182 pending F11 stable updates. WTF?

Michael Schwendt mschwendt at gmail.com
Mon May 11 19:00:49 UTC 2009


On Mon, 11 May 2009 12:51:15 -0400, Orcan wrote:

> On Mon, May 11, 2009 at 12:19 PM, Michael Schwendt wrote:
> > On Mon, 11 May 2009 15:08:49 +0200, Kevin wrote:
> >
> > [buildroot overrides / directly pushing to stable]
> >
> >> It has also hit a few packages with Qt 4.5 in F9/F10 updates (even though I
> >> did warn about this issue on fedora-devel-announce! So we also need to make
> >> sure that people read those announcements before submitting builds!)
> >
> > Of course. People doing updates must be careful in general. If all
> > ordinary updates were pushed to updates-testing first, this would reduce
> > the breakage.
> >
> 
> Some packages pull ~200 other packages (on top of the minimal build
> environment) through BR's during building. Do you really expect
> packagers to check the versions of all packages being pulled and
> compare them with the actual packages in the repos?

What kind of question is that?

I expect them to examine the update builds with tools like rpmdiff, to
skim over the build log, and to install their own package releases and
updates as soon as they appear in the updates-testing repo. And if that is
asked for too much, I expect them to reduce the number of mass-updates
that enter "stable" without community feedback.

Don't you realise that the lower the number of version upgrades, the
lower the risk of undetected breakage? Crucial changes enter a package
with version upgrades (unless a package applies huge/many patches).
Blindly pushing upstream's version upgrades bears a risk.

Library SONAME changes, for example, cannot poison the buildroots unless a
package is marked stable or gets tagged by releng. Buildroot stability
reduces the potential for breakage.




More information about the fedora-devel-list mailing list