Implicit conversion from float -123.70967741935485 to int loses precision