Implicit conversion from float -1.6853932584269558 to int loses precision