[Pkg-openldap-devel] Bug#665199: slapd: fails to install, remove, distupgrade, and install again

[Ryan Tandy]

Hi Niels,

On Sun, Apr 09, 2017 at 08:54:00AM +0000, Niels Thykier wrote:
>A quick review on my part found nothing and I am going to ask the SRMs
>if this approach is ok (usually we are not much for relying on people to
>upgrading to the latest point release before a major upgrade).

The only case where a successful upgrade to stretch would require 
updating the jessie package first is if the user has the ppolicy schema 
installed *and* they follow the install-remove-upgrade-install path.

Any other upgrade case should work without a point release (or at least 
not regress).

> * Can we trivially document the steps so the user / admin can perform
>   this manually (assuming they do not upgrade their stable version
>   first)?
>   - If so, then I will add that to the release notes.

Ah. But the manual step is to slapcat the config database and save it in 
a specific location. If slapd was removed, they no longer have slapcat! 
And if they install it again before upgrading, they may as well get the 
point release version.

If we really want something for the release notes, I'd almost be more 
inclined to suggest that users who have removed (but not purged) slapd 
should back up their config (/etc/ldap/slapd.d) and data (/var/lib/ldap) 
and purge slapd before they try to install it again. And as above, this 
is only needed at all for ppolicy users - non-ppolicy setups should be 
fine.

(There's a different set of manual steps required for users who have 
ppolicy *and* replication or overlays, not related to this bug. My next 
priority after this is to send in some release notes text for those.)

> * The next point release is probably going to be later this month.  The
>   current (but not yet announced) dates means that the stable fix
>   probably has to be uploaded before the end of the next weekend at
>   the latest.

I would be happier if the patch can soak in stretch for a little bit 
first, even if that means missing this point release - again, due to the 
very tiny portion of users actually affected by the bug compared to the 
risk of the areas the patch touches. Considering my points above, do you 
still think it's urgent to have it in this point release?

Thanks!