The resolution of CWG issue 1321 clarified that when deciding whether two
expressions involving template parameters are equivalent, two dependent
function calls where the function is named with an unqualified-id are
considered to be equivalent if the name is the same, even if unqualified
lookup finds different sets of functions. We were wrongly treating
qualified-ids the same way, so that EXISTS and test::EXISTS were considered
to be equivalent even though they are looking up the name in different
scopes. This also causes a mangling bug, but I don't think it's safe to fix
that for GCC 10; this patch just fixes the comparison.
gcc/cp/ChangeLog
2020-03-30 Jason Merrill <jason@redhat.com>
PR c++/90711
* tree.c (cp_tree_equal) [CALL_EXPR]: Compare KOENIG_LOOKUP_P.
(called_fns_equal): Check DECL_CONTEXT.
The following testcase ICEs, because the FE when processing the statement
expression changes the .VEC_CONVERT internal fn CALL_EXPR into .PHI call.
That is because the internal fn call is recorded in the base.u.ifn
field, which overlaps base.u.bits.lang_flag_1 which is used for
STMT_IS_FULL_EXPR_P, so this essentially does ifn |= 2 on little-endian.
STMT_IS_FULL_EXPR_P bit is used in:
cp-gimplify.c- if (STATEMENT_CODE_P (code))
cp-gimplify.c- {
cp-gimplify.c- saved_stmts_are_full_exprs_p = stmts_are_full_exprs_p ();
cp-gimplify.c- current_stmt_tree ()->stmts_are_full_exprs_p
cp-gimplify.c: = STMT_IS_FULL_EXPR_P (*expr_p);
cp-gimplify.c- }
and
pt.c- if (STATEMENT_CODE_P (TREE_CODE (t)))
pt.c: current_stmt_tree ()->stmts_are_full_exprs_p = STMT_IS_FULL_EXPR_P (t);
so besides being wrong on some other codes, it actually isn't beneficial at
all to set it on anything else, so the following patch restricts it to
trees with STATEMENT_CODE_P TREE_CODE.
2020-03-30 Jakub Jelinek <jakub@redhat.com>
PR c++/94385
* semantics.c (add_stmt): Only set STMT_IS_FULL_EXPR_P on trees with
STATEMENT_CODE_P code.
* c-c++-common/pr94385.c: New test.
Disable the code that limits initialization of builtins based
on the rs6000_builtin_mask. This allows all built-ins to be
properly referenced when building code using #pragma for cpu
targets newer than what was specified by the -mcpu default.
The use of built-ins is still properly limited by logic within
altivec_resolve_overloaded_builtin().
2020-03-30 Will Schmidt <will_schmidt@vnet.ibm.com>
gcc/
* config/rs6000/rs6000-call.c altivec_init_builtins(): Remove code
to skip defining builtins based on builtin_mask.
gcc/testsuite/
* gcc.target/powerpc/pragma_power6.c: New.
* gcc.target/powerpc/pragma_power7.c: New.
* gcc.target/powerpc/pragma_power8.c: New.
* gcc.target/powerpc/pragma_power9.c: New.
* gcc.target/powerpc/pragma_misc9.c: New.
* gcc.target/powerpc/vsu/vec-all-nez-7.c: Update error message.
* gcc.target/powerpc/vsu/vec-any-eqz-7.c: Update error message.
Update existing testcase powerpc/bswap64-4.c to
reflect that we generate ldbrx and stdbrx instructions
for newer cpu targets, versus the pair of lwbrx and stwbrx
instructions as seen on previous cpu targets.
2020-03-24 Will Schmidt <will_schmidt@vnet.ibm.com>
gcc/testsuite/
* gcc.target/powerpc/bswap64-4.c: Update scan-assembler
expected results.
This define_insn has two issues.
One is that with -mavx512f -mno-avx512vl it can emit an AVX512VL-only insn
- 128-bit or 256-bit EVEX encoded vpternlog{d,q}.
Another one is that because there is no vpternlog{b,w}, we emit vpternlogd
instead, but then we shouldn't pretend we support masking of that, because
we don't.
The first one can be fixed by forcing the use of %zmm* registers instead of
%xmm* or %ymm* if AVX512F but not AVX512VL, like we do for a couple of other
insns (although that is primarily done in order to support %xmm16+ regs).
But we need to make sure that in that case the input operand isn't memory,
because while we can read and store the higher bits of registers, we don't
want to read from memory more bytes than what we should read.
A variant to these two if_then_else set attrs, condition in the output and
larger condition would be 4 different define_insns (one with something like
VI48_AVX512VL iterator, masking, no g modifiers and "vm" input constraint,
another one with VI48_AVX iterator, !TARGET_AVX512VL in condition,
no masking, g modifiers and "v" input constraint, one with VI12_AVX512VL
iterator, no masking, no g modifiers and "vm" input constraint and last one with
VI12_AVX2 iterator, !TARGET_AVX512VL in condition, no masking, g modifiers
and "v" input constraint, but I think having one pattern is shorter than
that.
