Implicit conversion from float -28.518518518518505 to int loses precision