Implicit conversion from float -3.729189789123197 to int loses precision