Since the addition of DCE after complex lowering, reassoc1 has no
longer been impeded by COMPLEX_EXPRs, so it changes (a - (b + c)),
where (b + c) is the imaginary part of an intermediate complex value,
to ((a - b) - c), and this prevents the vectorizer from recognizing
the expected COMPLEX_ADD_ROT270 pattern in fms_elemconjsnd.
Stop expecting that pattern to be found on that test.
David Malcolm [Sat, 31 Jan 2026 04:27:40 +0000 (23:27 -0500)]
analyzer: fix ICE on rethrow without throw [PR123880]
gcc/analyzer/ChangeLog:
PR analyzer/123880
* engine.cc (throw_custom_edge::update_model): Gracefully handle
not knowing the current exception in a rethrow.
(exploded_node::on_throw): Likewise.
gcc/testsuite/ChangeLog:
PR analyzer/123880
* g++.dg/analyzer/exception-rethrow-3.C: New test.
Signed-off-by: David Malcolm <dmalcolm@redhat.com>
Jason Merrill [Wed, 28 Jan 2026 06:17:13 +0000 (14:17 +0800)]
c++: defer DECL_ONE_ONLY vs consteval-only [PR122785]
The failure in the given PR occurs because when setting up an
imported vague-linkage variable, we currently call 'maybe_commonize_var'
which for -freflection checks 'consteval_only_p'. Unfortunately this
latter function needs to call 'complete_type_p' which can perform
recursive loading of the (possibly yet-to-be-streamed) class type,
breaking modules assumptions.
If we just remove the consteval_only_p early exit from maybe_commonize_var,
we end up crashing at EOF while trying to mangle its comdat group, so we
need to undo maybe_commonize_var along with setting DECL_EXTERN.
Pietro Monteiro [Sat, 31 Jan 2026 01:22:49 +0000 (20:22 -0500)]
libitm: testsuite: don't include libstdc++ paths in test names [PR69018]
Libitm C++ tests pass -B/full/path/to/build/libstc++/src/.libs on the
options argument to dg-runtest and the libstdc++ include paths as
the default-extra-options. This causes 2 problems:
1) If a test uses `dg-options' then the libstdc++ include paths are
not passed to the compiler.
2) The `-B/full/path/to/build/libstc++/src/.libs' gets added to the
test name. This makes comparing build done in different directories
harder because the hardcoded full path.
Stop passing options and default-extra-options to dg-runtest and move
dealing with the path-releated flags to libitm_target_compile in
libitm.exp, where they are added to additional_flags.
Also change the FSF address to the website in the license text.
libitm/ChangeLog:
PR libitm/69018
* testsuite/lib/libitm.exp (libitm_target_compile): Add
lang_include_flags and `-B${blddir}/${lang_library_paths}' to
additional_flags.
* testsuite/libitm.c++/c++.exp: Set lang_library_paths and lang_include_flags.
Signed-off-by: Pietro Monteiro <pietro@sociotechnical.xyz>
Jerry DeLisle [Fri, 30 Jan 2026 21:25:23 +0000 (13:25 -0800)]
Fortran: Fix PR123868
When copying derived types with allocatable array components where the
array element type also has allocatable components, the condition at
line 11071 was incorrectly triggering a call to gfc_duplicate_allocatable.
However, for allocatable arrays with nested allocatables (where
cmp_has_alloc_comps && c->as is true), the add_when_allocated code
already includes a gfc_duplicate_allocatable call (generated by the
recursive structure_alloc_comps call at lines 10290-10293).
This caused the outer array to be allocated twice: first by the explicit
gfc_duplicate_allocatable call at line 11099, and then again by the
gfc_duplicate_allocatable embedded in add_when_allocated. The first
allocation was leaked when the second allocation overwrote the data
pointer.
PR121628 added "add_when_allocated != NULL_TREE ||" to the condition,
which was redundant for scalars (already handled by !c->as) and wrong
for arrays (caused double allocation). Simply removing this clause
restores the correct pre-PR121628 behavior.
PR fortran/123868
gcc/fortran/ChangeLog:
* trans-array.cc (structure_alloc_comps): For COPY_ALLOC_COMP,
remove the add_when_allocated != NULL_TREE clause that PR121628
added. This clause was redundant for scalars and caused double
allocation for arrays with nested allocatable components.
gcc/testsuite/ChangeLog:
* gfortran.dg/array_memcpy_2.f90: Update expected memcpy count
from 4 to 3, as the double allocation bug is now fixed.
* gfortran.dg/pr123868.f90: New test.
Signed-off-by: Christopher Albert <albert@alumni.tugraz.at>
Patrick Palka [Fri, 30 Jan 2026 20:25:43 +0000 (15:25 -0500)]
c++: non-empty constexpr constructor bodies in C++11 [PR123845]
This patch makes us support C++14 non-empty constexpr constructor bodies
in C++11, as an extension. This will make it trivial to safely fix the
C++11 library regression PR114865 that requires us to do
__builtin_clear_padding after initializing _M_i in std::atomic's
single-parameter constructor, and that's not really possible with the
C++11 constexpr restrictions.
Since we lower member initializers to constructor body statements
internally, and so constructor bodies are already effectively non-empty
internally even in C++11, supporting non-empty bodies in user code is
mostly a matter of relaxing the parse-time error.
But constexpr-ex3.C revealed that by accepting the non-empty body of A's
constructor, build_data_member_initialization goes on to mistake the
'i = _i' assignment as a member initializer, and we incorrectly accept
the constructor in C++11 mode (even though omitting mem-inits is only
valid since C++20). Turns out this is caused by that function
recognizing MODIFY_EXPR only in C++11 mode, logic that was last changed
by r5-5013 (presumably to limit impact of the patch at the time) but I
reckon could just be removed outright. This should be safe because the
result of build_data_member_initialization is only used by
cx_check_missing_mem_inits for validation; evaluation is in terms of
the entire lowered constructor body.
PR c++/123845
PR libstdc++/114865
gcc/cp/ChangeLog:
* constexpr.cc (build_data_member_initialization): Remove
C++11-specific recognition of MODIFY_EXPR.
(check_constexpr_ctor_body): Relax error diagnostic to a
pedwarn and don't clear DECL_DECLARED_CONSTEXPR_P upon
error. Return true if complaining.
gcc/testsuite/ChangeLog:
* g++.dg/cpp0x/constexpr-ex3.C: Adjust C++11 non-empty
constexpr constructor dg-error to a dg-warning. Expect
a follow-up missing member initializer diagnostic in C++11 mode.
* g++.dg/cpp2a/constexpr-try1.C: Expect a follow-up
compound-statement in constexpr function diagnostic in C++11
mode.
* g++.dg/cpp2a/constexpr-try2.C: Likewise. Adjust C++11
non-empty constexpr constructor dg-error to a dg-warning.
* g++.dg/cpp2a/constexpr-try3.C: Adjust C++11 non-empty
constexpr constructor dg-error to a dg-warning.
* g++.dg/cpp0x/constexpr-ctor23.C: New test.
The following properly checks expr_no_side_effects_p on two patterns
that turns a conditionally evaluated operand into unconditonal.
PR middle-end/123887
* match.pd ((zero_one ==/!= 0) ? .. z <op> y .. -> zero_one * z ..):
Check evaluating z unconditionally has no side-effects, like
trapping.
Richard Biener [Fri, 30 Jan 2026 13:55:58 +0000 (14:55 +0100)]
debug/123886 - remove GTY((skip)) from external_die_map entry
The following avoids GCing IDENTIFIER_POINTERs referenced from
external_die_map after we clear ggc_protect_identifiers. I never
expected those to be GCed.
Patrick Palka [Fri, 30 Jan 2026 18:16:20 +0000 (13:16 -0500)]
libstdc++/regex: add [[gnu::always_inline]] to _Executor::_M_node
The compiler understandably doesn't know that _M_node only ever has a
single call site, _M_dfs, (and is not directly called from other library
headers or user code) and so decides not to inline it. So use the
always_inline attribute to force the inlining. This seems sufficient to
make all _M_dfs subroutines get inlined away, and speeds up the executor
by 30% on some microbenchmarks.
