Than McIntosh [Fri, 9 Sep 2016 18:27:42 +0000 (18:27 +0000)]
compiler: add abstraction layer for sha1 checksums.
Add new interface for the front end code to use when computing SHA1
checksums; the intent is to allow the different implementation in
different back ends.
No change in functionality for gccgo; this is an enabling change to
permit the front end to be used with other back ends (e.g. LLVM).
Joseph Myers [Fri, 9 Sep 2016 16:59:43 +0000 (17:59 +0100)]
Define TS 18661-1 type width macros in <limits.h> and <stdint.h>.
TS 18661-1 defines <limits.h> and <stdint.h> macros for widths of
standard integer types and the typedefs defined in, or whose limits
are defined in, <stdint.h>. (The connection to the main
floating-point subject matter of TS 18661-1 is that these are intended
to be used with the fromfp functions to convert from floating point to
integer types of any width in any rounding direction, though these
macros may be of more general use.)
This patch implements support for these macros in GCC's <limits.h> and
<stdint.h>. To avoid needing to implement fixincludes for system
headers where GCC wraps the system libc's <stdint.h> in hosted mode,
the test for the <stdint.h> macros uses -ffreestanding (I'll add the
macros to glibc's headers separately) - but as usual for new features
in these headers, platforms (primarily OpenBSD) that use USER_H to
avoid using GCC's headers at all will have failures until the system
headers have the feature added or appropriate fixincludes are
implemented.
The header macros are implemented using appropriate new predefined
macros, with the code avoiding defining more such macros than
necessary (so one predefined macro suffices for corresponding signed
and unsigned types, while no such predefined macros are needed for the
exact-width types such as int8_t).
Bootstrapped with no regressions on x86_64-pc-linux-gnu.
runtime: remove remaining use of MAKECONTEXT_STACK_TOP macro
The definition and most uses of MAKECONTEXT_STACK_TOP were removed in
https://golang.org/cl/88660043, which removed support for Solaris 8/9.
One use of MAKECONTEXT_STACK_TOP was accidentally left in the source
code. Remove it now.
runtime: align ucontext_t argument to 16 byte boundary
Some systems, such as ia64 and PPC, require that a ucontext_t pointer
passed to getcontext and friends be aligned to a 16-byte boundary.
Currently the ucontext_t fields in the g structure are defined in Go,
and Go has no way to ensure a 16-byte alignment for a struct field.
The fields are currently represented by an array of unsafe.Pointer.
Enforce the alignment by making the array larger, and picking an offset
into the array that is 16-byte aligned.
[Patch libgcc] Enable HCmode multiply and divide (mulhc3/divhc3)
This patch arranges for half-precision complex multiply and divide
routines to be built if __LIBGCC_HAS_HF_MODE__. This will be true
if the target supports the _Float16 type.
libgcc/
PR target/63250
* Makefile.in (lib2funcs): Build _mulhc3 and _divhc3.
* libgcc2.h (LIBGCC_HAS_HF_MODE): Conditionally define.
(HFtype): Likewise.
(HCtype): Likewise.
(__divhc3): Likewise.
(__mulhc3): Likewise.
* libgcc2.c: Support _mulhc3 and _divhc3.
Steven G. Kargl [Thu, 8 Sep 2016 22:33:10 +0000 (22:33 +0000)]
re PR fortran/69514 (ICE with nested array constructor)
2016-09-08 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/69514
* array.c (gfc_match_array_constructor): If type-spec is present,
walk the array constructor performing possible conversions for
numeric types.
2016-09-08 Steven G. Kargl <kargl@gcc.gnu.org>
Louis Krupp <lkrupp@gcc.gnu.org>
PR fortran/69514
* gfortran.dg/pr69514_1.f90: New test.
* gfortran.dg/pr69514_2.f90: New test.
Co-Authored-By: Louis Krupp <lkrupp@gcc.gnu.org>
From-SVN: r240039
Jakub Jelinek [Thu, 8 Sep 2016 12:19:21 +0000 (14:19 +0200)]
re PR fortran/77500 (ICE in gfc_trans_omp_atomic, at fortran/trans-openmp.c:2822)
PR fortran/77500
* trans-openmp.c (gfc_trans_omp_atomic): For atomic write or
swap, don't try to look through GFC_ISYM_CONVERSION. In other cases,
check that value.function.isym is non-NULL before dereferencing it.
libgcc complex multiply is meant to eliminate excess
precision from certain internal values by forcing them to memory in
exactly those cases where the type has excess precision. But in
https://gcc.gnu.org/ml/gcc-patches/2014-09/msg01894.html I
accidentally inverted the logic so that values get forced to memory in
exactly the cases where it's not needed. (This is a pessimization in
the no-excess-precision case, in principle could lead to bad results
depending on code generation in the excess-precision case. Note: I do
not have a test demonstrating bad results.)
