(gfc_get_array_type_bounds): Same.
(gfc_sym_type): Same.
(gfc_get_derived_type): Same.
(gfc_get_caf_reference_type): Declare the caf_reference_type.
* trans-types.h: Prototype changes only.
* trans.c (gfc_allocate_using_lib): Use the updated caf_register
signature.
(gfc_allocate_allocatable): Same.
(gfc_deallocate_with_status): Same.
* trans.h: Defined the runtime types for caf_reference_t and the enums.

From-SVN: r240231

re PR fortran/77584 (Unclassifiable statement error with procedure pointer using template named "structure_")

2016-09-19  Fritz Reese  <fritzoreese@gmail.com>

PR fortran/77584
* gcc/fortran/decl.c (match_record_decl, gfc_match_decl_type_spec):
Fixes to handling of structure/record from declaration-type-spec.

* gcc/testsuite/gfortran.dg/dec_structure_15.f90: New testcase.

From-SVN: r240230

Fix zero-length parameter bug in pex_run.

2016-09-19  Andrew Stubbs  <ams@codesourcery.com>

libiberty/
* pex-win32.c (argv_to_cmdline): Quote zero-length parameters.

libiberty/testsuite/
* test-pexecute.c (main): Insert check for zero-length parameters.

From-SVN: r240229

dwarf2out.c (early_dwarf_finished): New global.

2016-09-16  Richard Biener  <rguenther@suse.de>

* dwarf2out.c (early_dwarf_finished): New global.
(set_early_dwarf::set_early_dwarf): Assert early_dwarf_finished
is false.
(dwarf2out_early_finish): Set early_dwarf_finished at the end,
if called from LTO exit early.
(dwarf2out_late_global_decl): When being during the early
debug phase do not add locations but only const value attributes.
Adjust the way we generate early DIEs for LTO.

lto/
* lto.c (lto_main): Invoke early_finish debug hook.

From-SVN: r240228

re PR tree-optimization/77605 (wrong code at -O3 on x86_64-linux-gnu)

2016-09-19  Richard Biener  <rguenther@suse.de>

PR middle-end/77605
* tree-data-ref.c (analyze_subscript_affine_affine): Use the
proper niter to bound the loops.

* gcc.dg/torture/pr77605.c: New testcase.

From-SVN: r240227

re PR tree-optimization/77514 (ICE in VN_INFO_GET, at tree-ssa-sccvn.c:406 w/ -O2 (and above))

2016-09-19  Richard Biener  <rguenther@suse.de>

PR tree-optimization/77514
* tree-ssa-pre.c (create_expression_by_pieces): Optimize
search for folded stmt.

From-SVN: r240226

pr68078.f90: XFAIL run on Darwin

2016-09-18  Louis Krupp  <louis.krupp@zoho.com>

* gfortran.dg/pr68078.f90: XFAIL run on Darwin

From-SVN: r240225

Daily bump.

From-SVN: r240224

passes.def (pass_early_thread_jumps): Schedule after forwprop.

* passes.def (pass_early_thread_jumps): Schedule after forwprop.
* tree-pass.h (make_pass_early_thread_jumps): Declare.
* tree-ssa-threadbackward.c (fsm_find_thread_path,
fsm_find_thread_path, profitable_jump_thread_path,
fsm_find_control_statement_thread_paths,
find_jump_threads_backwards): Add speed_p parameter.
(pass_data_early_thread_jumps): New pass.
(make_pass_early_thread_jumps): New function.

* g++.dg/predict-loop-exit-1.C: Disable early jump threading.
* g++.dg/predict-loop-exit-2.C: Disable early jump threading.
* g++.dg/predict-loop-exit-3.C: Disable early jump threading.
* gcc.dg/tree-ssa/pr69196-1.c: Disable early jump threading.
* gcc.dg/tree-ssa/vrp01.c: Disable early jump threading.
* gcc.dg/tree-ssa/ssa-dom-thread-2b.c: Disable early jump threading.
* gcc.dg/tree-ssa/pr68198.c: Scan ethread dump.
* gcc.dg/tree-ssa/ssa-thread-13.c: Scan ethread dump.
* gcc.dg/tree-ssa/vrp56.c: Scan ethread dump.
* gcc.dg/tree-ssa/vrp92.c: Scan ethread dump.
* gcc.dg/uninit-15.c: Swap xfailed and non-xfailed alternative.

From-SVN: r240221

pr68633.c: Fix expected result.

gcc/testsuite/
* gcc.target/i386/pr68633.c: Fix expected result.

From-SVN: r240220

re PR fortran/68078 (segfault with allocate and stat for derived types with default initialization)

2016-09-17  Louis Krupp  <louis.krupp@gmail.com>

PR fortran/68078
* gfortran.dg/pr68078.f90: New test.
* gfortran.dg/set_vm_limit.c: New, called by pr68078.

2016_09_17  Louis Krupp  <louis.krupp@zoho.com>

PR fortran/68078
* resolve.c (resolve_allocate_expr): Check that derived type
pointer, object or array has been successfully allocated before
initializing.

From-SVN: r240219

Daily bump.

From-SVN: r240218

Define pretty printers for C++17 library components

* python/libstdcxx/v6/printers.py (StdVariantPrinter): Define.
(StdExpAnyPrinter, StdExpOptionalPrinter, StdExpStringViewPrinter):
Register for C++17 components in namespace std. Strip inline namespace
from typename.

From-SVN: r240215

ia64.h (ASM_OUTPUT_DWARF_OFFSET): Add cast.

