Implicit conversion from float -2.231404958677686 to int loses precision