Implicit conversion from float -31.676300578034684 to int loses precision