On a posteriori error estimation for constant obstacle problems