2020-03-30 Jakub Jelinek <jakub@redhat.com>
PR target/94343
* config/i386/sse.md (<mask_codefor>one_cmpl<mode>2<mask_name>): If
!TARGET_AVX512VL, use 512-bit vpternlog and make sure the input
operand is a register. Don't enable masked variants for V*[QH]Imode.
* gcc.target/i386/avx512f-pr94343.c: New test.
* gcc.target/i386/avx512vl-pr94343.c: New test.
The AVX512F documentation clearly states that in instructions where the
destination is a memory only merging-masking is possible, not zero-masking,
and the assembler enforces that.
The testcase in this patch fails to assemble because of
Error: unsupported masking for `vextracti32x8'
on
vextracti32x8 $0x0, %zmm1, -64(%rsp){%k1}{z}
For the vector extraction patterns, we apparently have 7 *_maskm patterns
that only accept memory destinations and rtx_equal_p merge-masking source
for it, 7 *<mask_name> corresponding patterns that allow memory destination
only for the non-masked cases (through <store_mask_constraint>), then 2
*<mask_name> patterns (lo ssehalf V16FI and lo ssehalf VI8F_256 ones) which
do allow memory destination even for masked cases and are the cause of the
testsuite failure, because we must not allow C constraint if the destination
is m, and finally one pair of patterns (separate * and *_mask, hi ssehalf
VI4F_256), which has another issue (for which I don't have a testcase
though), where if it would match zero-masking with register destination,
it wouldn't emit the needed {z} into assembly.
The attached patch fixes those 3 issues only, perhaps more suitable for
backporting.
2020-03-30 Jakub Jelinek <jakub@redhat.com>
PR target/93069
* config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Use
<store_mask_constraint> instead of m in output operand constraint.
(vec_extract_hi_<mode><mask_name>): Use <mask_operand2> instead of
%{%3%}.
* gcc.target/i386/avx512vl-pr93069.c: New test.
* gcc.dg/vect/pr93069.c: New test.
-mlongcall -mno-pltseq is supposed to emit long calls by using
indirect calls. It differs from -mlongcall -mpltseq in that the
function addresses are not placed in the PLT and thus lazy PLT
resolution is not available, affecting programs that dlopen shared
libraries.
In the case of -mcpu=future -mpcrel -mlongcall -mno-pltseq we see an
indirect call being generated, but combine merrily optimises the
sequence back to a direct call. call_indirect_pcrel is enough like
call_nonlocal_aix that this can happen.
This patch puts the call cookie back in the call rtl, removed by git
commit f90f960ca8, in order to disable the optimisation for long
calls. When that is done for call_local_aix the pattern becomes the
same as call_local32/64, so I merged them. The only difference
besides mode between call_local32 and call_local64, dating back to
1998 commit a260abc996, is that call_local64 has TARGET_64BIT in the
predicate. That alone doesn't seem reason enough to need separate
patterns; The P mode iterator selects DI on TARGET_64BIT anyway.
* config/rs6000/rs6000.c (rs6000_call_aix): Emit cookie to pattern.
(rs6000_indirect_call_template_1): Adjust to suit.
* config/rs6000/rs6000.md (call_local): Merge call_local32,
call_local64, and call_local_aix.
(call_value_local): Simlarly.
(call_nonlocal_aix, call_value_nonlocal_aix): Adjust rtl to suit,
and disable pattern when CALL_LONG.
(call_indirect_aix, call_value_indirect_aix): Adjust rtl.
(call_indirect_elfv2, call_indirect_pcrel): Likewise.
(call_value_indirect_elfv2, call_value_indirect_pcrel): Likewise.
Change -falign-functions=N to
Align the start of functions to the next power-of-two greater than
or equal to N.
Add
If '-falign-labels' is greater than this value, then its value is
used instead.
to -falign-loops=N and -falign-jumps=N.
PR driver/94381
* doc/invoke.texi: Update -falign-functions, -falign-loops and
-falign-jumps documentation.
This patch adds cases for JSON and D header file generation flags in the
testsuite, and sets up the test accordingly to only compile, then check
that the expected output file exists.
Support has also been added for the DFLAGS test directive, which
disables testing with all default flag permutations.
gcc/testsuite/ChangeLog:
* lib/gdc-utils.exp: (gdc-convert-args): Handle compilation test
arguments -D, -H, -X, and set compilable_output_file_ext.
(gdc-convert-test): Document and add cases DFLAGS and
EXTRA_CPP_SOURCES. Add dg-final entry for output generation tests,
and set compilable_do_what to compile.
(gdc-do-test): Define compilable_do_what, default to assemble.
Tests have been moved into runnable_cxx as part of upstream dmd 3e10e2dd2.
The extra flags required for tests that mix C++ and D are now limited to
only a small subset of tests, rather than applied to all tests across
gdc.dg and gdc.test.
Reviewed-on: https://github.com/dlang/dmd/pull/10980
gcc/testsuite/ChangeLog:
* gdc.test/runnable_cxx/runnable_cxx.exp: New file.
