https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109618
Roger Sayle <roger at nextmovesoftware dot com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |roger at nextmovesoftware dot com --- Comment #5 from Roger Sayle <roger at nextmovesoftware dot com> --- A patch for this was posted at https://gcc.gnu.org/pipermail/gcc-patches/2024-April/650205.html back in April, but the review got side tracked into a discussion on poisoning of conflicting declarations. My opinion is that this is a necessary evil if the front-end wishes to continue compilation after encountering an error (that should otherwise be fatal). I'd prefer that front-ends not call the middle-end with generic containing deeply embedded error_mark_nodes, but as they (currently) do, we should avoid ICEing.