A DESCRIPTION OF THE PROBLEM :
The current implementation of Math.floorMod(int, int) reads:

    public static int floorMod(int x, int y) {
        int mod = x % y;
        // if the signs are different and modulo not zero, adjust result
        if ((mod ^ y) < 0 && mod != 0) {
            mod += y;
        }
        return mod;
    }

while Math.floorMod(long, long) reads:

    public static int floorMod(long x, long y) {
        long mod = x % y;
        // if the signs are different and modulo not zero, adjust result
        if ((x ^ y) < 0 && mod != 0) {
            mod += y;
        }
        return mod;
    }

There's a difference in how the signs are checked.

Prefer the floorMod(long, long) variant even for floorMod(int, int), that is, rewrite

    public static int floorMod(int x, int y) {
        int mod = x % y;
        // if the signs are different and modulo not zero, adjust result
        if ((x ^ y) < 0 && mod != 0) {
            mod += y;
        }
        return mod;
    }