law [Tue, 27 Nov 2018 15:22:01 +0000 (15:22 +0000)]
* config/microblaze/microblaze.c (microblaze_block_move_straight): Use
RETURN_BEGIN in call to move_by_pieces.
(microblaze_expand_block_move): Likewise.
marxin [Tue, 27 Nov 2018 13:06:48 +0000 (13:06 +0000)]
Support simd function declarations via a pre-include.
2018-11-27 Martin Liska <mliska@suse.cz>
* config/gnu-user.h (TARGET_F951_OPTIONS): New.
* gcc.c (find_fortran_preinclude_file): New function
to handle Fortran pre-include.
2018-11-27 Martin Liska <mliska@suse.cz>
* decl.c (gfc_match_gcc_builtin): New function.
* gfortran.h (struct vect_builtin_tuple): New.
(gfc_adjust_builtins): Likewise.
* lang-specs.h (TARGET_F951_OPTIONS): New.
(F951_OPTIONS): Use it.
* lang.opt: Add new option -fpre-include.
* match.h (gfc_match_gcc_builtin): Declare new function.
* parse.c (decode_gcc_attribute): Handle builtin.
(parse_progunit): Call gfc_adjust_builtins.
* scanner.c (gfc_new_file): Load pre-included header file
when provided.
* trans-intrinsic.c (add_simd_flag_for_built_in): New.
(gfc_adjust_builtins): Likewise.
2018-11-27 Martin Liska <mliska@suse.cz>
* gfortran.dg/simd-builtins-1.f90: New test.
* gfortran.dg/simd-builtins-1.h: New test.
* gfortran.dg/simd-builtins-2.f90: New test.
* gfortran.dg/simd-builtins-3.f90: New test.
* gfortran.dg/simd-builtins-3.h: New test.
* gfortran.dg/simd-builtins-4.f: New test.
* gfortran.dg/simd-builtins-4.h: New test.
* gfortran.dg/simd-builtins-5.f: New test.
* gfortran.dg/simd-builtins-6.f90: New test.
marxin [Tue, 27 Nov 2018 12:30:59 +0000 (12:30 +0000)]
Come up with memop_ret enum instead of int endp for memory operations.
2018-11-27 Martin Liska <mliska@suse.cz>
* asan.c (asan_emit_stack_protection): Use new enum values
instead of int constants.
* builtins.c (expand_builtin_memory_copy_args): Replace int
type with memop_ret enum type.
(expand_builtin_mempcpy_args): Likewise.
(expand_builtin_memcpy): Use new enum values
instead of int constants. Likewise.
(expand_builtin_mempcpy): Likewise.
(expand_movstr): Likewise.
(expand_builtin_strcpy_args): Likewise.
(expand_builtin_stpcpy_1): Likewise.
(expand_builtin_strncpy): Likewise.
(expand_builtin_memset_args): Likewise.
* expr.c (move_by_pieces_d::finish_endp): Rename to ...
(move_by_pieces_d::finish_retmode): ... this.
(move_by_pieces): Change last argument type to memop_ret.
(store_by_pieces): Use new enum values
instead of int constants.
(emit_block_move_hints): Likewise.
(emit_push_insn): Likewise.
(store_expr): Likewise.
* expr.h (store_by_pieces): Change int to newly added enum
type.
* rtl.h (enum memop_ret): Define.
(move_by_pieces): Use the enum type.
ebotcazou [Tue, 27 Nov 2018 10:37:20 +0000 (10:37 +0000)]
* c-ada-spec.c: Include stringpool.h.
(has_static_fields): Return false for incomplete types.
(is_tagged_type): Likewise.
(has_nontrivial_methods): Likewise.
(dump_ada_node) <INTEGER_TYPE>: Deal specifically with __int128.
(struct overloaded_name_hash): New structure.
(struct overloaded_name_hasher): Likewise.
(overloaded_names): New global variable.
(init_overloaded_names): New static function.
(overloaded_name_p): New predicate.
(dump_ada_declaration) <TYPE_DECL>: Tidy up and set TREE_VISITED
on the TYPE_STUB_DECL of the original type of a typedef, if any.
<FUNCTION_DECL>: Bail out for an unsupported overloaded name.
Remove always-true condition and dump forward types.
(dump_ada_specs): Delete overloaded_names.
amodra [Tue, 27 Nov 2018 01:59:56 +0000 (01:59 +0000)]
[RS6000] libgcc cfi
There are a few places in libgcc assembly where we don't emit call
frame information for functions, potentially breaking unwinding from
asynchronous signal handlers. This patch fixes them. Although I
patch tramp.S there is no attempt made to provide CFI for the actual
trampoline on the stack. Doing that would require generating CFI at
run time and both registering and deregistering it, which is probably
not worth doing since it would significantly slow down the call.
