Implicit conversion from float -2.323097463284384 to int loses precision