Implicit conversion from float -43.497884344146684 to int loses precision