Re: #6144: Document the semantics of "comps"