* config/rs6000/morestack.S (__stack_split_initialize),
(__morestack_get_guard, __morestack_set_guard),
(__morestack_make_guard): Provide CFI covering these functions.
* config/rs6000/tramp.S (__trampoline_setup): Likewise.
amodra [Tue, 27 Nov 2018 01:56:06 +0000 (01:56 +0000)]
[RS6000] PowerPC -mcpu=native support
The -mcpu=native support has bit-rotted a little, in particular the
fallback when the native cpu couldn't be determined. This patch fixes
the bit-rot and reorganizes ASM_CPU_SPEC so that it should be a little
easier to keep the -mcpu=native data up to date.
The patch also changes the fix for PR63177 (-mpower9-vector being
passed by the user when the default is -mpower8) to also apply when
-mcpu=powerpc64le and -mcpu=native is given. I'll note that the hack
for PR63177 should probably be extended to lots of other options, if
we're going to continue supporting all those sub-architecture options
(-mpower9-vector, -mpower8-vector, -mcrypto, -mdirect-move, -mhtm,
-mvsx and others) in the positive sense. I think those should have
only been supported in their -mno- variants..
* config/rs6000/aix71.h (ASM_SPEC): Don't select default -maix64
cpu here.
(ASM_CPU_SPEC): Do so here. Rewrite using if .. else if .. specs
form. Error on missing -mcpu case.
* config/rs6000/driver-rs6000.c (asm_names <_AIX>): Update NULL case.
(asm_names <!_AIX>): Add missing cpus. Update NULL case. Apply
PR63177 fix for -mcpu=power8 and -mcpu=powerpc64le.
* config/rs6000/rs6000.h (ASM_CPU_SPEC): Rewrite using if ..
else if .. specs form. Error on missing -mcpu case. Don't output
duplicate -maltivec. Apply PR63177 fix for -mcpu=powerpc64le.
dmalcolm [Tue, 27 Nov 2018 00:55:28 +0000 (00:55 +0000)]
Fix missing dump_impl_location_t values, using a new dump_metadata_t
The dump_* API attempts to capture emission location metadata for the
various dump messages, but looking in -fsave-optimization-record shows
that many dump messages are lacking useful impl_location values, instead
having this location within dumpfile.c:
The problem is that the auto-capturing of dump_impl_location_t is tied to
dump_location_t, and this is tied to the dump_*_loc calls. If a message
comes from a dump_* call without a "_loc" suffix (e.g. dump_printf), the
current code synthesizes the dump_location_t within
dump_context::ensure_pending_optinfo, and thus saves the useless
impl_location seen above.
This patch fixes things by changing the dump_* API so that, rather than
taking a dump_flags_t, they take a new class dump_metadata_t, which is
constructed from a dump_flags_t, but captures the emission location.
Hence e.g.:
dump_printf (MSG_NOTE, "some message\n");
implicitly builds a dump_metadata_t wrapping the MSG_NOTE and the
emission location. If there are several dump_printf calls without
a dump_*_loc call, the emission location within the optinfo is that
of the first dump call within it.
The patch updates selftest::test_capture_of_dump_calls to verify
that the impl location of various dump_* calls is captured. I also
verified things manually.
gcc/ChangeLog:
* dump-context.h (dump_context::dump_loc): Convert 1st param from
dump_flags_t to const dump_metadata_t &. Convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_context::dump_loc_immediate): Convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_context::dump_gimple_stmt): Convert 1st param from
dump_flags_t to const dump_metadata_t &.
(dump_context::void dump_gimple_stmt_loc): Likewise; convert
2nd param from const dump_location_t & to
const dump_user_location_t &.
(dump_context::dump_gimple_expr): Convert 1st param from
dump_flags_t to const dump_metadata_t &.
(dump_context::dump_gimple_expr_loc): Likewise; convert
2nd param from const dump_location_t & to
const dump_user_location_t &.
(dump_context::dump_generic_expr): Convert 1st param from
dump_flags_t to const dump_metadata_t &.
(dump_context::dump_generic_expr_loc): Likewise; convert
2nd param from const dump_location_t & to
const dump_user_location_t &.
(dump_context::dump_printf_va): Convert 1st param from
dump_flags_t to const dump_metadata_t &.
(dump_context::dump_printf_loc_va): Likewise; convert
2nd param from const dump_location_t & to
const dump_user_location_t &.