* config/ia64/ia64.h (ASM_OUTPUT_DWARF_OFFSET): Add cast.
* config/i386/cygming.h (ASM_OUTPUT_DWARF_OFFSET): Likewise.

From-SVN: r240214

Daily bump.

From-SVN: r240213

Remove bogus entry in ChangeLog added by git-svn

From-SVN: r240206

Fix documentation for overflow-checking builtins

* doc/extend.texi (Integer Overflow Builtins): Fix type of out
parameters for functions taking long long arguments.

From-SVN: r240205

Fix typo in Libstdc++ Profile Mode docs

* doc/xml/manual/profile_mode.xml: Fix typo.
* doc/html/manual/profile_mode_devel.html: Regenerate.

From-SVN: r240204

recog.c (rest_of_handle_split_after_reload): Delete.

* recog.c (rest_of_handle_split_after_reload): Delete.
(pass_split_after_reload::gate): New method.
(pass_split_after_reload::execute): Call split_all_insns directly.

From-SVN: r240203

Fix documentation for overflow-checking builtins

* doc/extend.texi (Integer Overflow Builtins): Fix type of out
parameters for functions taking long long arguments.

From-SVN: r240200

re PR target/77613 (Powerpc64le: redundant swaps in autovectorised loop)

[gcc]

2016-09-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

PR target/77613
* config/rs6000/rs6000.c (rtx_is_swappable_p): Add support for
splat with truncate.

[gcc/testsuite]

2016-09-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

PR target/77613
* gcc.target/powerpc/swaps-p8-25.c: New.

From-SVN: r240199

re PR c++/77482 (Segfault when compiling ill-formed constexpr code)

PR c++/77482
* error.c (dump_simple_decl): Only check DECL_DECLARED_CONCEPT_P
if DECL_LANG_SPECIFIC is non-NULL.  Fix up formatting.

* g++.dg/cpp0x/constexpr-77482.C: New test.

From-SVN: r240198

re PR c++/77379 (incorrect mangling for non-virtual adjustor thunk (missing abi tag))

PR c++/77379
* g++.dg/abi/abi-tag23.C: Adjust scan-assembler regex for differing
thunk offsets.
* g++.dg/abi/abi-tag23a.C: Likewise.

From-SVN: r240197

re PR c++/77338 (ICE on invalid C++11 code on x86_64-linux-gnu: Segmentation fault)

PR c++/77338
* constexpr.c (cxx_eval_constant_expression) <case PARM_DECL>: Only
call is_really_empty_class on complete types.

* g++.dg/cpp0x/decltype-77338.C: New test.

From-SVN: r240196

re PR c++/77375 (constant object with mutable subobject allocated in read-only memory)

PR c++/77375
* class.c (check_bases): Set CLASSTYPE_HAS_MUTABLE if any TYPE_HAS_MUTABLE_P
for any bases.

* g++.dg/cpp0x/mutable1.C: New test.

From-SVN: r240195

Add inline functions for various bitwise operations.

* hwint.h (least_bit_hwi, pow2_or_zerop, pow2p_hwi, ctz_or_zero):
New.
* hwint.c (exact_log2): Use pow2p_hwi.
(ctz_hwi, ffs_hwi): Use least_bit_hwi.
* alias.c (memrefs_conflict_p): Use pow2_or_zerop.
* builtins.c (get_object_alignment_2, get_object_alignment)
(get_pointer_alignment, fold_builtin_atomic_always_lock_free): Use
least_bit_hwi.
* calls.c (compute_argument_addresses, store_one_arg): Use
least_bit_hwi.
* cfgexpand.c (expand_one_stack_var_at): Use least_bit_hwi.
* combine.c (force_to_mode): Use least_bit_hwi.
* emit-rtl.c (set_mem_attributes_minus_bitpos, adjust_address_1):
Use least_bit_hwi.
* expmed.c (synth_mult, expand_divmod): Use ctz_or_zero, ctz_hwi.
(init_expmed_one_conv): Use pow2p_hwi.
* fold-const.c (round_up_loc, round_down_loc): Use pow2_or_zerop.
(fold_binary_loc): Use pow2p_hwi.
* function.c (assign_parm_find_stack_rtl): Use least_bit_hwi.
* gimple-fold.c (gimple_fold_builtin_memory_op): Use pow2p_hwi.
* gimple-ssa-strength-reduction.c (replace_ref): Use least_bit_hwi.
* hsa-gen.c (gen_hsa_addr_with_align, hsa_bitmemref_alignment):
Use least_bit_hwi.
* ipa-cp.c (ipcp_alignment_lattice::meet_with_1): Use least_bit_hwi.
* ipa-prop.c (ipa_modify_call_arguments): Use least_bit_hwi.
* omp-low.c (oacc_loop_fixed_partitions)
(oacc_loop_auto_partitions): Use least_bit_hwi.
* rtlanal.c (nonzero_bits1): Use ctz_or_zero.
* stor-layout.c (place_field): Use least_bit_hwi.
* tree-pretty-print.c (dump_generic_node): Use pow2p_hwi.
* tree-sra.c (build_ref_for_offset): Use least_bit_hwi.
* tree-ssa-ccp.c (ccp_finalize): Use least_bit_hwi.
* tree-ssa-math-opts.c (bswap_replace): Use least_bit_hwi.
* tree-ssa-strlen.c (handle_builtin_memcmp): Use pow2p_hwi.
* tree-vect-data-refs.c (vect_analyze_group_access_1)
(vect_grouped_store_supported, vect_grouped_load_supported)
(vect_permute_load_chain, vect_shift_permute_load_chain)
(vect_transform_grouped_load): Use pow2p_hwi.
* tree-vect-generic.c (expand_vector_divmod): Use ctz_or_zero.
* tree-vect-patterns.c (vect_recog_divmod_pattern): Use ctz_or_zero.
* tree-vect-stmts.c (vectorizable_mask_load_store): Use
least_bit_hwi.
* tsan.c (instrument_expr): Use least_bit_hwi.
* var-tracking.c (negative_power_of_two_p): Use pow2_or_zerop.

