Implicit conversion from float -21.17166212534059 to int loses precision