Implicit conversion from float -39.10798122065728 to int loses precision