[BPO] debhelper/13.6ubuntu1 from jammy to bionic, focal, impish

[Paride Legovini]

Please backport debhelper (>= 13.6ubuntu1) to Bionic and Focal. Currently we have

 debhelper | 13.5.2ubuntu1~bpo18.04.1 | bionic-backports | source, all
 debhelper | 13.5.2ubuntu1~bpo20.04.1 | focal-backports | source, all

Those versions are affected by LP: #1959054, which is fixed in 13.6ubuntu1. Especially bad about that bug is that it requires rebuilding packages which were built against the buggy debhelper to update their maintainer scripts.

I'm not preparing the upload myself as [1] specifies debhelper is handled by the Backports team.

[1] https://wiki.ubuntu.com/UbuntuBackports#Special_Cases

[Christian Ehrhardt  (paelzer)]

Subscribing Dave Jones who looked at this for Jammy.

[Dave Jones (waveform)]

Hmm, good point. This change ought to be back-ported and the same set of packages rebuilt otherwise an LTS->LTS upgrade risks stopping vital services like dbus too.

Assigning myself to have a look at this; looks like bionic and focal are only one minor version behind so a back-port shouldn't be hugely invasive.

[Thomas Ward (teward)]

Whoops looks like we had some collission in terms of assignments. Reassigning to waveform.

Once this is looked at and looks OK at the Canonical level, then the Backports Team will take over backporting this. Keep in mind that since the original Backports process is mostly scrapped, the Backports Team does have primary jurisdiction for these packages in the backports pocket under the new process under "Special Cases".

[Mattia Rizzolo (mapreri)]

Oh, that's probably a good reason to update the backport indeed, that I've been procrastinating.

Unfortunately at this time I'm quite busy so I wouldn't be able to do it before next Monday at the earliest. Personally I'm fine if somebody else take it and I'll later review the diff; I don't expect anything "interesting" in this task either.

[Thomas Ward (teward)]

waveform: Paride: given this bug, wouldn't this be more suited for full SRU and rebuilds? I'm wondering whether this is truly a backportable case because this requires rebuild of all affected packages, and if anyone builds with debhelper in backports (which is NOT available in standard build envs I believe easily) that will lead to the upgrade breakage.

[Mattia Rizzolo (mapreri)]

tbf, I don't think this is actually *urgent* and can really just stay there for another week.

[Mattia Rizzolo (mapreri)]

@teward: I'm pretty sure that bug only affects -backports: according to the debian bug the commit introducing it is 6067bc2f, which was first available in debhelper 13.4, which is only in jammy and -backports.

[Paride Legovini (paride)]

@teward: Mattia is right, at the moment only the -backports are affected, as the bug was introduced in Jammy.

[Thomas Ward (teward)]

Then this should be trivial to backport, if you need it backported waveform and you don't want to handle the actual backport uploads, I'll take this on, I have some spare cycles today this afternoon.

[Dave Jones (waveform)]

Having looked at the versions of debhelper in play, and the versions that introduced / fixed various bugs, my current reading of the situation is as follows:

All current bionic and focal versions in release, updates, and backports (11.1.6, 12.10, and 13.5.2) will be producing packages where the old version's prerm is responsible for stopping services when --no-restart-after-upgrade is selected. In other words, these are affected by debian bug 989155 (https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=989155) which was the ultimate issue that (indirectly) caused the fix in 13.6ubuntu1. *If* want to fix this in bionic and focal, a full SRU will be required (plus additional rebuilds, see below).

Additionally, the bionic and focal -backport versions (both 13.5.2) will be restarting services with --no-stop-on-upgrade, i.e. affected by debian bug 994204 (https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=994204) as presumably they include the original (problematic) fix for 989155 that was included in 13.4. This should definitely be fixed, so at a minimum a backport of 13.6ubuntu1 is warranted.

In the former case (SRU), we will additionally want to perform a no-change rebuild on all affected packages (all packages using --no-restart-after-upgrade and/or --no-{restart,stop}-on-upgrade). For reference, this was ~90 packages in jammy (I would assume similar numbers in bionic/focal).

In the latter case (backport only) we will want to perform a no-change rebuild on all packages that were built with the backport version although it sounds (given teward's comment #5 above) that this may be a minimal number.

Personally, I suspect this comes down to how many packages would be seriously affected by 989155. I know openldap was but are we planning on any SRUs for that in bionic/focal? If we are, we may wish to consider the full SRU option (however painful that might be). If not, the backport option is almost certainly preferable.

[Dan Streetman (ddstreet)]

> *If* want to fix this in bionic and focal, a full SRU will be required

agreed, and that would require its own [SRU] bug separate from this [BPO] bug

> I'll take this on, I have some spare cycles today this afternoon.

@teward did you get a chance to prepare this? I didn't see it in the upload queues yet

[Dan Streetman (ddstreet)]

as impish is a non-LTS release and will EOL next month, marking as wontfix.

[Dan Streetman (ddstreet)]

uploaded to -backports queues for bionic and focal.

[Mattia Rizzolo (mapreri)]

Thank you Dan!

I accepted both uploads!

[Mattia Rizzolo (mapreri)]

I think, looking at the list of packages currently in the backports pockets, nothing need rebuilding for that bug, right?

[Dan Streetman (ddstreet)]

Right, I don't think anything needs rebuilding.