> Now that we're on final approach, I've attempted, today, to enforce the new > keyword policy, only to find that I don't really know what it is.
Heh. Indeed. > I know SCCS keywords are going away. The RTI nits page, and Val in IRC, focus > on the "output string" aspect, viz: remove them from modinfo strings. > > However: > > 1) should we not also be removing them from #ident and #pragma ident? Aren't > those lines going to be gibberish in the face of Mercurial? Yes, but they're harmless gibberish. The modinfo (or other user-visible) uses are potentially more problematic. > 2) Have we agreed on a new form for the two problem children, #ident/#pragma > ident and the modinfo line? And if so, can we specify that exactly to avoid > as > much thinking as possible when the new requirements are thrust on gatelings > innocent of all the discussion and background? We've floated suggestions as long ago as Steve Lau's KTD, and that was to use the output of "hg id" as a unique (indeed, much better than %I%, because it applies to the entire workspace, not just one file) identifier. That said, we've still got some work to do in this regard, tying that into the VERSION macro in nightly/make. Invoking "hg id" is expensive, and we don't want folks to do that in their individual Makefiles. So for now, the policy really is what Val said. Don't let new, non-ident keyword usage in, and we'll have a policy soon for the unsightly-but-harmless vestiges in ident lines. --Mark
