[Pkg-openldap-devel] Package status

[Russ Allbery]

Steve Langasek <vorlon at debian.org> writes:
> On Sun, Jun 03, 2007 at 03:04:28PM -0700, Russ Allbery wrote:

>> We should probably talk about whether it would help anything with all
>> our important bugs about SSL/TLS problems to add symbol versioning to
>> libldap, at least the 2.3 version (I don't really want to touch the 2.1
>> packages very badly).

> I think that we should, for 2.3.  Adding symbol versioning to libldap2
> is only going to help if we think libldap2 will still be with us when
> lenny releases (or if we think it's important enough to resolve right
> now for unstable), since the symbols will only benefit software that's
> rebuilt against the new version of libldap2.

If we add symbol versioning only to 2.3 but not to libldap2, will that
help at all in resolving the weird problems we're seeing around TLS?  Or
do both of them have to be versioned to get much benefit?

I'm guessing upstream isn't going to be horribly thrilled with the idea of
symbol versioning, although I could be wrong.

>> For 2.3.35-3, it would be good to look at our backward compatibility
>> code and upgrade support and consider removing anything that only
>> applies to versions of OpenLDAP from prior to oldstable.  In a quick
>> look, I think we're carrying a lot of that code around that could be
>> dropped, along with a bunch of debconf templates.  We should decide
>> whether we're keeping fix_ldif around as well.

> Agreed.  I think we should drop fix_ldif from the package, fwiw; if we
> have a future need of it, it's always available in the svn history.

That's the course of action that makes the most sense to me, since it
would also let us close a bunch of bugs and I don't think it's doing a
great deal for us any more.

>> I'd also like to revisit patches/use-lpthread and see if we can get rid
>> of it now by fixing whatever other bug made it necessary in the first
>> place (or observing that the bug is already fixed), since it's the only
>> reason right now why we have to run all the autotools during build.

> Bug in the mips* toolchain; -pthread does not imply -lpthread when used
> together with -shared, so shared libs wind up with a missing ELF
> dependency.

> The bug in gcc remains unfixed after several years, I have no idea why
> that would be the case given that Debian's mips porters have been aware
> of it since the beginning.

Bleh, so this is still there.

I wonder if we can come up with a workaround that doesn't require patching
configure.  But I guess the current patch is known to be working.

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