(dump_context::dump_dec): Convert 1st param from
dump_flags_t to const dump_metadata_t &.
(dump_context::dump_symtab_node): Likewise.
(dump_context::begin_scope): Split out 2nd param into
user and impl locations.
(dump_context::ensure_pending_optinfo): Add metadata param.
(dump_context::begin_next_optinfo): Replace dump_location_t param
with metadata and user location.
* dumpfile.c (dump_context::dump_loc): Convert 1st param from
dump_flags_t to const dump_metadata_t &. Convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_context::dump_loc_immediate): Convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_context::dump_gimple_stmt): Convert 1st param from
dump_flags_t to const dump_metadata_t &.
(dump_context::void dump_gimple_stmt_loc): Likewise; convert
2nd param from const dump_location_t & to
const dump_user_location_t &.
(dump_context::dump_gimple_expr): Convert 1st param from
dump_flags_t to const dump_metadata_t &.
(dump_context::dump_gimple_expr_loc): Likewise; convert
2nd param from const dump_location_t & to
const dump_user_location_t &.
(dump_context::dump_generic_expr): Convert 1st param from
dump_flags_t to const dump_metadata_t &.
(dump_context::dump_generic_expr_loc): Likewise; convert
2nd param from const dump_location_t & to
const dump_user_location_t &.
(dump_context::dump_printf_va): Convert 1st param from
dump_flags_t to const dump_metadata_t &.
(dump_context::dump_printf_loc_va): Likewise; convert
2nd param from const dump_location_t & to
const dump_user_location_t &.
(dump_context::dump_dec): Convert 1st param from
dump_flags_t to const dump_metadata_t &.
(dump_context::dump_symtab_node): Likewise.
(dump_context::begin_scope): Split out 2nd param into
user and impl locations.
(dump_context::ensure_pending_optinfo): Add metadata param.
(dump_context::begin_next_optinfo): Replace dump_location_t param
with metadata and user location.
(dump_gimple_stmt): Convert 1st param from dump_flags_t to
const dump_metadata_t &.
(dump_gimple_stmt_loc): Likewise; convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_gimple_expr): Convert 1st param from dump_flags_t to
const dump_metadata_t &.
(dump_gimple_expr_loc): Likewise; convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_generic_expr): Convert 1st param from dump_flags_t to
const dump_metadata_t &.
(dump_generic_expr_loc): Likewise; convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_printf): Convert 1st param from dump_flags_t to
const dump_metadata_t &.
(dump_printf_loc): Likewise; convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_dec): Convert 1st param from dump_flags_t to
const dump_metadata_t &.
(dump_symtab_node): Likewise.
(dump_begin_scope): Split out 2nd param into user and impl
locations.
(selftest::assert_impl_location_eq): New function.
(ASSERT_IMPL_LOCATION_EQ): New macro.
(selftest::test_impl_location): Update to use
ASSERT_IMPL_LOCATION_EQ.
(selftest::test_capture_of_dump_calls): Convert "loc" to
dump_user_location_t. Add ASSERT_IMPL_LOCATION_EQ throughout,
verifying line numbers of dump emissions.
* dumpfile.h (class dump_metadata_t): New class.
(dump_printf): Convert 1st param from dump_flags_t to
const dump_metadata_t &.
(dump_printf_loc): Likewise; convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_generic_expr_loc): Likewise.
(dump_generic_expr): Convert 1st param from dump_flags_t to
const dump_metadata_t &.
(dump_gimple_stmt_loc): Likewise; convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_gimple_stmt): Convert 1st param from dump_flags_t to
const dump_metadata_t &.
(dump_gimple_expr_loc): Likewise; convert 2nd param from
const dump_location_t & to const dump_user_location_t &.
(dump_gimple_expr): Convert 1st param from dump_flags_t to
const dump_metadata_t &.
(dump_symtab_node): Likewise.
(dump_dec): Likewise.
(dump_begin_scope): Split out 2nd param into user and impl
locations.
(auto_dump_scope::auto_dump_scope): Split "loc" param into a user
location and impl_location, and capture the impl_location.
(AUTO_DUMP_SCOPE): Rename param from LOC to USER_LOC.
* loop-unroll.c (report_unroll): Update for changes to
dump_printf_loc and dump_printf.
* opt-problem.cc (opt_problem::opt_problem): Update for change to
dump_loc.
* optinfo-emit-json.cc
(selftest::test_building_json_from_dump_calls): Convert "loc" from
dump_location_t to dump_user_location_t.
