Implicit conversion from float -8.422712933753942 to int loses precision