Implicit conversion from float -19.22746781115879 to int loses precision