Implicit conversion from float -5.568561872909697 to int loses precision