Implicit conversion from float -17.803468208092482 to int loses precision