Implicit conversion from float -5.98393574297188 to int loses precision