Tom Tromey writes: > This patch is really horrible. I'm still trying to reproduce the problem. BTW, have you tried to find out whether parse.y:java_complete_tree() should catch this and substitute it for a sane value? ./A