* optinfo.cc (optinfo::emit_for_opt_problem): Update for change in
dump_loc_immediate.
* profile.c (compute_branch_probabilities): Update for change to
dump_printf_loc.
* selftest.h (ASSERT_STR_CONTAINS_AT): New macro.
* tree-vect-slp.c (vect_print_slp_tree): Update for change to
dump_printf_loc.
amodra [Tue, 27 Nov 2018 00:06:45 +0000 (00:06 +0000)]
[RS6000] rs6000/sysv4.h using gnu-user.h
This patch removes some duplication in rs6000/sysv4.h of macros found
in gnu-user.h that we want for linux. Including gnu-user.h will mean
powerpc doesn't miss updates to that file.
amodra [Mon, 26 Nov 2018 23:53:56 +0000 (23:53 +0000)]
[RS6000] Use config/linux.h for powerpc*-*-linux*
Using the macros in config/linux.h rather than duplicating them helps
stop future bitrot, and repairs existing bitrot (4 choices for libc in
linux.h, fewer in the rs6000 files not that it matters much). Also
fixes the fact that __gnu_linux__ was always defined rather than just
when glibc was the libc of choice.
* config.gcc (powerpc*-*-linux*): Add linux.h to tm_file.
* config/rs6000/linux.h (TARGET_OS_CPP_BUILTINS): Use
GNU_USER_TARGET_OS_CPP_BUILTINS.
(RS6000_ABI_NAME): Define.
* config/rs6000/linux64.h (TARGET_OS_CPP_BUILTINS): Use
GNU_USER_TARGET_OS_CPP_BUILTINS.
(MUSL_DYNAMIC_LINKER32): Undef before defining.
(UCLIBC_DYNAMIC_LINKER32, UCLIBC_DYNAMIC_LINKER64): Don't define.
(CHOOSE_DYNAMIC_LINKER): Don't define.
(GNU_USER_DYNAMIC_LINKER32, GNU_USER_DYNAMIC_LINKER64): Don't define.
* config/rs6000/sysv4.h (MUSL_DYNAMIC_LINKER): Undef before defining.
(CHOOSE_DYNAMIC_LINKER, GNU_USER_DYNAMIC_LINKER): Only define when
not already defined.
(CPP_OS_LINUX_SPEC): Remove defines and asserts handled by
TARGET_OS_CPP_BUILTINS.
ibuclaw [Mon, 26 Nov 2018 21:59:04 +0000 (21:59 +0000)]
Add filters for D language sources to contrib/update-copyright.py
2018-11-26 Iain Buclaw <ibuclaw@gdcproject.org>
* update-copyright.py (TestsuiteFilter): Skip .d tests.
(LibPhobosFilter): Add filter for upstream D sources.
(GCCCopyright): Add D Language Foundation as external author.
(GCCCmdLine): Add libphobos.
ian [Mon, 26 Nov 2018 21:44:20 +0000 (21:44 +0000)]
compiler: initial support for exporting function bodies
Create a framework for putting function bodies in export data. At
present only empty functions will be put there, and they will be
ignored on import. Later patches will get this to the point of
supporting inlining of (some) functions defined in other packages.
jozefl [Mon, 26 Nov 2018 21:39:51 +0000 (21:39 +0000)]
2018-11-26 Jozef Lawrynowicz <jozef.l@mittosystems.com>
Fix tests for 16-bit int.
gcc/testsuite/ChangeLog:
* c-c++-common/Warray-bounds-3.c (test_strcpy_bounds): Use long instead
of int if __SIZEOF_INT__ == 2.
* c-c++-common/Wrestrict.c: Test memcpy range with smaller length when
__SIZEOF_SIZE_T < 4.
* c-c++-common/rotate-8.c: Define smaller "large" constant when
__SIZEOF_INT__ == 2.
* gcc.dg/pr53037-1.c: Add dg-require-effective-target int32.
* gcc.dg/pr53037-2.c: Likewise.
* gcc.dg/pr53037-3.c: Likewise.
* gcc.dg/pr85512.c: Likewise.
* gcc.dg/pr59963-2.c: Add dg-warning for int16.
* gcc.dg/sancov/cmp0.c: Explicitly use __INT32_TYPE__ instead of int.
* gcc.dg/tree-ssa/addadd.c: Fix dg-final directives for int16.
* gcc.dg/tree-ssa/pr79327-2.c: Likewise.
* gcc.dg/tree-ssa/builtin-sprintf-2.c: Filter out invalid tests for
int16.
