Implicit conversion from float -17.773584905660357 to int loses precision