Implicit conversion from float -4.875717017208416 to int loses precision