Bug#410695: zope2.7 causqe upgrade failure

[Steve Langasek]

On Fri, Mar 02, 2007 at 07:49:53PM +0100, Josselin Mouette wrote:
> I have already encountered similar problems in the past, with GConf, and
> they were triggered by a circular dependency.

> It seems that APT is unable to deal with such dependencies correctly, as
> it removes packages depending on each other in random order (which is
> fair) but can also remove one of them before packages depending on the
> other are removed.

Sorry, can you point to the dependency loop at work here?  I didn't notice
one in my examination.

> The solution is, in such cases, to make APT remove all packages
> depending on either of those in the dependency loop, then remove the
> loop itself. This is unfortunately far from trivial to implement, and -
> let's be realistic - impossible to do before the release.

Agreed.

As the recommended upgrade path from sarge->etch is aptitude rather than
apt, the main reason I've left this as 'serious' is concern that the cause
is common to both apt-get and aptitude and just triggers sooner with apt-get
(which seems to be the case if the problem is the dependency loop bug), and
that, in the absence of clear understanding of the origin of the bug, there
could be a significant number of other upgrade scenarios where apt would
fail.

But if this only affects circular dependencies, that seems unlikely.

> An acceptable workaround would be to upload a dummy zope2.7 package,
> which doesn't depend on python2.3, and would make the transition to
> zope2.9. At least, such upgrades wouldn't fail.

Based on discussions on the -python list, where users insist that it's
unacceptable even for zope2.7's removal to be forced on upgrade to etch, it
seems unlikely that this would be accepted either.

-- 
Steve Langasek                   Give me a lever long enough and a Free OS
Debian Developer                   to set it on, and I can move the world.
vorlon at debian.org                                   http://www.debian.org/