Implicit conversion from float -11.953125 to int loses precision