From-SVN: r240194

re PR libgcc/71744 (Concurrently throwing exceptions is not scalable)

PR libgcc/71744
* unwind-dw2-fde.c (ATOMIC_FDE_FAST_PATH): Define if __register_frame*
is not the primary registry and atomics are available.
(any_objects_registered): New variable.
(__register_frame_info_bases, __register_frame_info_table_bases):
Atomically store 1 to any_objects_registered after registering first
unwind info.
(_Unwind_Find_FDE): Return early if any_objects_registered is 0.

From-SVN: r240193

Replace modulus with mask operation in over-aligned new

2016-09-16  Jonathan Wakely  <jwakely@redhat.com>
    Marc Glisse  <marc.glisse@inria.fr>

* libsupc++/new_opa.cc [_GLIBCXX_HAVE_ALIGNED_ALLOC]
(operator new(size_t, align_val_t)): Replace modulus operator with
mask.

Co-Authored-By: Marc Glisse <marc.glisse@inria.fr>
From-SVN: r240192

re PR fortran/77612 (ICE on invalid character len in contained procedure)

2016-09-16  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/77612
* decl.c (char_len_param_value): Check parent namespace for
seen_implicit_none.

2016-09-16  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/77612
* gfortran.dg/pr77612.f90: New test.

From-SVN: r240191

Adjust arguments to aligned_alloc or posix_memalign

* libsupc++/new_opa.cc [_GLIBCXX_HAVE_POSIX_MEMALIGN] (aligned_alloc):
Increase alignment if less than sizeof(void*).
[_GLIBCXX_HAVE_ALIGNED_ALLOC] (operator new(size_t, align_val_t)):
Increase size if not a multiple of alignment.

From-SVN: r240187

Fix testsuite failure for avr target

The test has integer literals that only fit on a target with int size 32+.

gcc/testsuite/

2016-09-16  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>

* gcc.dg/torture/pr70421.c: Require int32plus.

From-SVN: r240186

* c-typeck.c (lvalue_p): Use true and false instead of 1 and 0.

From-SVN: r240185

ia64.h (ASM_OUTPUT_DWARF_OFFSET): Use parameter OFFSET, not offset.

* config/ia64/ia64.h (ASM_OUTPUT_DWARF_OFFSET): Use parameter
OFFSET, not offset.
* config/i386/cygming.h (ASM_OUTPUT_DWARF_OFFSET): Likewise.

From-SVN: r240176

re PR middle-end/77475 (unnecessary or misleading context in reporting command line problems)

PR middle-end/77475
* gcc.target/i386/pr65990.c: Adjust dg-error regexp.

From-SVN: r240175

re PR target/77526 (ICE: in verify_dominators, at dominance.c:1039 with -mstringop-strategy=byte_loop)

PR target/77526
* combine.c (rest_of_handle_combine): If any edges have been purged,
free dominators if available.

* gcc.target/i386/pr77526.c: New test.

From-SVN: r240174

re PR middle-end/77594 (double computation for __builtin_sub_overflow (0, ...))

PR middle-end/77594
* internal-fn.c (expand_arith_overflow) <case MINUS_EXPR>: Don't fall
through into expand_addsub_overflow after expand_neg_overflow.

* gcc.target/i386/pr77594.c: New test.

Co-Authored-By: Eric Botcazou <ebotcazou@adacore.com>
From-SVN: r240173

Daily bump.

From-SVN: r240172

fix-it hints can't contain newlines

I hope to implement newline support within fix-it hints at some point,
but currently it's not supported, and leads to misleading diagnostic
output, so for now, fail gracefully.

gcc/ChangeLog:
* diagnostic-show-locus.c
(selftest::test_fixit_insert_containing_newline): New function.
(selftest::test_fixit_replace_containing_newline): New function.
(selftest::diagnostic_show_locus_c_tests): Call the above.

libcpp/ChangeLog:
* include/line-map.h (class rich_location): Note that newlines
aren't supported in fix-it text.
* line-map.c (rich_location::add_fixit_insert_before): Reject
attempts to add fix-its containing newlines.
(rich_location::add_fixit_replace): Likewise.

From-SVN: r240169

re PR fortran/69963 (ICE out of memory on displaced implicit character)

2016-09-15  Louis Krupp  <louis.krupp@zoho.com>
PR fortran/69963
* parse.c (reject_statement): Clear charlen pointers in implicit
character typespecs before those charlen structures are freed.

2016-09-15  Louis Krupp  <louis.krupp@zoho.com>

PR fortran/69963
* gfortran.dg/misplaced_implicit_character.f90: New test.

From-SVN: r240168

c-common.c (check_cxx_fundamental_alignment_constraints): Check DECL_EXTERNAL.

* c-common.c (check_cxx_fundamental_alignment_constraints): Check
DECL_EXTERNAL.

From-SVN: r240167

re PR tree-optimization/77503 (ICE in vect_transform_stmt compiling postgresql)

PR tree-optimization/77503
* tree-vect-loop.c (vectorizable_reduction): Record reduction
code for CONST_COND_REDUCTION at analysis stage and use it at
transform stage.
* tree-vectorizer.h (struct _stmt_vec_info): New field.
(STMT_VINFO_VEC_CONST_COND_REDUC_CODE): New macro.
* tree-vect-stmts.c (new_stmt_vec_info): Initialize above new
field.

gcc/testsuite
* gcc.dg/vect/pr77503.c: New test.

