https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118876
Bug ID: 118876
Summary: [15 Regression][OpenMP] ICE on valid code
Product: gcc
Version: 15.0
Status: UNCONFIRMED
Keywords: ice-on-valid-code, openmp
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: burnus at gcc dot gnu.org
Target Milestone: ---
omptests/xt-declare-target-ctors/test.cpp: In function ‘void __tcfsd(void*)’:
omptests/xt-declare-target-ctors/test.cpp:103:5: internal compiler error: in
add_stmt, at cp/semantics.cc:551
103 | SSD sd[3];
| ^~
0x28a93cf internal_error(char const*, ...)
gcc/diagnostic-global-context.cc:517
0xabbf21 fancy_abort(char const*, int, char const*)
gcc/diagnostic.cc:1722
0x86906e add_stmt(tree_node*)
gcc/cp/semantics.cc:551
0xba6c55 register_dtor_fn(tree_node*)
gcc/cp/decl.cc:10494
0xbb26b7 one_static_initialization_or_destruction
gcc/cp/decl2.cc:4518
* * *
The assert is at:
/* Add T to the statement-tree. Non-side-effect statements need to be
recorded during statement expressions. */
gcc_checking_assert (!stmt_list_stack->is_empty ());
I am not completely sure whether that's a true regression as it is a
gcc_checking_assert.
An educated guess, only, is that it has been caused by
r15-2799-gf1bfba3a9b3f31
OpenMP: Constructors and destructors for "declare target" static aggregates
which would make it a GCC 15 regression (and unsupported in GCC 14).
* * *
The unreduced test case is at
https://github.com/doru1004/omptests/blob/main/xt-declare-target-ctors/test.cpp#L103
