Implicit conversion from float -2.7892561983471182 to int loses precision