Implicit conversion from float -17.080291970802932 to int loses precision