From-SVN: r240166

re PR tree-optimization/77544 (segfault at -O0 - infinite loop in simplification)

2016-09-15  Richard Biener  <rguenther@suse.de>

PR middle-end/77544
* fold-const.c (split_tree): Do not split constant ~X.

* c-c++-common/torture/pr77544.c: New testcase.

From-SVN: r240164

re PR rtl-optimization/77425 (Pointer test follows dereference in sched-int.h)

PR rtl-optimization/77425
* sched-int.h (sd_iterator_cond): Don't update it_ptr->linkp if list
is NULL.

From-SVN: r240163

re PR middle-end/77475 (unnecessary or misleading context in reporting command line problems)

PR middle-end/77475
* config/i386/i386.c (ix86_parse_stringop_strategy_string): Simplify,
use %qs instead of %s where desirable, use argument instead of arg in
the diagnostic wording, add list of supported strategies and
spellcheck hint.
(ix86_option_override_internal): Emit target("m...") instead of
option("m...") in the diagnostic.  Use %qs instead of %s in invalid
-march/-mtune option diagnostic.  Add list of supported arches/tunings
and spellcheck hint.  Remove prefix, suffix and sw variables, use
main_args_p ? "..." : "..." in diagnostics to make translation
possible.

* gcc.target/i386/pr65990.c: Adjust expected diagnostics.
* gcc.dg/march-generic.c: Likewise.
* gcc.target/i386/spellcheck-options-1.c: New test.
* gcc.target/i386/spellcheck-options-2.c: New test.
* gcc.target/i386/spellcheck-options-3.c: New test.
* gcc.target/i386/spellcheck-options-4.c: New test.

From-SVN: r240162

Editorial fixes to libstdc++ debug mode docs

* doc/xml/manual/debug_mode.xml: Minor editorial fixes.
* doc/html/*: Regenerate.

From-SVN: r240161

Fix testsuite to not increment bool

* testsuite/23_containers/vector/debug/insert6_neg.cc: Remove
-Wno-deprecated.
* testsuite/util/debug/checks.h (generate_unique<bool>): Specialize.

From-SVN: r240160

Reformat libstdc++-v3/ChangeLog to fit in 80 columns

From-SVN: r240159

dwarf2asm.h (dw2_asm_output_offset): Add overload with extra offset argument.

2016-09-15  Richard Biener  <rguenther@suse.de>

* dwarf2asm.h (dw2_asm_output_offset): Add overload with
extra offset argument.
* dwarf2asm.c (dw2_asm_output_offset): Implement that.
* doc/tm.texi.in (ASM_OUTPUT_DWARF_OFFSET): Adjust documentation
to reflect new offset parameter.
* doc/tm.texi: Regenerate.
* config/darwin.h (ASM_OUTPUT_DWARF_OFFSET): Adjust.
* config/darwin-protos.h (darwin_asm_output_dwarf_delta): Add
offset argument.
(darwin_asm_output_dwarf_offset): Likewise.
* config/darwin.c (darwin_asm_output_dwarf_delta): Add offset
argument.
(darwin_asm_output_dwarf_offset): Pass offset argument through.
* config/ia64/ia64.h (ASM_OUTPUT_DWARF_OFFSET): Adjust.
* config/i386/cygmin.h (ASM_OUTPUT_DWARF_OFFSET): Likewise.

From-SVN: r240158

Fix testsuite to not increment bool

* testsuite/23_containers/vector/debug/insert6_neg.cc: Remove
-Wno-deprecated.
* testsuite/util/debug/checks.h (generate_unique<bool>): Specialize.

From-SVN: r240157

re PR fortran/72743 (ICE in get_constraint_for_ssa_var, at tree-ssa-structalias.c:2958)

2016-09-15  Chung-Lin Tang  <cltang@codesourcery.com>

PR fortran/72743
* ipa-icf.c (set_alias_uids): New function.
(sem_variable::merge): Use set_alias_uids to set DECL_PT_UID of
all the merged variable's referring aliases.

testsuite/
* gfortran.dg/goacc/pr72743.f90: New test.

From-SVN: r240155

fold-reassoc-2.c: Fix dump scan.

2016-09-15  Richard Biener  <rguenther@suse.de>

* gcc.dg/fold-reassoc-2.c: Fix dump scan.

From-SVN: r240154

re PR tree-optimization/77514 (ICE in VN_INFO_GET, at tree-ssa-sccvn.c:406 w/ -O2 (and above))

2016-09-15  Richard Biener  <rguenther@suse.de>

PR tree-optimization/77514
* tree-ssa-pre.c (create_expression_by_pieces): Handle garbage
only forced_stmts sequence.

* gcc.dg/torture/pr77514.c: New testcase.

