The version of nvprof in CUDA 9.0 causes a hang when used to profile an
OpenACC program. This is because it calls acc_get_device_type from
a callback called during device initialization, which then attempts
to acquire acc_device_lock while it is already taken, resulting in
deadlock. This works around the issue by returning acc_device_none
from acc_get_device_type without attempting to acquire the lock when
initialization has not completed yet.
2020-07-14 Tom de Vries <tom@codesourcery.com>
Cesar Philippidis <cesar@codesourcery.com>
Thomas Schwinge <thomas@codesourcery.com>
Kwok Cheung Yeung <kcy@codesourcery.com>
libgomp/
* oacc-init.c (acc_init_state_lock, acc_init_state, acc_init_thread):
New variable.
(acc_init_1): Set acc_init_thread to pthread_self (). Set
acc_init_state to initializing at the start, and to initialized at the
end.
(self_initializing_p): New function.
(acc_get_device_type): Return acc_device_none if called by thread that
is currently executing acc_init_1.
* libgomp.texi (acc_get_device_type): Update documentation.
(Implementation Status and Implementation-Defined Behavior): Likewise.
* testsuite/libgomp.oacc-c-c++-common/acc_prof-init-2.c: New.
libiberty/ChangeLog:
PR demangler/96143
* cp-demangle.c (d_lambda): Don't add substitution candidate.
* testsuite/demangle-expected: Update a few existing test cases
accordingly, and add a new test case.
Supports conversion of tabs to spaces when outputting diagnostics. Also
adds -fdiagnostics-column-unit and -fdiagnostics-column-origin options to
control how the column number is output, thereby resolving the two PRs.
gcc/c-family/ChangeLog:
PR other/86904
* c-indentation.c (should_warn_for_misleading_indentation): Get
global tabstop from the new source.
* c-opts.c (c_common_handle_option): Remove handling of -ftabstop, which
is now a common option.
* c.opt: Likewise.
gcc/ChangeLog:
PR preprocessor/49973
PR other/86904
* common.opt: Handle -ftabstop here instead of in c-family
options. Add -fdiagnostics-column-unit= and
-fdiagnostics-column-origin= options.
* opts.c (common_handle_option): Handle the new options.
* diagnostic-format-json.cc (json_from_expanded_location): Add
diagnostic_context argument. Use it to convert column numbers as per
the new options.
(json_from_location_range): Likewise.
(json_from_fixit_hint): Likewise.
(json_end_diagnostic): Pass the new context argument to helper
functions above. Add "column-origin" field to the output.
(test_unknown_location): Add the new context argument to calls to
helper functions.
(test_bad_endpoints): Likewise.
* diagnostic-show-locus.c
(exploc_with_display_col::exploc_with_display_col): Support
tabstop parameter.
(layout_point::layout_point): Make use of class
exploc_with_display_col.
(layout_range::layout_range): Likewise.
(struct line_bounds): Clarify that the units are now always
display columns. Rename members accordingly. Add constructor.
(layout::print_source_line): Add support for tab expansion.
(make_range): Adapt to class layout_range changes.
(layout::maybe_add_location_range): Likewise.
(layout::layout): Adapt to class exploc_with_display_col changes.
(layout::calculate_x_offset_display): Support tabstop parameter.
(layout::print_annotation_line): Adapt to struct line_bounds changes.
(layout::print_line): Likewise.
(line_label::line_label): Add diagnostic_context argument.
(get_affected_range): Likewise.
(get_printed_columns): Likewise.
(layout::print_any_labels): Adapt to struct line_label changes.
(class correction): Add m_tabstop member.
(correction::correction): Add tabstop argument.
(correction::compute_display_cols): Use m_tabstop.
(class line_corrections): Add m_context member.
(line_corrections::line_corrections): Add diagnostic_context argument.
(line_corrections::add_hint): Use m_context to handle tabstops.
(layout::print_trailing_fixits): Adapt to class line_corrections
changes.
(test_layout_x_offset_display_utf8): Support tabstop parameter.
(test_layout_x_offset_display_tab): New selftest.
(test_one_liner_colorized_utf8): Likewise.
(test_tab_expansion): Likewise.
(test_diagnostic_show_locus_one_liner_utf8): Call the new tests.
(diagnostic_show_locus_c_tests): Likewise.
(test_overlapped_fixit_printing): Adapt to helper class and
function changes.
(test_overlapped_fixit_printing_utf8): Likewise.
(test_overlapped_fixit_printing_2): Likewise.
* diagnostic.h (enum diagnostics_column_unit): New enum.
(struct diagnostic_context): Add members for the new options.
(diagnostic_converted_column): Declare.
(json_from_expanded_location): Add new context argument.
* diagnostic.c (diagnostic_initialize): Initialize new members.
(diagnostic_converted_column): New function.
(maybe_line_and_column): Be willing to output a column of 0.
(diagnostic_get_location_text): Convert column number as per the new
options.
