Implicit conversion from float -248.3458646616541 to int loses precision