Fix build failure with MPFR 2.4.x

* gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
GMP_RNDx for compatiblity.

From-SVN: r244858

Fix false positive for -Walloc-size-larger-than (PR bootstrap/79132).

2017-01-24  Martin Liska  <mliska@suse.cz>

PR bootstrap/79132
* tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
that would prevent us to call alloca with -1 as argument.

From-SVN: r244857

asan_test.C: Enable on all *-*-linux* targets that support asan...

* g++.dg/asan/asan_test.C: Enable on all *-*-linux* targets that
support asan, only on i?86/x86_64 require sse2_runtime.

From-SVN: r244856

dwarf2out.c (output_compilation_unit_header, [...]): Avoid -Wformat-security warning.

* dwarf2out.c (output_compilation_unit_header, output_file_names):
Avoid -Wformat-security warning.

From-SVN: r244855

* configure.tgt: Enable tsan and lsan on powerpc64{,le}-*-linux*.

From-SVN: r244854

aarch64.c (thunderx2t99_addrcost_table): Improve cost table.

2017-01-23  Andrew Pinski  <apinski@cavium.com>

        * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
        cost table.

From-SVN: r244851

re PR rtl-optimization/66669 (FAIL: gcc.dg/loop-8.c)

gcc/testsuite/ChangeLog:

2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>

PR target/66669
* gcc.dg/loop-8.c: Modify dg-skip-if directive to exclude this
test on powerpc targets.

From-SVN: r244850

Amend recent entry.

From-SVN: r244849

builtin-sprintf.c: Add test cases.

gcc/testsuite/ChangeLog:
* gcc.dg/tree-ssa/builtin-sprintf.c: Add test cases.

From-SVN: r244848

PR middle-end/78703 - -fprintf-return-value floating point handling incorrect...

PR middle-end/78703 - -fprintf-return-value floating point handling incorrect in locales with a mulltibyte decimal point
* gimple-ssa-sprintf.c (warn_level): New global.
(format_integer): Use it here and throughout the rest of the file.
Use the same switch to compute sign as base.
(maybe_warn): New function.
(format_directive): Factor out warnings into maybe_warn.
Add debugging output.  Use warn_level.
(add_bytes): Use warn_level.
(pass_sprintf_length::compute_format_length): Add debugging output.
(try_substitute_return_value): Same.
(pass_sprintf_length::handle_gimple_call): Set and use warn_level.

From-SVN: r244847

PR middle-end/78703 - -fprintf-return-value floating point handling incorrect...

PR middle-end/78703 - -fprintf-return-value floating point handling incorrect in locales with a mulltibyte decimal point
* gimple-ssa-sprintf.c (struct format_result): Remove constant member.
(struct fmtresult, format_integer, format_floating): Adjust.
(get_string_length, format_string,format_directive): Same.
(pass_sprintf_length::compute_format_length): Same.
(try_substitute_return_value): Simplify slightly.

From-SVN: r244846

PR middle-end/78703 - -fprintf-return-value floating point handling incorrect...

PR middle-end/78703 - -fprintf-return-value floating point handling incorrect in locales with a mulltibyte decimal point
* gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
(fmtresult::operator+=): Outlined.
(struct fmtresult): Add ctors.
(struct conversion_spec): Rename...
(struct directive): ...to this.  Add and remove data members.
(directive::set_width, directive::set_precision): New functions.
(format_percent): Use fmtresult ctor.
(get_width_and_precision): Remove.
(format_integer): Make naming changes.  Avoid computing width and
precision.
(format_floating): Same.  Adjust indentation.
(format_character, format_none): New functions.
(format_string): Moved character handling to format_character.
(format_directive): Remove arguments, change return type.
(parse_directive): New function.
(pass_sprintf_length::compute_format_length): Move directive
parsing to parse_directive.

From-SVN: r244845

re PR sanitizer/79168 (libtsan fails to link when cross compiling GCC tip for Aarch64 target)

PR sanitizer/79168
* merge.sh (change_comment_headers): Don't remove 2nd and 3rd line
if the 3rd line doesn't contain 'The LLVM Compiler Infrastructure'
text.
* sanitizer_common/sanitizer_linux_mips64.S: Regenerated.
* sanitizer_common/sanitizer_linux_x86_64.S: Likewise.
* tsan/tsan_ppc_regs.h: Likewise.
* tsan/tsan_rtl_aarch64.S: Likewise.
* tsan/tsan_rtl_mips64.S: Likewise.
* tsan/tsan_rtl_ppc64.S: Likewise.

From-SVN: r244844

Daily bump.

From-SVN: r244843

Restore <functional> header to test

* testsuite/experimental/array/make_array.cc: Restore <functional>
inclusion.

From-SVN: r244840

tree.h (assign_assembler_name_if_neeeded): Rename to ...

* tree.h (assign_assembler_name_if_neeeded): Rename to ...
(assign_assembler_name_if_needed): ... this.
* tree.c (assign_assembler_name_if_neeeded): Rename to ...
(assign_assembler_name_if_needed): ... this.
(free_lang_data_in_cgraph): Adjust callers.
* cgraphunit.c (cgraph_node::analyze): Likewise.
* omp-expand.c (expand_omp_taskreg, expand_omp_target):
Likewise.

From-SVN: r244839

re PR tree-optimization/79088 (wrong code at -O2 on x86_64-linux-gnu)

2017-01-23  Richard Biener  <rguenther@suse.de>

PR tree-optimization/79088
PR tree-optimization/79188
* tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
resetting loop bounds after last path deletion.  Reset loop
bounds of the target loop, make code match the comments.
* tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
Make sure loops need no fixups.

* gcc.dg/torture/pr79088.c: New testcase.
* gcc.dg/torture/pr79188.c: Likewise.

From-SVN: r244837

* da.po: Update.

From-SVN: r244836

libgo: update to go1.8rc2

    Fix a bug in the generation of the hash value in reflect.FuncOf.

    The merge script missed a bunch of testdata files over the years.
    Copy them over.

    Reviewed-on: https://go-review.googlesource.com/35570

From-SVN: r244835

scalar-insert-exp-3.c: New test.

gcc/testsuite/ChangeLog:

