[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Removing orphaned packages from testing



On 16/02/09 at 15:44 -0600, Raphael Geissert wrote:
> [No CC please, thank]
> 
> Philipp Kern wrote:
> 
> > On Sun, Feb 15, 2009 at 10:23:37PM -0600, Raphael Geissert wrote:
> >> Lenny is now out, so I think it is time to decide how to proceed with
> >> what was discussed during DC8. Is the release team still ok with the idea
> >> of keeping orphaned packages out of testing? how should it be done? via
> >> severity: serious bugs and, possibly automated, auto removal hints? some
> >> other way?
> >> 
> >> In any case, I think waiting a week since a package was orphaned before
> >> it is removed should be enough time in case a package is marked as
> >> orphaned by "mistake".
> > 
> > This was *not* decided during DC8 or anything. 
> 
> It was agreed (as in some agreed, nobody objected) on the first or second QA
> meeting during DC8. IIRC Luk was there (although I don't intend to mean
> that he represents the whole team or anything else). That's why I *asked*
> the RT directly, to confirm that the idea could be considered.

As the one who organized the BOFs, I think that we should have a
discussion using the result of the QA BOFs as a basis, not just decide
to implement the result of the BOFs.

Since you are interested in the topic, could you watch the videos from
the BOFs, and post a summary of the various points that were made? That
would be very helpful.

Now, back to the topic. We have a problem, which is:
   We have too many orphaned packages.
Those orphaned packages are orphaned either:
   (A) because they are 'crap' (poor quality/useless software, or
       software for which better alternatives exist)
   (B) because nobody knows they are orphaned
   (C) because nobody is interested in maintaining them

What we need to do is:
1) improve our ways to detect 'crap', to remove those packages from the
   archive (it doesn't make sense to keep them in unstable). We need
safeguards to make mistakes less frequent.
2) improve awareness of orphaned packages. During the QA BOF, the idea of
   a script taking as input a list of packages (the list of locally
installed packages on a DD's system, for example), and outputting the
list of "problems" (RC bugs, O/RFA bugs) was raised. There was
opposition to providing this as something enabled by default in
devscripts, but it would still make sense to provide such a script.

Having a policy of not releasing orphaned packages (which would be
implemented by removing orphaned packages from testing early in the
release process, for example by making O bugs RC) could, in addition, be
useful as a rule. But it doesn't solve either (1) or (2).
-- 
| Lucas Nussbaum
| lucas@lucas-nussbaum.net   http://www.lucas-nussbaum.net/ |
| jabber: lucas@nussbaum.fr             GPG: 1024D/023B3F4F |


Reply to: