Implicit conversion from float -30.30303030303031 to int loses precision