https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124850
--- Comment #8 from GCC Commits <cvs-commit at gcc dot gnu.org> --- The releases/gcc-14 branch has been updated by Jakub Jelinek <[email protected]>: https://gcc.gnu.org/g:68b42e35b48d1f9254df05603c92d62f71f6d4df commit r14-12648-g68b42e35b48d1f9254df05603c92d62f71f6d4df Author: Jakub Jelinek <[email protected]> Date: Wed Apr 15 08:55:48 2026 +0200 c++: Mark in cxx_mark_addressable DECL_VALUE_EXPR of DECL_ANON_UNION_VAR_P too [PR124850] The following testcase ICEs since my PR53932 change in checking, because it sees address of the unnamed anon union VAR_DECL taken but without having TREE_ADDRESSABLE set. The following patch sets it during cxx_mark_addressable. I'm wondering how we can get away with not marking other DECL_VALUE_EXPR cases, but couldn't come up with a testcase where it would be needed (e.g. for structured bindings, etc.). So just doing it for anon union vars seems safer to me at least for now. 2026-04-15 Jakub Jelinek <[email protected]> PR c++/124850 * typeck.cc (cxx_mark_addressable): For DECL_ANON_UNION_VAR_P vars also mark their DECL_VALUE_EXPR. * g++.dg/other/anon-union8.C: New test. Reviewed-by: Jason Merrill <[email protected]> (cherry picked from commit 8315b21457d740336295b58d04f0497329e94a67)
