182 pending F11 stable updates. WTF?

Michael Schwendt mschwendt at gmail.com
Mon May 11 10:20:57 UTC 2009


On Sun, 10 May 2009 23:04:23 +0200, Kevin wrote:

> Jussi Lehtola wrote:
> > Why not? The purpose of testing is to find out if the package breaks
> > something. In the case of a new package there is no functionality of it
> > to break; the only possibility of getting into trouble is if the package
> > itself hasn't been made correctly (clashing file names or so on), which
> > should be in any case picked up in the review.
> 
> But you still risk pushing out a completely broken new package, which
> doesn't reflect well on you. ;-)
> 
> One of the ways this can happen is if you inadvertently built your new
> package against a buildroot override which is not in stable yet. Or of
> course the package itself could be broken.

Just to acknowledge that this is not only theory, it has happened before
actually. Most recently with qtorrent in Fedora 11 stable updates, with
the needed rb_libtorrent only sitting in the koji buildroot. The opposite
has happened before, too, however. ABI-incompatible library updates marked
stable for one or more dists -- after not using rpmdiff, or deliberately
as a result of not knowing the releng buildroot override procedure.

Explicit Requires (with and without versions) also cause unresolved
dependencies regularly (on one or all archs) when pushing updates to
multiple target dists.

Other breakage is caused by automatically disabled features at
%build/%configure time, if a package is mass-built for multiple dists
without accurate/safe/fatal checks of build requirements and without the
packager skimming over build logs.




More information about the fedora-devel-list mailing list