Ah, good point! I was thinking of unsigned overflow, but signed overflow is a different matter, as pointed out elsewhere in this thread. I retract my statement about trapping overflow.