* gcc.dg/tree-ssa/builtin-sprintf-warn-10.c: Update sizes in dg-warning
directives for int16.
jakub [Mon, 26 Nov 2018 21:25:22 +0000 (21:25 +0000)]
PR testsuite/88090
* obj-c++.dg/attributes/method-nonnull-1.mm (my_size_t): New typedef.
(MyArray::removeObjectAtIndex): Use my_size_t instead of size_t and
expect it in diagnostics.
jakub [Mon, 26 Nov 2018 21:24:00 +0000 (21:24 +0000)]
PR c++/86900
* dwarf2out.c (secname_for_decl): For functions with
DECL_SECTION_NAME if in_cold_section_p, try to return
current_function_section's name if it is a named section.
jakub [Mon, 26 Nov 2018 21:03:41 +0000 (21:03 +0000)]
* testsuite/Makefile.am (AUTOMAKE_OPTIONS): Drop dejagnu.
(RUNTEST): Don't define.
(RUNTESTDEFAULTFLAGS): Add.
(check-DEJAGNU, site.exp, distclean-DEJAGNU): New goals.
(distclean-am): Depend on distclean-DEJAGNU.
(check-am): If -j% option is present in MFLAGS and if
`getconf _NPROCESSORS_ONLN` is more than 8, export OMP_NUM_THREADS=8.
(.PHONY): Add check-DEJAGNU and distclean-DEJAGNU.
* testsuite/Makefile.in: Regenerated.
acsawdey [Mon, 26 Nov 2018 20:59:06 +0000 (20:59 +0000)]
2018-11-26 Aaron Sawdey <acsawdey@linux.ibm.com>
* config/rs6000/rs6000-string.c (expand_cmp_vec_sequence): Rename
and modify expand_strncmp_vec_sequence.
(emit_final_compare_vec): Rename and modify emit_final_str_compare_vec.
(generate_6432_conversion): New function.
(expand_block_compare): Add support for vsx.
(expand_block_compare_gpr): New function.
* config/rs6000/rs6000.opt (rs6000_block_compare_inline_limit): Increase
default limit to 63 because of more compact vsx code.
uros [Mon, 26 Nov 2018 20:50:57 +0000 (20:50 +0000)]
PR target/88178
* config/i386/i386.c (dbx_register_map): Use IGNORED_DWARF_REGNUM
for registers for which no debug information can be generated.
Use INVALID_REGNUM for non-existent registers.
(dbx64_register_map): Ditto.
(svr4_dbx_register_map): Ditto.
ktkachov [Mon, 26 Nov 2018 17:50:03 +0000 (17:50 +0000)]
[arm][1/3] Rename mul64 attr to widen_mul64
The mul64 attribute in types.md causes some confusion as it is used to represent
aarch32 instructions that do widening multiplication to generate 32->64 bit results.
But these types are shared with aarch64, which has native 64-bit multiplication operations.
Those are currently not properly represented, which I will fix in follow-up patches.
For now, this patch renames the mul64 attribute to widen_mul64 to more clearly communicate its meaning.
It mechanically updates all users of that name in config/arm/ (there are no users in config/aarch64).
There is thus no change in behaviour.
ibuclaw [Mon, 26 Nov 2018 17:27:34 +0000 (17:27 +0000)]
libphobos: Fix backtraces in Fibers on AArch64.
When throwing an Exception in the Fiber the backtrace generation
crashes. This happens because backtrace does not func the stack bottom.
Using '.cfi_undefined x30' tells the debug info that the value in the lr
is unknown, which seems to be the nicest way to stop the unwinder.
Setting x30 to 0 is another option, however it still creates one invalid
frame in gdb, so the .cfi variant is used here instead.
rguenth [Mon, 26 Nov 2018 15:37:35 +0000 (15:37 +0000)]
2018-11-26 Richard Biener <rguenther@suse.de>
PR tree-optimization/88182
* tree-vect-loop.c (vectorizable_reduction): Pick up single
correct reduc_def_info.
* tree-vect-slp.c (vect_analyze_slp_instance): Set
STMT_VINFO_REDUC_DEF of the first stmt.
libgomp/
* testsuite/libgomp.c++/pr88182.C: New testcase.
gcc/
* doc/extend.texi (Common Variable Attributes): Use preferred
placement of type attributes in examples, plus whitespace fixes.
(Type Attributes): Clarify why placement of attributes
immediately after struct/union/enum keyword is preferred.
(Common Type Attributes): Use preferred placement of type
attributes in examples, plus more whitespace fixes.
ro [Sun, 25 Nov 2018 15:59:44 +0000 (15:59 +0000)]
Disable gdc execution tests without libphobos
gcc/testsuite:
* lib/target-supports.exp (check_compile): Handle D.
