[PATCH 0 of 4] Build on latesttag for archival and mercurial builds

[Gilles Moris]

On Tuesday 06 October 2009 12:35:10 pm Mads Kiilerich wrote:
> But the root of the problem is that latesttaginfo not only is
> interesting for the templater. I think it would fit more natural as a
> method on the repo objects, but I don't know if such high-level
> functionality is welcome in the repo class?
>
> > I had to make the latesttaginfo a global function, that I choose to make
> > reentrant. May be it would be better to make latesttagcache a global
> > variable. Please advise.
>
> The cache _is_ a function of the repo graph, so I think it fits more
> natural on the repo objects. Also, AFAICS the algorithm will work with
> long-lived caches as long as revisions only are added and not stripped.
> I do however not know _how_ long-lived the repo objects are.
>
> "Just say no to global variables". But caches might by design be an
> exception to that.

Here is an alternate patch queue using show_changeset, so that we do not
need to make latesttaginfo a global function.

I have also splitted the last patch to setup.py in 2. Tell me if I need
to split it further down to ease review.

Regards.
Gilles.