Gary Dismukes [Fri, 30 Jan 2026 16:55:16 +0000 (17:55 +0100)]
Ada: Fix regression in visibility for nested use_type_clauses
An earlier fix for wrong handling of visibility when a use_type_clause
with "all" occurs within a nested scope when there's a use_type_clause
in an enclosing scope without "all" was incorrect and led to primitives
being visible in a later nested scope that has a use_type_clause with
"all" for the same type. We now properly restore Current_Use_Clause
to refer to the outer use_type_clause when ending the scope of nested
use_type_clauses.
gcc/ada/
* sem_ch8.adb (End_Use_Type): Remove test of Prev_Use_Clause as
as a condition for resetting In_Use and Current_Use_Clause, and
change setting of that flag and field based on Prev_Use_Clause
rather than setting them to Empty. Revise preceding comment.
Patrick Palka [Fri, 30 Jan 2026 17:29:17 +0000 (12:29 -0500)]
libstdc++/regex: Defer _M_current restoration to backtracking nodes [PR86164]
The incrementing of the current input string position (_M_current) is
done by _M_handle_match, which also makes sure to restore it afterwards,
via a restore_current frame. But restoring _M_current is naturally only
necessary when backtracking is involved, not after every single match.
So this patch moves the responsibility of saving/restoring _M_current
from _M_handle_match to the branching nodes _M_handle_alternative and
_M_handle_repeat. This is done by storing _M_current within the
fallback_next, fallback_rep_once_more and posix_alternative frames.
In turn we can get rid of the now unused restore_current frame kind.
This reduces the maximum size of the _M_frames stack by 15% for
regex_match(string(200000, 'a'), "(a|b|c)*")
PR libstdc++/86164
libstdc++-v3/ChangeLog:
* include/bits/regex_executor.tcc (__detail::_ExecutorFrameOpcode):
Remove _S_fopcode_restore_current.
(__detail::_Executor::_M_handle_repeat): Pass _M_current when
pushing a fallback_next or fallback_rep_once_more frame.
(__detail::_Executor::_M_handle_match): Don't push a
restore_current frame.
(__detail::_Executor::_M_handle_backref): Likewise and simplify
accordingly.
(__detail::_Executor::_M_handle_alternative): Pass _M_current when
pushing a fallback_next or posix_alternative frame.
(__detail::_Executor::_M_dfs) <case _S_fopcode_fallback_next>:
Restore _M_current.
<case _S_fopcode_fallback_rep_once_more>: Likewise.
<case _S_fopcode_posix_alternative>: Likewise.
<case _S_fopcode_restore_current>: Remove.
Reviewed-by: Tomasz Kamiński <tkaminsk@redhat.com> Reviewed-by: Jonathan Wakely <jwakely@redhat.com>
Patrick Palka [Fri, 30 Jan 2026 17:23:40 +0000 (12:23 -0500)]
libstdc++/regex: Make DFS executor non-recursive [PR86164]
This patch replaces the recursive implementation of the DFS executor
with an iterative one using an explicit heap-based stack. System
stack usage of the executor is now constant with respect to input size
rather than linear, avoding stack overflow errors when processing
long inputs.
PR libstdc++/86164
libstdc++-v3/ChangeLog:
* include/bits/regex.h (__detail::_Executor): Use inline
namespace _V2.
* include/bits/regex_executor.h (__detail::_ExecutorFrame):
Declare.
(__detail::_Executor): Use inline namespace _V2.
(__detail::_Executor::_M_node): Declare.
(__detail::_Executor::_M_frames): New data member.
* include/bits/regex_executor.tcc (__detail::_ExecutorFrameOpcode):
New.
(__detail::_ExecutorFrameBase): New.
(__detail::_ExecutorFrame): New.
(__detail::_Executor): Use inline namespace _V2.
(__detail::_Executor::_M_rep_once_more): Replace recursive
_M_dfs calls with an _S_opcode_next frame push, and any work
after such calls with an appropriate frame push.
(__detail::_M_handle_repeat): Likewise.
(__detail::_M_handle_subexpr_begin): Likewise.
(__detail::_M_handle_subexpr_end): Likewise.
(__detail::_M_handle_line_begin_assertion): Likewise.
(__detail::_M_handle_line_end_assertion): Likewise.
(__detail::_M_handle_word_boundary): Likewise.
(__detail::_M_handle_subexpr_lookahead): Likewise.
(__detail::_M_handle_match): Likewise.
(__detail::_M_handle_backref): Likewise.
(__detail::_M_handle_accept): Likewise.
(__detail::_M_handle_alternative): Likewise.
(__detail::_M_node): Factored out from _M_dfs.
(__detail::_M_dfs): Push an initial frame to _M_frames that
visits the starting node and pass this stack each subroutine.
Pop the latest _ExecutorFrame from _M_frames and handle
appropriately according to its _ExecutorFrameOpcode. Loop until
_M_frames is empty.
Reviewed-by: Tomasz Kamiński <tkaminsk@redhat.com> Reviewed-by: Jonathan Wakely <jwakely@redhat.com>
Marek Polacek [Wed, 28 Jan 2026 22:05:46 +0000 (17:05 -0500)]
c++/reflection: check TYPE_BEING_DEFINED in define_aggregate
As discussed in
<https://gcc.gnu.org/pipermail/gcc-patches/2026-January/705756.html>,
we should check TYPE_BEING_DEFINED along with COMPLETE_TYPE_P in
eval_define_aggregate. It seems that with this check added, we don't
need this code anymore:
if (c == type)
{
auto_diagnostic_group d;
error_at (loc, "%<define_aggregate%> evaluated from "
"%<consteval%> block enclosed by %qT being "
"defined", type);
//...
}
so I'm removing that in this patch.
gcc/cp/ChangeLog:
* reflect.cc (eval_define_aggregate): Also give an error when
TYPE_BEING_DEFINED is true for the first argument. Remove code
that did the same.
Marek Polacek [Wed, 28 Jan 2026 21:33:34 +0000 (16:33 -0500)]
c++/reflection: tweak eval_can_substitute
As discussed in
<https://gcc.gnu.org/pipermail/gcc-patches/2026-January/705756.html>,
we should check for undeduced_auto_decl after performing
resolve_nondeduced_context_or_error. Also add a test to exercise
the new check.
gcc/cp/ChangeLog:
* reflect.cc (eval_can_substitute): Check undeduced_auto_decl after
resolve_nondeduced_context_or_error.
gcc/testsuite/ChangeLog:
* g++.dg/reflect/substitute3.C: New test.
Reviewed-by: Jason Merrill <jason@redhat.com> Reviewed-by: Jakub Jelinek <jakub@redhat.com>
Eric Botcazou [Fri, 30 Jan 2026 10:58:58 +0000 (11:58 +0100)]
Ada: Fix spurious visibility error from limited_with clause in hierarchy
The problem is that the compiler installs the limited view of a package that
is already installed by the virtue of being an ancestor of the main unit.
gcc/ada/
PR ada/123867
* sem_ch10.adb (Analyze_Compilation_Unit): Output info message
when -gnatdi is specified.
(Install_Parents): Likewise. Set the Is_Visible_Lib_Unit flag
on the unit.
(Install_Private_With_Clauses): Do not output info message here.
(Remove_Parents): Output info message when -gnatdi is specified
and clear the Is_Visible_Lib_Unit flag on the unit.
gcc/testsuite/
* gnat.dg/specs/limited_with3.ads: New test.
* gnat.dg/specs/limited_with3-child.ads: New helper.
* gnat.dg/specs/limited_with3-child-grandchild.ads: Likewise.
* gnat.dg/specs/limited_with3-child-grandchild-grandgrandchild.ads:
Likewise.
The following fixes a link failure due to missing skeleton
ancestors.
PR debug/110885
* dwarf2out.cc (generate_skeleton_bottom_up): Generate the
skeleton ancestor tree when moving a new child to the parent
even for template instantiations.
Jakub Jelinek [Fri, 30 Jan 2026 10:43:49 +0000 (11:43 +0100)]
libatomic: Don't install libatomic_asneeded.{so,a} if corresponding libatomic.{so,a} wasn't installed
Filip complained on IRC that libatomic_asneeded.a is installed as
stale symlink in --disable-static build.
The following patch makes sure to install the libatomic_asneeded.{so,a}
script and/or symlink only if the corresponding libatomic.{so,a} has been
installed.
2026-01-30 Jakub Jelinek <jakub@redhat.com>
* Makefile.am (all-local, install-asneeded): Only create
libatomic_asneeded.so script if libatomic.so exist, only
create libatomic_asneeded.a symlink if libatomic.a exist.
* Makefile.in: Regenerate.
Jakub Jelinek [Fri, 30 Jan 2026 10:42:41 +0000 (11:42 +0100)]
match.pd: Fix up __imag__ .MUL_OVERFLOW simplification [PR123864]
The integer_minus_onep case is really meant for multiplication by
-1, not by (unsigned char)0xff or (unsigned short)0xffff or 0xffffffffU etc.
We've already tested that the first operand is signed (otherwise the
earlier case is used) and also that int_fits_type_p (@1, TREE_TYPE (@0)),
but if @0 is signed and @1 is unsigned all ones with smaller precision
than that, it fits into @0's type, integer_minus_onep will be true
and it still should be handled through the ranges, not as @0 == min.
2026-01-30 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/123864
* match.pd (__builtin_mul_overflow_p (x, cst, (stype) 0) ->
x > stype_max / cst || x < stype_min / cst): Only check
integer_minus_onep for signed types.
* gcc.c-torture/execute/pr123864.c: New test.
* gcc.dg/torture/pr123864.c: New test.
Jakub Jelinek [Fri, 30 Jan 2026 10:40:17 +0000 (11:40 +0100)]
fold-const: Fix implicit cast to nullptr_t typedef [PR123790]
The following testcase is incorrectly rejected, because since r15-6744
it adds build_nop to preserve the exact typedef type and
cxx_eval_constant_expression can't fold NOP_EXPR from integer_zerop of one
NULLPTR_TYPE to another NULLPTR_TYPE, while cxx_eval_constant_expression
relies on fold to do such folding.
I see 3 options to fix that, one is deal with this in the r15-6744 spot
and special case NULLPTR_TYPE there and build_zero_cst instead of build_nop
(with similar condition like in the patch below), another possibility is
special case this in cxx_eval_constant_expression, and
another one (done in this patch) is to handle this in fold-const.cc -
fold_convert_loc and also in fold_convert_const.
2026-01-30 Jakub Jelinek <jakub@redhat.com>
PR c++/123790
* fold-const.cc (fold_convert_const): Handle conversion of
integer_zerop to NULLPTR_TYPE.
(fold_convert_loc): Likewise.
Jørgen Kvalsvik [Wed, 28 Jan 2026 19:33:21 +0000 (20:33 +0100)]
Permit const counters in flush_on_edges [PR123855]
Relax the (accidental) requirement and permit that function-local
counters don't have SSA names. We really only look up the def
statement to check if it is a phi node, in which case we need to
resolve the counter from there. This obviously doesn't apply when it
is a constant.
PR gcov-profile/123855
gcc/ChangeLog:
* path-coverage.cc (flush_on_edges): Only look up SSA name def
stmt when counter is non-const.
Jørgen Kvalsvik [Mon, 3 Nov 2025 23:05:04 +0000 (00:05 +0100)]
gcov-dump: Print PATHS tag
Print the record associated with the PATHS tag. While it just prints
the number of prime paths for the function, this is more useful than
the UNKNOWN it would otherwise print.
Jørgen Kvalsvik [Mon, 3 Nov 2025 22:48:25 +0000 (23:48 +0100)]
gcc: -f*-coverage implies -ftest-coverage
The --coverage flag is a useful alias for -fprofile-arcs
-ftest-coverage where the latter makes gcc output the .gcno file.
While it makes sense to use -fprofile-arcs without a .gcno file (pgo,
maybe more), it never really makes sense to request -fpath-coverage or
-fcondition-coverage without also wanting the .gcno to interpret the
results.
gcc/ChangeLog:
* common.opt: ftest-coverage enabled by fpath-coverage or
fcondition-coverage
That patch adds discriminators to edge->goto_locus which will map to
the right source location, but breaks the equality comparison used in
the hash set.
Jakub Jelinek [Fri, 30 Jan 2026 09:56:10 +0000 (10:56 +0100)]
libcpp: Fix up comment handling in -fdirectives-only preprocessing [PR123273]
Back in 2020 Nathan rewrote the -E -fdirectives-only preprocessing.
In PR103130 a year and half later I've fixed the handling of comments
so that /* \*/ is considered as full comment even when * is escaped,
to match the normal preprocessing.
The following testcases shows further bugs in the comment handling.
One is that /* *\/ should not be considered as full comment (i.e.
when the / after * is escaped). And another one is that the code
was treating any number of backslashes as escape, which is wrong,
only a single backslash is an escape, two backslashes preprocess as
one backslash, three as one backslash and one escape, etc.
So, while /* *\
/ is a full comment, /* *\\
/ or /* *\\\\\\\\\\\\\
/ is not.
2026-01-30 Jakub Jelinek <jakub@redhat.com>
PR preprocessor/123273
* lex.cc (cpp_directive_only_process): Only go to done_comment
for '/' if star is true and esc is false. When seeing '\\' with
esc set to true, clear esc as well as star instead of keeping esc
set.
* c-c++-common/cpp/dir-only-10.c: New test.
* c-c++-common/cpp/dir-only-11.c: New test.
the problem being that nonstreaming_only expects an instance of
aarch64_feature_flags which (statically) describes the required
extensions for the intrinsics. Instead, we were passing TARGET_SIMD,
which computes a boolean telling us dynamically whether the simd feature
is currently available.
This patch fixes the issue by simply doing
s/TARGET_SIMD/AARCH64_FL_SIMD/ in that file. We also add a simple test
which checks that we now diagnose the missing extension instead of
ICEing when compiling with +nosimd.
As an additional conservative hardening step (to prevent a similar issue
from re-occurring), this patch adjusts the aarch64_pragma_builtins table
to make it constexpr. This makes the bug a compile-time error.
The next patch in the series adjusts the ctor of bbitmap to make it
explicit (and deals with the fallout), this patch however is
deliberately a minimal fix which is suitable for backporting.
gcc/ChangeLog:
PR target/123206
* config/aarch64/aarch64-builtins.cc (struct aarch64_pragma_builtins_data):
Declare array as CONSTEXPR.
* config/aarch64/aarch64-simd-pragma-builtins.def: Update
incorrect uses of TARGET_SIMD to use AARCH64_FL_SIMD instead.
gcc/testsuite/ChangeLog:
PR target/123206
* gcc.target/aarch64/pr123206.c: New test.
Iain Buclaw [Fri, 30 Jan 2026 08:06:39 +0000 (09:06 +0100)]
d: Refactor for changes to dmd front-end inteface
There are a number of interface changes being made to the dmd front-end
interface. This makes any necessary refactorings ahead of the merge to
reduce the size of the diff.
gcc/d/ChangeLog:
* decl.cc (DeclVisitor::visit (EnumDeclaration *)): Treat sinit member
field as a generic pointer.
(enum_initializer_decl): Likewise.
* expr.cc (ExprVisitor::visit (ArrayLiteralExp *)): Compute static
array length separately from creating type.
* modules.cc (struct module_info): Add ctor_decl, dtor_decl,
sharedctor_decl, shareddtor_decl, standalonector_decl, and
unittest_decl.
(layout_moduleinfo_fields): Add mi argument. Use it to check whether
module helpers have been generated.
(layout_moduleinfo): Likewise.
(build_module_tree): Cache generated module helpers in module_info.
* typeinfo.cc (cpp_type_info_ptrs): New variable.
(get_cpp_typeinfo_decl): Cache generated C++ type_info references in
cpp_type_info_ptrs.
* types.cc (TypeVisitor::visit (TypeEnum *)): Separate getting
front-end member value from building its CST tree.