(check_effective_target_d_runtime): New proc.
* lib/gdc-dg.exp (gdc-dg-test): Demote link and run tests to
compile unless d_runtime.
* gdc.dg/runnable.d: Skip unless d_runtime.
* gdc.dg/lto/lto.exp: Require d_runtime to run tests.
jb [Fri, 23 Nov 2018 20:42:03 +0000 (20:42 +0000)]
Make recursion_check work for multiple threads
With multiple threads, using an unprotected static variable to check
whether recursion has occured isn't valid, as one thread might have
modified the variable, thus causing another thread to incorrectly
conclude that recursion has occured. This patch avoids this problem
by using a thread-specific variable for the recursion check.
Regtested on x86_64-pc-linux-gnu.
libgfortran/ChangeLog:
2018-11-23 Janne Blomqvist <jb@gcc.gnu.org>
* runtime/error.c (MAGIC): Remove.
(recursion_key): New variable.
(recursion_check): Use thread-specific variable for recursion
check if threads are active.
(constructor_recursion_check): New function.
(destructor_recursion_check): New funcion.
msebor [Fri, 23 Nov 2018 18:45:45 +0000 (18:45 +0000)]
PR tree-optimization/87756 - missing unterminated argument warning using address of a constant character
gcc/ChangeLog:
PR tree-optimization/87756
* expr.c (string_constant): Synthesize a string literal from
the address of a constant character.
* tree.c (build_string_literal): Add an argument.
* tree.h (build_string_literal): Same.
PR testsuite/88098
* c-typeck.c (convert_arguments): Call builtin_decl_explicit instead.
(maybe_warn_builtin_no_proto_arg): Handle short enum to int promotion.
redi [Fri, 23 Nov 2018 15:48:56 +0000 (15:48 +0000)]
PR libstdc++/87308 adjust regex used in std::any pretty printer
The pretty printer for std::any fails when the contained value is a
locally-defined type, because the name in the debuginfo has
cv-qualifiers and ptr-declarators in different positions. The unexpected
format confuses the printer. This makes the printer's regex handle
either format.
This isn't a complete fix because looking up the contained type fails
when there are two types with the same name (defined in different local
scopes). This applies to all closure types defined in a given function,
as they all appear as "func()::lambda" in the debuginfo names.
PR libstdc++/87308 (partial)
* python/libstdcxx/v6/printers.py (StdExpAnyPrinter): Adjust regex to
work around PR 88166.
* testsuite/libstdc++-prettyprinters/cxx17.cc: Test std::any
containing a local type.
rguenth [Fri, 23 Nov 2018 12:53:39 +0000 (12:53 +0000)]
2018-11-23 Richard Biener <rguenther@suse.de>
PR tree-optimization/88149
* tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
the case where there are two different def types for the
same operand at different operand position in the same stmt.
ktkachov [Fri, 23 Nov 2018 10:48:52 +0000 (10:48 +0000)]
[PATCH, ARM] Clean up arm backend using the @ construct for MD patterns
This patch removes some of the machine mode checks from the arm backend when
emitting instructions by using the '@' construct (Parameterized Names[2]). It
is based on the previous AArch64 patch[1].
jakub [Fri, 23 Nov 2018 09:12:16 +0000 (09:12 +0000)]
PR tree-optimization/86614
* gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early
if TREE_NO_WARNING is set on ref.ref.
* c-c++-common/Warray-bounds-2.c (wrap_strncpy_dstarray_diff_neg,
call_strncpy_dstarray_diff_neg): Don't expect late -Warray-bounds
warnings, just early ones from FE. Remove dg-prune-output.
* c-c++-common/Warray-bounds-6.c: New test.
jakub [Fri, 23 Nov 2018 00:11:11 +0000 (00:11 +0000)]
* lang.opt (fpad-source): New option.
* scanner.c (load_line): Don't pad fixed form lines if
!flag_pad_source.
* invoke.texi (-fno-pad-source): Document.
* gfortran.dg/pad_source_1.f: New test.
* gfortran.dg/pad_source_2.f: New test.
* gfortran.dg/pad_source_3.f: New test.
* gfortran.dg/pad_source_4.f: New test.
* gfortran.dg/pad_source_5.f: New test.
hubicka [Thu, 22 Nov 2018 23:10:57 +0000 (23:10 +0000)]
PR lto/88142
* ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned
parameters; do not warn here.
(odr_subtypes_equivalent_p): Likewise.
(warn_odr): Fix typo.
(warn_types_mismatch): Do not output confused warnings on integer types.
