Implicit conversion from float -17.839388145315482 to int loses precision