182 pending F11 stable updates. WTF?

Michael Schwendt mschwendt at gmail.com
Mon May 11 18:21:03 UTC 2009


On Mon, 11 May 2009 17:46:48 +0200, Ralf wrote:

> >>>>>> (1) New package.
> >>>>>> [cut]
> >>>>> Is it a good practice to push a new package to stable?
> >>> No.
> >> I strongly disagree. Adding "stable" packages to stable is the primary 
> >> interest contributors are after. It's the #1 reason, why Fedora Extras 
> >> one had been launched.
> > 
> > Cool down and slow down, at least a bit, please.
> 
> Why should I? 

Because your reply above looks a lot as if you misunderstood the
question. And in case you didn't, it doesn't become clear why "stable
packages" should not spend 1-2 weeks in updates-testing?
Why rush? What do we win by doing it? Impatient users can get
the released (!) packages from the updates-testing repo.

There's nothing wrong with disagreeing. I just see that skipping
updates-testing leads to more problems.

> * Fact is: In the overwhelming majority of all cases, testing is 
> effectively a delay queue, and a PITA to developers.
> 
> * Fact also is: "Testing" doesn't help wrt. packagers doing a poor job.

Depends on the level of poorness. Ignoring feedback in bodhi and
ignoring negative karma in bodhi is bad, yes. Fortunately, there are
only very few half-assed packagers who don't take -1 karma votes serious
and who would show other PITA attitudes.
 
> * Fact also is: "Testing" occasionally is the cause of broken package deps.

How? Or, what do you mean?

> > Fact is that one packager plus one reviewer are not enough
> > to ensure that new packages really work good enough to call them "stable"
> > for one or more dist targets.
> So you are demanding for even more bureaucracy and for more @RH deitism? 
> This would be truely silly.

If packagers, who try to maintain a hundred packages, are concerned
about artificial hurdles and "bureaucracy", it's not my role to disagree
with them. Voice your concerns as much as you like.

I expect a committee of Fedora leaders to influence [or decide on] the
road to take. It is beyond my energy to fight endlessly and with
continuing intensity. It's easier to reduce activity and wish them
luck. With F9, for example, I asked myself "why spend any time at all on
running repository checking tools if some issues, such as broken deps or
conflicts, won't be fixed?" In many cases I'm willing to compromise.

A new Fedora dist release is made every six months. That's a rather short
period. I don't see what we win by skipping updates-testing. Give the
community what it deserves. If there's not a single user who gives
positive feedback about a new package [or a version upgrade], then why
rush and skip updates-testing? You may be confident enough to perform the
necessary testing yourself and take responsibility for breakage. You may
be competent enough to publish sane updates only. You may be experienced
enough to avoid poor mistakes. Still, there are other packagers who
release broken updates or who don't understand how to fix them. What can
be done to increase the quality of updates? Maybe it needs a privilege
bit like "provenpackager" to skip updates-testing? I don't know. I just
wish some people would be more conservative (or "shy") in what they
choose to push to multiple dists without any testing.

> As far as I am concerned, it's this freaking freeze and run-down shape 
> the fedora repso currently is in, which is causing the mess.

In case it's not known, I've never backed up the post-Core freeze
process. The long time of inactivity in Rawhide, the growing pile of
pending updates, the massive amount of zero-day updates, and the
increasing number of unknown broken deps in the uninstallable FN+1 Rawhide
have confused me and caused me to raise an eyebrow. Something's wrong
with how package developers follow Rawhide/Alpha/Beta (or the Test
releases previously) and when/how they start to prepare their packages
for the next Fedora while Rawhide is frozen. The thing simply is,
in a community of volunteers (with some Fedora staff being paid by
Red Hat and a muddy situation with Red Hat developers who happen
to maintain Fedora packages) I don't see any organisational structures
and hierarchies that would be possible in a company. I can't tell
any other Fedora person to do this or that, and I don't want to
deal with 





More information about the fedora-devel-list mailing list