binary NEW processing: warning signs, pitfalls, best practices?

[Martin Pitt]

Steve Kowalik [2010-02-17 16:13 +1100]:
> I think the queue tool (at least on the command line) could be a little
> smarter about imposing the OGRE model, if accepting the package is going
> to break it -- that is, the source is in universe, but one or more of
> the built binary packages are in main, it should not allow that
> particular binary package out of NEW and error out.

That would be very helpful indeed. It's a situation which can easily
be detected automatically.

> I think we can attack this problem from two vectors, one in making queue
> a little smarter, and two, changing our procedures to make sure this
> doesn't happen again.

We probably still need to do the latter, since the OGRE check won't
stop us from accepting sources which do the same thing in universe, i.
e. clobber another package's binaries.

> sync-source is (correctly) more brutal about this, and won't allow a
> package to be synced if it is going to take over a binary package built
> by another source package.

Right, it could just behave the same way. THen you can check the
situation, and if it's correct, --force it.

Martin

-- 
Martin Pitt                        | http://www.piware.de
Ubuntu Developer (www.ubuntu.com)  | Debian Developer  (www.debian.org)