(odr_types_equivalent_p): Update.
* g++.dg/lto/odr-5_0.C: New testcase.
* g++.dg/lto/odr-5_1.C: New testcase.
jakub [Thu, 22 Nov 2018 21:34:49 +0000 (21:34 +0000)]
* config/i386/i386.c (ix86_option_override_internal): For
stack_protector_guard related options, use opts_set->x_ instead
of global_options_set. and prefix options with opts->x_ . Move
defaults for offset and reg into else block.
vmakarov [Thu, 22 Nov 2018 17:25:57 +0000 (17:25 +0000)]
2018-11-22 Vladimir Makarov <vmakarov@redhat.com>
PR rtl-optimization/87718
* ira-costs.c: Remove trailing white-spaces.
(record_operand_costs): Add a special treatment for moves
involving a hard register.
2018-11-22 Vladimir Makarov <vmakarov@redhat.com>
PR rtl-optimization/87718
* gcc.target/i386/pr82361-1.c: Check only the first operand of
moves.
marxin [Thu, 22 Nov 2018 15:09:55 +0000 (15:09 +0000)]
Fix option values for -march.
2018-11-22 Martin Liska <mliska@suse.cz>
* common/config/i386/i386-common.c (processor_names): Add
static assert and add missing "znver2".
(ix86_get_valid_option_values): Add checking assert for null
values and add "native" value if feasible.
* config/i386/i386.h: Do not declare size of processor_names.
* common/config/i386/i386-common.c:
* config/i386/i386.c: Add static assert for size
of processor_cost_table.
thopre01 [Thu, 22 Nov 2018 14:46:17 +0000 (14:46 +0000)]
PR85434: Prevent spilling of stack protector guard's address on ARM
In case of high register pressure in PIC mode, address of the stack
protector's guard can be spilled on ARM targets as shown in PR85434,
thus allowing an attacker to control what the canary would be compared
against. ARM does lack stack_protect_set and stack_protect_test insn
patterns, defining them does not help as the address is expanded
regularly and the patterns only deal with the copy and test of the
guard with the canary.
This problem does not occur for x86 targets because the PIC access and
the test can be done in the same instruction. Aarch64 is exempt too
because PIC access insn pattern are mov of UNSPEC which prevents it from
the second access in the epilogue being CSEd in cse_local pass with the
first access in the prologue.
The approach followed here is to create new "combined" set and test
standard pattern names that take the unexpanded guard and do the set or
test. This allows the target to use an opaque pattern (eg. using UNSPEC)
to hide the individual instructions being generated to the compiler and
split the pattern into generic load, compare and branch instruction
after register allocator, therefore avoiding any spilling. This is here
implemented for the ARM targets. For targets not implementing these new
standard pattern names, the existing stack_protect_set and
stack_protect_test pattern names are used.
To be able to split PIC access after register allocation, the functions
had to be augmented to force a new PIC register load and to control
which register it loads into. This is because sharing the PIC register
between prologue and epilogue could lead to spilling due to CSE again
which an attacker could use to control what the canary gets compared
against.
2018-11-22 Thomas Preud'homme <thomas.preudhomme@linaro.org>
gcc/
PR target/85434
* target-insns.def (stack_protect_combined_set): Define new standard
pattern name.
(stack_protect_combined_test): Likewise.
* cfgexpand.c (stack_protect_prologue): Try new
stack_protect_combined_set pattern first.
* function.c (stack_protect_epilogue): Try new
stack_protect_combined_test pattern first.
* config/arm/arm.c (require_pic_register): Add pic_reg and compute_now
parameters to control which register to use as PIC register and force
reloading PIC register respectively. Insert in the stream of insns if
possible.
(legitimize_pic_address): Expose above new parameters in prototype and
adapt recursive calls accordingly. Use pic_reg if non null instead of
cached one.
(arm_load_pic_register): Add pic_reg parameter and use it if non null.
(arm_legitimize_address): Adapt to new legitimize_pic_address
prototype.
(thumb_legitimize_address): Likewise.
(arm_emit_call_insn): Adapt to require_pic_register prototype change.
(arm_expand_prologue): Adapt to arm_load_pic_register prototype change.
(thumb1_expand_prologue): Likewise.
* config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype
change.
(arm_load_pic_register): Likewise.
* config/arm/predicated.md (guard_addr_operand): New predicate.
(guard_operand): New predicate.
* config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address
prototype change.
(builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue
prototype change.
(stack_protect_combined_set): New expander..
(stack_protect_combined_set_insn): New insn_and_split pattern.
