Implicit conversion from float -1.7441860465116292 to int loses precision