2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>

* gcc.target/powerpc/bfp/scalar-insert-exp-3.c: New test.
* gcc.target/powerpc/bfp/scalar-insert-exp-4.c: New test.
* gcc.target/powerpc/bfp/scalar-insert-exp-5.c: New test.
* gcc.target/powerpc/bfp/scalar-test-data-class-0.c: Adjust return
type of test function to reflect change in built-in function's
return type.
* gcc.target/powerpc/bfp/scalar-test-data-class-1.c: Likewise.
* gcc.target/powerpc/bfp/scalar-test-data-class-2.c: Likewise.
* gcc.target/powerpc/bfp/scalar-test-data-class-3.c: Likewise.
* gcc.target/powerpc/bfp/scalar-test-data-class-4.c: Adjust return
type and second argument type to reflect change in built-in
function's type signature.
* gcc.target/powerpc/bfp/scalar-test-data-class-5.c: Likewise.
* gcc.target/powerpc/bfp/scalar-test-data-class-6.c: Adjust return
type of test function to reflect change in built-in function's
return type.
* gcc.target/powerpc/bfp/scalar-test-data-class-7.c: Likewise.
* gcc.target/powerpc/bfp/scalar-test-neg-0.c: Likewise.
* gcc.target/powerpc/bfp/scalar-test-neg-1.c: Likewise.
* gcc.target/powerpc/bfp/scalar-test-neg-2.c: Likewise.
* gcc.target/powerpc/bfp/scalar-test-neg-3.c: Likewise.
* gcc.target/powerpc/bfp/vec-extract-exp-0.c: Likewise.
* gcc.target/powerpc/bfp/vec-extract-exp-1.c: Likewise.
* gcc.target/powerpc/bfp/vec-extract-exp-2.c: Likewise.
* gcc.target/powerpc/bfp/vec-extract-exp-3.c: Likewise.
* gcc.target/powerpc/bfp/vec-extract-sig-0.c: Likewise.
* gcc.target/powerpc/bfp/vec-extract-sig-1.c: Likewise.
* gcc.target/powerpc/bfp/vec-extract-sig-2.c: Likewise.
* gcc.target/powerpc/bfp/vec-extract-sig-3.c: Likewise.
* gcc.target/powerpc/bfp/vec-insert-exp-4.c: New test.
* gcc.target/powerpc/bfp/vec-insert-exp-5.c: New test.
* gcc.target/powerpc/bfp/vec-insert-exp-6.c: New test.
* gcc.target/powerpc/bfp/vec-insert-exp-7.c: New test.
* gcc.target/powerpc/bfp/vec-test-data-class-0.c: Adjust return
type of test function to reflect change in built-in function's
return type.
* gcc.target/powerpc/bfp/vec-test-data-class-1.c: Likewise.
* gcc.target/powerpc/bfp/vec-test-data-class-2.c: Likewise.
* gcc.target/powerpc/bfp/vec-test-data-class-3.c: Likewise.
* gcc.target/powerpc/bfp/vec-test-data-class-4.c: Likewise.
* gcc.target/powerpc/bfp/vec-test-data-class-5.c: Likewise.
* gcc.target/powerpc/bfp/vec-test-data-class-6.c: Adjust types of
test function's result and second argument to reflect change in
built-in function's type signature.
* gcc.target/powerpc/bfp/vec-test-data-class-7.c: Likewise.

gcc/ChangeLog:

2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>

* config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
exponent support with double type for first argument.
* config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
type returned by __builtin_vec_extract_sig,
__builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
functions from "vector int" to "vector unsigned int" or from
"vector long long int" to "vector unsigned long long int".
Changed type returned by __builtin_vec_extract_exp,
__builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
functions from "vector int" to "vector unsigned int" or from
"vector long long int" to "vector unsigned long long int".
Changed return type of __builtin_vec_test_data_class,
__builtin_vec_test_data_class_sp, and
__builtin_vec_test_data_class_dp from "vector int" to
"vector bool int" or from "vector long long int" to "vector bool
long long int" and changed second argument type from "unsigned
int" to "int".  Added new overloaded function forms "vector float
__builtin_vec_insert_exp (vector float, vector unsigned int)" and
"vector float __builtin_vec_insert_exp_sp (vector float, vector
unsigned int)" and "vector double __builtin_vec_insert_exp (vector
double, vector unsigned long long int)" and "vector double
__builtin_vec_insert_exp_dp (vector double, vector unsigned long
long int)".  Changed return type of
__builtin_scalar_test_data_class and
__builtin_scalar_test_data_class_sp and
__builtin_scalar_test_data_class_dp from "unsigned int" to "bool
int" and changed second argument from "unsigned int" to "int".
Changed type returned by __builtin_scalar_test_neg,
__builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
from "int" to "bool int".  Added new overloaded function form
"double __builtin_scalar_insert_exp (double, unsigned long long int)".
* config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
exponent double-precision with floating point first argument.
* doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
documentation of scalar_test_data_class, scalar_test_neg,
scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
vector_extract_exp, vec_extract_sig, vec_insert_exp, and
vec_test_data_class built-in functions to reflect refinements in
their type signatures.

From-SVN: r244834

PR c++/71710 - template using directive of field

PR c++/71710 - template using directive of field
* pt.c (tsubst_copy_and_build [COMPONENT_REF]): Move FIELD_DECL
check earlier.

PR C++/71710
* g++.dg/template/pr71710.C: New.

From-SVN: r244833

PR c++/71406 - ICE with scope-ref'd template id exprs

PR c++/71406 - ICE with scope-ref'd template id exprs
PR c++/77508
* typeck.c (finish_class_member_access_expr): Break up SCOPE_REF
before breaking up TEMPLATE_ID_EXPR.

PR c++/71406
PR c++/77508
* g++.dg/template/pr71406.C: New.

From-SVN: r244832

arith.c (arith_power): If simplifying integer power expression to zero...

2017-01-23  Thomas Koenig  <tkoenig@netcologne.de>

* arith.c (arith_power):  If simplifying integer power expression
to zero, warn if -Winteger-division is given.

2017-01-23  Thomas Koenig  <tkoenig@netcologne.de>

* gfortran.dg/integer_exponentiation_7.f90:  New test.

From-SVN: r244830

aarch64.c (aarch64_elf_asm_constructor): Increase size of buf.

2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>

    * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
    size of buf.
    (aarch64_elf_asm_destructor): Likewise.

From-SVN: r244828

re PR go/79037 (gccgo: Binaries crash with parforsetup: pos is not aligned on m68k)

PR go/79037
    compiler, runtime: align gc data for m68k

    The current GC requires that the gc data be aligned to at least a 4
    byte boundary, because it uses the lower two bits of the address for
    flags (see LOOP and PRECISE in runtime/mgc0.c).  As the gc data is
    stored as a [...]uintptr, that is normally always true.  However, on
    m68k, that only guarantees 2 byte alignment.  Fix it by forcing the
    alignment.

    The parfor code used by the current GC requires that the parfor data
    be aligned to at least an 8 byte boundary.  The code in parfor.c
    verifies this.  This is normally true, as the data uses uint64_t
    values, but, again, this must be enforced explicitly on m68k.

    Fixes GCC PR 79037.

    Reviewed-on: https://go-review.googlesource.com/35478

From-SVN: r244824

libgo: bump library version to 11

    As usual I forgot to bump the library version when I updated the
    library to 1.8rc1.

    Reviewed-on: https://go-review.googlesource.com/35568

From-SVN: r244823

69301.cc: Require atomic builtins.

2017-01-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>

    libstdc++-v3/
    * testsuite/29_atomics/atomic/69301.cc: Require atomic builtins.

From-SVN: r244819

* fr.po: Update.

From-SVN: r244818

re PR rtl-optimization/71724 (ICE: Segmentation fault, deep recursion between combine_simplify_rtx and subst)

PR rtl-optimization/71724
* combine.c (if_then_else_cond): Look for situations where it is
beneficial to undo the work of one of the recursive calls.

From-SVN: r244817

re PR rtl-optimization/78634 (30% performance drop after r242832.)

PR rtl-optimization/78634
* config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
(TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
* ifcvt.c (noce_try_cmove): Add missing cost check.

testsuite/
PR rtl-optimization/78634
* gcc.target/i386/funcspec-11.c: Also pass -mtune=i686.

From-SVN: r244816

re PR tree-optimization/70754 (ICE during predictive commoning)

PR tree-optimization/70754
* tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
(reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
combined stmt before it if not NULL.
(combine_chains): Process refs reversely and compute dominance point
for root ref.

gcc/testsuite
PR tree-optimization/70754
* gfortran.dg/pr70754.f90: New test.

