Implicit conversion from float -40.30303030303031 to int loses precision