Implicit conversion from float -12.317880794701992 to int loses precision