Implicit conversion from float -68.88888888888889 to int loses precision