Iain Buclaw [Fri, 30 Jan 2026 07:14:34 +0000 (08:14 +0100)]
d: Fix ICE in dwarf2out_abstract_function, at dwarf2out.cc:23771 [PR123263]
Emit artificial functions as being part of the module context, so that
they are unaffected by dwarf early_finish pass removing the parent type
that they were generated for.
PR d/123263
gcc/d/ChangeLog:
* d-codegen.cc (d_decl_context): Set DECL_CONTEXT of compiler
generated functions to that of parent module.
Richard Biener [Thu, 29 Jan 2026 14:43:01 +0000 (15:43 +0100)]
tree-optimization/109410 - ICE with noreturn and -g
This fixes the -g variant of the original testcase.
PR tree-optimization/109410
* tree-ssa-reassoc.cc (build_and_add_sum): Use
gsi_start_nondebug_after_labels_bb to look for a possible
returns-twice call.
Andrew Pinski [Wed, 28 Jan 2026 20:32:12 +0000 (12:32 -0800)]
ifcvt: Improve noce_can_force_operand in ifcvt [PR122170]
Currently if the rtl is either an arithmetic or an unary
rtl, noce_can_force_operand checks to see if there is an
optab for that rtl code. This works for more things except
on some targets they have a ss_minus instruction but don't
implement the optab for it. In the case of arm you can
generate a ss_minus with a builtin and then when it comes
to trying to do ifcvt, force_operand fails over.
In this case the optab, sssub was only supported for
fixed-point modes before and it was working as code_to_optab
would return there was not optabs. But after r15-1030-gabe6d39365476e,
the optab will be return. What the backend is doing is correct and will
most likely happen with other rtl codes/optabs later on.
To fix this instead of just returning true if the optab exists, we need
to check if the optab entry for the mode exists.
PR rtl-optimization/122170
gcc/ChangeLog:
* ifcvt.cc (noce_can_force_operand): Don't only check if
there is an optab for the code check the entry for the
mode is non-null. Handle non integral div by checking
optab like force_operand does.
Signed-off-by: Andrew Pinski <andrew.pinski@oss.qualcomm.com>
Lulu Cheng [Tue, 27 Jan 2026 02:31:36 +0000 (10:31 +0800)]
LoongArch: Fix bug123766.
The pointer parameter type for the original store class builtin
functions is CVPOINTER (const volatile void *).
Taking the following test as an example:
```
v4i64 v = {0, 0, 0, 0};
void try_store() {
long r[4];
__lasx_xvst(v, r, 0);
}
```
At this point, the type of r is CVPOINTER, which means data in memory
can only be read through r. Therefore, if the array r is not
initialized, an uninitialized warning will be issued.
This patch changes the pointer type of store-class builtin functions
from CVPOINTER to VPOINTER (volatile void *).
PR target/123766
gcc/ChangeLog:
* config/loongarch/loongarch-builtins.cc
(loongarch_build_vpointer_type): New function. Return a type
for 'volatile void *'.
(LARCH_ATYPE_VPOINTER): New macro.
* config/loongarch/loongarch-ftypes.def: Change the pointer
type of the store class function from CVPOINTER to VPOINTER.
gcc/testsuite/ChangeLog:
* gcc.target/loongarch/vector/lasx/pr123766.c: New test.
* gcc.target/loongarch/vector/lsx/pr123766.c: New test.
Lulu Cheng [Mon, 26 Jan 2026 03:38:59 +0000 (11:38 +0800)]
LoongArch: Fix bug123807.
In the function loongarch_expand_vector_init_same, if same is MEM and
the mode of same does not match imode, it will cause an ICE
in force_reg (imode, same).
PR target/123807
gcc/ChangeLog:
* config/loongarch/loongarch.cc
(loongarch_expand_vector_init_same): When same is MEM and
GET_MODE(same) != imode, first load the data from memory
and then process it further.
gcc/testsuite/ChangeLog:
* gcc.target/loongarch/vector/lsx/pr123807.c: New test.
mengqinggang [Fri, 23 Jan 2026 07:00:58 +0000 (15:00 +0800)]
LoongArch: Fix movsf in lp64s abi
When adding LoongArch32 ilp32s abi support, add TARGET_HARD_FLOAT condition for
movsf to prevent matching in target without FPU.
But movsf also needs to be used in lp64s abi, it can expand to some
non float instructions.
Delete TARGET_HARD_FLOAT condition.
Andrew Pinski [Thu, 29 Jan 2026 00:50:52 +0000 (16:50 -0800)]
optabs: Fix expansion of abs and neg for Float16 [PR123869]
The problem here is we try to use the widening type before
doing the bitwise expansion of neg/and for floating point types.
This moves the code around to try the bitwise expansion first.
Note this mostly matters for NaNs where you widening (promotion)
would cause a NaN to be slightly different when doing the rounding
back.
Bootstrapped and tested on x86_64-linux-gnu.
PR middle-end/123869
gcc/ChangeLog:
* optabs.cc (expand_unop): Move the NEG optab
handling before the widening code.
Move the ABS bitwise expansion from expand_abs_nojump
to before the widening code.
(expand_abs_nojump): Remove the bitwise expansion trial
since expand_unop is called right above.
Signed-off-by: Andrew Pinski <andrew.pinski@oss.qualcomm.com>
Jonathan Wakely [Wed, 28 Jan 2026 12:33:46 +0000 (12:33 +0000)]
libstdc++: Make std::expected trivially copy/move assignable (LWG 4026)
This is the subject of two NB comments on C++26 which seem likely to be
approved. We're allowed to make this change as QoI anyway, even if it
isn't approved for the standard, and it should apply to C++23 as well to
avoid ABI changes between C++23 and C++26.
As shown in the updates to the test, defaulted special members can have
noexcept(false) even if they would be noexcept(true) by default. The new
defaulted operator= overloads added by this commit have conditional
noexcept-specifiers that match the conditions of the non-trivial
assignments, propagating any noexcept(false) on trivial special members
of the T and E types. We could strengthen the noexcept for the trivial
operators, but propagating the conditions from the underlying types is
probably what users expect, if they've bothered to put noexcept(false)
on their defaulted special members.
libstdc++-v3/ChangeLog:
* include/std/expected (__expected::__trivially_replaceable)
(__expected::__usable_for_assign)
(__expected::__usable_for_trivial_assign)
(__expected::__can_reassign_type): New concepts.
(expected::operator=): Adjust constraints
on existing overloads and add defaulted overload.
(expected<cv void, E>::operator=): Likewise.
* testsuite/20_util/expected/requirements.cc: Check for trivial
and nothrow properties of assignments.
Andrew Pinski [Wed, 28 Jan 2026 20:16:43 +0000 (12:16 -0800)]
testsuite: Move pr116353.c from being x86 specific to being generic
While working on ifcvt's noce_can_force_operand, I noticed that
the testcase pr116353.c was added as x86 specific testcase but
it has nothing in it that is x86 specific. It is currently
compiled at -O2 but there is no reason why it can't be compiled
at any other optimization level.
So this moves the testcase to be a torture testcase.
The original issue showed up on x86 with respect to ifcvt
where there was a vec_select which does not have an optab for it
but it might show up on other targets too.
Roger Sayle [Thu, 29 Jan 2026 18:48:14 +0000 (18:48 +0000)]
Allow CONCATs in emit_group_load_1.
This one line patch is a pre-requisite to a solution to PR target/123506;
an effort to improve middle-end code for returning structures. Currently,
emit_group_load_1 contains code to handle a CONCAT returned by force_reg.
This tweak avoids the call for force_reg if the source is already a CONCAT,
which allows this subroutine to be re-used by target-specific implementations
of emit_group_load. Many thanks to Jeff Law for cross-platform testing.
2026-01-29 Roger Sayle <roger@nextmovesoftware.com>
gcc/ChangeLog
* expr.cc (emit_group_load_1): Don't call force_reg if orig_src
is already a CONCAT.
