Arithmetic operations are never done in a type narrower than int. int8_t is necessarily narrower than int (int must be able to represent at least the range -32

Arithmetic operations are never done in a type narrower than int. int8_t is necessarily narrower than int (int must be able to represent at least the range -32767..32767), so in -x the x is promoted to int, then negated.