Implicit conversion from float -32.30046948356808 to int loses precision