Implicit conversion from float -39.06403940886702 to int loses precision