In the JDK 1.8 source code, why is A-B>0 used to determine which one is bigger instead of A>B?

private void ensureExplicitCapacity(int minCapacity) {
    modCount++;

    // overflow-conscious code
    if (minCapacity - elementData.length > 0)
        grow(minCapacity);
}

public static void main(String[] args) {
    int elementDataLength = 10;
    int minCapacity = Integer.MAX_VALUE; // very big

    minCapacity += 1; // accidentally overflow it

    if (minCapacity > elementDataLength) {
        System.out.println("Will not work: minCapacity=-2147483647 elementDataLength=10");
    }
    if (minCapacity - elementDataLength > 0) {
        System.out.println("Will work: minCapacity - elementDataLength = 2147483638");
    }
}