Implicit conversion from float -22.464985994397765 to int loses precision