(diagnostic_report_current_module): Likewise.
(assert_location_text): Add origin and column_unit arguments for
testing the new functionality.
(test_diagnostic_get_location_text): Test the new functionality.
* doc/invoke.texi: Document the new options and behavior.
* input.h (location_compute_display_column): Add tabstop argument.
* input.c (location_compute_display_column): Likewise.
(test_cpp_utf8): Add selftests for tab expansion.
* tree-diagnostic-path.cc (default_tree_make_json_for_path): Pass the
new context argument to json_from_expanded_location().
libcpp/ChangeLog:
PR preprocessor/49973
PR other/86904
* include/cpplib.h (struct cpp_options): Removed support for -ftabstop,
which is now handled by diagnostic_context.
(class cpp_display_width_computation): New class.
(cpp_byte_column_to_display_column): Add optional tabstop argument.
(cpp_display_width): Likewise.
(cpp_display_column_to_byte_column): Likewise.
* charset.c
(cpp_display_width_computation::cpp_display_width_computation): New
function.
(cpp_display_width_computation::advance_display_cols): Likewise.
(compute_next_display_width): Removed and implemented this
functionality in a new function...
(cpp_display_width_computation::process_next_codepoint): ...here.
(cpp_byte_column_to_display_column): Added tabstop argument.
Reimplemented in terms of class cpp_display_width_computation.
(cpp_display_column_to_byte_column): Likewise.
* init.c (cpp_create_reader): Remove handling of -ftabstop, which is now
handled by diagnostic_context.
gcc/testsuite/ChangeLog:
PR preprocessor/49973
PR other/86904
* c-c++-common/Wmisleading-indentation-3.c: Adjust expected output
for new defaults.
* c-c++-common/Wmisleading-indentation.c: Likewise.
* c-c++-common/diagnostic-format-json-1.c: Likewise.
* c-c++-common/diagnostic-format-json-2.c: Likewise.
* c-c++-common/diagnostic-format-json-3.c: Likewise.
* c-c++-common/diagnostic-format-json-4.c: Likewise.
* c-c++-common/diagnostic-format-json-5.c: Likewise.
* c-c++-common/missing-close-symbol.c: Likewise.
* g++.dg/diagnostic/bad-binary-ops.C: Likewise.
* g++.dg/parse/error4.C: Likewise.
* g++.old-deja/g++.brendan/crash11.C: Likewise.
* g++.old-deja/g++.pt/overload2.C: Likewise.
* g++.old-deja/g++.robertl/eb109.C: Likewise.
* gcc.dg/analyzer/malloc-paths-9.c: Likewise.
* gcc.dg/bad-binary-ops.c: Likewise.
* gcc.dg/format/branch-1.c: Likewise.
* gcc.dg/format/pr79210.c: Likewise.
* gcc.dg/plugin/diagnostic-test-expressions-1.c: Likewise.
* gcc.dg/plugin/diagnostic-test-string-literals-1.c: Likewise.
* gcc.dg/redecl-4.c: Likewise.
* gfortran.dg/diagnostic-format-json-1.F90: Likewise.
* gfortran.dg/diagnostic-format-json-2.F90: Likewise.
* gfortran.dg/diagnostic-format-json-3.F90: Likewise.
* go.dg/arrayclear.go: Add a comment explaining why adding a
comment was necessary to work around a dejagnu bug.
* c-c++-common/diagnostic-units-1.c: New test.
* c-c++-common/diagnostic-units-2.c: New test.
* c-c++-common/diagnostic-units-3.c: New test.
* c-c++-common/diagnostic-units-4.c: New test.
* c-c++-common/diagnostic-units-5.c: New test.
* c-c++-common/diagnostic-units-6.c: New test.
* c-c++-common/diagnostic-units-7.c: New test.
* c-c++-common/diagnostic-units-8.c: New test.
Storing CLASSTYPE_AS_BASE in a local variable makes some code clearer
(and textually no longer). For some reason we store a DECL in a variable
called 'value', which is confusing.
gcc/cp/
* class.c (build_base_field_1): Cache CLASSTYPE_AS_BASE.
(build_self_reference): Rename value -> decl.
(dump_class_hierarchy_1): Cache CLASSTYPE_AS_BASE.
The FAT libraries config fragments need to know which library is native
and which is a multilib to choose the correct multilib from which to
append the additional object file or shared object file. Testing the
top-level archive is fragile because it will fail if rebuilding. This
patch tests the compiler preprocessing macros for the 64 bit AIX specific
__64BIT__ to determine the native mode of the compiler in MULTILIBTOP.
2020-07-14 David Edelsohn <dje.gcc@gmail.com>
libatomic/ChangeLog
* config/t-aix: Set BITS from compiler cpp macro.
libgcc/ChangeLog
* config/rs6000/t-slibgcc-aix: Set BITS from compiler cpp macro.
libgfortran/ChangeLog
* config/t-aix: Set BITS from compiler cpp macro.
libgomp/ChangeLog
* config/t-aix: Set BITS from compiler cpp macro.
libstdc++-v3/ChangeLog
* config/os/aix/t-aix: Set BITS from compiler cpp macro.
