Re: Issue 802: menubar-x.c assertion failure

I believe that returning NULL is all it can do if it is handed a broken menu specification. Ben's error wrapping stuff is right in the middle of the call stack, so probably this code used to (unsafely) throw Lisp errors right past compute_menubar_data, so the assert was correct. With the current code structure, though, we need to keep unraveling the stack upwards until we reach a point where the delayed error can be rethrown and handled properly (by displaying it in *Warnings*). I think. At least it seems to work that way when handed the broken ESS menu specification.