Eric Botcazou [Thu, 29 Jan 2026 15:16:08 +0000 (16:16 +0100)]
Ada: Fix internal error on equality test with empty container
This is a regression present on the mainline and 15 branch, although the
root cause has been present for years: the Sem_Type.Covers predicate
returns true for the type of an aggregate (Any_Composite) and any type
declared with the Aggregate aspect (when invoked in Ada 2022 or later),
but its companion function Sem_Type.Specific_Type punts when it is called
on the same combination.
gcc/ada/
PR ada/123861
* sem_type.adb (Covers): Fix couple of typos in comment.
(Specific_Type): Adjust to Covers' handling of types declared
with the Aggregate aspect in Ada 2022.
gcc/testsuite/
* gnat.dg/specs/aggr11.ads: New test.
Richard Earnshaw [Thu, 29 Jan 2026 10:25:41 +0000 (10:25 +0000)]
testsuite: arm: rework some target-supports checks to use -mfpu=auto
Several target-supports checks for Arm are still using the antiquated
-mfpu=... setting rather than picking up FPU extensions via the
architecture specification. This causes problems when tests are
layered because they do not consistently override each other.
Arguably, that is incorrect anyway: you can't test two sets of flag
combinations independently and then expect to be able to apply both
of them, but this change at least makes this behave in a reasonable
way provided the second set of flags fully overrides the first.
gcc/testsuite/ChangeLog:
* lib/target-supports.exp:
(check_effective_target_arm_v8_3a_complex_neon_ok_nocache):
Split and fill in arm and aarch64 compile options. Remove the
cpu_unset variable.
(check_effective_target_arm_v8_2a_fp16_neon_ok_nocache): Likewise.
(check_effective_target_arm_v8_3a_fp16_complex_neon_ok_nocache):
Likewise.
(check_effective_target_arm_neon_ok_nocache): Rework to use
-mfpu=auto.
(check_effective_target_arm_neon_fp16_ok_nocache): Likewise.
Artemiy Volkov [Thu, 8 Jan 2026 11:51:06 +0000 (11:51 +0000)]
testsuite: require arm_v8_3a_fp16_complex_neon for complex fp16 tests
Changes to target-supports.exp from r16-6519-g56dd47c073cabf introduced
some new failures for complex fp16 tests with some arm-eabi targets
configured with -mfpu=auto. This patch fixes those by (a) requiring the
full arm_v8_3a_fp16_complex_neon_ok target instead of just float16 in
vect/complex fp16 tests, and (b) simplifying down to one
dg-require-effective-target and one dg-add-options and removing the
explicit -march in advsimd-intrinsics/vector-complex_f16.c.
Regtested on aarch64, and the Linaro CI also delivers a positive verdict
for the relevant cortex-m33/-m3/-m55/-m7 configurations (after the
accompanying patch to target-supports.exp that follows). Huge thanks to
Christophe for testing this on arm.
vspefs [Thu, 29 Jan 2026 16:58:19 +0000 (16:58 +0000)]
libstdc++: fix a wrong export of a contracts facility in std module
This patch fixes a wrong export name in std module. std module currently
exports std::contracts::invoke_default_violation_handler, which is wrong.
The correct name is std::contracts::invoke_default_contract_violation_handler.
libstdc++-v3/ChangeLog:
* src/c++23/std.cc.in (invoke_default_violation_handler): Change
to invoke_default_contract_violation_handler.
Sandra Loosemore [Thu, 29 Jan 2026 16:41:07 +0000 (16:41 +0000)]
doc: Add "user experience" documentation for options
After working through PR122243, it seems appropriate to write down
some guidelines for adding/maintaining options and options
documentation so that we can at least point to procedures to keep
things from getting out of sync again.
Thanks to jemarch@gnu.org and dmalcolm@redhat.com for their suggestions
to improve this patch.
gcc/ChangeLog
* doc/options.texi (Options): Point to the "user experience"
documentation.
* doc/ux.texi (Guidelines for Options): Add some.
Robin Dapp [Mon, 26 Jan 2026 14:24:10 +0000 (15:24 +0100)]
RISC-V: Handle VL-setting FoF loads. [PR123806]
For PR122869 I thought I fixed the issue of VL-spills clobbering
explicit VL reads after fault-only-first (FoF) loads but it turns
out the fix is insufficient. Even though it avoided the original
issue, we can still have spills that clobber VL before the read_vl
RTL pattern. That's mostly due to us hiding the VL data flow from
the optimizers so a regular spill to memory can and will introduce
a VL clobber. In vsetvl we catch all the regular cases but not the
FoF-load case of PR123806 and PR122869.
This patch adds specific FoF patterns that emit the same instruction but
have a register-setting VL pattern inside the insn's PARALLEL.
It serves as a marker for the vsetvl pass that can recognize that we
clobber VL before reading its value. In that case we now emit an
explicit csrr ..,vl.
After vsetvl it's safe to emit the read_vls because at that point the
VL dataflow has been established and we can be sure to not clobber VL
anymore.
Thus, the main changes are:
- Unify read_vl si and di and make it an UNSPEC. We don't optimize
it anyway so a unified one is easier to include in the new FoF
VL-setter variants.
- Introduce VL-setting variants of FoF loads and handle them like
read_vl()s in the vsetvl pass.
- Emit read_vl()s after vsetvl insertion is done.
What this doesn't get rid of is the XFAIL in ff-load-3.c that I
introduced for PR122869. The code is still "good" at -O1 and
"bad" at -O2 upwards.
PR target/123806
gcc/ChangeLog:
* config/riscv/riscv-string.cc (expand_rawmemchr): Use unified
vl_read.
(expand_strcmp): Ditto.
* config/riscv/riscv-vector-builtins-bases.cc:
* config/riscv/riscv-vector-builtins.cc (function_expander::use_fof_load_insn):
Only emit the store and not the VL read.
* config/riscv/riscv-vsetvl.cc (get_fof_set_vl_reg): New
function.
(init_rtl_ssa): New wrapper.
(finish_rtl_ssa): Ditto.
(emit_fof_read_vls): Emit read_vl after each fault-only-first
load.
(pass_vsetvl::simple_vsetvl): Call emit_fof_read_vls ().
(pass_vsetvl::lazy_vsetvl): Ditto.
* config/riscv/vector-iterators.md: Add read_vl unspec.
* config/riscv/vector.md (read_vlsi): Unify.
(@read_vl<mode>): Ditto.
(read_vldi_zero_extend): Ditto.
(@pred_fault_load_set_vl<V_VLS:mode><P:mode>): New FoF variant
that saves VL in a register.
(@pred_fault_load_set_vl<VT:mode><P:mode>): Ditto.
gcc/testsuite/ChangeLog:
* g++.target/riscv/rvv/base/pr123806.C: New test.
* g++.target/riscv/rvv/base/pr123808.C: New test.
* g++.target/riscv/rvv/base/pr123808-2.C: New test.
Robin Dapp [Wed, 28 Jan 2026 16:54:42 +0000 (17:54 +0100)]
RISC-V: testsuite: Add zvl requirement to PR123626.
This adds a new require-effective-target check to pr123626.c.
As the test is a run test compiled with _zvl256b we need
to ensure the target actually supports 256b vectors.
We can only check for exactly 256b right now
(rvv_zvl256b_ok), i.e. "VLS". Therefore, the patch also adds
a new target check rvv_zvl_ge_256b_ok where ge means greater
or equal.
gcc/testsuite/ChangeLog:
* lib/target-supports.exp: Add rvv_zvl_ge_256b_ok.
* gcc.target/riscv/rvv/base/pr123626.c: Use new target check.
Filip Kastl [Thu, 29 Jan 2026 15:01:33 +0000 (16:01 +0100)]
pta: Fix comment of intra_create_variable_infos
The comment says that this function creates varinfos for *all*
variables in a function. Meanwhile, it only creates varinfos for
parameters, the return value and the static chain.
Commiting as obvious.
gcc/ChangeLog:
* gimple-ssa-pta-constraints.cc (make_param_constraints): Adjust
comment: The function only creates varinfos for the parameters,
the return value and the static chain.
I notice that we don't emit the diag_array_subscript diagnostic because
slice calls fold_non_dependent_expr which calls _eval_outermost with
allow_non_constant=true. I wonder if we want to change this.
