Implicit conversion from float -17.215619694397276 to int loses precision