Re: mumble and celt, #682010, TC

[Ian Jackson <ijackson@chiark.greenend.org.uk>]

Neil McGovern writes ("Re: mumble and celt, #682010, TC"):
> On Thu, Jul 19, 2012 at 12:24:12AM +0100, Ian Jackson wrote:
> > The problem with mumble and the celt codec has been referred to the TC
> > - see the bug mentioned above.  I would be interested to hear from the
> > security and release teams.
...
> I consider there to be two issues which would concern me from a release
> point of view:
> * That, given the potential problems and issues as discussed in the log,
> a sufficiently experienced maintainer can be found to deal with the
> package for the supported lifetime of a stable+oldstable release.

Right.

> * That the package is likely to be able to communicate with non-debian
> derived distributions.

The information we seem to be getting is that it is indeed likely that
the mumble we have will be able to do so, because mumble upstream have
somehow nominated or blessed 0.7.1.

> If both of these cannot be satisfied, I would consider the package to be
> unreleaseable and would look for removal.

I think we need to find a maintainer, then.

> >  * Would the release team be happy with a reintroduction for wheezy of
> >    the celt package containing the 0.7.1 codec ?  I don't know yet
> >    whether mumble would need to be updated too.  Obviously this would
> >    have to be done promptly.
> 
> Hrm. I'd point out at this point that celt 0.7.1-1 currently exists in
> wheezy.  It's not going anywhere because mumble in testing currently
> depends on it. For any updates, we would apply the freeze criteria as
> specified at http://release.debian.org/wheezy/freeze_policy.html

Ah I had missed that the removal of celt (and corresponding mumble
change to desupport celt and remove the dependency) hadn't made it to
wheezy.

> I'm also a little disappointed that the version in unstable would cause
> communication problems which means that any updates need to go through
> t-p-u. I haven't looked at depth in this, but would re-adding the celt
> dependancy to the sid package, and getting it back in unstable make the
> world happy again?

Yes, it might well.  We would have to re-add the celt package to sid.

> Finally, it would be wonderful if this could all be sorted out in
> unstable and then someone can let us know what's going on with a
> coherent message, that would be fantastic.

Great.

So in summary, the conclusion seems to be:

 * If we cannot find a maintainer for celt who looks like they'll be
   able to handle it for the lifetime of wheezy then we need to allow
   the current mumble (and perhaps other rdepends) in sid to propagate
   and will then be able to remove celt from wheezy.

 * If we can find such a maintainer we need to reintroduce celt into
   sid, and update mumble in sid.  When we have done that everything
   will be fine.

 * We need to decide whether any other rdepends on celt in wheezy need
   to have the celt support removed in wheezy.

Thanks,
Ian.