(stack_protect_set_insn): New insn pattern.
(stack_protect_combined_test): New expander.
(stack_protect_combined_test_insn): New insn_and_split pattern.
(arm_stack_protect_test_insn): New insn pattern.
* config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern.
* config/arm/unspecs.md (UNSPEC_SP_SET): New unspec.
(UNSPEC_SP_TEST): Likewise.
* doc/md.texi (stack_protect_combined_set): Document new standard
pattern name.
(stack_protect_set): Clarify that the operand for guard's address is
legal.
(stack_protect_combined_test): Document new standard pattern name.
(stack_protect_test): Clarify that the operand for guard's address is
legal.
gcc/testsuite/
PR target/85434
* gcc.target/arm/pr85434.c: New test.
The implementations of std::make_shared for -frtti and -fno-rtti are not
compatible, because they pass different arguments to
_Sp_counted_ptr_inplace::_M_get_deleter and so can't interoperate.
Either the argument doesn't match the expected value, and so the
shared_ptr::_M_ptr member is never set, or the type-punned reference is
treated as a real std::type_info object and gets dereferenced.
This patch removes the differences between -frtti and -fno-rtti, so that
typeid is never used, and the type-punned reference is used in both
cases. For backwards compatibility with existing code that passes
typeid(_Sp_make_shared_tag) that still needs to be handled, but only
after checking that the argument is not the type-punned reference (so
it's safe to treat as a real std::type_info object). The reference is
bound to an object of literal type, so that it doesn't need a guard
variable to make its initialization thread-safe.
This patch also fixes 87520 by ensuring that the type-punned reference
is bound to "a region of storage of suitable size and alignment to
contain an object of the reference's type" (as per the proposed
resolution of Core DR 453).
If all objects are built with the fixed version of GCC then -frtti and
-fno-rtti can be mixed freely and std::make_shared will work correctly.
If some objects are built with unfixed GCC versions then problems can
still arise, depending on which template instantiations are kept by the
linker.
PR libstdc++/85930
PR libstdc++/87520
* include/bits/shared_ptr_base.h (_Sp_make_shared_tag::_S_ti)
[__cpp_rtti]: Define even when RTTI is enabled. Use array of
sizeof(type_info) so that type-punned reference binds to an object
of the correct size as well as correct alignment.
(_Sp_counted_ptr_inplace::_M_get_deleter) [__cpp_rtti]: Check for
_S_ti() reference even when RTTI is enabled.
(__shared_ptr(_Sp_make_shared_tag, const _Alloc&, _Args&&...))
[__cpp_rtti]: Pass _S_ti() instead of typeid(_Sp_make_shared_tag).
jakub [Thu, 22 Nov 2018 09:48:43 +0000 (09:48 +0000)]
PR target/85644
PR target/86832
* config/i386/i386.c (ix86_option_override_internal): Default
ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
is defined.
* config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
stack_protect_test, stack_protect_test_<mode>): Use empty condition
instead of TARGET_SSP_TLS_GUARD.
jakub [Thu, 22 Nov 2018 09:26:29 +0000 (09:26 +0000)]
PR c++/87386
* parser.c (cp_parser_operator): Use str.get_value () instead of just
str in USERDEF_LITERAL_VALUE and USERDEF_LITERAL_SUFFIX_ID arguments.
jb [Thu, 22 Nov 2018 07:58:29 +0000 (07:58 +0000)]
Replace sync builtins with atomic builtins
The old __sync builtins have been deprecated for a long time now in
favor of the __atomic builtins following the C++11/C11 memory model.
This patch converts libgfortran to use the modern __atomic builtins.
At the same time I weakened the consistency to relaxed for
incrementing and decrementing the counter, and acquire-release when
decrementing to check whether the counter is 0 and the unit can be
freed. This is similar to e.g. std::shared_ptr in C++.
Regtested on x86_64-pc-linux-gnu.
libgfortran/ChangeLog:
2018-11-22 Janne Blomqvist <jb@gcc.gnu.org>
* acinclude.m4 (LIBGFOR_CHECK_ATOMIC_FETCH_ADD): Rename and test
presence of atomic builtins instead of sync builtins.
* configure.ac (LIBGFOR_CHECK_ATOMIC_FETCH_ADD): Call new test.
* io/io.h (inc_waiting_locked): Use __atomic_fetch_add.
(predec_waiting_locked): Use __atomic_add_fetch.
(dec_waiting_unlocked): Use __atomic_fetch_add.
* config.h.in: Regenerated.
* configure: Regenerated.
* Makefile.in: Regenerated.