Implicit conversion from float -56.19999999999999 to int loses precision