* lib/gdc-utils.exp (gdc-do-test): Add case for runnable_cxx.
* lib/gdc.exp (gdc_include_flags): Only add flags for libstdc++-v3 if
GDC_INCLUDE_CXX_FLAGS is true.
(gdc_link_flags): Likewise.
(gdc_init): Move setting of default gdc test flags to...
(gdc_target_compile): ...here.
Use multiple test scripts, one for each subdirectory containing test
files, instead of having one test script to manage them all.
This allows removing some workarounds, such as the need to create
symlinks in the test run directory.
gcc/testsuite/ChangeLog:
* gdc.test/compilable/compilable.exp: New file.
* gdc.test/fail_compilation/fail_compilation.exp: New file.
* gdc.test/gdc-test.exp: Remove.
* gdc.test/runnable/runnable.exp: New file.
* lib/gdc-utils.exp: ...this. Remove load_lib gdc-dg.exp.
(dmd2dg): Rename to...
(gdc-convert-test): ...this.
(gdc-do-test): Add testcases parameter and remove subdir handling.
* testsuite/20_util/is_constructible/value-2.cc: Fix test to account
for changes due to parenthesized aggregate-initialization in C++20.
* testsuite/20_util/time_point/cons/81468.cc: Fix test to not clash
with std::chrono::sys_time in C++20.
My recent changes to reverse_iterator's comparisons was not the version
of the code (or tests) that I meant to commit, and broke the relational
operators. This fixes them to reverse the order of the comparisons on
the base() iterators.
This also replaces the SFINAE constraints in the return type of the
reverse_iterator and move_iterator comparisons with a requires-clause.
This ensures the constrained overloads are preferred to unconstrained
ones. This means the non-standard same-type overloads can be omitted for
C++20 because they're not needed to solve the problem with std::rel_ops
or the testsuite's greedy_ops::X type.
* include/bits/stl_iterator.h (reverse_iterator): Use requires-clause
to constrain C++20 versions of comparison operators. Fix backwards
logic of relational operators.
(move_iterator): Use requires-clause to constrain comparison operators
in C++20. Do not declare non-standard same-type overloads for C++20.
* testsuite/24_iterators/move_iterator/rel_ops_c++20.cc: Check result
of comparisons and check using greedy_ops type.
* testsuite/24_iterators/reverse_iterator/rel_ops_c++20.cc: Likewise.
* testsuite/24_iterators/move_iterator/greedy_ops.cc: Remove redundant
main function from compile-only test.
* testsuite/24_iterators/reverse_iterator/greedy_ops.cc: Likewise.
Without this patch tortured tests eventually fail because of another
known issue which I plan to fix later in stage1.
* gcc.dg/torture/pr87197-debug-sms.c: Move to ...
* gcc.dg/sms-compare-debug-1.c: ... this. Add -O2.
* gcc.c-torture/execute/pr70127-debug-sms.c: Move to ...
* gcc.dg/sms-compare-debug-2.c: ... this. Add -O2.
This adds support to detect and recover from the case where an opening brace
immediately follows the start of a requires-clause. So rather than emitting the
error
error: expected primary-expression before '{' token
followed by a slew of irrevelant errors, we now assume the user had intended to
write "requires requires {" and diagnose and recover accordingly.
gcc/cp/ChangeLog:
PR c++/94306
* parser.c (cp_parser_requires_clause_opt): Diagnose and recover from
"requires {" when "requires requires {" was probably intended.
gcc/testsuite/ChangeLog:
PR c++/94306
* g++.dg/concepts/diagnostic8.C: New test.
This PR shows that a REQUIRES_EXPR outside of a template can sometimes be
misevaluated. This happens because the evaluation routine tsubst_requires_expr
(and diagnose_requires_expr) assumes the REQUIRES_EXPR's subtrees are templated
trees and that therefore it's safe to call tsubst_expr on them. But this
assumption isn't valid when we've parsed a REQUIRES_EXPR outside of a template
context. In order to make this assumption valid here, this patch sets
processing_template_decl to non-zero before parsing the body of a REQUIRES_EXPR
so that its subtrees are indeed always templated trees.
gcc/cp/ChangeLog:
PR c++/94252
* constraint.cc (tsubst_compound_requirement): Always suppress errors
from type_deducible_p and expression_convertible_p, as they're not
substitution errors.
(diagnose_atomic_constraint) <case INTEGER_CST>: Remove this case so
that we diagnose INTEGER_CST expressions of non-bool type via the
default case.
* cp-gimplify.c (cp_genericize_r) <case REQUIRES_EXPR>: New case.
* parser.c (cp_parser_requires_expression): Always parse the requirement
body as if we're processing a template, by temporarily incrementing
processing_template_decl. Afterwards, if we're not actually in a
template context, perform semantic processing to diagnose any invalid
types and expressions.
* pt.c (tsubst_copy_and_build) <case REQUIRES_EXPR>: Remove dead code.