Iain Buclaw [Thu, 29 Jan 2026 13:06:14 +0000 (14:06 +0100)]
d: Fix ICE in gimplify_expr with const ref noreturn parameters [PR123046]
The ICE was caused by references to const/immutable qualified `noreturn'
declarations that were being leaked to the code generation when they
should have been omitted or replaced with `assert(0)'.
PR d/123046
gcc/d/ChangeLog:
* d-codegen.cc (build_address): Return `null' when generating the
address of a `noreturn' declaration.
(d_build_call): Compare TYPE_MAIN_VARIANT of type with `noreturn'.
* decl.cc (get_fndecl_arguments): Likewise.
* types.cc (finish_aggregate_mode): Likewise.
(TypeVisitor::visit (TypeFunction *)): Likewise.
Richard Biener [Thu, 29 Jan 2026 12:56:11 +0000 (13:56 +0100)]
tree-optimization/122537 - do not elide maybe_zero condition for wrapping IV
The following removes the optimization eliding the maybe_zero condition
from number_of_iterations_lt_to_ne when the IV can overflow since the
IV delta input is not accurately reflecting this.
PR tree-optimization/122537
* tree-ssa-loop-niter.cc (number_of_iterations_lt_to_ne): Register
may_be_zero condition when the IV may overflow.
Marek Polacek [Wed, 28 Jan 2026 20:43:35 +0000 (15:43 -0500)]
c++/reflection: tweak for eval_has_template_arguments
As discussed in
<https://gcc.gnu.org/pipermail/gcc-patches/2026-January/705756.html>:
> For reflection it's probably best to go with
>
> if (TYPE_P (r) && typedef_variant_p (r))
> return alias_template_specialization_p (r, nt_opaque);
>
> and not get into primary_template_specialization_p at all.
Here in a patch form.
gcc/cp/ChangeLog:
* reflect.cc (eval_has_template_arguments): Return immediately after
checking alias_template_specialization_p.
Fix AIX build break in gcc/diagnostics/sarif-sink.cc.
In AIX, build fails while trying to compile gcc/diagnostics/sarif-sink.cc.
Here is the snapshot of the error we are getting.
In file included from /opt/freeware/src/packages/BUILD/gcc/gcc/system.h:46,
from /opt/freeware/src/packages/BUILD/gcc/gcc/diagnostics/sarif-sink.cc:34:
/opt/freeware/lib/gcc/powerpc-ibm-aix7.3.0.0/13/include-fixed/stdio.h:593:12: error: conflicting declaration of C function 'int fgetpos64(FILE*, fpos64_t*)'
593 | extern int fgetpos64(FILE *, fpos64_t *);
| ^~~~~~~~~
This happens when we include sys/types.h before defining _LARGE_FILES.
We can see similar errors with this sample program.
In gcc/diagnostics/sarif-sink.cc we are including config.h after #include <sys/un.h> which intern
including sys/types.h and causing the conflicting errors.
Including config.h before including sys/un.h should solve this error.
The patch is to fix the build break on AIX.
gcc/ChangeLog:
* diagnostics/sarif-sink.cc: Move config.h at the begining.
Richard Biener [Thu, 29 Jan 2026 09:41:16 +0000 (10:41 +0100)]
tree-optimization/123596 - fix partial virtual SSA update in eh_cleanup
The following replaces the not quite correct use of
mark_virtual_operand_for_renaming by an appropriate way of dealing
with a possibly partially up-to-date virtual SSA form. Namely
when we just move stmts and not remove a VDEF we should arrange
for missing virtual PHIs to be created and just queue its arguments
for possible renaming. For the testcase at hand there's no renaming
necessary in the end when done this way.
PR tree-optimization/123596
* tree-eh.cc (sink_clobbers): Create a virtual PHI when
one is required but not present, queuing arguments
for renaming.
Richard Biener [Thu, 29 Jan 2026 07:47:44 +0000 (08:47 +0100)]
tree-optimization/116747 - ICE in cselim due to duplicate sinking
The following avoids queueing duplicate stmts in the set of sinkings
to consider as well as pick candidates in an order that ensures
we don't unnecessarily re-order stores. While we currently only can
trigger the ICE with out-of-bound accesses future enhancements
to how we deal with dependence analysis in this pass could expose
the issue to a wider range of testcases, so this makes it future-proof.
PR tree-optimization/116747
* tree-ssa-phiopt.cc (cond_if_else_store_replacement): Avoid
duplicate stmts in the set of store pairs to process.
Jakub Jelinek [Thu, 29 Jan 2026 11:34:59 +0000 (12:34 +0100)]
c++: Fix ICE in eval_annotations_of [PR123866]
eval_annotations_of throws if the passed in reflection handle is not
eval_is_function (and various others). Now, eval_is_function uses
maybe_get_first_fn to look through BASELINK/OVERLOAD etc., but
eval_annotations_of wasn't doing that and ICEd on
else if (TYPE_P (r))
r = TYPE_ATTRIBUTES (r);
else if (DECL_P (r))
r = DECL_ATTRIBUTES (r);
else
gcc_unreachable ();
because r isn't a decl nor type (nor REFLECT_BASE earlier).
2026-01-29 Jakub Jelinek <jakub@redhat.com>
PR c++/123866
* reflect.cc (eval_annotations_of): Use maybe_get_first_fn.
Jose E. Marchesi [Thu, 29 Jan 2026 02:33:17 +0000 (03:33 +0100)]
a68: implement GNU68-2026-001-short-of-symbol
This patch implements the GNU extension:
GNU68-2026-001-brief-selection - Brief style for selection
which adds the preferred brief style for selection recommended by
Hansen in "ALGOL 68 Hardware Represenatation Recommendations"
published in the Algol Bulletin issue 42.
This extension is already listed in https://algol68-lang.org.
Signed-off-by: Jose E. Marchesi <jemarch@gnu.org>
gcc/algol68/ChangeLog
* ga68.vw: Update formal grammar to express the GNU extension.
* a68-parser.cc (a68_dont_mark_here): Likewise.
* a68-parser-scanner.cc (SINGLE_QUOTE_CHAR): Define.
(get_next_token): Recognize ' as QUOTE_SYMBOL.
(tokenise_source): Acknowledge QUOTE_SYMBOL.
* a68-parser-keywords.cc (a68_set_up_tables): Likewise.
* a68-parser-bottom-up.cc (reduce_primary_parts): Adjust parser to
brief form of selection.
* a68-parser-attrs.def (QUOTE_SYMBOL): New attribute.
* ga68.texi (Brief selection): New section.
Jakub Jelinek [Thu, 29 Jan 2026 08:49:51 +0000 (09:49 +0100)]
c++: Implement CWG3131 - Value categories and types for the range in iterable expansion statements
For the https://gcc.gnu.org/pipermail/gcc/2025-November/246977.html
issues I've filed https://github.com/cplusplus/CWG/issues/805
which resulted in two CWG issues,
https://cplusplus.github.io/CWG/issues/3131.html
and
https://cplusplus.github.io/CWG/issues/3140.html
This patch implements the former, changing
the iterating expansion statement http://eel.is/c++draft/stmt.expand#5.2
line from
constexpr auto&& range = expansion-initializer;
to
constexpr decltype(auto) range = (expansion-initializer);
(for our partly pre-CWG3044 implementation with static before it).
2026-01-29 Jakub Jelinek <jakub@redhat.com>
* cp-tree.h (build_range_temp): Implement CWG3131 - Value
categories and types for the range in iterable expansion statements.
Add bool argument defaulted to false.
* parser.cc (build_range_temp): Add expansion_stmt_p argument,
if true build const decltype(auto) __for_range = range_expr instead of
auto &&__for_range = range_expr.
(cp_build_range_for_decls): For expansion stmts build __for_range as
static constexpr decltype(auto) __for_range = (range_expr);
rather than static constexpr auto &&__for_range = range_expr;.
* g++.dg/cpp26/expansion-stmt1.C (N::begin, N::end, O::begin,
O::end): Change argument type from B & to const B & or from
D & to const D &.
