https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128765
--- Comment #1 from Drea Pinski <pinskia at gcc dot gnu.org> --- So I think the order of maybe_duplicate_ssa_info_at_copy/replace_uses_by needs to be swapped. Let me check that. I think the LLM produced possible fix is bogus as the gimple_bb will be set correctly before the call to replace_uses_by and skipping those who is null for replace_uses_by won't change anything here.