The current testsuite regex looks for THUNK0. AIX generates THUNK..0.
This patch expands the regex to allow 0 or more dots between THUNK
and the 0.
gcc/testsuite/ChangeLog
2020-07-14 David Edelsohn <dje.gcc@gmail.com>
* g++.dg/ipa/pr83667.C: Allow 0 or more dots between THUNK and 0.
> > The store to the whole of each volatile object was picked apart
> > like there had been an individual assignment to each of the
> > fields. Reads were added as part of that; see PR for details.
> > The reads from volatile memory were a clear bug; individual
> > stores questionable. A separate patch clarifies the docs.
This breaks building of mesa on both the trunk and 10 branch.
The problem is that the middle-end may never create temporaries of non-POD
(TREE_ADDRESSABLE) types, those can be only created when the language says
so and thus only the FE is allowed to create those.
This patch just reverts the behavior to what we used to do before for the
stores to volatile non-PODs. Perhaps we want to do something else, but
definitely we can't create temporaries of the non-POD type. It is up to
discussions on what should happen in those cases.
2020-07-14 Jakub Jelinek <jakub@redhat.com>
PR middle-end/96194
* expr.c (expand_constructor): Don't create temporary for store to
volatile MEM if exp has an addressable type.
* g++.dg/opt/pr96194.C: New test.
This is an ICE-on-invalid but I've been seeing it when reducing
various testcases, so it's more important for me than usually.
splice_late_return_type now checks that if we've seen a late return
type, the function return type was auto. That's a fair assumption
but grokdeclarator/cdk_function wasn't giving errors for function
pointers and similar. So we want to perform various checks not only
when funcdecl_p || inner_declarator == NULL. But only give the
!late_return_type errors when funcdecl_p, to accept e.g.
auto (*fp)() = f;
in C++11. Here's a diff -w to ease the review:
--- a/gcc/cp/decl.c
+++ b/gcc/cp/decl.c
@@ -12102,14 +12102,9 @@ grokdeclarator (const cp_declarator *declarator,
/* Handle a late-specified return type. */
tree late_return_type = declarator->u.function.late_return_type;
- if (funcdecl_p
- /* This is the case e.g. for
- using T = auto () -> int. */
- || inner_declarator == NULL)
- {
if (tree auto_node = type_uses_auto (type))
{
- if (!late_return_type)
+ if (!late_return_type && funcdecl_p)
{
if (current_class_type
&& LAMBDA_TYPE_P (current_class_type))
@@ -12201,7 +12196,6 @@ grokdeclarator (const cp_declarator *declarator,
"type specifier", name);
return error_mark_node;
}
- }
type = splice_late_return_type (type, late_return_type);
if (type == error_mark_node)
return error_mark_node;
gcc/cp/ChangeLog:
PR c++/95820
* decl.c (grokdeclarator) <case cdk_function>: Check also
pointers/references/... to functions.
gcc/testsuite/ChangeLog:
PR c++/95820
* g++.dg/cpp1y/auto-fn58.C: New test.
One of hash_map's getters returns a pointer to T or null, The comment
said it returned T or null. Also an unaligned pair of comments.
gcc/
* hash-map.h (hash_map::get): Note it is a pointer to value.
* incpath.h (incpath_kind): Align comments.
I found some bad formatting and misleading or incomplete comments
during my spelunking around the c++FE. May as well clean up trunk and
record what I noted.
gcc/cp/
* cp-tree.h: Correct some tree lang flag comments,
reformat some structure definitions. Note some structure
sizes. Clarify some comments.
(yyungetc): Delete. Not been a thing for some time.
* class.c (copy_fndecl_with_name): Comment.
(check_bases_and_members): Unnecessary {}.
(layout_class_type): Comment.
* cp-tree.def (UNBOUND_CLASS_TEMPLATE): Adjust comment.
* decl.c: Fix some formatting & whitespace issues.
(function_requirements_equivalent_p): Note why
substitutions are needed.
* decl2.c (no_linkage_error): Note that heroics about
'typedef struct { ... };' are no longer needed.
* method.c: Whitespace.
* name-lookup.c: Whitespace.
(add_decl_to_level): Reformat a line.
(print_binding_stack): Mark as DEBUG_FUNCTION.
(has_using_namespace_std_directive_p): Delete comment.
* pt.c: Whitespace
* ptree.c: Whitespace.
* rtti.c: Whitespace & comment.
* tree.c: Comment.
* typeck.c (structural_comptypes): Add comment.
As we've moved to 64-bit systems, the padding information has become
conditionally inaccurate. I also hit cases where invalid tree codes
did not get flagged as invalid.
gcc/
* tree-core.h (tree_decl_with_vis, tree_function_decl):
Note additional padding on 64-bits
* tree.c (cache_integer_cst): Note why no caching of enum literals.