* semantics.c (finish_static_assert): Explain an assertion failure
when the condition is a REQUIRES_EXPR like we do when it is a concept
check.
gcc/testsuite/ChangeLog:
PR c++/94252
* g++.dg/concepts/diagnostic7.C: New test.
* g++.dg/concepts/pr94252.C: New test.
* g++.dg/cpp2a/concepts-requires18.C: Adjust to expect an additional
diagnostic.
The previous patch tries to avoid changing our current default diagnostics. But
for the sake of consistency we arguably should also respect
current_constraint_diagnosis_depth in diagnose_compound_requirement() like we do
in the other error-replaying diagnostic routines. But doing so would be a
change to our default diagnostics behavior, so the change has been split out
into this separate patch for separate consideration.
gcc/cp/ChangeLog:
* constraint.cc (diagnose_compound_requirement): When diagnosing a
compound requirement, maybe replay the satisfaction failure, subject to
the current diagnosis depth.
gcc/testsuite/ChangeLog:
* g++.dg/concepts/diagnostic1.C: Pass -fconcepts-diagnostics-depth=2.
* g++.dg/concepts/diagnostic5.C: Adjust expected diagnostics.
* g++.dg/cpp2a/concepts-iconv1.C: Pass -fconcepts-diagnostics-depth=2.
* g++.dg/cpp2a/concepts-requires5.C: Likewise.
This patch adds a new flag -fconcepts-diagnostics-depth to the C++ frontend
which controls how deeply we replay errors when diagnosing a constraint
satisfaction failure. The default is -fconcepts-diagnostics-depth=1 which
diagnoses only the topmost constraint satisfaction failure and is consistent
with our behavior before this patch. By increasing this flag's value, the user
can control how deeply they want the compiler to explain a constraint
satisfaction error.
For example, if the unsatisfied constraint is a disjunction, then the default
behavior is to just say "no branch in the disjunction is satisfied", but with
-fconcepts-diagnostics-depth=2 we will additionally replay and diagnose the
error in each branch of the disjunction. And if the unsatisfied constraint is a
requires expression, then we will replay the error in the requires expression,
etc. This proceeds recursively until there is nothing more to replay or we
exceeded the maximum depth specified by the flag.
Implementation wise, this patch essentially just uncomments the existing
commented-out code that performs the error-replaying, and along the way adds
logic to keep track of and limit the current replay depth. Besides that, there
is a new routine collect_operands_of_disjunction which flattens a disjunction
and collects all of its operands into a vector.
The extra diagnostics enabled by this flag are at times longer than they need to
be (e.g. "the operand is_array_v<...> is unsatisfied because \n the expression
is_array_v<...> [with ...] evaluated to false") and not immediately easy to
follow (especially when there are nested disjunctions), but the transparency
provided by these optional diagnostics seems to be pretty helpful in practice.
gcc/c-family/ChangeLog:
* c.opt: Add -fconcepts-diagnostics-depth.
gcc/cp/ChangeLog:
* constraint.cc (finish_constraint_binary_op): Set the location of EXPR
as well as its range, because build_x_binary_op doesn't always do so.
(current_constraint_diagnosis_depth): New.
(concepts_diagnostics_max_depth_exceeded_p): New.
(collect_operands_of_disjunction): New.
(satisfy_disjunction): When diagnosing a satisfaction failure, maybe
replay each branch of the disjunction, subject to the current diagnosis
depth.
(diagnose_valid_expression): When diagnosing a satisfaction failure,
maybe replay the substitution error, subject to the current diagnosis
recursion.
(diagnose_valid_type): Likewise.
(diagnose_nested_requiremnet): Likewise.
(diagnosing_failed_constraint::diagnosing_failed_constraint): Increment
current_constraint_diagnosis_depth when diagnosing.
(diagnosing_failed_constraint::~diagnosing_failed_constraint): Decrement
current_constraint_diagnosis_depth when diagnosing.
(diagnosing_failed_constraint::replay_errors_p): New static member
function.
(diagnose_constraints): Don't diagnose if concepts_diagnostics_max_depth
is 0. Emit a one-off note to increase -fconcepts-diagnostics-depth if
the limit was exceeded.
* cp-tree.h (diagnosing_failed_constraint::replay_errors_p): Declare.
gcc/testsuite/ChangeLog:
* g++.dg/concepts/diagnostic2.C: Expect "no operand" instead of
"neither operand".
* g++.dg/concepts/diagnostic5.C: New test.
After we report various errors about array size, we set for error-recovery
the size to be 1, but because size_int_const is false, it still means we
pretend the array is a VLA, can emit a second diagnostics in that case etc.
E.g.
$ ./cc1.unpatched -quiet a.c
a.c:1:5: error: size of array ‘f’ has non-integer type
1 | int f[100.0];
| ^
a.c:1:1: warning: variably modified ‘f’ at file scope
1 | int f[100.0];
| ^~~
$ ./cc1 -quiet a.c
a.c:1:5: error: size of array ‘f’ has non-integer type
1 | int f[100.0];
| ^
2020-03-28 Jakub Jelinek <jakub@redhat.com>
PR c/93573
* c-decl.c (grokdeclarator): After issuing errors, set size_int_const
to true after setting size to integer_one_node.