From-SVN: r244815

PR libstdc++/79195 fix make_array type deduction

PR libstdc++/79195
* include/experimental/array (__make_array_elem): New class template
and partial specialization.
(__is_reference_wrapper): Move into __make_array_elem specialization.
(make_array): Use __make_array_elem to determine element type and move
static assertion into specialization. Qualify std::forward call.
(to_array): Add exception specifiation.
* testsuite/experimental/array/make_array.cc: Test argument types
without a common type.
* testsuite/experimental/array/neg.cc: Adjust expected error message.

From-SVN: r244813

Fix strstr folding (PR tree-optimization/79196).

2017-01-23  Martin Liska  <mliska@suse.cz>

PR tree-optimization/79196
* tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
(fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
instead of memcmp.
(strlen_optimize_stmt): Call the renamed function.
2017-01-23  Martin Liska  <mliska@suse.cz>

PR tree-optimization/79196
* gcc.dg/asan/pr79196.c: New test.
* gcc.dg/strlenopt-30.c: Update scanned pattern.

From-SVN: r244812

fix pr78384

PR tree-optimization/78384
* tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.

testsuite/
PR tree-optimization/78384
* gcc.dg/pr78384.c: New test.

From-SVN: r244811

sem_util.adb (New_Copy_Tree): Code cleanup: removal of the internal map (ie.

2017-01-23  Javier Miranda  <miranda@adacore.com>

* sem_util.adb (New_Copy_Tree): Code cleanup:
removal of the internal map (ie. variable Actual_Map, its
associated local variables, and all the code handling it).
* sem_ch9.adb (Analyze_Task_Type_Declaration): in GNATprove mode
force loading of the System package when processing a task type.
(Analyze_Protected_Type_Declaration): in GNATprove mode force
loading of the System package when processing a protected type.
* sem_ch10.adb (Analyze_Compilation_Unit): in GNATprove mode
force loading of the System package when processing compilation
unit with a main-like subprogram.
* frontend.adb (Frontend): remove forced loading of the System
package.

From-SVN: r244810

sem_prag.adb (Default_Initial_Condition): If the desired type declaration is a derived type declaration with discriminants...

2017-01-23  Ed Schonberg  <schonberg@adacore.com>

* sem_prag.adb (Default_Initial_Condition): If the desired type
declaration is a derived type declaration with discriminants,
it is rewritten as a private type declaration.
* sem_ch13.adb (Replace_Type_References_Generic,
Visible_Component): A discriminated private type with descriminnts
has components that must be rewritten as selected components
if they appear as identifiers in an aspect expression such as
a Default_Initial_Condition.
* sem_util.adb (Defining_Entity): support N_Iterator_Specification
nodes.

From-SVN: r244809

ghost.ads, ghost.adb (Is_Ignored_Ghost_Unit): New routine.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* ghost.ads, ghost.adb (Is_Ignored_Ghost_Unit): New routine.
* gnat1drv.adb Generate an empty object file for an ignored
Ghost compilation unit.
* inline.adb, sem_util.adb, sem_ch4.adb: Minor reformatting.

From-SVN: r244808

[multiple changes]

2017-01-23  Yannick Moy  <moy@adacore.com>

* sem_ch4.adb (Analyze_Indexed_Component_Form):
Adapt to inlined prefix with string literal subtype.
* inline.adb (Expand_Inlined_Call): Keep unchecked
conversion inside inlined call when formal type is constrained.

2017-01-23  Javier Miranda  <miranda@adacore.com>

* sem_util.adb (New_Copy_Tree): Code cleanup:
removal of global variables. All the global variables, global
functions and tables of this subprogram are now declared locally.

From-SVN: r244807

[multiple changes]

2017-01-23  Gary Dismukes  <dismukes@adacore.com>

* exp_strm.ads: Minor reformatting and typo fixes.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* sem_aggr.adb, par_sco.adb, exp_util.adb, sem.adb, sem_ch4.adb,
exp_aggr.adb: Minor reformatting.
* g-diopit.adb: minor grammar/punctuation fix in comment.
* g-byorma.ads: minor fix of unbalanced parens in comment.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* par.adb: Update the documentation of component Labl.
* par-ch6.adb (P_Return_Statement): Set the expected label of
an extended return statement to Error.

2017-01-23  Tristan Gingold  <gingold@adacore.com>

* s-boustr.ads, s-boustr.adb (Is_Full): New function.

2017-01-23  Ed Schonberg  <schonberg@adacore.com>

* expander.adb: Handle N_Delta_Aggregate.

2017-01-23  Javier Miranda  <miranda@adacore.com>

* exp_ch6.adb (Expand_Call): Improve the code that
checks if some formal of the called subprogram is a class-wide
interface, to handle subtypes of class-wide interfaces.

2017-01-23  Javier Miranda  <miranda@adacore.com>

* checks.adb (Apply_Parameter_Aliasing_Checks):
Remove side effects of the actuals before generating the overlap
check.

From-SVN: r244806

[multiple changes]

2017-01-23  Justin Squirek  <squirek@adacore.com>

* exp_strm.ads, exp_strm.ads
(Build_Record_Or_Elementary_Input_Function): Add an extra parameter so
as to avoid getting the underlying type by default.
* exp_attr.adb (Expand_N_Attribute_Reference): Remove use of
underlying type in the Iiput and output attribute cases when
building their respective functions.

2017-01-23  Gary Dismukes  <dismukes@adacore.com>

* scng.adb: Minor reformatting of error message.

2017-01-23  Ed Schonberg  <schonberg@adacore.com>

* sem_ch6.adb (Analyze_Expression_Function): Do not attempt
to freeze the return type of an expression funxtion that is a
completion, if the type is a limited view and the non-limited
view is available.

From-SVN: r244805

re PR tree-optimization/79186 (ICE on valid code at -O2 and -O3 on x86_64-linux-gnu:  Segmentation fault (in VRP))

2017-01-23  Richard Biener  <rguenther@suse.de>

PR tree-optimization/79186
* tree-vrp.c (register_new_assert_for): Make sure we've seen
both incoming edges before moving an assert.

* gcc.dg/torture/pr79186.c: New testcase.
* gcc.dg/torture/pr79187.c: Likewise.

From-SVN: r244804

Fix compute_complex_assign_jump_func

2017-01-23  Martin Jambor  <mjambor@suse.cz>

* ipa-prop.c (load_from_param_1): Removed.
(load_from_unmodified_param): Bits from load_from_param_1 put back
here.
(load_from_param): Removed.
(compute_complex_assign_jump_func): Removed stmt2 and just replaced it
with stmt.  Reverted back to use of load_from_unmodified_param.

testsuite/
* gcc.dg/ipa/vrp8.c: New test.

From-SVN: r244803

[PR 79108] Put ipa_node_params to GC memory

2017-01-23  Martin Jambor  <mjambor@suse.cz>

PR ipa/79108
* ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
(ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
field a pointer to garbage collected vector, mark lattices and
ipcp_orig_node with GTY((skip)).
(ipa_get_param_count): Adjust to descriptors being a pointer.
(ipa_get_param): Likewise.
(ipa_get_type): Likewise.
(ipa_get_param_move_cost): Likewise.
(ipa_set_param_used): Likewise.
(ipa_get_controlled_uses): Likewise.
(ipa_set_controlled_uses): Likewise.
(ipa_is_param_used): Likewise.
(ipa_node_params_t): Move into garbage collector.  New methods insert
and remove.
(ipa_node_params_sum): Annotate wth GTY(()).
(ipa_check_create_node_params): Adjust to ipa_node_params_sum being
garbage collected.
(ipa_load_from_parm_agg): Adjust declaration.
* ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
* ipa-profile.c (ipa_profile): Likewise.
* ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
(ipa_populate_param_decls): Make descriptors parameter garbage
collected.
(ipa_dump_param): Adjust to descriptors being a pointer.
(ipa_alloc_node_params): Likewise.
(ipa_initialize_node_params): Likewise.
(load_from_param_1): Make descriptors parameter garbage collected.
(load_from_unmodified_param): Likewise.
(load_from_param): Likewise.
(ipa_load_from_parm_agg): Likewise.
(ipa_node_params::~ipa_node_params): Removed.
(ipa_free_all_node_params): Remove call to delete operator.
(ipa_node_params_t::insert): New.
(ipa_node_params_t::remove): Likewise.
(ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
copy known_csts and known_contexts vectors.
(ipa_read_node_info): Adjust to descriptors being a pointer.
(ipcp_modif_dom_walker): Make m_descriptors field garbage
collected.
(ipcp_transform_function): Make descriptors variable garbage
collected.

testsuite/
* gfortran.dg/lto/pr79108_0.f90: New test.