(get_tree_code_name): Robustify error case.
gty calls gt_clear*e*_cache not gt_clear_cache. I know not why it is
named so, but at least document it correctly. invoke.texi had a duplicate opindex.
gcc/
* doc/gty.texi: Fic gt_cleare_cache name.
* doc/invoke.texi: Remove duplicate opindex Wabi-tag.
Output an error if the right hand value is a zero sized array or
does not have a symbol tree otherwise continue checking.
2020-07-14 Steven G. Kargl <kargl@gcc.gnu.org>
gcc/fortran/
PR fortran/95612
* expr.c (gfc_check_pointer_assigb): Output an error if
rvalue is a zero sized array or output an error if rvalue
doesn't have a symbol tree.
2020-07-14 Mark Eggleston <markeggleston@gcc.gnu.org>
gcc/testsuite/
PR fortran/95612
* gfortran.dg/pr95612.f90: New test.
gcc/fortran/ChangeLog:
PR fortran/67311
* trans-openmp.c (gfc_has_alloc_comps): Return false also for
pointers to arrays.
libgomp/ChangeLog:
PR fortran/67311
* testsuite/libgomp.fortran/target-map-1.f90: New test.
In loops like:
#pragma omp parallel for collapse(2)
for (i = -4; i < 8; i++)
for (j = 3 * i; j > 2 * i; j--)
for some outer loop iterations there are no inner loop iterations at all,
the condition is false. In order to use Summæ Potestate to count number
of iterations or to transform the logical iteration number to actual
iterator values using quadratic non-equation root discovery the outer
iterator range needs to be adjusted, such that the inner loop has at least
one iteration for each of the outer loop iterator value in the reduced
range. Sometimes this adjustment is done at the start of the range,
at other times at the end.
This patch implements it during the compile time number of loop computation
(if all expressions are compile time constants).
2020-07-14 Jakub Jelinek <jakub@redhat.com>
* omp-general.h (struct omp_for_data): Add adjn1 member.
* omp-general.c (omp_extract_for_data): For non-rect loop, punt on
count computing if n1, n2 or step are not INTEGER_CST earlier.
Narrow the outer iterator range if needed so that non-rect loop
has at least one iteration for each outer range iteration. Compute
adjn1.
* omp-expand.c (expand_omp_for_init_vars): Use adjn1 if non-NULL
instead of the outer loop's n1.
* testsuite/libgomp.c/loop-21.c: New test.
If a paramter to declared and initialised before its type is
declared a bogus error is output at the type declaration
idicating that initialisation is missing.
2020-07-14 Steven G. Kargl <kargl@gcc.gnu.org>
gcc/fortran/
PR fortran/96038
* decl.c (add_init_expr_sym): For a symbol that is a
parameter accept an initialisation if it does not have a
value otherwise output a error and reject.
2020-07-14 Mark Eggleston <markeggleston@gcc.gnu.org>
gcc/testsuite/
PR fortran/96038
* gfortran.dg/pr96038.f90: New test.
This patch consolidates four insn patterns into two.
* config/rs6000/rs6000.md (sibcall_local): Merge sibcall_local32
and sibcall_local64.
(sibcall_value_local): Similarly.
Because the check for power10_hw is not called
check_effective_target_power10_hw, it needs to be looked
for by is-effective-target-keyword. Also reorder things
in is-effective-target to put power10_hw with the other
ppc stuff.
gcc/testsuite/
* lib/target-supports.exp (is-effective-target):
Reorder to put powerpc stuff together.
(is-effective-target-keyword): Add power10_hw.
Sigh, last week's success at not breaking things failed with an
incorrect 'fix' this morning. Let's reduce my confusion by making
lib/scanlang.exp the same on trunk as modules.
gcc/testsuite/
* lib/scanlang.exp (scan-lang-dump): Fix breakage.
(scan-lang-dump-times, scan-lang-dump-not): New.
I discovered we were attempting to delete some no-longer generated
files.
gcc/
* Makefile.in (distclean): Remove long gone cxxmain.c
gcc/cp/
* Make-lang.in (c++.disclean): Likewise.
Add a test for dejagnu to determine if execution of MMA instructions is
supported in the test environment. Add an execution test to make sure
that __builtin_cpu_supports("mma") is true if we can execute MMA
instructions.
gcc/testsuite/
* lib/target-supports.exp (check_ppc_mma_hw_available):
New function.
(is-effective-target): Add ppc_mma_hw.
(is-effective-target-keyword): Add ppc_mma_hw.
* gcc.target/powerpc/mma-supported.c: New file.
* gcc.target/powerpc/mma-single-test.c: Require ppc_mma_hw.
