Implicit conversion from float -8.676748582230616 to int loses precision