Implicit conversion from float -7.049999999999997 to int loses precision