* gcc.target/powerpc/mma-double-test.c: Require ppc_mma_hw.
cmpstrnsi expander may pass the actual string length directly to cmpstrnqi
patterns. For cmpstrnsi, one of the strings must be a constant and
expand_builtin_strncmp rewrites the length argument to be the minimum of
the const string length and the actual string length. But it is not the
case for cmpmemsi. Pass a copy of the string length to cmpstrnqi patterns
to avoid changing the actual string length by cmpstrnqi patterns.
gcc/
PR target/95443
* config/i386/i386.md (cmpstrnsi): Pass a copy of the string
length to cmpstrnqi patterns.
gcc/testsuite/
PR target/95443
* gcc.target/i386/pr95443-1.c: New test.
* gcc.target/i386/pr95443-2.c: Likewise.
Somewhat improved by r11-2064, though we still generate junk that seems
redundant. But at least it says
error: expected ‘}’ before ‘.’ token
PR c++/95288
* g++.dg/diagnostic/enum2.C: New test.
The following testcase ICEs since r10-3199.
There is a switch with default label, where the controlling expression has
range just 0..7 and there are case labels for all those 8 values, but
nothing has yet optimized away the default.
Since r10-3199, set_switch_stmt_execution_predicate sets the switch to
default label's edge's predicate to a false predicate and then
compute_bb_predicates propagates the predicates through the cfg, but false
predicates aren't really added. The caller of compute_bb_predicates
in one place handles NULL bb->aux as false predicate:
if (fbi.info)
{
if (bb->aux)
bb_predicate = *(predicate *) bb->aux;
else
bb_predicate = false;
}
else
bb_predicate = true;
but then in two further spots that the patch below is changing
it assumes bb->aux must be non-NULL. Those two spots are guarded by a
condition that is only true if fbi.info is non-NULL, so I think the right
fix is to treat NULL aux as false predicate in those spots too.
2020-07-13 Jakub Jelinek <jakub@redhat.com>
PR ipa/96130
* ipa-fnsummary.c (analyze_function_body): Treat NULL bb->aux
as false predicate.
* gcc.dg/torture/pr96130.c: New test.
Initialisation of a variable results in an implicit save attribute
being added to the variable. The save attribute is not allowed for
variables with the dummy attribute set. Initialisation should be
rejected for dummy variables.
2020-07-13 Mark Eggleston <markeggleston@gcc.gnu.org>
gcc/fortran/
PR fortran/45337
* resolve.c (resolve_fl_variable): Remove type and intent
checks from the check for dummy.
2020-07-13 Mark Eggleston <markeggleston@gcc.gnu.org>
gcc/testsuite/
PR fortran/45337
* gfortran.dg/pr45337_1.f90: New test.
* gfortran.dg/pr45337_2.f90: New test.
Here's an interesting issue: in this code a ) is missing:
enum { E = (2 } e;
but we compile the code anyway, and E is set to 0 in build_enumerator,
which is sneaky.
The problem is that cp_parser_enum_specifier parses tentatively, because
when we see the enum keyword, we don't know yet if we'll find an
enum-specifier, opaque-enum-declaration, or elaborated-enum-specifier.
In this test when we call cp_parser_enumerator_list we're still parsing
tentatively, and as a consequence, parens.require_close (parser) in
cp_parser_primary_expression doesn't report any errors. But we only go
on to parse the enumerator-list after we've seen a {, at which point we
might as well commit -- we know we're dealing with an enum-specifier.
gcc/cp/ChangeLog:
PR c++/96077
* parser.c (cp_parser_enum_specifier): Commit to tentative parse
after we've seen an opening brace.
gcc/testsuite/ChangeLog:
PR c++/96077
* g++.dg/parse/enum14.C: New test.
turned out scan-lang-dump-not was broken in the 3 argument case -- I'd missed a
necessary empty arg. Fixed thusly.
gcc/testsuite/
* lib/scanlang.exp (scan-lang-dump-not): Fix 3-arg case.
This avoids placing stmts beyond the vectorizer region begin which
confuses vect_stmt_dominates_stmt_p.
2020-07-13 Richard Biener <rguenther@suse.de>
PR tree-optimization/96163
* tree-vect-slp.c (vect_schedule_slp_instance): Put new stmts
at least after region begin.
* g++.dg/vect/pr96163.cc: New testcase.
When looking for an array reference allow NULL references. If
no array reference is found dim_rank_check should return false.
2020-07-13 Steven G. Kargl <kargl@gcc.gnu.org>
gcc/fortran/
PR fortran/95981
* check.c (dim_rank_check): Allow NULL references in call to
gfc_find_array_ref and return false if no reference is found.
2020-07-13 Mark Eggleston <markeggleston@gcc.gnu.org>
gcc/testsuite/
PR fortran/95981
* gfortran.dg/pr95981.f90: New test.
Define the __ARM_FEATURE_PAC_DEFAULT feature test
macro when PAC-RET branch protection is enabled.
2020-07-13 Szabolcs Nagy <szabolcs.nagy@arm.com>
gcc/ChangeLog:
* config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Add
__ARM_FEATURE_PAC_DEFAULT support.
The expected semantics and valid usage of __builtin_return_address is
not clear since it exposes implementation internals that are normally
not meaningful to portable c code.
