Please merge unattended-upgrades 0.96 (main) from Debian unstable (main)

Bug #1714019 reported by Balint Reczey
12
This bug affects 1 person
Affects Status Importance Assigned to Milestone
unattended-upgrades (Ubuntu)
Fix Released
Undecided
Unassigned

Bug Description

I'm also asking for a Feature Freeze Exception following the rules for new upstream version per https://wiki.ubuntu.com/FreezeExceptionProcess.

[Reason]
Unattended-upgrades is planned to be enabled by default but the versions present in Ubuntu can break the system by not protecting the upgrade process from termination during shutdown (see LP: #1690980).
The newer version in Debian fixes the shutdown-related issue and contains several additional reliability and debuggability fixes.

[Logs]
They are attached to this bug.

[Testing]
The autopkgtest log is attached and I also tested the version in Debian several times while preparing the fixes for various bugs.

Note that the package in the PPA does not close this bug from the changelog while the patches attaches do.

Balint Reczey (rbalint)
Changed in unattended-upgrades (Ubuntu):
assignee: nobody → Balint Reczey (rbalint)
Balint Reczey (rbalint)
Changed in unattended-upgrades (Ubuntu):
status: New → In Progress
Revision history for this message
Balint Reczey (rbalint) wrote :
Revision history for this message
Balint Reczey (rbalint) wrote :
Revision history for this message
Balint Reczey (rbalint) wrote :
Changed in unattended-upgrades (Ubuntu):
status: In Progress → Confirmed
description: updated
Revision history for this message
Balint Reczey (rbalint) wrote :
Revision history for this message
Balint Reczey (rbalint) wrote :
Balint Reczey (rbalint)
description: updated
Changed in unattended-upgrades (Ubuntu):
assignee: Balint Reczey (rbalint) → nobody
description: updated
Revision history for this message
Steve Langasek (vorlon) wrote :

https://bugs.launchpad.net/ubuntu/+source/unattended-upgrades/+bug/1714019/+attachment/4942097/+files/unattended-upgrades_0.93.1ubuntu8-to-0.96ubuntu1.patch appears to be a reversed patch, and doesn't have a changelog entry for 0.96ubuntu1.

Which parts of this change is it that you believe violate feature freeze and require an FFe? (Bugfixes do not require an FFe.)

If an FFe is required for the merge, why do a merge rather than cherry-picking of the specific bugfix changes?

From NEWS.Debian, I understand that a key change is to turn on the minimal steps handling by default. What is the impact of this change on the time it takes to do a typical update? Are we in any way increasing the risk of an unclean shutdown due to a laptop running out of battery during the upgrade, where this wouldn't have happened if we weren't doing minimal steps?

debian/postinst changes the minimum version at which installinit upgrade handling is applied, and will be reapplied to systems which have the existing artful package installed. Could this be a problem? E.g., does this forcibly override the user's preference on upgrade if they have disabled the service in systemd?

Changed in unattended-upgrades (Ubuntu):
status: Confirmed → Incomplete
Revision history for this message
Balint Reczey (rbalint) wrote :
Revision history for this message
Balint Reczey (rbalint) wrote : Re: [Bug 1714019] Re: Please merge unattended-upgrades 0.96 (main) from Debian unstable (main)

On Fri, Sep 1, 2017 at 2:07 AM, Steve Langasek
<email address hidden> wrote:
> https://bugs.launchpad.net/ubuntu/+source/unattended-
> upgrades/+bug/1714019/+attachment/4942097/+files/unattended-
> upgrades_0.93.1ubuntu8-to-0.96ubuntu1.patch appears to be a reversed
> patch, and doesn't have a changelog entry for 0.96ubuntu1.

I have fixed the patch it was reversed indeed, but contains the
changelog, just not at the beginning like it is usual for diffs.

>
> Which parts of this change is it that you believe violate feature freeze
> and require an FFe? (Bugfixes do not require an FFe.)

There are some changes which don't fix bugs but can be seen as
improvements, like:
   * Update the signal used to SIGTERM in the documentation
   * Fix incorrect example for Update-Days.
   * add pep484 type hints
  ...

I also prefer being cautious and asking for such permissions even they
may not be strictly required.

>
> If an FFe is required for the merge, why do a merge rather than cherry-
> picking of the specific bugfix changes?

The part which would be cherry-picked is like ~90% of the changes if
we want to have u-u be in a reasonably good shape in Ubuntu and having
the big delta was painful enough to maintain already.

I aimed at having a sync but dropping/integrating LP: #1649709 needs
more discussion at least with infinity, it seems.
See: https://irclogs.ubuntu.com/2017/08/01/%23ubuntu-devel.html ,
starting from 11:06

>
> >From NEWS.Debian, I understand that a key change is to turn on the
> minimal steps handling by default. What is the impact of this change on
> the time it takes to do a typical update? Are we in any way increasing

I have not tested the additional time because without enabling minimal
steps u-u can't be stopped gracefully after it started the actual
upgrade.
Thus i saw no decision point here.
It is worth measuring the impact and checking if we can speed u-u up
in that area, but IMO this optimization should not delay the
acceptance of this upload.

> the risk of an unclean shutdown due to a laptop running out of battery
> during the upgrade, where this wouldn't have happened if we weren't
> doing minimal steps?

When InstallOnShutdown not set apt-daily-upgrade.service does not run
u-u when running on battery:
http://sources.debian.net/src/apt/1.5%7Ebeta2/debian/apt-daily-upgrade.service/#L4

When InstallOnShutdown is set this is a valid concern and thank you
for raising it.
I've just prepared a fix to mitigate this risk:
https://github.com/mvo5/unattended-upgrades/pull/75/commits/fd65bb54556624aacac5212d9d21e3a66df251c3
>
> debian/postinst changes the minimum version at which installinit upgrade
> handling is applied, and will be reapplied to systems which have the
> existing artful package installed. Could this be a problem? E.g., does
> this forcibly override the user's preference on upgrade if they have
> disabled the service in systemd?

I have not changed the version check compared to 0.93.1ubuntu8, thus I
see no problem in this area related to FFe.

Balint Reczey (rbalint)
Changed in unattended-upgrades (Ubuntu):
status: Incomplete → Confirmed
Revision history for this message
Balint Reczey (rbalint) wrote :

I suggest not delaying the merge by waiting for the ConditionACPower=true change either.
I will cherry-pick the change if it is needed.

Revision history for this message
Balint Reczey (rbalint) wrote :

According to the discussion with apw there is no need for FFe for this merge.
https://irclogs.ubuntu.com/2017/09/04/%23ubuntu-release.html

Revision history for this message
Gianfranco Costamagna (costamagnagianfranco) wrote :

sponsored debdiff attached, please double check

Changed in unattended-upgrades (Ubuntu):
status: Confirmed → Fix Committed
status: Fix Committed → Fix Released
tags: added: id-596781a4fd7307546010c061
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.