Implicit conversion from float -10.26737967914437 to int loses precision