* gcc.dg/pr93573-1.c: New test.
* gcc.dg/pr93573-2.c: New test.
The following testcase FAILs with -fcompare-debug, because reassociate_bb
mishandles the case when the last stmt in a bb has zero uses. In that case
reassoc_remove_stmt (like gsi_remove) moves the iterator to the next stmt,
i.e. gsi_end_p is true, which means the code sets the iterator back to
gsi_last_bb. The problem is that the for loop does gsi_prev on that before
handling the next statement, which means the former penultimate stmt, now
last one, is not processed by reassociate_bb.
Now, with -g, if there is at least one debug stmt at the end of the bb,
reassoc_remove_stmt moves the iterator to that following debug stmt and we
just do gsi_prev and continue with the former penultimate non-debug stmt,
now last non-debug stmt.
The following patch fixes that by not doing the gsi_prev in this case; there
are too many continue; cases, so I didn't want to copy over the gsi_prev to
all of them, so this patch uses a bool for that instead. The second
gsi_end_p check isn't needed anymore, because when we don't do the
undesirable gsi_prev after gsi = gsi_last_bb, the loop !gsi_end_p (gsi)
condition will catch the removal of the very last stmt from a bb.
2020-03-28 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/94329
* tree-ssa-reassoc.c (reassociate_bb): When calling reassoc_remove_stmt
on the last stmt in a bb, make sure gsi_prev isn't done immediately
after gsi_last_bb.
* gfortran.dg/pr94329.f90: New test.
Define the feature test macro now that ranges support is complete.
This also changes the preprocessor checks for the __cpp_concepts macro
so that library components depending on concepts are only enabled when
C++20 concepts are supported, and not just for the Concepts TS (which
uses different syntax in places).
* include/bits/range_cmp.h (__cpp_lib_ranges): Define.
* include/bits/stl_iterator.h: Check value of __cpp_concepts so that
C++20 concepts are required.
* include/bits/stl_iterator_base_types.h: Likewise.
* include/std/concepts: Likewise.
* include/std/version: Likewise.
* testsuite/std/ranges/headers/ranges/synopsis.cc: Check feature test
macro.
This adds the missing parts of P0896R4 to reverse_iterator and
move_iterator, so that they meet the C++20 requirements. This should be
the last piece of P0896R4, meaning ranges support is now complete.
The PR 94354 bug with reverse_iterator's comparisons is fixed for C++20
only, but that change should be extended to C++11, C++14 and C++17 modes
in stage 1.
* include/bits/stl_iterator.h (reverse_iterator::iterator_concept)
(reverse_iterator::iterator_category): Define for C++20.
(reverse_iterator): Define comparison operators correctly for C++20.
(__normal_iterator): Add constraints to comparison operators for C++20.
(move_iterator::operator++(int)) [__cpp_lib_concepts]: Define new
overload for input iterators.
(move_iterator): Add constraints to comparison operators for C++20.
Define operator<=> for C++20.
* testsuite/24_iterators/move_iterator/input_iterator.cc: New test.
* testsuite/24_iterators/move_iterator/move_only.cc: New test.
* testsuite/24_iterators/move_iterator/rel_ops_c++20.cc: New test.
* testsuite/24_iterators/reverse_iterator/rel_ops_c++20.cc: New test.
std::insert_iterator and std::inserter need to be adjusted for C++20, so
that they use ranges::iterator_t. That alias template requires
ranges::begin to be defined. Rather than moving the whole of
ranges::begin (and related details like ranges::enable_borrowed_range)
into <iterator>, this defines a new, simpler version of ranges::begin
that is sufficient for ranges::iterator_t to be defined. This works
because ranges::iterator_t uses an lvalue reference type, so the logic
in ranges::begin for non-lvalue ranges (i.e. borrowed ranges) isn't
needed.
This also adds the missing constexpr specifiers to the other insert
iterators.
* include/bits/iterator_concepts.h (__detail::__decay_copy)
(__detail::__member_begin, __detail::__adl_begin): Move here from
<bits/range_access.h>.
(__detail::__ranges_begin, __detail::__range_iter_t): Define.
* bits/range_access.h (__cust_access::__decay_copy)
(__cust_access::__member_begin, __cust_access::__adl_begin): Move to
<bits/iterator_concepts.h>.
(ranges::iterator_t): Use __detail::__range_iter_t.
* include/bits/stl_iterator.h (back_insert_iterator): Simplify
conditional compilation. Add _GLIBCXX20_CONSTEXPR to all members.
(front_insert_iterator): Likewise.
(insert_iterator): Implement changes from P0896R4 for C++20.
* testsuite/24_iterators/back_insert_iterator/constexpr.cc: New test.
* testsuite/24_iterators/front_insert_iterator/constexpr.cc: New test.
* testsuite/24_iterators/headers/iterator/synopsis_c++17.cc: Adjust
for inclusion in synopsis_c++20.cc which expects different signatures
for some function templates.
