FESCo wants to ban direct stable pushes in Bodhi (urgent call for feedback)

Michael Schwendt mschwendt at gmail.com
Sat Feb 27 19:18:22 UTC 2010


On Fri, 26 Feb 2010 16:45:36 -0500, Bill wrote:

> > > To phrase a strawman differently:
> > > 
> > > "No update is pushed to users without verification and testing from entities
> > > other than the packager."
> > 
> > No, thanks. The "popular"/"high profile" packages will get their usual
> > rushed +1 votes in bodhi (from people who even download from koji without
> > waiting for an entire package set to be published in a repo - from people
> > who regularly vote +1 even when something is clearly broken). And
> > less-popular packages will suffer.
> 
> Again, that's just a "testing isn't working as well as we want right
> now, so let's just not bother with it at all because it might
> incovenience me" response.

Not at all. Go and take a look at how I've used testing/stable pushes
before. I just fear that I will be degraded to a package monkey, who must
obey more and more rules - arbitrary rules - just to please an updates
system OR the people who love needless bureaucracy (such as updates-testing
for F13 development, IMO).

When I submit an update request, I want to be done with it, so I can move
on and focus on more important stuff. I don't care whether any integrated
package sanity checks are run on the submitted packages, or whether it
takes a day or several days for a push to happen. If something is wrong
with the package and a person or tool reports it to me, fine, I'll take a
look. Provided that it's no silly rules like Notes field is empty (even if
bugzilla ticket links are present and %changelog contains good entries),
a missing link to an online ChangeLog, enforced delays before something
may be marked stable, or mandatory positive karma from testers.

> If that's the sort of defeatist attitude people really want to take,
> it's sad.

Sad?  It's sad if one cannot earn more trust and retain the freedom to
make the most out of existing tools and infrastructure.

Some of the off-topic parts of this thread are unbelievable.

> > > Consider it a second eye.
> > 
> > If there is a volunteer tester, who also takes responsibility when not
> > noticing regression caused by an update, fine. If there is no volunteer,
> > who will lend the update submitter a second eye? Either there are
> > resources or there aren't.
> 
> I said entities above. Could be testers. Could be releng. Could be a
> battery of autoQA tests.

Three times "Could". Let's talk about it when you know something definite,
please, but before it will become another hurdle.


More information about the devel mailing list