Implicit conversion from float -107.76315789473685 to int loses precision