With dmd 2.048 this program gives (on Windows) a runtime error:

int foo(int x, int y) {
    return x % y;
}
void main() {
    int r = foo(int.min, -1);
}



But the same operation done at compile time gives no errors, and foo returns 0:

int foo(int x, int y) {
    return x % y;
}
static assert(foo(int.min, -1) == 0);
void main() {}


So one of the two cases is wrong (or both).

While floating point operations done at compile-time may give slightly different results, I'd like integral operations to give the same results at compile-time and run-time.