The difference is that the constraint of using a signed int as the length of the array is known and available to users of the function. What happens when you try to do that is you get a compiler error, because it won't implicitly cast from, say, bigint or unsigned int down to signed int, for example. That's not necessarily a defect, that's just a constraint.

This contrasts with the overflow defect which would cause the mergesort routine to fail when given an otherwise perfectly valid input.