Implicit conversion from float -68.34990059642146 to int loses precision