Bootstrapped with no regressions on x86_64-pc-linux-gnu. Code size
went down on x86_64 as expected; old sizes:
text data bss dec hex filename
887 0 0 887 377 _muldc3.o
810 0 0 810 32a _mulsc3.o
2032 0 0 2032 7f0 _multc3.o
983 0 0 983 3d7 _mulxc3.o
David Malcolm [Wed, 7 Sep 2016 16:56:23 +0000 (16:56 +0000)]
Move class substring_loc from c-family into gcc
gcc/ChangeLog:
* Makefile.in (OBJS): Add substring-locations.o.
* langhooks-def.h (class substring_loc): New forward decl.
(lhd_get_substring_location): New decl.
(LANG_HOOKS_GET_SUBSTRING_LOCATION): New macro.
(LANG_HOOKS_INITIALIZER): Add LANG_HOOKS_GET_SUBSTRING_LOCATION.
* langhooks.c (lhd_get_substring_location): New function.
* langhooks.h (class substring_loc): New forward decl.
(struct lang_hooks): Add field get_substring_location.
* substring-locations.c: New file, taking definition of
format_warning_va and format_warning_at_substring from
c-family/c-format.c, making them non-static.
* substring-locations.h (class substring_loc): Move class here
from c-family/c-common.h. Add and rewrite comments.
(format_warning_va): New decl.
(format_warning_at_substring): New decl.
(get_source_location_for_substring): Add comment.
gcc/c-family/ChangeLog:
* c-common.c (get_cpp_ttype_from_string_type): Handle being passed
a POINTER_TYPE.
(substring_loc::get_location): Move to substring-locations.c,
keeping implementation as...
(c_get_substring_location): New function, from the above, reworked
to use accessors rather than member lookup.
* c-common.h (class substring_loc): Move to substring-locations.h,
replacing with a forward decl.
(c_get_substring_location): New decl.
* c-format.c: Include "substring-locations.h".
(format_warning_va): Move to substring-locations.c.
(format_warning_at_substring): Likewise.
gcc/c/ChangeLog:
* c-lang.c (LANG_HOOKS_GET_SUBSTRING_LOCATION): Use
c_get_substring_location for this new langhook.
gcc/testsuite/ChangeLog:
* gcc.dg/plugin/diagnostic_plugin_test_string_literals.c: Include
"substring-locations.h".
Improve aarch64_legitimize_address - avoid splitting the offset if it is supported.
Improve aarch64_legitimize_address - avoid splitting the offset if it is
supported. When we do split, take the mode size into account. BLKmode
falls into the unaligned case but should be treated like LDP/STP.
This improves codesize slightly due to fewer base address calculations:
Jerry DeLisle [Tue, 6 Sep 2016 23:22:26 +0000 (23:22 +0000)]
re PR libfortran/77393 (Revision r237735 changed the behavior of F0.0)
2016-09-06 Jerry DeLisle <jvdelisle@gcc.gnu.org>
PR libgfortran/77393
* io/write_float.def (build_float_string): Recognize when the
result will not fit in the user provided, star fill, and exit
early.
* gfortran.dg/fmt_f0_2.f90: Update test.
* gfortran.dg/fmt_f0_3.f90: New test.
PR c/77336
* c-format.c (check_function_format): Avoid issuing warnings for
functions unless they call format functions with non-constant
format strings.
The avr backend supports __builtin_return_address only for the current
stack frame. The test expects it to work for caller frames as well and
therefore fails for avr.
Add avr to the list of targets already excluded with dg-skip-if.
Jakub Jelinek [Mon, 5 Sep 2016 19:45:21 +0000 (21:45 +0200)]
re PR target/77476 ([AVX-512] illegal kmovb instruction on KNL)
PR target/77476
* config/i386/i386.md (isa): Add x64_avx512bw.
(*zero_extendsidi2): For alternative 11 use x64_avx512bw isa.
(kmov_isa): New mode attr.
(zero_extend<mode>di2): Use <kmov_isa> isa for the last alternative.
(*zero_extend<mode>si2): Likewise.
