Implicit conversion from float -10.049999999999997 to int loses precision