This documentation change tries to clarify the semantics in case the
return address is stored in a mangled form. This affects AArch64 when
pointer authentication is used for the return address signing (i.e.
-mbranch-protection=pac-ret).
2020-07-13 Szabolcs Nagy <szabolcs.nagy@arm.com>
gcc/ChangeLog:
PR target/94891
* doc/extend.texi: Update the text for __builtin_return_address.
Mangling, currently only used on AArch64 for return address signing,
is an internal representation that should not be exposed via
__builtin_return_address return value,
__builtin_eh_return handler argument,
_Unwind_DebugHook handler argument.
Note that a mangled address might not even fit into a void *, e.g.
with AArch64 ilp32 ABI the return address is stored as 64bit, so
the mangled return address cannot be accessed via _Unwind_GetPtr.
This patch changes the unwinder hooks as follows:
MD_POST_EXTRACT_ROOT_ADDR is removed: root address comes from
__builtin_return_address which is not mangled.
MD_POST_EXTRACT_FRAME_ADDR is renamed to MD_DEMANGLE_RETURN_ADDR,
it now operates on _Unwind_Word instead of void *, so the hook
should work when return address signing is enabled on AArch64 ilp32.
(But for that __builtin_aarch64_autia1716 should be fixed to operate
on 64bit input instead of a void *.)
MD_POST_FROB_EH_HANDLER_ADDR is removed: it is the responsibility of
__builtin_eh_return to do the mangling if necessary.
2020-07-13 Szabolcs Nagy <szabolcs.nagy@arm.com>
libgcc/ChangeLog:
PR target/94891
* config/aarch64/aarch64-unwind.h (MD_POST_EXTRACT_ROOT_ADDR): Remove.
(MD_POST_FROB_EH_HANDLER_ADDR): Remove.
(MD_POST_EXTRACT_FRAME_ADDR): Rename to ...
(MD_DEMANGLE_RETURN_ADDR): This.
(aarch64_post_extract_frame_addr): Rename to ...
(aarch64_demangle_return_addr): This.
(aarch64_post_frob_eh_handler_addr): Remove.
* unwind-dw2.c (uw_update_context): Demangle return address.
(uw_frob_return_addr): Remove.
Currently __builtin_eh_return takes a signed return address, which can
cause ABI and API issues: 1) pointer representation problems if the
address is passed around before eh return, 2) the source code needs
pac-ret specific changes and needs to know if pac-ret is used in the
current frame, 3) signed address may not be representible as void *
(with ilp32 abi).
Using address signing to protect eh return is ineffective because the
instruction sequence in the unwinder that starts from the address
signing and ends with a ret can be used as a return to anywhere gadget.
Using indirect branch istead of ret with bti j landing pads at the
target can reduce the potential of such gadget, which also implies
that __builtin_eh_return should not take a signed address.
This is a big hammer fix to the ABI and API issues: it turns pac-ret
off for the caller completely (not just on the eh return path). To
harden the caller against ROP attacks, it should use indirect branch
instead of ret, this is not attempted so the patch remains small and
backportable.
2020-07-13 Szabolcs Nagy <szabolcs.nagy@arm.com>
gcc/ChangeLog:
PR target/94891
* config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
Disable return address signing if __builtin_eh_return is used.
gcc/testsuite/ChangeLog:
PR target/94891
* gcc.target/aarch64/return_address_sign_1.c: Update test.
* gcc.target/aarch64/return_address_sign_b_1.c: Likewise.
This is a big hammer fix for __builtin_return_address (PR target/94891)
returning signed addresses (sometimes, depending on wether lr happens
to be signed or not at the time of call which depends on optimizations),
and similarly -pg may pass signed return address to _mcount
(PR target/94791).
At the time of return address expansion we don't know if it's signed or
not so it is done unconditionally.
2020-07-13 Szabolcs Nagy <szabolcs.nagy@arm.com>
gcc/ChangeLog:
PR target/94891
PR target/94791
* config/aarch64/aarch64-protos.h (aarch64_return_addr_rtx): Declare.
* config/aarch64/aarch64.c (aarch64_return_addr_rtx): New.
(aarch64_return_addr): Use aarch64_return_addr_rtx.
* config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
The testcase has failed since r9-5035, because obj_type_ref_class
tries to look up an ODR type when no ODR type information is
available. (The information was available earlier in the
compilation, but was freed during pass_ipa_free_lang_data.)
We then crash dereferencing the null get_odr_type result.
The test passes with -O2. However, it fails again if -fdump-tree-all
is used, since obj_type_ref_class is called indirectly from the
dump routines.
Other code creates ODR type entries on the fly by passing “true”
as the insert parameter. But obj_type_ref_class can't do that
unconditionally, since it should have no side-effects when used
from the dumping code.
