Implicit conversion from float -7.0909090909090935 to int loses precision