Implicit conversion from float -8.53556485355648 to int loses precision