* testsuite/24_iterators/insert_iterator/constexpr.cc: New test.
This moves __is_array_convertible so it's not between
__is_nothrow_convertible and its helper, since it isn't related to
those.
* include/std/type_traits (__is_array_convertible): Move definition
to immediately after is_convertible.
The PLT is volatile. On PowerPC it is a bss style section which the
dynamic loader initialises to point at resolver stubs (called glink on
PowerPC64) to support lazy resolution of function addresses. The
first call to a given function goes via the dynamic loader symbol
resolver, which updates the PLT entry for that function and calls the
function. The second call, if there is one and we don't have a
multi-threaded race, will use the updated PLT entry and thus avoid
the relatively slow symbol resolver path.
Calls via the PLT are like calls via a function pointer, except that
no initialised function pointer is volatile like the PLT. All
initialised function pointers are resolved at program startup to point
at the function or are left as NULL. There is no support for lazy
resolution of any user visible function pointer.
So why does any of this matter to gcc? Well, normally the PLT call
mechanism happens entirely behind gcc's back, but since we implemented
inline PLT calls (effectively putting the PLT code stub that loads the
PLT entry inline and making that code sequence scheduled), the load of
the PLT entry is visible to gcc. That load then is subject to gcc
optimization, for example in
/* -S -mcpu=future -mpcrel -mlongcall -O2. */
int foo (int);
void bar (void)
{
while (foo(0))
foo (99);
}
we see the PLT load for foo being hoisted out of the loop and stashed
in a call-saved register. If that happens to be the first call to
foo, then the stashed value is that for the resolver stub, and every
call to foo in the loop will then go via the slow resolver path. Not
a good idea. Also, if foo turns out to be a local function and the
linker replaces the PLT calls with direct calls to foo then gcc has
just wasted a call-saved register.
This patch teaches gcc that the PLT loads are volatile. The change
doesn't affect other loads of function pointers and thus has no effect
on normal indirect function calls. Note that because the
"optimization" this patch prevents can only occur over function calls,
the only place gcc can stash PLT loads is in call-saved registers or
in other memory. I'm reasonably confident that this change will be
neutral or positive for the "ld -z now" case where the PLT is not
volatile, in code where there is any register pressure. Even if gcc
could be taught to recognise cases where the PLT is resolved, you'd
need to discount use of registers to cache PLT loads by some factor
involving the chance that those calls would be converted to direct
calls.
PR target/94145
* config/rs6000/rs6000.c (rs6000_longcall_ref): Use unspec_volatile
for PLT16_LO and PLT_PCREL.
* config/rs6000/rs6000.md (UNSPEC_PLT16_LO, UNSPEC_PLT_PCREL): Remove.
(UNSPECV_PLT16_LO, UNSPECV_PLT_PCREL): Define.
(pltseq_plt16_lo_, pltseq_plt_pcrel): Use unspec_volatile.
With the PR94346 fix in, we can revert the attr-copy-2.C workaround.
2020-03-27 Jakub Jelinek <jakub@redhat.com>
PR c++/94339
* g++.dg/ext/attr-copy-2.C: Revert the last changes.
gcc/c-family/ChangeLog:
PR c++/94346
* c-attribs.c (handle_copy_attribute): Avoid passing expressions
to decl_attributes. Make handling of different kinds of entities
more robust.
gcc/c-c++-common/ChangeLog:
PR c++/94346
* c-c++-common/attr-copy.c: New test.
The iterative addition of 8 and 16 bit vectors has left the mode iterators in a
bit of a mess. Also, the original names were rather verbose leading to
formatting difficulties.
This patch renames all the vector modes such that they are shorter and tidier.
It does not change the output machine description at all.
2020-03-27 Andrew Stubbs <ams@codesourcery.com>
gcc/
* config/gcn/gcn-valu.md:
(VEC_SUBDWORD_MODE): Rename to V_QIHI throughout.
(VEC_1REG_MODE): Delete.
(VEC_1REG_ALT): Delete.
(VEC_ALL1REG_MODE): Rename to V_1REG throughout.
(VEC_1REG_INT_MODE): Delete.
(VEC_ALL1REG_INT_MODE): Rename to V_INT_1REG throughout.
(VEC_ALL1REG_INT_ALT): Rename to V_INT_1REG_ALT throughout.
(VEC_2REG_MODE): Rename to V_2REG throughout.
(VEC_REG_MODE): Rename to V_noHI throughout.
(VEC_ALLREG_MODE): Rename to V_ALL throughout.
(VEC_ALLREG_ALT): Rename to V_ALL_ALT throughout.
(VEC_ALLREG_INT_MODE): Rename to V_INT throughout.
(VEC_INT_MODE): Delete.
(VEC_FP_MODE): Rename to V_FP throughout and move to top.
(VEC_FP_1REG_MODE): Rename to V_FP_1REG throughout and move to top.
(FP_MODE): Delete and replace with FP throughout.
(FP_1REG_MODE): Delete and replace with FP_1REG throughout.