From-SVN: r240153

Iterator to visit SSA

Iterator to visit SSA
gcc/ChangeLog:

2016-09-15  Kugan Vivekanandarajah  <kuganv@linaro.org>

* tree-ssanames.h (FOR_EACH_SSA_NAME): New.
* cfgexpand.c (update_alias_info_with_stack_vars): Use
FOR_EACH_SSA_NAME to iterate over SSA variables.
(pass_expand::execute): Likewise.
* omp-simd-clone.c (ipa_simd_modify_function_body): Likewise.
* tree-cfg.c (dump_function_to_file): Likewise.
* tree-into-ssa.c (pass_build_ssa::execute): Likewise.
(update_ssa): Likewise.
* tree-ssa-alias.c (dump_alias_info): Likewise.
* tree-ssa-ccp.c (ccp_finalize): Likewise.
* tree-ssa-coalesce.c (build_ssa_conflict_graph): Likewise.
(create_outofssa_var_map): Likewise.
(coalesce_ssa_name): Likewise.
* tree-ssa-operands.c (dump_immediate_uses): Likewise.
* tree-ssa-pre.c (compute_avail): Likewise.
* tree-ssa-sccvn.c (init_scc_vn): Likewise.
(scc_vn_restore_ssa_info): Likewise.
(free_scc_vn): Likwise.
(run_scc_vn): Likewise.
* tree-ssa-structalias.c (compute_points_to_sets): Likewise.
* tree-ssa-ter.c (new_temp_expr_table): Likewise.
* tree-ssa-copy.c (fini_copy_prop): Likewise.
* tree-ssa.c (verify_ssa): Likewise.

From-SVN: r240152

Daily bump.

From-SVN: r240151

re PR c++/77549 (ICE on invalid C++ code that references undeclared variable)

PR c++/77549
* name-lookup.c (consider_binding_level): Look through TREE_LIST
and OVERLOAD.

* g++.dg/lookup/pr77549.C: New test.

From-SVN: r240148

libgo: fix typo in configure.ac (PCQUANTUm -> PCQUANTUM)

    Reviewed-on: https://go-review.googlesource.com/29154

From-SVN: r240146

Infer architecture from ABI for mips-mti* and mips-img* triplets.

gcc/
* config.gcc (mips*-mti-elf*, mips*-mti-linux*): Set mips32r2
and mips64r2 as default 32-bit and 64-bit architectures.
(mips*-img-elf*, mips*-img-linux*): Set mips32r6 and mips64r6
as default 32-bit and 64-bit architectures.

From-SVN: r240145

simplify.c (gfc_simplify_repeat): Fix a misplaced closing ')'.

2016-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        * simplify.c (gfc_simplify_repeat): Fix a misplaced closing ')'.

From-SVN: r240142

typeck.c (cp_build_unary_op): Diagnose incrementing boolean expressions.

* typeck.c (cp_build_unary_op): Diagnose incrementing boolean
expressions.  Tweak an error message.

* c-c++-common/gomp/atomic-12.c: Use -Wno-deprecated.
* c-c++-common/gomp/atomic-13.c: Likewise.
* c-c++-common/gomp/atomic-14.c: Likewise.
* g++.dg/cpp1y/lambda-init11.C: Remove invalid code.
* g++.dg/cpp1z/bool-increment1.C: New test.
* c-c++-common/pr60439.c: Add dg-warning.
* g++.dg/expr/bitfield4.C: Likewise.
* g++.dg/expr/bitfield5.C: Likewise.
* g++.dg/expr/bitfield6.C: Likewise.
* g++.dg/expr/bool1.C: Likewise.
* g++.dg/expr/bool3.C: Likewise.
* g++.dg/expr/lval3.C: Likewise.
* g++.dg/expr/lval4.C: Likewise.
* g++.old-deja/g++.jason/bool5.C: Likewise.
* g++.dg/expr/bitfield3.C: Adjust dg-error.
* g++.dg/other/error18.C: Likewise.
* g++.dg/gomp/atomic-14.C: Likewise.
libgomp/
* testsuite/libgomp.c++/atomic-3.C: Use -Wno-deprecated.
libstdc++-v3/
* testsuite/23_containers/vector/debug/insert6_neg.cc: Use
-Wno-deprecated.

From-SVN: r240141

loop-unroll.c (unroll_loop_runtime_iterations): Set probability of succ edge.

* loop-unroll.c (unroll_loop_runtime_iterations): Set probability of succ edge.

From-SVN: r240140

c-common.c (check_cxx_fundamental_alignment_constraints): Don't limit FIELD_DECL, either.

* c-common.c (check_cxx_fundamental_alignment_constraints): Don't
limit FIELD_DECL, either.

From-SVN: r240139

re PR c++/77539 (gcc-5/6: comparison of array to nullptr failure in constexpr (fixed by r235506 on trunk))

PR c++/77539
* g++.dg/cpp1y/pr77539.C: New.

From-SVN: r240138

c-common.c (c_common_truthvalue_conversion): Use false instead of 0.

* c-common.c (c_common_truthvalue_conversion): Use false instead of 0.
* c-common.h (build_unary_op): Change nonconvert parameter type to bool.
* c-omp.c (c_finish_omp_atomic): Use false instead of 0.

* c-array-notation.c (create_cmp_incr): Use false instead of 0.
(fix_array_notation_expr): Likewise.
* c-decl.c (finish_decl): Likewise.
* c-parser.c (c_parser_postfix_expression_after_primary): Likewise.
* c-typeck.c (array_to_pointer_conversion): Use true instead of 1.
(function_to_pointer_conversion): Use false instead of 0.
(convert_lvalue_to_rvalue): Likewise.
(parser_build_unary_op): Likewise.
(build_atomic_assign): Likewise.
(build_unary_op): Change nonconvert parameter type to bool, use
true/false instead of 1/0.
(build_binary_op): Use true instead of 1.

