Implicit conversion from float -8.936170212765973 to int loses precision