Dividing INT_MIN by -1 or taking INT_MIN modulo -1 can sometimes
cause floating-point exceptions or otherwise misbehave.

Fabien Coelho and Michael Paquier