182 pending F11 stable updates. WTF?

Kevin Kofler kevin.kofler at chello.at
Mon May 11 21:03:01 UTC 2009


Michael Schwendt wrote:
> So, with the help of buildroot override you've had a complete set of Qt
> 4.5 builds in testing (as repoclosure has not detected related breakage),
> but somebody pushed only parts of the builds to stable? (instead of
> grouping all builds in a single bodhi ticket)

Qt 4.5 was not a soname bump, it was a basically backwards-compatible
release, only some stuff had to be rebuilt against it (thankfully;
otherwise it wouldn't have been possible to do the upgrade, there's too
much stuff using Qt!), and we needed the buildroot override for that stuff.

What happened is that people inadvertently built Qt-based packages which did
not require a rebuild and thus weren't on our radar against the Qt 4.5
buildroot override, then pushed them to stable while Qt 4.5 was still in
testing (which doesn't work because the compatibility is only
unidirectional: you can build against 4.4 and run on 4.5, except for a few
special packages which required the buildroot override, but you can't build
on 4.5 and run on 4.4).

What needed the buildroot override:
* some packages (mostly KDE packages, and we had to rebuild those anyway for
the 4.2.2 upgrade) which had build-time conditionals on the Qt version
controlling some Qt-version-specific workarounds (some of those required
some new API in Qt 4.5 to fix issues with Qt 4.5, others just dropped
Qt-4.4-specific workarounds when built against 4.5),
* 1 package which segfaulted when built on 4.4 and run on 4.5, rebuilding
fixed it (apparent silent Qt ABI breakage, but it only affected 1 single
package),
* a few packages which require Qt >= 4.5 to build (current Arora which uses
the latest QtWebKit features, Qt Creator etc.),
but by no means all Qt-using packages needed a rebuild!

        Kevin Kofler




More information about the fedora-devel-list mailing list