Implicit conversion from float -170.1428571428571 to int loses precision