Desktop Team Updates - Monday 3rd May 2021

Hi everyone, below you will find the updates from the Desktop team from the last week. If you’re interested in discussing a topic please start a thread in the Desktop area of Discourse .

Last week’s notes are here: Desktop Team Updates - Monday 26th April 2021

1 Like

General Stuff

  • Setup Jira to track Enterprise Desktop projects
  • Research to integrate WSL2 with AD
  • Customer calls
  • OEM Stats


  • Adding support for command “adsysctl status”
  • investigated a bit an update-notifier segfault in hirsute, reported a bug with details and suggestions, reviewed and tested fix
  • SRUed an evolution-data-server cherry pick to fix imapx with the new version if iCloud
  • worked a bit on the opensc MIR review feedback comments (needed for smartcard)
  • created a git repository for the new installer design so we can browse back previous iteration easily
  • GNOME updates + SRUs: epiphany, gnome-calculator, evolution-data-server, gnome-disk-utility, evolution, evolution-ews, gnome-system-monitor
  • ddebs
    • investigated missing ddebs on focal, we don’t have enough log information to know what happened exactly but it seems the focal-updates archive got wiped out in february. Could be that it failed to read the packages index for some reasons and didn’t handle the error correctly then considered the list empty which led the cleaning job to remove anything that it though was not needed anymore.
    • added some debugging prints to the ddebs cleanup function for next time there is a such issue
    • tried to re-import the missing ddebs but failed so far because the query needed timeouts on launchpad
  • new desktop installer
    • started poking at installing the ubuntu-desktop-install snap on the canary image
  • others
    • Merged the SRU yaml report
    • Investigate vanished gjs/hirsute SRU, turned out to be a side effect of the new archive opening, restored the SRU and check if we had other similar cases, apparently not
    • Continued the post release launchpad triaging

General Stuff

  • Setup Jira to track Enterprise Desktop projects
  • Research to integrate WSL2 with AD
  • Customers calls
  • Prepare and uploaded SRU for ubuntu-settings on focal.


  • Reorganize config/consts split
  • Adding support for command “adsysctl status”
  • Attended sprint sessions (today, rest of the week)
  • Worked a lot more on impish opening, deployed the new autopkgtest (see many past status updates) and switched everything over to that. Quite a few followups were required but we’re more or less stable now.
  • The archive is open now for uploads.
    • but we’ve got a big arm64 backlog. There are a few issues there
      • A possible cloud misconfiguration (in our user’s quotas) is causing spawns to fail sometimes. I’ve got a ticket with Canonical IS to get that fixed, so that will hopefully stop soon.
      • Instances are often failng to reboot, could be a bug in the kernel or edk2. This means that packages which include many autopkgtests - a lot of rebooting - have a higher chance of failing like this. When that happens we requeue the test, but it might happen again multiple times. That’s affecting throughput somewhat. The kernel team is looking into it for us, hopefully we’ll have a resolution soon.
      • The links between the cloud where we request / process test results and the one where the tests are actually run went down multiple times on Friday, over the weekend and today (Monday). When that happens, the network connection is broken and autopkgtest fails the test run. It needs to be restarted then, which can set long running tests back a long way.
  • Spent most of my time focused on the network-manager snap, testing Lukas’ new patched keyfile plugin (test cases/results) with various locally built snapd snaps. There has been a lot of iteration and testing to determine the following:
    • User can’t make new interfaces, and turns out that NM is failing to call ‘netplan generate’.
    • netplan generate needs perms to write /run/systemd/system because last year netplan introduced a new OVS backend which always needs to write one single file, independently of the backend renderer: /run/systemd/system/netplan-ovs-cleanup.service. This is a permissions problem and causes network manager to fail to call netplan generate.
      • the correct solution (to write in a dbus call directly to generate) is hacky and would take too long to get the nework-manager snap with this fixed netplan landed in core20/stable by the end of May (the plan is to deliver an image with the netplan functionality via a patched keyfile plugin to customers by the end of May).
      • A quick and easy solution would be for the snapd team to just whitelist this binary in the apparmor rules but then that increases the attack surface so this method is receiving a lot of pushback.
      • Including all of libnetplan in the network-manager snap doesn’t work either because netplan generate still needs to be able to write to /run/systemd/system
      • So we are still working on a path forward.
  • cherrytree: trying to updated it to gnome-3-38 extension and core20. led to gnome-3-38-2004-sdk needing glibmm added so that’s been merged and is available in candidate now. but now we need gtksourceviewmm in the gnome-3-38-2004-sdk snap too and I’m working on getting this building as part of the build snap
  • LO bug triage: 1923305, 1926087


  • More thread safety for gjs, fixing a deadlock and further crashes when using objects using threads under the hood [MR]
  • Found and fixed a threads race on GObject toggle references notifications [MR]
  • Reiteration on GPatternSpec as boxed type and adapted to be fully introspectable [MR, merged]
  • Some cleanups on meson tests in GLib [MR]
  • Some more tiny fixes to gdbus tool [MR, merged]
  • Fixed a mozjs linking issue due to missing exported symbol [patch]
  • Tested Flutter input cursor issues due to GTK bug on 21.04.
  • Worked on Avahi Dart library.

firefox24 firefox

chromium22 chromium

flutter24 Ubuntu Desktop Installer

package24 other



  • core20 PR #102 was merged, adding mDNS hostname resolution support (on systems with Avahi installed). The snap on the channel seems to work correctly.
  • Opened snapd PR #10219, implementing a polkit interface. I’ve left it marked draft, as I’m working on spread tests to show the feature works. It is still in need of review.