Implicit conversion from float -4.869402985074629 to int loses precision