> Well, this goes back to the question what TYPE_IS_SIZETYPE semantics > are ;) I guess it would be a sort-of "sticky" 'does not overflow' > flag? No. I would say it's closer to "don't even *think* about overflow on operations of this type".