From-SVN: r244802

Add AVX512 k-mask intrinsics

gcc/
* config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
* config/i386/avx512dqintrin.h: Ditto.
* config/i386/avx512fintrin.h: Ditto.
* gcc/config/i386/i386.c: Handle new builtins.
* config/i386/i386-builtin.def: Add new builtins.
* config/i386/sse.md (ktest<mode>, kortest<mode>): New.
(UNSPEC_KORTEST, UNSPEC_KTEST): New.

gcc/testsuite/
* gcc.target/i386/avx512bw-ktestd-1.c: New test.
* gcc.target/i386/avx512bw-ktestq-1.c: Ditto.
* gcc.target/i386/avx512dq-ktestb-1.c: Ditto.
* gcc.target/i386/avx512f-ktestw-1.c: Ditto.
* gcc.target/i386/avx512bw-kortestd-1.c: Ditto.
* gcc.target/i386/avx512bw-kortestq-1.c: Ditto.
* gcc.target/i386/avx512dq-kortestb-1.c: Ditto.
* gcc.target/i386/avx512f-kortestw-1.c: Ditto.
* gcc.target/i386/avx512bw-ktestd-2.c: Ditt
* gcc.target/i386/avx512bw-ktestq-2.c: Ditto.
* gcc.target/i386/avx512dq-ktestb-2.c: Ditto.
* gcc.target/i386/avx512f-ktestw-2.c: Ditto.
* gcc.target/i386/avx512bw-kortestd-2.c: Ditto.
* gcc.target/i386/avx512bw-kortestq-2.c: Ditto.
* gcc.target/i386/avx512dq-kortestb-2.c: Ditto.
* gcc.target/i386/avx512f-kortestw-2.c: Ditto.

From-SVN: r244801

re PR testsuite/78421 (vect-strided-a-u8-i2-gap.c fails on armeb)

PR testsuite/78421
* lib/target-supports.exp (check_effective_target_vect_hw_misalign):
If the target is ARM return the result of the
check_effective_target_arm_vect_no_misalign proc.
* gcc.dg/vect/vect-strided-a-u8-i2-gap.c: If the target does not
support unaligned vectors then only expect one of the loops to be
unrolled.

Co-Authored-By: Nick Clifton <nickc@redhat.com>
From-SVN: r244796

Minor editing.

From-SVN: r244795

[multiple changes]

2017-01-23  Ed Schonberg  <schonberg@adacore.com>

* par-ch4.adb (P_Aggregate_Or_Parent_Expr): Recognize delta
aggregate construct.
(P_Record_Or_Array_Component_Association): An array aggregate
can start with an Iterated_Component_Association.
* scng.adb: Modify error message on improper use of @ in earlier
versions of the language.
* sinfo.ads: New node kind N_Delta_Aggregate.
* sinfo.adb: An N_Delta_Aggregate has component associations and
an expression.
* sem_res.adb (Resolve): Call Resolve_Delta_Aggregate.
* sem_aggr.ads, sem_aggr.adb (Resolve_Iterated_Component_Association):
Create a new index for each one of the choices in the association,
to prevent spurious homonyms in the scope.
(Resolve_Delta_Aggregate): New.
* sem.adb: An N_Delta_Aggregate is analyzed like an aggregate.
* exp_util.adb (Insert_Actions): Take into account
N_Delta_Aggregate.
* exp_aggr.ads: New procedure Expand_N_Delta_Aggregate.
* exp_aggr.adb: New procedure Expand_N_Delta_Aggregate,
and local procedures Expand_Delta_Array_Aggregate and
expand_Delta_Record_Aggregate.
* sprint.adb: Handle N_Delta_Aggregate.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch11.adb (Expand_N_Exception_Declaration): Generate an
empty name when the exception declaration is subject to pragma
Discard_Names.
(Null_String): New routine.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* par-ch9.adb (P_Protected_Definition): Parse
any optional and potentially illegal pragmas which appear in
a protected operation declaration list.
(P_Task_Items): Parse
any optional and potentially illegal pragmas which appear in a
task item list.

From-SVN: r244794

use-after-scope: handle writes to a poisoned variable

2017-01-23  Martin Liska  <mliska@suse.cz>

* gcc.dg/asan/use-after-scope-10.c: New test.
* gcc.dg/asan/use-after-scope-11.c: New test.
* g++.dg/asan/use-after-scope-5.C: New test.
2017-01-23  Jakub Jelinek  <jakub@redhat.com>
    Martin Liska  <mliska@suse.cz>

* asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
* asan.c (asan_expand_poison_ifn): Support stores and use
appropriate ASAN report function.
* internal-fn.c (expand_ASAN_POISON_USE): New function.
* internal-fn.def (ASAN_POISON_USE): Declare.
* tree-into-ssa.c (maybe_add_asan_poison_write): New function.
(maybe_register_def): Create ASAN_POISON_USE when sanitizing.
* tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
ASAN_POISON calls w/o LHS.
* tree-ssa.c (execute_update_addresses_taken): Create clobber
for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
* gimplify.c (asan_poison_variables): Add attribute
use_after_scope_memory to variables that really needs to live
in memory.
* tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
having the attribute.

From-SVN: r244793

[multiple changes]

2017-01-23  Pascal Obry  <obry@adacore.com>

* s-taprop-mingw.adb (Enter_Task): Initialize the Thread handle which
is needed when a foreign thread call a Win32 API using a thread handle
like GetThreadTimes() for example.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* sem_ch13.adb (Analyze_Attribute_Definition_Clause): Do not
allow an 'Address clause to be specified on a prefix of a
class-wide type.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* checks.adb (Insert_Valid_Check): Ensure that the prefix of
attribute 'Valid is a renaming of the original expression when
the expression denotes a name. For all other kinds of expression,
use a constant to capture the value.
* exp_util.adb (Is_Name_Reference): Moved to Sem_Util.
* sem_util.ads, sem_util.adb (Is_Name_Reference): Moved from Exp_Util.

2017-01-23  Justin Squirek  <squirek@adacore.com>

* sem_eval.adb (Eval_Integer_Literal): Add special
case to avoid optimizing out check if the literal appears in
an if-expression.

