Implicit conversion from float -7.467438494934882 to int loses precision