* g++.dg/cpp26/expansion-stmt2.C (N::begin, N::end, O::begin,
O::end): Likewise.
* g++.dg/cpp26/expansion-stmt3.C (N::begin, N::end, O::begin,
O::end): Likewise.
* g++.dg/cpp26/expansion-stmt16.C: Expect different diagnostics
for C++11.
* g++.dg/cpp26/expansion-stmt18.C (N::begin, N::end): Change
argument type from B & to const B &.
* g++.dg/cpp26/expansion-stmt25.C (N::begin, N::end): Likewise.
* g++.dg/cpp26/expansion-stmt26.C: New test.
* g++.dg/reflect/p3491-2.C (baz): Move workaround to a new
function garply, use the previously #if 0 guarded implementation.
(garply): New function.
Pietro Monteiro [Thu, 29 Jan 2026 01:16:04 +0000 (20:16 -0500)]
libffi: Add missing GCC patches to LOCAL_PATCHES
Add all the patches to GCC's copy of libffi to LOCAL_PATCHES. I
skipped patches that only regenerate the configure script because it
doesn't come from upstream.
Jose E. Marchesi [Sat, 24 Jan 2026 19:10:10 +0000 (20:10 +0100)]
a68: support FFI with C via formal holes
This commit adds FFI support to access C variables and call C
functions from Algol 68, using the construct specified by the Modules
and Separated compilation facilities intended to that effect: formal
holes (`nest' constructs).
Briefly put, an `int foo' C variable can be accessed from Algol 68
like:
int foo = nest C "foo";
if foo > 10 then ... fi;
A ref to a C variable `int counter', that allows to modify it as well
as read it:
ref int counter = nest C "&counter";
counter +:= 1;
The address stored in an `int *ptr' variable, as an Algol 68
reference:
ref int ptr = nest C "ptr";
ptr := 100; { Changes the integer pointed by the C ptr }
Calling the standard C `random' function:
proc long int random = nest C "random";
if random < 100 then ... fi;
Some tests and an sketchy section in the manual are also included.
Signed-off-by: Jose E. Marchesi <jemarch@gnu.org>
gcc/algol68/ChangeLog
* ga68.vw: Fix grammar for formal holes.
* ga68.texi (Particular programs): New section.
(Comments): Likewise.
* a68.h: Prototypes for a68_is_c_mode, a68_make_fomal_hole_decl,
a68_lower_formal_hole.
* a68-parser-attrs.def (FORMAL_HOLE): New attribute.
(LANGUAGE_INDICANT): Likewise.
* a68-parser-extract.cc (a68_elaborate_bold_tags): Elaborate
language indicants.
* a68-parser-moids-check.cc (mode_check_unit): Check mode in
formal holes.
* a68-low-units.cc (a68_lower_formal_hole): New function.
* a68-low.cc (a68_make_formal_hole_decl): Likewise.
* a68-moids-misc.cc (a68_is_c_mode): Likewise.
* a68-parser-bottom-up.cc (reduce_formal_holes): Likewise.
Eric Botcazou [Wed, 28 Jan 2026 22:52:41 +0000 (23:52 +0100)]
Ada: Fix crash on Unchecked_Union parameter with -gnateV -gnata
The problem is that the compiler generates 'Valid_Scalars for a formal
parameter of an Unchecked_Union type, which cannot work because it is not
possible to find out where the scalars are in it, given that the parameter
does not contain the discriminants of its Unchecked_Union type. This also
changes -gnateV to work without the need for -gnata, as there is no mention
of this dependence in the documentation.
gcc/ada/
PR ada/123857
* checks.adb (Apply_Parameter_Validity_Checks.Add_Validity_Check):
Set Is_Checked on the generated {Pre,Post}_Condition pragma and
bail out if the parameter is of an Unchecked_Union type.
gcc/testsuite/
* gnat.dg/unchecked_union4.adb: New test.
Eric Botcazou [Wed, 28 Jan 2026 22:31:49 +0000 (23:31 +0100)]
Ada: Fix stack corruption with concatenation and 'Image of composite type
The issue is that the expansion of 'Image for composite types is heavyweight
and involves a mix of Expression_With_Actions and controlled object that
does not work properly when it is the argument of a call to a subprogram,
so this replaces it by the canonical scheme used for controlled temporaries.
gcc/ada/
PR ada/123832
* exp_imgv.adb: Add with and use clauses for Exp_Ch7.
(Expand_Image_Attribute): Establish a transient scope before
rewriting the attribute as a call to Put_Image.
(Expand_Wide_Image_Attribute): Likewise.
(Expand_Wide_Wide_Image_Attribute): Likewise.
* exp_put_image.ads (Build_Image_Call): Add note about the
need for a transient scope when the function is invoked.
* exp_put_image.adb (Build_Image_Call): Call Insert_Actions
to immediately insert the actions instead of wrapping them
in an Expression_With_Actions node.
gcc/testsuite/
* gnat.dg/put_image2.adb: New test.
Iain Buclaw [Wed, 28 Jan 2026 22:06:23 +0000 (23:06 +0100)]
d: Fix ICE in ExprVisitor::visit, at d/expr.cc:2224 [PR123419]
The original assert expected the type of `__traits(initSymbol)' to be
exactly `const(void[])', but because D strips const from arrays to allow
passing slices as mutable ranges to template functions, so it got turned
into `const(void)[]'.
Iain Buclaw [Wed, 28 Jan 2026 21:09:37 +0000 (22:09 +0100)]
d: Fix ICE: in output_constructor_regular_field, at varasm.cc:5500 [PR123798]
PR d/123798
gcc/d/ChangeLog:
* types.cc (insert_aggregate_bitfield): Set DECL_NONADDRESSABLE_P and
DECL_PADDING_P on bit-field decls.
(finish_aggregate_type): Pass the aligned bit offset to layout_decl.
Uros Bizjak [Wed, 28 Jan 2026 20:57:47 +0000 (21:57 +0100)]
i386: Use x >> ~y for x >> 31-y [PR36503]
x86 targets mask 32-bit shifts with a 5-bit mask (and 64-bit with 6-bit mask),
so they can use x >> ~y instead of x >> 31-y.
The optimization converts:
movl $31, %ecx
subl %esi, %ecx
sall %cl, %eax
to:
notl %ecx
sall %cl, %eax
PR target/36503
gcc/ChangeLog:
* config/i386/i386.md (*<insn:any_shift><mode:SWI48>3_sub):
Also allow operands[3] & (<mode_bitsize>-1) == (<mode_bitsize>-1)
in insn condition. Emit NOT RTX instead of NEG RTX in this case.
(*<insn:any_shift><mode:SWI48>3_sub_1): Ditto.
gcc/testsuite/ChangeLog:
* gcc.target/i386/pr36503-5.c: New test.
* gcc.target/i386/pr36503-6.c: New test.
Robin Dapp [Mon, 26 Jan 2026 16:59:58 +0000 (17:59 +0100)]
RISC-V: Fix ABI vector passing on stack and GPR [PR123824].
Krister reported that we violate the psABI when one vector argument
halfway fits into a register:
"Aggregates whose total size is no more than 2×XLEN bits are passed in a
pair of registers; if only one register is available, the first XLEN
bits are passed in a register and the remaining bits are passed on the
stack. If no registers are available, the aggregate is passed on the
stack."
This patch fixes this oversight and adds a few tests.
Regtested on rv64gcv_zvl512b.
PR target/123824
gcc/ChangeLog:
* config/riscv/riscv.cc (riscv_vls_mode_fits_in_gprs_p): New
helper.
(riscv_pass_vls_aggregate_in_gpr): Use helper and distribute
half-fitting vector to GPR and stack.
(riscv_pass_aggregate_in_vr): Reformat comment.
(riscv_get_arg_info): Use helper.
(riscv_pass_by_reference): Ditto.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/abi/vls-gpr-1.c: New test.
* gcc.target/riscv/abi/vls-gpr-10.c: New test.
* gcc.target/riscv/abi/vls-gpr-11.c: New test.
* gcc.target/riscv/abi/vls-gpr-12.c: New test.
* gcc.target/riscv/abi/vls-gpr-13.c: New test.