* cp-tree.h (cp_build_unary_op): Change nonconvert parameter type to
bool.
* decl2.c (one_static_initialization_or_destruction): Use true instead
of 1.
* init.c (build_vec_init): Use false instead of 0.
* pt.c (tsubst_copy_and_build): Likewise.
* semantics.c (simplify_loop_decl_cond): Likewise.
* typeck.c (rationalize_conditional_expr): Likewise.
(cp_build_binary_op): Use true instead of 1.
(cp_build_unary_op): Change nonconvert parameter type to bool.  Use true
instead of 1.
(build_unary_op): Change nonconvert parameter type to bool.
(unary_complex_lvalue): Use false instead of 0.

From-SVN: r240137

Fix dg-require before dg-run directives in testsuite.

libstdc++-v3/ChangeLog:

2016-09-14  Christophe Lyon  <christophe.lyon@linaro.org>

* testsuite/23_containers/vector/bool/modifiers/insert/31370.cc:
Move dg-do directive before dg-skip.
* testsuite/21_strings/debug/iterator_self_move_assign_neg.cc:
Move dg-do directive before dg-require.
* testsuite/21_strings/debug/self_move_assign_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/57779_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/60587_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/assign1_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/assign2_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/assign3_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/construct1_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/construct2_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/construct3_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/insert1_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/insert2_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/insert3_neg.cc: Likewise.
* testsuite/23_containers/vector/debug/insert5_neg.cc: Likewise.
* testsuite/25_algorithms/lower_bound/debug/irreflexive.cc: Likewise.
* testsuite/25_algorithms/partial_sort_copy/debug/irreflexive_neg.cc:
Likewise.
* testsuite/25_algorithms/pop_heap/empty2_neg.cc: Likewise.
* testsuite/25_algorithms/pop_heap/empty_neg.cc: Likewise.
* testsuite/27_io/objects/char/12048-5.cc: Likewise.
* testsuite/ext/special_functions/conf_hyperg/check_nan.cc: Likewise.
* testsuite/ext/special_functions/hyperg/check_nan.cc: Likewise.

gcc/testsuite/ChangeLog:

2016-09-14  Christophe Lyon  <christophe.lyon@linaro.org>

* g++.dg/cpp0x/lambda/lambda-mangle.C: Move dg-do directive before
dg-require.
* g++.dg/ext/builtin_alloca.C: Likewise.
* g++.dg/template/spec35.C: Likewise.
* gcc.dg/builtins-68.c: Likewise.
* gcc.misc-tests/gcov-13.c: Likewise.
* gcc.misc-tests/gcov-14.c: Likewise.

From-SVN: r240135

dec_structure_14.f90: New testcase.

2016-09-13  Jim MacArthur  <jim.macarthur@codethink.co.uk>

* gcc/testsuite/gfortran.dg/dec_structure_14.f90: New testcase.

From-SVN: r240134

Fix TARGET_LRA_P description

2016-09-14  Segher Boessenkool  <segher@kernel.crashing.org>

* target.def (lra_p): Change commentary (for the manual) for the
new default.
* doc/tm.texi: Regenerate.

From-SVN: r240133

Delete TARGET_LRA_P from those targets that set it to "true"

A few targets already always want LRA; those then do not need to override
the default anymore.

2016-09-14  Segher Boessenkool  <segher@kernel.crashing.org>

* config/aarch64/aarch64.c (TARGET_LRA_P): Delete macro.
* config/arm/arm.c (TARGET_LRA_P): Delete macro.
* config/i386/i386.c (TARGET_LRA_P): Delete macro.
* config/nds32/nds32.c (TARGET_LRA_P): Delete macro.

From-SVN: r240132

Change the default of TARGET_LRA_P to true

Since all targets now define TARGET_LRA_P, this does not change behaviour
for any existing target.  Newly added ports will by default use LRA now.

2016-09-14  Segher Boessenkool  <segher@kernel.crashing.org>

* targhooks.c (default_lra_p): Return true instead of false.

From-SVN: r240131

Put a TARGET_LRA_P into every target

This patch adds a TARGET_LRA_P (defined to hook_bool_void_false) to
every target that didn't yet override the hook.  No functional change.

2016-09-14  Segher Boessenkool  <segher@kernel.crashing.org>

* config/alpha/alpha.c (TARGET_LRA_P): New macro, defined to
hook_bool_void_false.
* config/avr/avr.c: Ditto.
* config/bfin/bfin.c: Ditto.
* config/c6x/c6x.c: Ditto.
* config/cr16/cr16.c: Ditto.
* config/cris/cris.c: Ditto.
* config/epiphany/epiphany.c: Ditto.
* config/fr30/fr30.c: Ditto.
* config/frv/frv.c: Ditto.
* config/h8300/h8300.c: Ditto.
* config/ia64/ia64.c: Ditto.
* config/iq2000/iq2000.c: Ditto.
* config/lm32/lm32.c: Ditto.
* config/m32c/m32c.c: Ditto.
* config/m32r/m32r.c: Ditto.
* config/m68k/m68k.c: Ditto.
* config/mcore/mcore.c: Ditto.
* config/microblaze/microblaze.c: Ditto.
* config/mmix/mmix.c: Ditto.
* config/mn10300/mn10300.c: Ditto.
* config/moxie/moxie.c: Ditto.
* config/msp430/msp430.c: Ditto.
* config/nios2/nios2.c: Ditto.
* config/nvptx/nvptx.c: Ditto.
* config/pa/pa.c: Ditto.
* config/pdp11/pdp11.c: Ditto.
* config/rl78/rl78.c: Ditto.
* config/sparc/sparc.c: Ditto.
* config/spu/spu.c: Ditto.
* config/stormy16/stormy16.c: Ditto.
* config/tilegx/tilegx.c: Ditto.
* config/tilepro/tilepro.c: Ditto.
* config/v850/v850.c: Ditto.
* config/vax/vax.c: Ditto.
* config/visium/visium.c: Ditto.
* config/xtensa/xtensa.c: Ditto.

