[Pkg-openldap-devel] Bug#409923: Etch deserves upgrade handling for the ldbm backend

[Steve Langasek]

tags 392747 pending
tags 409923 pending
thanks

On Thu, Feb 15, 2007 at 10:49:42AM +0100, Tim Dijkstra wrote:
> You downgraded this bug. That's your call. But I do think this should be
> fixed for etch. 

> I think it is unacceptable that an upgrade of such a critical pacakage
> (for systems that use it) silently fails. It would be better to have
> proper support for upgrading a database with the ldbm backend, but I
> understand it is a little late for that.  The least you can do is add a
> note to prepare people for the disaster to happen.

> I just skimmed through the buglist and I think this is actually a
> duplicate of #392747: slapd: failed upgrade with ldbm backend The replay
> form Matthijs seems to suggest that there is some support for upgrading
> the database. Well apparently that is broken, it didn't happen for me,
> neither for Filip.

Bug #392747 shows debconf information indicating the user refused to allow
the migration from ldbm to bdb, so yes, the upgrade failed.

Your own bug report didn't include any debconf information, so I don't know
why the upgrade failed for you.  But there most certainly *is* a note
telling users that ldbm is broken.

I've taken a look at the package, and apparently the changelog claim that
this happens in the preinst is false because the code was never added to the
right place.

I hate our slapd maintainer scripts. :P

I've committed a fix for that particular bug, but that still doesn't explain
why you claim you received *no* indication it would break.  Even if it only
happens in the postinst, you should be prompted about the fact that ldbm is
dead.  I've reviewed all of the code here, and I don't see any bugs -- you
would have to explicitly refuse the conversion to end up in the state you
describe.  That's not a release-critical bug.

There also is a NEWS.Debian file in place in the package, which documents
this exact issue -- but the file name in the source package is wrong, so
it's not getting installed in the binary package.  I've now fixed this in
svn as well.

I'm not planning to upload this anytime soon, though.  Maybe someone else on
the team has time to do so.

-- 
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/