as discussed in the ldap docs, we have diverged from upstream userdir-ldap. remerge with the latest release, ideally reducing our diff to zero by contributing our patches back upstream.
this might imply some code changes to override some templates locally; through puppet instead of through the package, for example.
note that in #40182 (closed), i reduced that diff quite a bit, and submitted a pile of changes upstream. i haven't heard from DSA about those changes, unfortunately, but at least that part of the work was done. when I did those changes, i documented the new branching model and diff with upstream in the LDAP docs. to quote that document:
usedir-ldap-cgi fork status
In the last sync, usedir-ldap-cgi was brought from 27 patches down to 16, 10 of which were sent upstream. Our diff there is now:
The large number of inserted lines is because we included the styleguidebootstrap.css which is 11561 lines on its own, so really, this is the diff stat if we ignore that stylesheet:
The only way forward here is either to make the "Debian" strings "variables" in the WML templates or completely remove the documentation from userdir-ldap-cgi (and move it to the project's respective wikis).
For now, we have changed the navigation to point to our wiki as much as possible. The next step is to remove our patches to the upstream documentation and make sure that documentation is not reachable to avoid confusion.
We have 4 patches there, and a handful were merged upstream. The remaining patches could probably live as configuration files in Puppet, reducing the diff to nil.
so the first target here should probably be to finish the latter, userdir-ldap. then the more difficult part is the -cgi stuff, where we should probably just strip the tricky docs out of the web interface, pointing to a (configurable) external document, since it's where most of the diff resides.
We seem incapable of getting our changes merged upstream at this
point. Numerous patches were sent to DSA only to be either ignored,
rewritten, or replaced without attribution. It has become such a
problem that we have effectively given up on merging the two
code bases.
We should acknowledge that some patches were actually merged, but
the patches that weren't were so demotivating that it seems easier to
just track this as a non-collaborating upstream, with our code as a
friendly fork, than pretending there's real collaboration happening.
i've actually done a merge of dsa/master in our master branch, but i'll deploy that in prod only tomorrow, and only after we merge the good work me and @lavamind did on fixing thunderbird support.
i've deployed the new merged version to production now and sent the tb/sequoia patch upstream. punting this back into the icebox until we hear from DSA.