(VCMP_MODE): Rename to V_noQI throughout and move to top.
(VCMP_MODE_INT): Rename to V_INT_noQI throughout and move to top.
* config/gcn/gcn.md (FP): New mode iterator.
(FP_1REG): New mode iterator.
-Wredundant-tags doesn't consider type declarations that are also
the first uses of the type, such as in 'void f (struct S);' and
issues false positives for those. According to the reported that's
making it harder to use the warning to clean up LibreOffice.
The attached patch extends -Wredundant-tags to avoid these false
positives by relying on the same class_decl_loc_t::class2loc mapping
as -Wmismatched-tags. The patch also improves the detection
of both issues in template declarations.
gcc/cp/ChangeLog
2020-03-27 Martin Sebor <msebor@redhat.com>
PR c++/94078
PR c++/93824
PR c++/93810
* cp-tree.h (most_specialized_partial_spec): Declare.
* parser.c (cp_parser_elaborated_type_specifier): Distinguish alias
from declarations.
(specialization_of): New function.
(cp_parser_check_class_key): Move code...
(class_decl_loc_t::add): ...to here. Add parameters. Avoid issuing
-Wredundant-tags on first-time declarations in other declarators.
Correct handling of template specializations.
(class_decl_loc_t::diag_mismatched_tags): Also expect to be called
when -Wredundant-tags is enabled. Use primary template or partial
specialization as the guide for uses of implicit instantiations.
* pt.c (most_specialized_partial_spec): Declare extern.
gcc/testsuite/ChangeLog
2020-03-27 Martin Sebor <msebor@redhat.com>
PR c++/94078
PR c++/93824
PR c++/93810
* g++.dg/warn/Wmismatched-tags-3.C: New test.
* g++.dg/warn/Wmismatched-tags-4.C: New test.
* g++.dg/warn/Wmismatched-tags-5.C: New test.
* g++.dg/warn/Wmismatched-tags-6.C: New test.
* g++.dg/warn/Wredundant-tags-3.C: Remove xfails.
* g++.dg/warn/Wredundant-tags-6.C: New test.
* g++.dg/warn/Wredundant-tags-7.C: New test.
Following DR2061, 'namespace F', looks for 'F's inside inline namespaces.
That can result in ambiguous lookups that we failed to diagnose early enough,
leading us to push a new namespace and ICE later. Diagnose the ambiguity
earlier, and then pick one.
PR c++/94257
* name-lookup.c (push_namespace): Triage ambiguous lookups that
contain namespaces.
Fixes to exploded_path::feasible_p exposed a pre-existing bug
with pointer NULL-ness for pointers to symbolic_region.
symbolic_region has an "m_possibly_null" flag which if set means
that a region_svalue pointing to that region is treated as possibly
NULL. Adding a constraint of "!= NULL" on an edge records that
the pointer is non-NULL, but doesn't affect other pointers (e.g.
if the first if a void *, but the other pointers are cast to other
pointer types). This showed up in the tests
gcc.dg/analyzer/data-model-5b.c and -5c.c, which malloc a buffer
and test for NULL, but then cast that to a struct * and later test
that struct *: a path for the first test being non-NULL and the
second being NULL was erroneously found to be feasible.
This patch clears the m_possibly_null flag when a "!= NULL" constraint
is added, fixing that erroneous path (but not yet fixing the false
positive in the above tests, which seems to go on to hit a different
issue). It also adds the field to dumps.
gcc/analyzer/ChangeLog:
* program-state.cc (selftest::test_program_state_dumping): Update
expected dump to include symbolic_region's possibly_null field.
* region-model.cc (symbolic_region::print_fields): New vfunc
implementation.
(region_model::add_constraint): Clear m_possibly_null from
symbolic_regions now known to be non-NULL.
(selftest::test_malloc_constraints): New selftest.
(selftest::analyzer_region_model_cc_tests): Call it.
* region-model.h (region::dyn_cast_symbolic_region): Add non-const
overload.
(symbolic_region::dyn_cast_symbolic_region): Implement it.
(symbolic_region::print_fields): New vfunc override decl.
gcc/testsuite/ChangeLog:
* gcc.dg/analyzer/data-model-5b.c: Add xfail for new false
positive leak.
* gcc.dg/analyzer/data-model-5c.c: Likewise.
* gcc.dg/analyzer/malloc-5.c: New test.
This patch extends -fdump-analyzer-supergraph so that rather than just
dumping a DUMP_BASE_NAME.supergraph.dot at the start of analysis, it
also dumps a DUMP_BASE_NAME.supergraph-eg.dot at the end.
The new dump file contains a concise dump of the exploded_graph,
organized with respect to the supergraph and its statements. The
exploded nodes are colorized to show sm-state, but no other state
is shown. Per exploded_node saved_diagnostics are also shown,
along with feasibility of the paths to reach them.
I've been finding this a useful way of tracking down issues in
exploded_graphs that are sufficiently large that the output of
-fdump-analyzer-exploded-graph becomes unwieldy.