From-SVN: r244792

Speed up use-after-scope (v2): rewrite into SSA

2017-01-23  Martin Liska  <mliska@suse.cz>

* asan.c (create_asan_shadow_var): New function.
(asan_expand_poison_ifn): Likewise.
* asan.h (asan_expand_poison_ifn): New declaration.
* internal-fn.c (expand_ASAN_POISON): Likewise.
* internal-fn.def (ASAN_POISON): New builtin.
* sanopt.c (pass_sanopt::execute): Expand
asan_expand_poison_ifn.
* tree-inline.c (copy_decl_for_dup_finish): Make function
external.
* tree-inline.h (copy_decl_for_dup_finish): Likewise.
* tree-ssa.c (is_asan_mark_p): New function.
(execute_update_addresses_taken): Rewrite local variables
(identified just by use-after-scope as addressable) into SSA.
2017-01-23  Martin Liska  <mliska@suse.cz>

* gcc.dg/asan/use-after-scope-3.c: Add additional flags.
* gcc.dg/asan/use-after-scope-9.c: Likewise and grep for
sanopt optimization for ASAN_POISON.

From-SVN: r244791

sem_ch4.adb (Try_Primitive_Operations, [...]): argument is valid if it is a derived type with unknown discriminants that...

2017-01-23  Ed Schonberg  <schonberg@adacore.com>

* sem_ch4.adb (Try_Primitive_Operations,
Is_Valid_First_Argument_Of): argument is valid if it is a derived
type with unknown discriminants that matches its underlying
record view.
* exp_util.adb (Expand_Subtype_From_Expr): Do not expand
expression if its type is derived from a limited type with
unknown discriminants, because the expansion (which is a call)
must be expanded in the enclosing context to add the proper build-
in-place parameters to the call.
* lib.ads, exp_ch9.adb: Minor fixes in comments.

From-SVN: r244790

[multiple changes]

2017-01-23  Yannick Moy  <moy@adacore.com>

* frontend.adb (Frontend): Do not load runtime
unit for GNATprove when parsing failed.
* exp_ch9.adb: minor removal of extra whitespace
* exp_ch6.adb: minor typo in comment
* sem_util.adb: Code cleanup.
* exp_ch9.ads, par-ch2.adb: minor style fixes in whitespace and comment
* a-ngcefu.adb: minor style fix in whitespace

2017-01-23  Thomas Quinot  <quinot@adacore.com>

* scos.ads: Document usage of 'd' as default SCO kind for
declarations.
* par_sco.adb (Traverse_Declarations_Or_Statements.
Traverse_Degenerate_Subprogram): New supporting routine for expression
functions and null procedures.
(Traverse_Declarations_Or_Statements.Traverse_One): Add
N_Expression_Function to the subprogram case; add required
support for null procedures and expression functions.

2017-01-23  Bob Duff  <duff@adacore.com>

* namet.ads (Bounded_String): Decrease the size of type
Bounded_String to avoid running out of stack space.
* namet.ads (Append): Don't ignore buffer overflow; raise
Program_Error instead.

From-SVN: r244789

[multiple changes]

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch5.adb, freeze.adb, par-ch4.adb, scng.adb, sem_ch13.adb,
sem_ch3.adb, sem_ch5.adb, sem_ch5.ads, sem_util.adb, sinfo.ads: Minor
reformatting.
* exp_ch9.adb: minor style fix in comment.

2017-01-23  Ed Schonberg  <schonberg@adacore.com>

* sem_ch4.adb (Analyze_Allocator): Handle properly a type derived
for a limited record extension with unknown discriminants whose
full view has no discriminants.

2017-01-23  Yannick Moy  <moy@adacore.com>

* exp_spark.adb: Alphabetize with clauses.

From-SVN: r244788

[multiple changes]

2017-01-23  Gary Dismukes  <dismukes@adacore.com>

* a-calend.adb, prep.adb, debug.adb, prj.ads, prepcomp.adb,
exp_disp.adb, s-imgrea.adb, g-socket.adb, g-socket.ads, sem_ch13.adb,
prj-tree.ads: Minor spelling change for consistency (behaviour ->
behavior).

2017-01-23  Ed Schonberg  <schonberg@adacore.com>

* scng.adb (Scan): Use Ada version Ada_2020 to flag use of
Target_Name.
* par-ch4.adb (P_Primary): Ditto.
* opt.ads: Add Ada_2020 (optimistically) to enumeration list of
Ada_Version_Type.
* switch-c.adb (Scan_Front_End_Switches): Recognize -gnat2020 for
new Ada version Ada_2020.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_attr.adb (Expand_Loop_Entry_Attribute): Force the generation
of a nominal type for the constant which captures the value of
the attribute prefix. Various clean ups.
* sem_attr.adb (Analyze_Attribute): Clean up the processing of
'Loop_Entry.

2017-01-23  Yannick Moy  <moy@adacore.com>

* sem_util.adb (Has_Enabled_Property): Treat
protected objects and variables differently from other variables.

From-SVN: r244787

sem_ch13.adb (Adjust_Record_For_Reverse_Bit_Order): Split original Ada 95 part off into new subprogram below.

2017-01-23  Thomas Quinot  <quinot@adacore.com>

* sem_ch13.adb (Adjust_Record_For_Reverse_Bit_Order):
Split original Ada 95 part off into new subprogram
below. Call that subprogram (instead of proceeding with
AI95-0133 behaviour) if debug switch -gnatd.p is in use.
(Adjust_Record_For_Reverse_Bit_Order_Ada_95): ... new subprogram
* debug.adb Document new switch -gnatd.p
* freeze.adb (Freeze_Entity.Freeze_Record_Type): Do not adjust
record for reverse bit order if an error has already been posted
on the record type.  This avoids generating extraneous "info:"
messages for illegal code.

From-SVN: r244786

sem_ch3.adb (Analyze_Declarations): Correct comments

2017-01-23  Justin Squirek  <squirek@adacore.com>

* sem_ch3.adb (Analyze_Declarations): Correct comments
* freeze.adb (Find_Constant): Add detection of deferred constants
so they are not incorrectly flagged as premature.

From-SVN: r244785

Minor reformatting.

From-SVN: r244784

scans.ads: New token At_Sign.

2017-01-23  Ed Schonberg  <schonberg@adacore.com>

* scans.ads: New token At_Sign. Remove '@' from list of illegal
characters for future version of the language. '@' is legal name.
* scng.ads, scng.adb (Scan):  Handle '@' appropriately.
* scn.adb (Scan_Reserved_Identifier): An occurrence of '@'
denotes a Target_Name.
* par-ch4.adb (P_Name, P_Primary): Handle Target_Name.
* sinfo.ads, sinfo.adb (N_Target_Name): New non-terminal node.
(Has_Target_Names): New flag on N_Assignment_Statement, to
indicate that RHS has occurrences of N_Target_Name.
* sem.adb: Call Analyze_Target_Name.
* sem_ch5.ads, sem_ch5.adb (Analyze_Target_Name): New subpogram.
(urrent_LHS): Global variable that denotes LHS of assignment,
used in the analysis of Target_Name nodes.
* sem_res.adb (Resolve_Target_Name): New procedure.
* exp_ch5.adb (Expand_Assign_With_Target_Names): (AI12-0125):
N is an assignment statement whose RHS contains occurences of @
that designate the value of the LHS of the assignment. If the
LHS is side-effect free the target names can be replaced with
a copy of the LHS; otherwise the semantics of the assignment
is described in terms of a procedure with an in-out parameter,
and expanded as such.
(Expand_N_Assignment_Statement): Call
Expand_Assign_With_Target_Names when needed.
* exp_util.adb (Insert_Actions): Take into account N_Target_Name.
* sprint.adb: Handle N_Target_Name.

From-SVN: r244783

[multiple changes]

2017-01-23  Eric Botcazou  <ebotcazou@adacore.com>

* checks.adb: Minor fix in comment.

