Possible overflow in imoper(.., OPER_SUB) ?

Hi, I was quickly reading the source code and I found this line inside imoper: https://gitlab.com/free-astro/siril/blob/master/src/core/siril.c#L177

gbuf[i] = round_to_WORD(gbuf[i] - buf[i]);

Isn't it buggy in case buf[i] is greater than gbuf[i]? In that case, we have an overflow; I see that round_to_WORD takes a double, but in this case the cast is done after the subtraction.