Implicit conversion from float -43.400000000000006 to int loses precision