Why do you assume it's so easy for the compiler to tell the difference between the two cases? From this optimization pass's perspective, it has proved that in

Why do you assume it's so easy for the compiler to tell the difference between the two cases? From this optimization pass's perspective, it has proved that in a legal program `Do` can only have one possible value, and simplified the code appropriately. Most of the time this sort of optimization is correct and the programmer wants it; how is it supposed to figure out that this time the programmer had some different semantics in mind?