Plan for tomorrow's FESCo meeting (2010-11-17)

Adam Williamson awilliam at redhat.com
Sat Nov 20 22:05:11 UTC 2010


On Sat, 2010-11-20 at 16:48 -0500, Tom Lane wrote:
> Adam Williamson <awilliam at redhat.com> writes:
> > I don't disagree with anything you say, but the question of what's more
> > important than testing an update is key. If an update's worth doing,
> > it's worth testing. This is pretty simple, and amply demonstrated by
> > Fedora history: if we allow people to push untested packages as official
> > updates to stable releases, we will break those stable releases, and
> > people who use them will be badly affected.
> 
> > If you don't think you can find a single person to test an update you're
> > about to make, ask yourself if it's actually worth doing the update.
> 
> I don't know about you, but I *do* try to test my packages before I ever
> push them.
> 
> The relevant question here is not whether package maintainers should
> spend time convincing themselves to their own satisfaction that their
> updates are OK.  One would hope they already do that.

Bitter experience suggests otherwise. If no-one pushed updates without
testing them, we'd never have had updates which broke the entire system
getting pushed.

>   The question is
> whether package maintainers should spend time trying to mobilize testing
> efforts above and beyond what they can do themselves.  As a maintainer,
> I do not consider that part of my job or my skill set.  I am *not* going
> to pester fedora-test-list for testers for every minor update.

That would be pretty useless, because people who read test list already
do all the discretionary testing they can. The only thing I'd recommend
sending to test list is an emergency poke for a *significant* update to
a *critical path* package which hasn't got any testing. 

What it would make sense for maintainers to do is talk to the people
they know use their packages. Also, as I said before, I think just
requiring maintainers to actually test their own packages would be
enough for many non-critpath packages, but that's my personal opinion
and not, apparently, current policy.

>   Even if
> I had the time and inclination to do that, I think if that list gets
> swamped by maintainers trying to drum up ad-hoc testing support for
> their updates, it'll probably lose its readership altogether.

Sure. As I said, it's not really what I'd expect maintainers to do.
People on test list already know what packages are available for
testing, and they test what they can.

> I don't know what the answer is, but telling maintainers they have to
> become test managers in addition to their other responsibilities isn't
> likely to improve anything.  It sounds to me like just another iteration
> of the same wishful thinking that created this policy to begin with.
> Inventing hoops for people to jump through does not result in more
> manpower becoming available ... more likely the opposite.

I don't think that's true. One of the goals of the policy was to reduce
the volume of updates that are pushed just because hey, it's easy to do,
right? So if the policy is discouraging people from pushing trivial
updates it's actually *achieving its goals*.

I don't think it's unreasonable to expect maintainers to share some of
the burden of testing code they want to submit to all Fedora users.
Again, if you're not willing to make minimal effort to find out if the
thing actually works, you probably shouldn't be shipping it. Part of the
idea here is that Fedora will be better off overall if we ship 500 well
tested updates than 2000 untested updates.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
http://www.happyassassin.net



More information about the devel mailing list