(*zero_extendqihi2): Use avx512dq isa for the last alternative.
* gcc.target/i386/avx512f-pr77476.c: New test.
* gcc.target/i386/avx512bw-pr77476.c: New test.
* gcc.target/i386/avx512dq-pr77476.c: New test.
re PR target/77452 (ICE: in plus_constant, at explow.c:87 with -fno-split-wide-types -mavx512f --param=max-combine-insns=2)
PR rtl-optimization/77452
* explow.c (plus_constant) <case MEM>: Extract scalar constant from
inner-mode reference to a CONST_VECTOR constant in the constant pool.
testsuite/ChangeLog:
PR rtl-optimization/77452
* gcc.target/i386/pr77452.c: New test.
* c-common.c (bool_promoted_to_int_p): New function.
(expr_has_boolean_operands_p): New function.
(warn_logical_not_parentheses): Return if expr_has_boolean_operands_p.
(maybe_warn_bool_compare): Use bool_promoted_to_int_p.
* c-c++-common/Wlogical-not-parentheses-3.c: New test.
ffi.c (ffi_struct_type): Put type declaration on separate line.
* src/pa/ffi.c (ffi_struct_type): Put type declaration on separate line.
(ffi_prep_args_pa32): Likewise.
(ffi_size_stack_pa32): Likewise.
(ffi_prep_cif_machdep): Likewise.
(ffi_call): Likewise. Rename to ffi_call_int. Add closure argument
and update call to ffi_call_pa32.
(ffi_call, ffi_call_go, ffi_prep_go_closure): New.
(ffi_closure_inner_pa32): Update to handle go closures.
* src/pa/ffitarget.h (FFI_GO_CLOSURES): Define.
* src/pa/hpux32.S (ffi_call_pa32): Pass go closure argument in static
chain register (%ret1).
(ffi_closure_pa32): Set closure type argument to zero.
(ffi_go_closure_pa32): New function. Add unwind data for it.
* src/pa/linux.S: Likewise. Use cfi directives for unwind data.
* testsuite/libffi.go/static-chain.h (STATIC_CHAIN_REG): Define for
hppa.
David Malcolm [Fri, 2 Sep 2016 19:41:17 +0000 (19:41 +0000)]
Add -fdiagnostics-generate-patch
gcc/ChangeLog:
* common.opt (fdiagnostics-generate-patch): New option.
* diagnostic.c: Include "edit-context.h".
(diagnostic_initialize): Initialize context->edit_context_ptr.
(diagnostic_finish): Delete context->edit_context_ptr.
(diagnostic_report_diagnostic): Add fix-it hints from the
diagnostic to context->edit_context_ptr, if any.
* diagnostic.h (class edit_context): Add forward decl.
(struct diagnostic_context): Add field "edit_context_ptr".
* doc/invoke.texi (Diagnostic Message Formatting Options): Add
-fdiagnostics-generate-patch.
(-fdiagnostics-generate-patch): New item.
* toplev.c: Include "edit-context.h".
(process_options): Set global_dc->edit_context_ptr to a new
edit_context if the options need one.
(toplev::main): Handle -fdiagnostics-generate-patch by using
global_dc->edit_context_ptr.
gcc/testsuite/ChangeLog:
* gcc.dg/plugin/diagnostic-test-show-locus-generate-patch.c: New
test case.
* gcc.dg/plugin/plugin.exp (plugin_test_list): Add
diagnostic-test-show-locus-generate-patch.c to the sources
for diagnostic_plugin_test_show_locus.c.
David Malcolm [Fri, 2 Sep 2016 18:00:57 +0000 (18:00 +0000)]
Introduce class edit_context
gcc/ChangeLog:
* Makefile.in (OBJS-libcommon): Add edit-context.o.
* diagnostic-color.c (color_dict): Add "diff-filename",
"diff-hunk", "diff-delete", and "diff-insert".
(parse_gcc_colors): Update default value of GCC_COLORS in comment
to reflect above changes.
* doc/invoke.texi (-fdiagnostics-color): Update description of
default GCC_COLORS, and of the supported capabilities.
* edit-context.c: New file.
* edit-context.h: New file.
* input.c (struct fcache): Add field "missing_trailing_newline".
(diagnostics_file_cache_forcibly_evict_file): Initialize it to
true.
(add_file_to_cache_tab): Likewise.
(fcache::fcache): Likewise.
(get_next_line): Update c->missing_trailing_newline.
