https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92765
--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> --- Note, determine_min_objsize calls compute_builtin_object_size with 2 rather than 3, which means it is in this regard conservative and uses whole object size rather than just subobject, we've been there in the past, punt on some optimizations before the first objsz pass and for subobject sizes compute them early now. But the rest of determine_min_objsize, if compute_builtin_object_size fails, ignores this and thinks it can use subobject sizes when it can't.
