Policy when dependencies break on updates?

[Thorsten Leemhuis]

BTW, some will have noticed already: I was traveling for two days (and
have been a bit busy right before that) and I'm making my way through
the mail backlog...

On 08.10.2008 00:18, Orcan Ogetbil wrote:
> Since I'm new with packaging for a 3rd part repo I need to get
> certain things clear.

Note that the problems you outline can happen similarly within the 
Fedora repos as well.

> What is the policy whenever Fedora pushes new packages to its updates
> repo and that breaks package dependencies of rpmfusion?

There are no real policies that handle this situation except "that 
should not happen and hence packagers should do everything to prevent 
that from happening".

> For instance, right now xmms-*.rpm packages are updated in Fedora and
> yum says:
> [...] 
> There might be cases where the package maintainer (xmms-mp3) is not
> around (sleeping, on a trip, in depression etc.). How do we solve
> such a problem then?

The usual answer is: File a bug, it's up to the maintainers to fix it.

But that of course is not the quickest way to fix this and thus no real 
solution :-/ Sometimes it takes days or weeks until the problem gets 
fixed that way.

In livna I now and then after a while simply went forwarded and just 
fixed the problem without waiting for the maintainer. Like I fixed the 
xmms-mp3 problem there (xmms-mp3 hasn't had a real owner there, so I 
could just do it directly) a few days ago with a updated spec file I got 
from Paul (the xmms maintainer in Fedora). I forwarded it to Xavier (who 
maintains xmms-mp3 for RPM Fusion), but afaics he didn't commit it for 
RPM Fusion yet (maybe I should have told him more about the backgrounds, 
but I was in a rush...).

But simply fixing a issue like that often might have side effects 
someone that is not really familiar with the package doesn't know about. 
Sometimes those are that bad that it might be better to leave things 
broken until the package maintainer (or someone else that is familiar 
with the package) has a proper (interim-) solution.

So each of those problems need to be fixed differently. Maybe we can 
work out policies to handle those situations more efficient in the 
future, but I'm not sure if that really is worth the work. I'd say for 
now we just wait and look how things work out.

> For example, in this case I can fix the problem
> in the cvs right away, since I'm one of the few avaliable people at
> this hour, but I don't know if I'm supposed to. Or if I do fix the
> problem in cvs, would it be considered rude?

Yes, it would.

> Well if I do, then there is also a package signing step. Is this done
> manually?

Yes.

> In fedora, there are ueberpackagers who can access/modify other
> people's packages. Do we have such a hierarchy system in rpmfusion?

Sort of. The cvsadmins (Xavier, hansg and I) can commit everywhere. I'd 
say that should do the trick for now, as those problems hopefully should 
not happen that often.

Cu
knurd