Implicit conversion from float -16.958393113342908 to int loses precision