Implicit conversion from float -17.400000000000006 to int loses precision