From-SVN: r240130

re PR sanitizer/68260 (false positive with tsan)

PR sanitizer/68260
* tsan.c: Include target.h.
(enum tsan_atomic_action): Add bool_clear and bool_test_and_set.
(BOOL_CLEAR, BOOL_TEST_AND_SET): Define.
(tsan_atomic_table): Add BUILT_IN_ATOMIC_CLEAR and
BUILT_IN_ATOMIC_TEST_AND_SET entries.
(instrument_builtin_call): Handle bool_clear and bool_test_and_set.

* c-c++-common/tsan/pr68260.c: New test.

From-SVN: r240129

Add braces to a condition in predict.c (PR middle-end/77574)

    Martin Liska  <mliska@suse.cz>

PR middle-end/77574
* predict.c (force_edge_cold): Add braces to a condition.

Co-Authored-By: Martin Liska <mliska@suse.cz>
From-SVN: r240128

Daily bump.

From-SVN: r240127

re PR rtl-optimization/77289 (ICE in extract_constrain_insn, at recog.c:2212 on powerpc64)

2016-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>

        PR rtl-optimization/77289
        * lra-constraints.c (get_final_hard_regno): Removed.
        (get_hard_regno): Add new parameter final_p.
        (get_reg_class): Directly call lra_get_elimination_hard_regno.
        (operands_match_p): Adjust call to get_hard_regno.
        (uses_hard_regs_p): Likewise.
        (process_alt_operands): Likewise.

From-SVN: r240124

re PR target/70713 (msp430 interrupt attribute prevents overriding weak symbols)

2016-09-13  Joe Seymour  <joe.s@somniumtech.com>

gcc/
PR target/70713
* config/msp430/msp430.c (msp430_start_function): Emit an error
if a function is both weak and specifies an interrupt number.

gcc/testsuite/
PR target/70713
* gcc.target/msp430/function-attributes-1.c: New test.
* gcc.target/msp430/function-attributes-2.c: New test.
* gcc.target/msp430/function-attributes-3.c: New test.

From-SVN: r240123

Implement P0040R3, Extending memory management tools.

* include/bits/stl_uninitialized.h (utility): New include
in C++17 mode.
(uninitialized_default_construct): New.
(uninitialized_default_construct_n): Likewise.
(uninitialized_value_construct): Likewise.
(uninitialized_value_construct_n): Likewise.
(uninitialized_move): Likewise.
(uninitialized_move_n): Likewise.
(destroy_at, destroy, destroy_n): Likewise.
* testsuite/20_util/specialized_algorithms/memory_management_tools/1.cc:
New.

From-SVN: r240122

Implement P0028R4, C++17 using attribute namespaces without repetition

Implement P0028R4, C++17 using attribute namespaces without repetition
* parser.c (cp_parser_std_attribute): Add ATTR_NS argument.  Diagnose
non-NULL ATTR_NS with scoped attribute token.  Handle non-NULL
ATTR_NS with non-scoped attribute tokens.  Allow named ops in
identifier after ::.
(cp_parser_std_attribute_list): Add ATTR_NS argument, pass it down
to cp_parser_std_attribute calls.
(cp_parser_std_attribute_spec): Parse optional C++17
attribute-using-prefix, adjust grammar in function comment.

* g++.dg/cpp0x/gen-attrs-61.C: New test.
* g++.dg/cpp1z/gen-attrs1.C: New test.

From-SVN: r240121

re PR tree-optimization/77454 (IMM ERROR w/ -O2 and above)

PR tree-optimization/77454
* tree-ssa-dom.c (optimize_stmt): Set modified flag on stmt after
changing GIMPLE_COND.  Move update_stmt_if_modified call after this.
Formatting fix.

* gcc.dg/pr77454.c: New test.

From-SVN: r240120

re PR c++/77553 (wrong code with post-increment operator in constexpr)

PR c++/77553
* constexpr.c (cxx_fold_pointer_plus_expression): New function.
(cxx_eval_binary_expression): Use it for POINTER_PLUS_EXPR.
(cxx_eval_pointer_plus_expression): Remove.
(cxx_eval_constant_expression) <case POINTER_PLUS_EXPR>: Don't
call cxx_eval_pointer_plus_expression.

* g++.dg/cpp1y/constexpr-77553.C: New test.

From-SVN: r240119

re PR fortran/77420 (gfortran and equivalence produces internal compiler error)

2016-09-13  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/77420
* module.c (load_equiv):  Revert revision 240063.

2016-09-13  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/77420
* gfortran.dg/pr77420.f90: Revert revision 240063 by removing test.

From-SVN: r240118

[PATCH 1/2][AArch64]  Add missing support for poly64x1_t

Committed on behalf of:

  Tamar Christina  <tamar.christina@arm.com>

gcc/

* config/aarch64/aarch64-builtins.c
(aarch64_init_simd_builtins): Fix builtin type signature printing.

From-SVN: r240117

alpha.c (alpha_pass_by_reference): Pass un-named SFmode and SCmode arguments by reference.

* config/alpha/alpha.c (alpha_pass_by_reference): Pass un-named
SFmode and SCmode arguments by reference.

From-SVN: r240116

fix-it hints: insert_before vs insert_after

The API for adding "insert text" fix-it hints was unclear
about exactly where the text should be inserted relative
to the given insertion point.