2017-01-23  Philippe Gil  <gil@adacore.com>

* g-debpoo.adb (Do_Report) remove freed chunks from chunks
count in Sort = Memory_Usage or Allocations_Count

2017-01-23  Justin Squirek  <squirek@adacore.com>

* sem_ch3.adb: Code cleanup.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* sem_prag.adb (Analyze_Refined_Depends_In_Decl_Part): Move all global
variables to the local variable section. Update the profile
of various nested routine that previously had visibility
of those globals. One the matching phase has completed,
remove certain classes of clauses which are considered noise.
(Check_Dependency_Clause): Properly detect a match between two
'Result attributes. Update the various post-match cases to use
Is_Already_Matched as this routine now automatically recognizes
a previously matched 'Result attribute.
(Is_Already_Matched): New routine.
(Remove_Extra_Clauses): New routine.
(Report_Extra_Clauses): Remove the detection of ... => null
clauses as this is now done in Remove_Extra_Clauses.

From-SVN: r244782

Fix typo.

From-SVN: r244781

[multiple changes]

2017-01-23  Ed Schonberg  <schonberg@adacore.com>

* sem_aggr.adb (Resolve_Array_Aggregate): In ASIS mode do not
report on spurious overlaps between values involving a subtype
with a static predicate, because the expansion of such a subtype
into individual ranges in inhibited in ASIS mode.

2017-01-23  Justin Squirek  <squirek@adacore.com>

* sem_ch3.adb (Analyze_Declarations): Add detection
of an edge case and delay freezing if it is present.

From-SVN: r244780

[testsuite] Fix FAIL: gcc.dg/lto/pr69188 on bare-metal targets

PR lto/69188
* gcc.dg/lto/pr69188_0.c: Require profiling support for testcase.

From-SVN: r244779

sem_ch3.adb, [...]: Minor reformatting.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* sem_ch3.adb, exp_spark.adb, exp_attr.adb, sem_ch9.adb, sem_prag.adb,
sem_util.adb, sem_warn.adb, exp_ch3.adb: Minor reformatting.

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* freeze.adb (Freeze_Subprogram): Ensure that all anonymous
access-to-subprogram types inherit the convention of the
associated subprogram. (Set_Profile_Convention): New routine.
* sem_ch6.adb (Check_Conformance): Do not compare the conventions
of the two entities directly, use Conventions_Match to account
for anonymous access-to-subprogram and subprogram types.
(Conventions_Match): New routine.

From-SVN: r244778

exp_spark.adb (Expand_SPARK_Attribute_Reference): For attributes which return Universal_Integer...

2017-01-23  Claire Dross  <dross@adacore.com>

* exp_spark.adb (Expand_SPARK_Attribute_Reference): For attributes
which return Universal_Integer, force the overflow check flag for
Length and Range_Length for types as big as Long_Long_Integer.

From-SVN: r244777

[multiple changes]

2017-01-23  Claire Dross  <dross@adacore.com>

* exp_spark.adb (Expand_SPARK_Attribute_Reference):  For
attributes which return Universal_Integer, introduce a conversion
to the expected type with the appropriate check flags set.
* sem_res.adb (Resolve_Range): The higher bound can be in Typ's
base type if the range is null. It may still be invalid if it
is higher than the lower bound. This is checked later in the
context in which the range appears.

2017-01-23  Pierre-Marie de Rodat  <derodat@adacore.com>

* scos.ads: Introduce a constant to represent ignored
dependencies in SCO_Unit_Table_Entry.

From-SVN: r244776

[multiple changes]

2017-01-23  Hristian Kirtchev  <kirtchev@adacore.com>

* exp_ch9.adb (Expand_N_Protected_Type_Declaration): Remove extra
spaces from error messages.

2017-01-23  Ed Schonberg  <schonberg@adacore.com>

* exp_ch3.adb (Check_Large_Modular_Array): New procedure,
subsidiary to Expand_N_Object_ Declaration, to compute a guard on
an object declaration for an array type with a modular index type
with the size of Long_Long_Integer. Special processing is needed
in this case to compute reliably the size of the object, and
eventually  to raise Storage_Error, when wrap-around arithmetic
might compute a meangingless size for the object.

2017-01-23  Justin Squirek  <squirek@adacore.com>

* a-wtenau.adb, par-endh.adb, sem_prag.adb,
sem_type.adb: Code cleanups.

From-SVN: r244775

sem_res.adb (Resolve_Call): In the part of the code where it is deciding whether to turn the call into an...

2017-01-23  Bob Duff  <duff@adacore.com>

* sem_res.adb (Resolve_Call): In the part of the code where
it is deciding whether to turn the call into an indexed
component, avoid doing so if the call is to an instance of
Unchecked_Conversion. Otherwise, the compiler turns it into an
indexed component, and resolution of that turns it back into a
function call, and so on, resulting in infinite recursion.
* sem_util.adb (Needs_One_Actual): If the first formal has a
default, then return False.

From-SVN: r244774

re PR lto/79061 ([LTO][ASAN] LTO plus ASAN fails with "AddressSanitizer: initialization-order-fiasco")

Revert fix for PR lto/79061 due to this regresses compile-time by 100%
on some fortran cases.

From-SVN: r244773

[ARM,AArch64][testsuite] Fix format string in AdvSIMD tests.

2017-01-23  Christophe Lyon  <christophe.lyon@linaro.org>

* gcc.target/aarch64/advsimd-intrinsics/arm-neon-ref.h (CHECK_POLY): New.
(CHECK_RESULTS_NAMED_NO_FP16): Call CHECK_POLY instead of CHECK
for poly*_t types.
* gcc.target/aarch64/advsimd-intrinsics/vcnt.c (FNNAME): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vcombine.c (void
exec_vcombine): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vcreate.c (FNNAME): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vget_high.c (void exec_vget_high): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vget_low.c (void exec_vget_low): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vldX.c (void exec_vldX): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vldX_dup.c (void exec_vldX_dup): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vldX_lane.c (void exec_vldX_lane): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vmul.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vmvn.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vqtbX.c (void exec_vqtbX): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vreinterpret_p128.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vrev.c (void exec_vrev): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vsXi_n.inc: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vshuffle.inc: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vsli_n.c (void vsli_extra): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vsri_n.c (void vsri_extra): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vstX_lane.c (void exec_vstX_lane): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vtbX.c (void exec_vtbX): Likewise.
* gcc.target/aarch64/advsimd-intrinsics/p64_p128.c (int main):
Likewise.
(TEST_VGET_LANE): Cast to uint to avoid warning.
* gcc.target/aarch64/advsimd-intrinsics/unary_sat_op.inc (void
FNNAME): Fix PRIx format for int16_t, int32_t, uint16_t, uint32_t.
* gcc.target/aarch64/advsimd-intrinsics/vfms_vfma_n.c (void
exec_vfma_vfms_n): Fix PRIx format for float64_t.
* gcc.target/aarch64/advsimd-intrinsics/vmovn.c (void exec_vmovn):
Fix PRIx format for int8_t, int16_t, uint8_t, uint16_t.
* gcc.target/aarch64/advsimd-intrinsics/vmul_lane.c (void
exec_vmul_lane): Fix PRIx format for int16_t, uint16_t.
* gcc.target/aarch64/advsimd-intrinsics/vmul_n.c: Likewise.
* gcc.target/aarch64/advsimd-intrinsics/vmull.c (void exec_vmull):
Fix PRIx format for int16_t, int64_t, uint16_t, uint64_t. Call
CHECK_POLY instead of CHECK for poly64_t types.
* gcc.target/aarch64/advsimd-intrinsics/vmull_lane.c (void
exec_vmull_lane): Fix PRIx format for int64_t, uint64_t.
* gcc.target/aarch64/advsimd-intrinsics/vpXXX.inc:
Fix PRIx format for int8_t, int16_t, uint8_t, uint16_t.
* gcc.target/aarch64/advsimd-intrinsics/vqabs.c (void
vqabs_extra): Fix PRIx format for int16_t, int32_t, uint16_t,
uint32_t.
* gcc.target/aarch64/advsimd-intrinsics/vqdmull.c: Fix PRIx format
for int32_t, int64_t.
* gcc.target/aarch64/advsimd-intrinsics/vqneg.c (void
vqneg_extra): Fix PRIx format for int16_t, int32_t.
* gcc.target/aarch64/advsimd-intrinsics/vreinterpret_p64.c
(TEST_VREINTERPRET_TO_POLY): New. (main): Call
TEST_VREINTERPRET_TO_POLY instead of TEST_VREINTERPRET where
needed.