The patch extends feasiblity-testing so that if the exploded_path
for a saved_diagnostic is found to be infeasible, the reason is
saved and written into the saved_diagnostic, so it can be shown in the
dump. I've found this very useful when tracking down feasibility
issues.
I'm keeping the initial dump file as it's useful when tracking down
ICEs within the analyzer (which would stop the second dump file being
written).
gcc/analyzer/ChangeLog:
* analyzer.h (class feasibility_problem): New forward decl.
* diagnostic-manager.cc (saved_diagnostic::saved_diagnostic):
Initialize new fields m_status, m_epath_length, and m_problem.
(saved_diagnostic::~saved_diagnostic): Delete m_problem.
(dedupe_candidate::dedupe_candidate): Convert "sd" param from a
const ref to a mutable ptr.
(dedupe_winners::add): Convert "sd" param from a const ref to a
mutable ptr. Record the length of the exploded_path. Record the
feasibility/infeasibility of sd into sd, capturing a
feasibility_problem when feasible_p fails, and storing it in sd.
(diagnostic_manager::emit_saved_diagnostics): Update for pass by
ptr rather than by const ref.
* diagnostic-manager.h (class saved_diagnostic): Add new enum
status. Add fields m_status, m_epath_length and m_problem.
(saved_diagnostic::set_feasible): New member function.
(saved_diagnostic::set_infeasible): New member function.
(saved_diagnostic::get_feasibility_problem): New accessor.
(saved_diagnostic::get_status): New accessor.
(saved_diagnostic::set_epath_length): New member function.
(saved_diagnostic::get_epath_length): New accessor.
* engine.cc: Include "gimple-pretty-print.h".
(exploded_path::feasible_p): Add OUT param and, if non-NULL, write
a new feasibility_problem to it on failure.
(viz_callgraph_node::dump_dot): Convert begin_tr calls to
begin_trtd. Convert end_tr calls to end_tdtr.
(class exploded_graph_annotator): New subclass of dot_annotator.
(impl_run_checkers): Add a second -fdump-analyzer-supergraph dump
after the analysis runs, using exploded_graph_annotator. dumping
to DUMP_BASE_NAME.supergraph-eg.dot.
* exploded-graph.h (exploded_node::get_dot_fillcolor): Make
public.
(exploded_path::feasible_p): Add OUT param.
(class feasibility_problem): New class.
* state-purge.cc (state_purge_annotator::add_node_annotations):
Return a bool, add a "within_table" param.
(print_vec_of_names): Convert begin_tr calls to begin_trtd.
Convert end_tr calls to end_tdtr.
(state_purge_annotator::add_stmt_annotations): Add "within_row"
param.
* state-purge.h ((state_purge_annotator::add_node_annotations):
Return a bool, add a "within_table" param.
(state_purge_annotator::add_stmt_annotations): Add "within_row"
param.
* supergraph.cc (supernode::dump_dot): Call add_node_annotations
twice: as before, passing false for "within_table", then again
with true when within the TABLE element. Convert some begin_tr
calls to begin_trtd, and some end_tr calls to end_tdtr.
Repeat each add_stmt_annotations call, distinguishing between
calls that add TRs and those that add TDs to an existing TR.
Add a call to add_after_node_annotations.
* supergraph.h (dot_annotator::add_node_annotations): Add a
"within_table" param.
(dot_annotator::add_stmt_annotations): Add a "within_row" param.
(dot_annotator::add_after_node_annotations): New vfunc.
gcc/ChangeLog:
* doc/invoke.texi (-fdump-analyzer-supergraph): Document that this
now emits two .dot files.
* graphviz.cc (graphviz_out::begin_tr): Only emit a TR, not a TD.
(graphviz_out::end_tr): Only close a TR, not a TD.
(graphviz_out::begin_td): New.
(graphviz_out::end_td): New.
(graphviz_out::begin_trtd): New, replacing the old implementation
of graphviz_out::begin_tr.
(graphviz_out::end_tdtr): New, replacing the old implementation
of graphviz_out::end_tr.
* graphviz.h (graphviz_out::begin_td): New decl.
(graphviz_out::end_td): New decl.
(graphviz_out::begin_trtd): New decl.
(graphviz_out::end_tdtr): New decl.
gcc/testsuite/ChangeLog:
* gcc.dg/analyzer/dot-output.c: Check that
dot-output.c.supergraph-eg.dot is valid.
gcc/analyzer/ChangeLog:
* diagnostic-manager.cc (dedupe_winners::add): Show the
exploded_node index in the log messages.
(diagnostic_manager::emit_saved_diagnostics): Log a summary of
m_saved_diagnostics at entry.
This avoids completing types for DINFO_LEVEL_TERSE by using
the should_emit_struct_debug machinery.
2020-03-27 Richard Biener <rguenther@suse.de>
PR debug/94273
* dwarf2out.c (should_emit_struct_debug): Return false for
DINFO_LEVEL_TERSE.
* g++.dg/debug/pr94273.C: New testcase.
