Implicit conversion from float -17.28218465539662 to int loses precision