Well, the suggestion you raised was that fatal errors weren't trappable if they happened at compile time. The only point I'm making is that you can actually trap them - by delaying the compile time.