Implicit conversion from float -40.7625649913345 to int loses precision