* gcc.target/riscv/abi/vls-gpr-14.c: New test.
* gcc.target/riscv/abi/vls-gpr-2.c: New test.
* gcc.target/riscv/abi/vls-gpr-3.c: New test.
* gcc.target/riscv/abi/vls-gpr-4.c: New test.
* gcc.target/riscv/abi/vls-gpr-5.c: New test.
* gcc.target/riscv/abi/vls-gpr-6.c: New test.
* gcc.target/riscv/abi/vls-gpr-7.c: New test.
* gcc.target/riscv/abi/vls-gpr-8.c: New test.
* gcc.target/riscv/abi/vls-gpr-9.c: New test.
Andrew Pinski [Wed, 15 Oct 2025 00:50:13 +0000 (17:50 -0700)]
C/C++: Better notification if misused a function like macro [PR102846]
Currently if assert is misused, e.g. like passing to a function like a
function pointer, GCC suggests to include assert.h/cassert. But since
it is already included that is just a bogus suggestion. Instead we should
see if a function-like macro is that same name (in the case of assert it will be),
and inform the user that it is being mis-used.
[PR121571, LRA]: Reprocess asm insn with different preferences when there are no enough regs for the asm-insn
The test for the PR contains asm insn requiring 7 general regs but
three operands can be in memory too ('g' constraint). There are only
6 available general regs. IRA in the test case assigns a mask reg to
one pseudo. When LRA reloads the pseudo assigned to mask reg, it
frees general reg assigned to another pseudo in the asm and assigns
mask reg to another pseudo. After a few iterations, we have asm all
operands of which are reload pseudos assigned to general regs and one
pseudo assigned to mask reg. After that LRA can do nothing and
reports "not enough regs". The patch recognizes situation when there
are not enough regs for an asm insn and makes 2nd attempt to find
reloads when memory for operands with 'g' or 'rm' constraint is
preferred.
gcc/ChangeLog:
PR target/121571
* lra-constraints.cc (process_alt_operands): Reprocess asm insn
when there are no enough regs for the asm-insn.
c: Fix parser error on address space names [PR 123583]
This patch fixes a regression introduced by PR 67784
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67784) - it broke
recognition of named address space qualifiers (such as __memx on
AVR or __seg_gs on x86) after certain statements.
The fix for PR 67784 attempts to reclassify tokens whose
token->id_kind could have been set incorrectly (because of possibly
wrong scope when peeking ahead). c_parser_maybe_reclassify_token
only skips reclassification for C_ID_CLASSNAME though - a token with
id_kind = C_ID_ADDRSPACE ends up getting reclassified as C_ID_ID,
eventually causing a ": error: '<address space name>' undeclared"
error later down the line.
Rather than explicitly excluding C_ID_ADDRSPACE, the patch modifies
the check to reclassify only tokens kinds that could potentially get
incorrectly classified - C_ID_ID and C_ID_TYPENAME.
Bootstrapped and regtested on x86_64-linux.
PR c/123583
gcc/c/ChangeLog:
* c-parser.cc (c_parser_maybe_reclassify_token): Reclassify only
C_ID_ID and C_ID_TYPENAME tokens.
gcc/testsuite/ChangeLog:
* gcc.target/avr/pr123583.c: New test.
* gcc.target/i386/pr123583.c: New test.
Jonathan Wakely [Wed, 28 Jan 2026 10:27:07 +0000 (10:27 +0000)]
libstdc++: Default to C++20 in src/experimental/Makefile.am
r16-7089-gfbde291af66e02 had a stray change to the Makefile.in but
actually we probably should change that directory to use C++20, now that
it's the default. So this updates the Makefile.am, and the Makefile.in
doesn't need to be regenerated now.
libstdc++-v3/ChangeLog:
* src/experimental/Makefile.am: Change AM_CXXFLAGS to use C++20
by default.
Richard Biener [Wed, 28 Jan 2026 09:55:56 +0000 (10:55 +0100)]
ipa/111036 - strip nop conversions around __builtin_constant_p arguments
The PR is about inconsistent behavior wrt inline predicate analysis
and later folding of __builtin_constant_p which ultimatively results
from fold_builtin_constant_p stripping nops off its argument but
this not being done on GIMPLE. The following adds a match.pd pattern
for this.
Richard Biener [Wed, 28 Jan 2026 09:16:29 +0000 (10:16 +0100)]
rtl-optimization/106859 - fix location dumping from var-tracking
We currently ICE when the locations include one without a setting insn.
Looking at cselib this seems to be a supported state so the following
arranges for this and dumps -1 as UID.
PR rtl-optimization/106859
* var-tracking.cc (val_store): Dump -1 as UID if setting_insn
is NULL.
Karl Meakin [Thu, 6 Nov 2025 16:34:06 +0000 (16:34 +0000)]
aarch64: FEAT_SVE_BFSCALE support
Add support for the `SVE_BFSCALE` architecture extension.
gcc/ChangeLog:
* doc/invoke.texi: Document `+sve-bfscale` flag.
* config/aarch64/aarch64.h (TARGET_SVE_BFSCALE): New macro.
* config/aarch64/aarch64-c.cc (aarch64_update_cpp_builtins):
Define `__AARCH64_FEATURE_SVE_BFSCALE`.
* config/aarch64/aarch64-sve-builtins-base.cc: Skip constant
folding for floating-point or unpredicated multiplications.
* config/aarch64/aarch64-sve-builtins-sve2.def: New `SVE_FUNCTION`s.
* config/aarch64/aarch64-sve.md: Modify insns for
`SVE_COND_FP_BINARY_INT` to handle BF16 modes.
* config/aarch64/aarch64-sve2.md
(@aarch64_sve_<optab><mode>, @aarch64_sve_<optab><mode>_single): New insn for `BFSCALE`.
Modify insns for `UNSPEC_FSCALE` to handle BF16 modes.
* config/aarch64/iterators.md (SVE_FULL_F_SCALAR): Add `VNx8BF` to iterator.
(SVE_FULL_F_BFSCALE): New iterator.
(SVE_Fx24_BFSCALE): New iterator.
(SVE_BFx24): New iterator.
(UNSPEC_FMUL): New unspec.
(V_INT_EQUIV): Add entries for BF16 modes.
(b): Add entries for scalar float modes.
(is_bf16): Add entries for BF16 modes and reformat.
(SVSCALE_SINGLE_INTARG): Likewise.
(SVSCALE_INTARG): Likewise.
(SVE_FP_MULL): New iterator.
gcc/testsuite/ChangeLog:
* lib/target-supports.exp: Add `sve-bfscale` to `sve_exts`.
* gcc.target/aarch64/pragma_cpp_predefs_4.c: Add test for `__ARM_SVE_FEATURE_BFSCALE`.
* gcc.target/aarch64/sme2/acle-asm/mul_bf16_x2.c: New test.
* gcc.target/aarch64/sme2/acle-asm/mul_bf16_x4.c: New test.
* gcc.target/aarch64/sme2/acle-asm/scale_bf16_x2.c: New test.
* gcc.target/aarch64/sme2/acle-asm/scale_bf16_x4.c: New test.
* gcc.target/aarch64/sve/acle/asm/scale_bf16.c: New test.
* gcc.target/aarch64/sve/acle/general-c/bfscale.c: New test.
The operand is a constant _Decimal64 vector with BLKmode, so expand has
to materialize it in memory. Current get_object_alignment() returns a
1-byte guaranteed alignment for this VECTOR_CST, as indicated by A8 in
the RTL dump above. However, with "-mstrict-align" enabled, later subreg
lowering pass expects at least 64-bit alignment when it constructs a new
RTX to decompose the store into pieces. Because the original alignment
is too small, simplify_gen_subreg() returns NULL_RTX and an assertion is
hit.
This patch increases the stack slot alignment for STRICT_ALIGNMENT
targets, when the operand is forced into memory. The increased alignment
is capped by MAX_SUPPORTED_STACK_ALIGNMENT so it won't be too large.
Bootstrapped and tested on aarch64-linux-gnu and x86_64-linux-gnu.
projects / thirdparty / gcc.git / log
