Implicit conversion from float -6.257961783439498 to int loses precision