Implicit conversion from float -1.8449197860962556 to int loses precision