Implicit conversion from float -12.367688022284113 to int loses precision