(location_missing_trailing_newline): New function.
* input.h (location_missing_trailing_newline): New decl.
* selftest-run-tests.c (selftest::run_tests): Call
edit_context_c_tests.
* selftest.h (edit_context_c_tests): New decl.
libcpp/ChangeLog:
* include/line-map.h (rich_location::seen_impossible_fixit_p): New
accessor.
Jakub Jelinek [Fri, 2 Sep 2016 17:11:42 +0000 (19:11 +0200)]
re PR sanitizer/77396 (address sanitizer crashes if all static global variables are optimized)
PR sanitizer/77396
* sanopt.c: Include gimple-ssa.h, tree-phinodes.h and ssa-iterators.h.
(sanopt_optimize_walker): Optimize away
__asan_before_dynamic_init (...) followed by
__asan_after_dynamic_init () without intervening memory loads/stores.
* ipa-pure-const.c (special_builtin_state): Handle
BUILT_IN_ASAN_BEFORE_DYNAMIC_INIT and
BUILT_IN_ASAN_AFTER_DYNAMIC_INIT.
* decl2.c (do_static_initialization_or_destruction): Only
call asan_dynamic_init_call if INITP is true.
David Malcolm [Fri, 2 Sep 2016 15:22:21 +0000 (15:22 +0000)]
Improvements to typed_splay_tree
This patch adds foreach, max and min methods to
class typed_splay_tree, along with the start of a selftest
suite.
gcc/ChangeLog:
* Makefile.in (OBJS): Add typed-splay-tree.o.
* selftest-run-tests.c (selftest::run_tests): Call
typed_splay_tree_c_tests.
* selftest.h (typed_splay_tree_c_tests): New decl.
* typed-splay-tree.c: New file.
* typed-splay-tree.h (typed_splay_tree::foreach_fn): New typedef.
(typed_splay_tree::max): New method.
(typed_splay_tree::min): New method.
(typed_splay_tree::foreach): New method.
(typed_splay_tree::closure): New struct.
(typed_splay_tree::inner_foreach_fn): New function.
Martin Sebor [Thu, 1 Sep 2016 22:47:49 +0000 (22:47 +0000)]
c-ada-spec.c (dump_ada_function_declaration): Increase buffer size to guarantee it fits the output of the formatted function...
gcc/c-family/ChangeLog:
* c-ada-spec.c (dump_ada_function_declaration): Increase buffer
size to guarantee it fits the output of the formatted function
regardless of its arguments.
gcc/cp/ChangeLog:
* mangle.c: Increase buffer size to guarantee it fits the output
of the formatted function regardless of its arguments.
gcc/go/ChangeLog:
* gofrontend/expressions.cc: Increase buffer size to guarantee
it fits the output of the formatted function regardless of its
arguments.
gcc/java/ChangeLog:
* decl.c (give_name_to_locals): Increase buffer size to guarantee
it fits the output of the formatted function regardless of its
arguments.
* mangle_name.c (append_unicode_mangled_name): Same.
gcc/ChangeLog:
* genmatch.c (parser::parse_expr): Increase buffer size to guarantee
it fits the output of the formatted function regardless of its
arguments.
* gcc/genmodes.c (parser::parse_expr): Same.
* gimplify.c (gimplify_asm_expr): Same.
* passes.c (pass_manager::register_one_dump_file): Same.
* print-tree.c (print_node): Same.
rs6000: Use LR_REGNO directly in the save/restore patterns
Various patterns use "register_operand" "l" (or "=l") although those
patterns are only created refering to LR_REGNO directly. This patch
changes those patterns to use the hard regs as well.
* config/rs6000/rs6000.md (*restore_gpregs_<mode>_r11,
*restore_gpregs_<mode>_r12, *restore_gpregs_<mode>_r1,
*return_and_restore_gpregs_<mode>_r11,
*return_and_restore_gpregs_<mode>_r12,
*return_and_restore_gpregs_<mode>_r1,
*return_and_restore_fpregs_<mode>_r11,
*return_and_restore_fpregs_<mode>_r12,
*return_and_restore_fpregs_<mode>_r1): Use the hard register LR_REGNO
directly instead of via the "l" constraint. Renumber operands.
Fix whitespace.
Many places still use 65 instead of the symbolic constant LR_REGNO. This
fixes them all (I looked for the string "65" only, in config/rs6000/ only,
I didn't read all code :-) )
I left it in *restore_world because Iain will remove it there soon.
