Implicit conversion from float -134.8901903367496 to int loses precision