This patch clarifies things by renaming the pertinent methods from
  richloc.add_fixit_insert
to
  richloc.add_fixit_insert_before
and adding:
  richloc.add_fixit_insert_after

The latter allows us to consolidate some failure-handling into
class rich_location, rather than having to have every such diagnostic
check for it.

The patch also adds a description of how fix-it hints work to the
comment for class rich_location within libcpp/include/line-map.h.

gcc/c-family/ChangeLog:
* c-common.c (warn_logical_not_parentheses): Replace
rich_location::add_fixit_insert calls with add_fixit_insert_before
and add_fixit_insert_after, eliminating the "next_loc" calculation.

gcc/c/ChangeLog:
* c-parser.c (c_parser_declaration_or_fndef): Update for renaming
of add_fixit_insert to add_fixit_insert_before.

gcc/cp/ChangeLog:
* parser.c (cp_parser_class_specifier_1): Update for renaming of
add_fixit_insert to add_fixit_insert_before.
(cp_parser_class_head): Likewise.

gcc/ChangeLog:
* diagnostic-show-locus.c (selftest::test_one_liner_fixit_insert):
Rename to...
(selftest::test_one_liner_fixit_insert_before): ...this, and update
for renaming of add_fixit_insert to add_fixit_insert_before.
(selftest::test_one_liner_fixit_insert_after): New function.
(selftest::test_one_liner_fixit_validation_adhoc_locations):
Update for renaming of add_fixit_insert to
add_fixit_insert_before.
(selftest::test_one_liner_many_fixits): Likewise.
(selftest::test_diagnostic_show_locus_one_liner): Update for
renaming, call new test function.
(selftest::test_diagnostic_show_locus_fixit_lines): Update for
renaming of add_fixit_insert to add_fixit_insert_before.
(selftest::test_fixit_consolidation): Likewise.
* diagnostic.c (selftest::test_print_parseable_fixits_insert):
Likewise.
* edit-context.c (selftest::test_applying_fixits_insert): Rename
to...
(selftest::test_applying_fixits_insert_before): ...this.
(selftest::test_applying_fixits_insert): Update for renaming of
add_fixit_insert to add_fixit_insert_before.
(selftest::test_applying_fixits_insert_after): New function.
(selftest::test_applying_fixits_insert_after_at_line_end): New
function.
(selftest::test_applying_fixits_insert_after_failure): New
function.
(selftest::test_applying_fixits_multiple): Update for renaming of
add_fixit_insert to add_fixit_insert_before.
(selftest::change_line): Likewise.
(selftest::test_applying_fixits_unreadable_file): Likewise.
(selftest::test_applying_fixits_line_out_of_range): Likewise.
(selftest::test_applying_fixits_column_validation): Likewise.
(selftest::test_applying_fixits_column_validation): Likewise.
(selftest::edit_context_c_tests): Update for renamed test
function; call new test functions.

gcc/testsuite/ChangeLog:
* gcc.dg/plugin/diagnostic_plugin_test_show_locus.c
(test_show_locus): Replace rich_location::add_fixit_insert calls
with add_fixit_insert_before and add_fixit_insert_after.

libcpp/ChangeLog:
* include/line-map.h (class rich_location): Add description of
fix-it hints to leading comment.
(rich_location::add_fixit_insert): Rename both overloaded methods
to..
(rich_location::add_fixit_insert_before): ...this, updating their
comments.
(rich_location::add_fixit_insert_after): Two new overloaded
methods.
(rich_location::stop_supporting_fixits): New method.
* line-map.c (rich_location::add_fixit_insert): Rename both
overloaded methods to..
(rich_location::add_fixit_insert_before): ...this, updating their
comments.
(rich_location::add_fixit_insert_after): Two new methods.
(rich_location::reject_impossible_fixit): Split out
failure-handling into...
(rich_location::stop_supporting_fixits): New method.

From-SVN: r240115

re PR tree-optimization/77536 (Vectorizer not maintaining relationship of relative block frequencies in absence of real profile data)

PR tree-optimization/77536
PR rtl-optimization/68212
* config/rs6000/rs6000.md (div->recip splitter): Remove
optimize_insn_for_speed_p condition.

From-SVN: r240113

Don't treat array as builtin type in set_underlying_type

2016-09-13  Jason Merrill  <jason@redhat.com>
    Tom de Vries  <tom@codesourcery.com>

PR c++/77427
* c-common.c (set_underlying_type): Don't treat array as builtin type.

* g++.dg/pr77427.C: New test.

Co-Authored-By: Tom de Vries <tom@codesourcery.com>
From-SVN: r240112

Add new effective target: profile_update_atomic

* g++.dg/gcov/gcov-threads-1.C: Use profile_update_atomic
effective target.
* gcc.dg/tree-prof/val-profiler-threads-1.c: Likewise.
* lib/target-supports.exp: Define the new target.

From-SVN: r240111

Trying again to fix aligned-new on m68k.

* c-common.c (check_cxx_fundamental_alignment_constraints): Don't
limit types at all.

From-SVN: r240110

c-typeck.c (build_unary_op): Rename FLAG parameter to NOCONVERT.

* c-typeck.c (build_unary_op): Rename FLAG parameter to NOCONVERT.  Use
it.

From-SVN: r240109

* optabs.c (prepare_cmp_insn): Update documentation comment.

From-SVN: r240108

re PR middle-end/77475 (unnecessary or misleading context in reporting command line problems)

PR middle-end/77475
* opts.h (candidates_list_and_hint): Declare.
* opts-common.c (candidates_list_and_hint): New function.
(cmdline_handle_error): Use it.

Co-Authored-By: Manuel López-Ibáñez <manu@gcc.gnu.org>
From-SVN: r240107