Implicit conversion from float -105.84139264990327 to int loses precision