This should fix r239866 for AIX. I missed two patterns that refer to LR
as "register_operand" "l" instead of as reg:P LR_REGNO.
* config/rs6000/rs6000.md
(define_insn "*return_and_restore_fpregs_aix_<mode>_r11"): Delete
the use of the link register.
(define_insn "*return_and_restore_fpregs_aix_<mode>_r1"): Ditto.
Bill Schmidt [Thu, 1 Sep 2016 14:43:55 +0000 (14:43 +0000)]
re PR target/72827 (gnat bootstrap broken on powerpc64le-linux-gnu)
2016-09-01 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
Michael Meissner <meissner@linux.vnet.ibm.com>
PR target/72827
* config/rs6000/rs6000.c (rs6000_legitimize_address): Avoid
reg+reg addressing for TImode.
(rs6000_legitimate_address_p): Only allow register indirect
addressing for TImode, even without TARGET_QUAD_MEMORY.
Co-Authored-By: Michael Meissner <meissner@linux.vnet.ibm.com>
From-SVN: r239938
This patch adds legitimize_address_displacement hook so that stack accesses...
This patch adds legitimize_address_displacement hook so that stack accesses
with large offsets are split into a more efficient sequence. Unaligned and
TI/TFmode use a 256-byte range, byte and halfword accesses use a 4KB range,
wider accesses use a 16KB range to maximise the available addressing range
and increase opportunities to share the base address.
* config/aarch64/aarch64.md (*ands<mode>_compare0): New pattern.
* config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_NZmode
for comparisons of integer ZERO_EXTEND against zero.
David Malcolm [Wed, 31 Aug 2016 18:54:55 +0000 (18:54 +0000)]
diagnostic-show-locus.c: handle fixits on lines outside the regular ranges
The diagnostic_show_locus implementation determines the set
of line spans that need printing based on the ranges within the
rich_location (in layout::calculate_line_spans).
Currently this doesn't take into account fix-it hints, and hence
we fail to print fix-it hints that are on lines outside of
those ranges.
This patch updates the implementation to take fix-it hints into
account when calculating the pertinent line spans, so that such fix-it
hints do get printed. It also adds some validation, to ensure that
we don't attempt to print fix-its hints affecting a different source
file.
gcc/ChangeLog:
* diagnostic-show-locus.c (class layout): Add field m_fixit_hints.
(layout_range::intersects_line_p): New method.
(test_range_contains_point_for_single_point): Rename to...
(test_layout_range_for_single_point): ...this, and add testing
for layout_range::intersects_line_p.
(test_range_contains_point_for_single_line): Rename to...
(test_layout_range_for_single_line): ...this, and add testing
for layout_range::intersects_line_p.
(test_range_contains_point_for_multiple_lines): Rename to...
(test_layout_range_for_multiple_lines): ...this, and add testing
for layout_range::intersects_line_p.
(layout::layout): Populate m_fixit_hints.
(layout::get_expanded_location): Handle the case of a line-span
for a fix-it hint.
(layout::validate_fixit_hint_p): New method.
(get_line_span_for_fixit_hint): New function.
(layout::calculate_line_spans): Add spans for fixit-hints.
(layout::should_print_annotation_line_p): New method.
(layout::print_any_fixits): Drop param "richloc", instead using
validated fixits in m_fixit_hints. Add "const" to hint pointers.
(diagnostic_show_locus): Avoid printing blank annotation lines.
(selftest::test_diagnostic_context::test_diagnostic_context):
Initialize show_column and start_span.
(selftest::test_diagnostic_context::start_span_cb): New static
function.
(selftest::test_diagnostic_show_locus_fixit_lines): New function.
(selftest::diagnostic_show_locus_c_tests): Update for function
renamings. Call test_diagnostic_show_locus_fixit_lines.
Jakub Jelinek [Wed, 31 Aug 2016 18:42:08 +0000 (20:42 +0200)]
re PR fortran/77374 (ICE in resolve_omp_atomic, at fortran/openmp.c:3949)
PR fortran/77374
* parse.c (parse_omp_oacc_atomic): Copy over cp->ext.omp_atomic
to cp->block->ext.omp_atomic.
* resolve.c (gfc_resolve_blocks): Assert block with one or two
EXEC_ASSIGNs for EXEC_*_ATOMIC.
* openmp.c (resolve_omp_atomic): Don't assert one or two
EXEC_ASSIGNs, instead return quietly for EXEC_NOPs and otherwise
error unexpected statements.