From-SVN: r244772

patch_tester.sh (TESTLOGS): Remove $TARGET/libjava/testsuite/libjava.sum.

* patch_tester.sh (TESTLOGS): Remove
$TARGET/libjava/testsuite/libjava.sum.

From-SVN: r244771

* doc/xml/manual/debug.xml: code.google.com uses https now.

From-SVN: r244770

Daily bump.

From-SVN: r244769

* warn_summary: Remove the -java option.

From-SVN: r244766

* doc/xml/manual/test.xml: Fix link into gccint online manual.

From-SVN: r244765

install.texi (Specific): opensource.apple.com uses https now.

* doc/install.texi (Specific): opensource.apple.com uses https
now. Remove trailing slash.

From-SVN: r244764

re PR fortran/79154 (omp declare simd in pure function?)

PR fortran/79154
* parse.c (matchs, matcho, matchds, matchdo): Replace return st;
with { ret = st; goto finish; }.
(decode_omp_directive): Allow declare simd, declare target and
simd directives in PURE/ELEMENTAL procedures.  Only call
gfc_unset_implicit_pure on successful match of other procedures.

* gfortran.dg/gomp/pr79154-1.f90: New test.
* gfortran.dg/gomp/pr79154-2.f90: New test.

From-SVN: r244763

2017-01-22  Matthias Klose  <doko@ubuntu.com>

        * Import zlib 1.2.11.
        * configure: Regenerate.

Changes in 1.2.11 (15 Jan 2017)
- Fix deflate stored bug when pulling last block from window
- Permit immediate deflateParams changes before any deflate input

From-SVN: r244762

* README.Portability: Remove note on an Irix compatibility issue.

From-SVN: r244759

gcov.c (INCLUDE_ALGORITHM): Define.

* gcov.c (INCLUDE_ALGORITHM): Define.
(INCLUDE_VECTOR): Define.
No longer include <vector> and <algorithm> directly.

From-SVN: r244758

* gcc.dg/tree-ssa/pr77445-2.c: Quote brackets.

From-SVN: r244757

Daily bump.

From-SVN: r244756

extend.texi (Thread-Local): Change akkadia.org reference to https.

* doc/extend.texi (Thread-Local): Change www.akkadia.org reference
to https.
* doc/invoke.texi (Code Gen Options): Ditto.

From-SVN: r244751

gfc-internals.texi (Symbol Versioning): Change references to akkadia.org to https.

* gfc-internals.texi (Symbol Versioning): Change references
to www.akkadia.org to https.

From-SVN: r244750

re PR lto/78407 (LTO breaks separate overriding of symbol aliases)

PR lto/78407
* cfg.c (update_bb_profile_for_threading): Fix updating of probablity.

From-SVN: r244749

Make poisoned hashes SFINAE away the call operator of the hash.

* include/bits/functional_hash.h
(__poison_hash::__enable_hash_call): New.
* include/std/optional (__optional_hash_call_base): New.
(hash<optional<_Tp>>): Derive from the new base,
move the hash function into that base.
* include/std/variant (__variant_hash_call_base_impl): New.
(__variant_hash_call_base): Likewise.
(hash<variant<_Types...>>): Derive from the new base,
move the hash function into that base.
* testsuite/20_util/optional/hash.cc: Add tests for is_callable.
* testsuite/20_util/variant/hash.cc: Likewise.

From-SVN: r244748

re PR testsuite/79169 (g++.dg/warn/Wduplicated-branches1.C fails on powerpc since its introduction in r244705)

PR testsuite/79169
* g++.dg/warn/Wduplicated-branches1.C: Use signed char.

From-SVN: r244747

* gcc.dg/tree-ssa/pr77445-2.c: Adjust template.

From-SVN: r244746

sem_eval.adb (Compile_Time_Compare): Reinstate the expr+literal (etc) optimizations when...

* sem_eval.adb (Compile_Time_Compare): Reinstate the expr+literal (etc)
optimizations when the type is modular and the offsets are equal.

From-SVN: r244745

gfortran.h (gfc_extract_int): Change return type to bool.

* gfortran.h (gfc_extract_int): Change return type to bool.  Add
int argument with = 0.
* decl.c (gfc_match_kind_spec): Adjust gfc_extract_int caller, pass
1 as new last argument to it, don't emit gfc_error.
(match_char_kind): Likewise.
(gfc_match_decl_type_spec): Use gfc_get_string ("%s", x) instead of
gfc_get_string (x).
(gfc_match_derived_decl, match_binding_attributes): Likewise.
(gfc_match_structure_decl): Don't sprintf back to name, call
get_struct_decl directly with gfc_dt_upper_string (name) result.
* trans-stmt.c (gfc_trans_allocate): Use gfc_get_string ("%s", x)
instead of gfc_get_string (x).
* module.c (gfc_dt_lower_string, gfc_dt_upper_string,
gfc_match_use, gfc_match_submodule, find_true_name, mio_pool_string,
mio_symtree_ref, mio_expr, mio_omp_udr_expr, load_generic_interfaces,
load_omp_udrs, load_needed, read_module, dump_module,
create_intrinsic_function, import_iso_c_binding_module,
create_int_parameter, create_int_parameter_array, create_derived_type,
use_iso_fortran_env_module): Likewise.
* error.c (gfc_diagnostic_starter, gfc_diagnostic_start_span): Use
pp_verbatim (context->printer, "%s", x) instead of
pp_verbatim (context->printer, x).
* match.c (gfc_match_small_int): Adjust gfc_extract_int caller, pass
1 as new last argument to it, don't emit gfc_error.
(gfc_match_small_int_expr): Likewise.
* iresolve.c (gfc_get_string): Optimize format "%s" case.
(resolve_bound): Use gfc_get_string ("%s", x) instead of
gfc_get_string (x).
(resolve_transformational): Formatting fix.
(gfc_resolve_char_achar): Change name argument to bool is_achar,
use a single format string and if is_achar add "a" before "char".
(gfc_resolve_achar, gfc_resolve_char): Adjust callers.
* expr.c (gfc_extract_int): Change return type to bool, return true
if some error occurred.  Add REPORT_ERROR argument, if non-zero
call either gfc_error or gfc_error_now depending on its sign.
* arith.c (arith_power): Adjust gfc_extract_int caller.
* symbol.c (gfc_add_component): Use gfc_get_string ("%s", x) instead
of gfc_get_string (x).
(gfc_new_symtree, gfc_delete_symtree, gfc_get_uop, gfc_new_symbol,
gfc_get_gsymbol, generate_isocbinding_symbol): Likewise.
* openmp.c (gfc_match_omp_clauses): Adjust gfc_extract_int caller, pass
-1 as new last argument to it, don't emit gfc_error_now.
(gfc_match_omp_declare_reduction): Use gfc_get_string ("%s", x)
instead of gfc_get_string (x).
* check.c (kind_check): Adjust gfc_extract_int caller.
* intrinsic.c (add_sym, find_sym, make_alias): Use
gfc_get_string ("%s", x) instead of gfc_get_string (x).
* simplify.c (get_kind, gfc_simplify_btest, gfc_simplify_maskr,
gfc_simplify_maskl, gfc_simplify_poppar, gfc_simplify_repeat,
gfc_simplify_selected_int_kind, gfc_simplify_selected_real_kind):
Adjust gfc_extract_int callers.
* trans-decl.c (gfc_find_module): Use gfc_get_string ("%s", x)
instead of gfc_get_string (x).
* matchexp.c (expression_syntax): Add const.
* primary.c (match_kind_param, match_hollerith_constant,
match_string_constant): Adjust gfc_extract_int callers.
(match_keyword_arg): Use gfc_get_string ("%s", x) instead of
gfc_get_string (x).
* frontend-passes.c (optimize_minmaxloc): Likewise.

