bugmap extension (was re: API question: ancestors of B that are not ancestors of A)

[Greg Ward]

On Wed, Dec 9, 2009 at 11:19 PM, Nicolas Dumazet
> I would not like, personally, "setbug" to be able to change bugmap
> associated to already propagated changesets. In fact, I imagined  bug
> information to be commit metadata, quite similar to tags in kind, with
> no way to directly manipulate remote history.
> But I guess that you _need_ this behavior.

Yep: that is absolutely essential.  I had to write up a "how do I
modify the CVS commit message?" document because we have to do it so
frequently, when people put the wrong bug # in the commit message.
(That's the starting point of how we let our documentation team, our
QA team, and our customers know what's in every new release.  We're
currently maintaining 3 or 4 release branches, pumping out a new
release of each one roughly monthly.  So this stuff really has to be
well automated.)

The fact that Hg does not let you modify commit messages made it
essential to move this info out of the commit message.  The side
benefits of having it in a relational DB will, I think, be very
interesting.

Greg