Implicit conversion from float -134.79564032697547 to int loses precision