Specifically I was referring to standard float representation which permits signed zeros. However, other comments provide some interesting examples also.