Implicit conversion from float -5.072886297376101 to int loses precision