Following a suggestion from Honza, this patch adds parameters
to say whether the routines are being called from dump routines
and uses those to derive the insert parameter.
gcc/
PR middle-end/95114
* tree.h (virtual_method_call_p): Add a default-false parameter
that indicates whether the function is being called from dump
routines.
(obj_type_ref_class): Likewise.
* tree.c (virtual_method_call_p): Likewise.
* ipa-devirt.c (obj_type_ref_class): Likewise. Lazily add ODR
type information for the type when the parameter is false.
* tree-pretty-print.c (dump_generic_node): Update calls to
virtual_method_call_p and obj_type_ref_class accordingly.
gcc/testsuite/
PR middle-end/95114
* g++.target/aarch64/pr95114.C: New test.
My previous fix for PR 94749 did fix the reported case, so that the next
character is not discarded if it happens to equal the delimiter when __n
characters have already been read. But it introduced a new bug, which is
that the delimiter character would *not* be discarded if the number of
characters discarded is numeric_limits<streamsize>::max() or more before
reaching the delimiter.
The new bug happens because I changed the code to check _M_gcount < __n.
But when __n == numeric_limits<streamsize>::max() that is false, and so
we don't discard the delimiter. It's not sufficient to check for the
delimiter when the __large_ignore condition is true, because there's an
edge case where the delimiter is reached when _M_gcount == __n and so
we break out of the loop without setting __large_ignore.
PR 96161 is a similar bug to the original PR 94749 report, where eofbit
is set after discarding __n characters if there happen to be no more
characters in the stream.
This patch fixes both cases (and the regression) by checking different
conditions for the __n == max case and the __n < max case. For the
former case, we know that we must have either reached the delimiter or
EOF, and the value of _M_gcount doesn't matter (except to avoid integer
overflow). For the latter case we need to check _M_gcount first and only
set eofbit or discard the delimiter if it didn't reach __n. For the
latter case overflow can't happen because _M_gcount <= __n < max.
libstdc++-v3/ChangeLog:
PR libstdc++/94749
PR libstdc++/96161
* include/bits/istream.tcc (basic_istream::ignore(streamsize))
[n == max]: Check overflow conditions on _M_gcount. Rely on
the fact that either EOF or the delimiter was reached.
[n < max]: Check _M_gcount < n before checking for EOF or
delimiter.
(basic_istream::ignore(streamsize, char_type): Likewise.
* src/c++98/compatibility.cc (istream::ignore(streamsize))
(wistream::ignore(streamsize)): Likewise.
* src/c++98/istream.cc (istream::ignore(streamsize, char_type))
(wistream::ignore(streamsize, char_type)): Likewise.
* testsuite/27_io/basic_istream/ignore/char/94749.cc: Check that
delimiter is discarded if the number of characters ignored
doesn't fit in streamsize.
* testsuite/27_io/basic_istream/ignore/wchar_t/94749.cc:
Likewise.
* testsuite/27_io/basic_istream/ignore/char/96161.cc: New test.
* testsuite/27_io/basic_istream/ignore/wchar_t/96161.cc: New test.
OpenACC 2.6 specifies that the array descriptor (when present) must be
copied to the target before attaching pointers in Fortran. This patch
reverses the stripping of GOMP_MAP_TO_PSET and GOMP_MAP_POINTER that
was introduced by the "OpenACC reference count overhaul" patch.
2020-07-10 Julian Brown <julian@codesourcery.com>
Thomas Schwinge <thomas@codesourcery.com>
gcc/
* gimplify.c (gimplify_scan_omp_clauses): Do not strip
GOMP_MAP_TO_PSET/GOMP_MAP_POINTER for OpenACC enter/exit data
directives (see also PR92929).
gcc/testsuite/
* gfortran.dg/goacc/finalize-1.f: Update expected dump output.
libgomp/
* testsuite/libgomp.oacc-fortran/dynamic-pointer-1.f90: New test.
Co-Authored-By: Thomas Schwinge <thomas@codesourcery.com>
The testcase was errorneously accessing the global variable via a
type that might require bigger alignment than provided. Fix that
via an appropriate attribute.
2020-07-13 Richard Biener <rguenther@suse.de>
PR testsuite/96180
* gcc.dg/torture/pr96133.c: Align global variable.
This patch eliminates a check of targetm.truly_noop_truncation from
the early middle-end, where the gimple/generic being generated by
GCC's front-ends is being inappropriately influenced by the target's
TRULY_NOOP_TRUNCATION. The (recent) intention of TRULY_NOOP_TRUNCATION
is to indicate that a backend requires explicit truncation instructions
rather than using SUBREGs to perform truncations. A long standing
(and probably unintentional) side-effect has been that this setting
also controls whether the middle-end narrows integer operations at
the tree-level. Understandably, GCC and its testsuite assume that
GIMPLE and GENERIC behave consistently across platforms, and alas
defining TRULY_NOOP_TRUNCATION away from the default triggers several
regressions (including gcc.dg/fold-rotate-1.c).