From-SVN: r244744

* config/i386/cygming-crtbegin.c (LIBGCJ_SONAME): No longer #define.

From-SVN: r244743

re PR other/79046 (g++ -print-file-name=plugin uses full version number in path)

PR other/79046
libatomic/
* testsuite/Makefile.am (gcc_version): Use @get_gcc_base_ver@ instead
of cat to get version from BASE-VER file.
* testsuite/Makefile.in: Regenerated.
libffi/
* configure.ac: Add GCC_BASE_VER.
* include/Makefile.am (gcc_version): Use @get_gcc_base_ver@ instead
of cat to get version from BASE-VER file.
* configure: Regenerated.
* testsuite/Makefile.in: Regenerated.
* include/Makefile.in: Regenerated.
* Makefile.in: Regenerated.
* man/Makefile.in: Regenerated.
libgomp/
* testsuite/Makefile.am (gcc_version): Use @get_gcc_base_ver@ instead
of cat to get version from BASE-VER file.
* testsuite/Makefile.in: Regenerated.
libitm/
* testsuite/Makefile.am (gcc_version): Use @get_gcc_base_ver@ instead
of cat to get version from BASE-VER file.
* testsuite/Makefile.in: Regenerated.
libmpx/
* mpxrt/Makefile.am (gcc_version): Use @get_gcc_base_ver@ instead
of cat to get version from BASE-VER file.
* mpxwrap/Makefile.am (gcc_version): Likewise.
* mpxrt/Makefile.in: Regenerated.
* mpxwrap/Makefile.in: Regenerated.
liboffloadmic/
* plugin/configure.ac: Add GCC_BASE_VER.
* plugin/Makefile.am (gcc_version): Use @get_gcc_base_ver@ instead
of cat to get version from BASE-VER file.
* plugin/configure: Regenerated.
* plugin/aclocal.m4: Regenerated.
* plugin/Makefile.in: Regenerated.
libsanitizer/
* interception/Makefile.am (gcc_version): Use @get_gcc_base_ver@
instead of cat to get version from BASE-VER file.
* asan/Makefile.am (gcc_version): Likewise.
* ubsan/Makefile.am (gcc_version): Likewise.
* sanitizer_common/Makefile.am (gcc_version): Likewise.
* lsan/Makefile.am (gcc_version): Likewise.
* tsan/Makefile.am (gcc_version): Likewise.
* interception/Makefile.in: Regenerated.
* asan/Makefile.in: Regenerated.
* ubsan/Makefile.in: Regenerated.
* sanitizer_common/Makefile.in: Regenerated.
* lsan/Makefile.in: Regenerated.
* tsan/Makefile.in: Regenerated.
libvtv/
* testsuite/Makefile.am (gcc_version): Use @get_gcc_base_ver@ instead
of cat to get version from BASE-VER file.
* testsuite/Makefile.in: Regenerated.

From-SVN: r244742

re PR rtl-optimization/79125 (ICE in rtl_verify_bb_insns, at cfgrtl.c:2661 (error: flow control insn inside a basic block))

2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>

rtl-optimization/79125
* cprop.c (local_cprop_pass): Handle cases where we make an
unconditional trap.

PR rtl-optimization/79125
* gcc.dg/torture/pr79125.c: New test.

From-SVN: r244741

rs6000: Small varargs for BE SVR4 (PR61729, PR77850)

The varargs code for SVR4 puts all (integer) arguments in 4-byte slots.
When it then reads an item from there as something not a multiple of 4
bytes, it needs to adjust the address if big endian.  We didn't yet do
that.

This fixes the g++.dg/abi/scoped1.C, gcc.dg/compat/scalar-by-value-4,
and gcc.dg/compat/scalar-return-4 testcases.

PR target/61729
PR target/77850
* config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
read from, for big endian.

From-SVN: r244740

Daily bump.

From-SVN: r244739

runtime, testing/internal/testdeps: fixes for cgo

    Some fixes that permit misc/cgo/test in the master gc repository to
    pass using the current gccgo.

    Install testing/internal/testdeps.gox; it is needed by `go test`.

    Export runtime.lockedOSThread to enable calling via go:linkname; it is
    used by misc/cgo/test.

    Loop on EAGAIN when creating a new thread; this is what the gc code
    does, and misc/cgo/test tests that it works.

    Reviewed-on: https://go-review.googlesource.com/35479

From-SVN: r244733

[AArch64] Only build & test pauth code for LP64

gcc/
* config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
register pauth builtins for LP64 only.

libgcc/
* config/aarch64/aarch64-unwind.h: Empty this file on ILP32.
* unwind-dw2.c (execute_cfa_program):  Only multiplexing
DW_CFA_GNU_window_save for AArch64 and LP64.

gcc/testsuite/
* testsuite/gcc.target/aarch64/return_address_sign_1.c: Enable on LP64
only.
* testsuite/gcc.target/aarch64/return_address_sign_2.c: Likewise.
* testsuite/gcc.target/aarch64/return_address_sign_3.c: Likewise.

From-SVN: r244732

re PR go/79146 (Bootstrapping go on s390x fails; redefined symbols)

PR go/79146
    math/big: fix build on s390x

    Don't build arith_decl_s390x.go for gccgo; it is only for assembly
    code that has not yet been ported to gccgo.

    For GCC PR 79146.

From-SVN: r244731

PR c++/78495 - wrong code inherited ctor and invisi-ref parm

PR c++/78495 - wrong code inherited ctor and invisi-ref parm
* cp-gimplify.c (cp_generize_r): Don't skip thunks.

PR c++/79495
* g++.dg/cpp1z/inh-ctor38.C: New.

From-SVN: r244728

Support unsigned __int20 in checks for size_t mangling

2017-01-20  Joe Seymour  <joe.s@somniumtech.com>

* acinclude.m4 (GLIBCXX_CHECK_SIZE_T_MANGLING): Support uint20_t.
* configure: Regenerate.

From-SVN: r244727

re PR c/79152 (-Wimplicit-fallthrough false positive triggered by goto statements)

PR c/79152
* gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
non-case labels.

* c-c++-common/Wimplicit-fallthrough-35.c: New test.

From-SVN: r244726