Implicit conversion from float -30.36082474226805 to int loses precision