2020-07-13 Roger Sayle <roger@nextmovesoftware.com>
gcc/ChangeLog
* convert.c (convert_to_integer_1): Narrow integer operations
even on targets that require explicit truncation instructions.
Delayed-branch-slot-filling a.k.a. reorg or dbr, often causes
opportunities for more compare-elimination than were visible for
the cmpelim pass. With cc0, these were caught by the
elimination pass run in "final", thus the missed opportunities
is a regression. A simple reorg-aware pass run just after reorg
handles most of them, if not all. I chose to keep the "mach2"
pass identifier string I copy-pasted from the SPARC port instead
of inventing one like "postdbr_cmpelim". Note the gap in numbers
in the test-case file names.
gcc:
PR target/93372
* config/cris/cris-passes.def: New file.
* config/cris/t-cris (PASSES_EXTRA): Add cris-passes.def.
* config/cris/cris.c: Add infrastructure bits and pass execute
function cris_postdbr_cmpelim.
* config/cris/cris-protos.h (make_pass_cris_postdbr_cmpelim): Declare.
gcc/testsuite:
* gcc.target/cris/pr93372-44.c, gcc.target/cris/pr93372-46.c: New.
Getting tired of:
make[1]: Entering directory 'x/gccobj/gcc'
Makefile:2682: warning: overriding recipe for target 'gt-cris.h'
xx/gcc/gcc/config/cris/t-cris:29: warning: ignoring old recipe for target 'gt-cris.h'
I'm just going to assume it is just stale cruft no longer (if
ever) needed since nothing else but sh/t-sh has it, and the
commit log shows just (x prepended to avoid commit-log parsing
confusion):
x Merge from pch-branch up to tag pch-commit-20020603.
x
x From-SVN: r54232
Building "works better"; the related warning is gone.
This effectively empties the t-cris file, but stuff will be
added soon enough that it's kept around.
gcc:
* config/cris/t-cris: Remove gt-cris.h-related excessive cargo.
Comparing to the cc0 version of the CRIS port, I ran a few
microbenchmarks, for example gcc.c-torture/execute/arith-rand.c,
where there's sometimes an addition between an operation of
interest and the test on the result.
Unfortunately this patch doesn't remedy all the performance
regression for that program. But, this patch by itself helps
and makes sense to commit separately: lots of addi.b in
previously empty delay-slots, with functions shortened by one or
a few insns, in libgcc. I had an experience with the
reload-related caveat of % on constraints, which is "fixed"
documentationwise since long (soon 15 years ago;
be3914df4cc8/r105517). I removed an even older related FIXME.
gcc:
PR target/93372
* config/cris/cris.md ("*add<mode>3_addi"): New splitter.
("*addi_b_<mode>"): New pattern.
("*addsi3<setnz>"): Remove stale %-related comment.
gcc/testsuite:
PR target/93372
* gcc.target/cris/pr93372-45.c: New test.
Whoops. This little gem had the effect of making the output
operand (0) constraints disappear but not the input operand (1)
constraints for define_subst:ed patterns, probably because
there's another (match_dup 1) in the output template (not
investigated).
That went surprisingly unnoticed until I added a pass leaning
just a little bit harder on the define_subst:ed patterns and
then only by the libgfortran library generating assembly with
nominally incorrect syntax. (There was a move to a special
register from a general register, and it incorrectly matched a
pattern affecting condition codes.)
gcc:
* config/cris/cris.md ("setnz_subst", "setnz_subst", "setcc_subst"):
Use match_dup in output template, not match_operand.
This eliminates the visited bitmap and makes whether a to be processed
block goes to the next or the current iteration only depend on its
position in RPO order rather than on whether it was visited in the
current iteration. As optimization single-BB iteration is processed
immediately.
2020-07-10 Richard Biener <rguenther@suse.de>
* var-tracking.c (bb_heap_node_t): Remove unused typedef.
(vt_find_locations): Eliminate visited bitmap in favor of
RPO order check. Dump statistics about the number of
local BB dataflow computes.
The store to the whole of each volatile object was picked apart
like there had been an individual assignment to each of the
fields. Reads were added as part of that; see PR for details.
The reads from volatile memory were a clear bug; individual
stores questionable. A separate patch clarifies the docs.
gcc:
2020-07-09 Richard Biener <rguenther@suse.de>
PR middle-end/94600
* expr.c (expand_constructor): Make a temporary also if we're
storing to volatile memory.
gcc/testsuite:
2020-07-09 Hans-Peter Nilsson <hp@axis.com>
PR middle-end/94600
* gcc.dg/pr94600-1.c, gcc.dg/pr94600-2.c, gcc.dg/pr94600-3.c,
gcc.dg/pr94600-4.c, gcc.dg/pr94600-5.c, gcc.dg/pr94600-6.c,
gcc.dg/pr94600-7.c, gcc.dg/pr94600-8.c: New tests.
