[Pkg-openldap-devel] 2.4.7-5 away

[Russ Allbery]

Steve Langasek <vorlon at debian.org> writes:

> I've gone ahead and uploaded 2.4.7-5 to unstable.  Sorry for not sending
> any sort of "last call" beforehand, but there were enough known, pending
> fixes in need of upload that I felt it needed to be pushed out so we
> could clear these off and see if there are any other regressions left.
> So if I've botched anything in -5 I'm happy to follow it with a -6 as
> necessary.

I was happy to see the release -- I think it's great to get it out.

> This also applies to the addition of the new debconf error template.
> The template hasn't been translated at all before upload, nor has it
> been reviewed by the English l10n team; nor, for that matter, has there
> been any feedback from anyone on the team regarding whether this should
> be a debconf error at all. :)

I'm okay with it being an error.  I think we should probably add something
like:

    If you were using this option to disable SSLv2, it is no longer
    necessary; GnuTLS does not support SSLv2 at all.

> So - is there anything that should be remedied in a -6?

At some point, I do think it would be good to rename the source package to
just openldap.  I don't really care when, though.

Looking at the remaining open bugs, I suspect #309026, #314793, #340601,
and #418997 are now fixed with the combination of a much newer upstream,
fully supported GnuTLS support, and not having multiple LDAP libraries
floating around conflicting with each other.  They should probably should
be pinged.

I'm not sure what's up with #419222 (see the very last message).  Quanah
was going to ping upstream about it.

The double-free in #458830 is intriguing, but it may be gone in 2.4.7.

I get the impression that BerkeleyDB 4.6 is becoming more and more
recommended by upstream and we really should switch.  I'd love to get the
package built with the flags previously discussed in #421946, but I think
maybe we should switch even if that doesn't happen.

Not sure what to do about #451536 (ownership of cert and key on upgrades).

We should probably at least look at incorporating the smbk5pwd patch in
#443073.

Also, we need to push the symbol versioning for the shared libraries
upstream or at least discuss upstream what to do about that.  I didn't get
any response from upstream when I asked about the libldap vs. libldap_r
collision issue.

-- 
Russ Allbery (rra at debian.org)               <http://www.eyrie.org/~eagle/>