Implicit conversion from float -86.58643326039387 to int loses precision