Staying on GTK3 and GNOME 3.38 this cycle

Hey there,

The topic of what to do about the new GNOME started being discussed and after some consideration we decided to stick to GTK3 and GNOME 3.38 this cycle.

There are quite some moving parts in GNOME 40

  • The new shell design, is it going to be fully ready in one cycle? what’s the impact on our desktop and extensions? it’s likely that we will have design questions to resolve and non trivial code changes
  • GTK4 is out, Debian is packaging it (currently waiting on NEW review) so that part shouldn’t create extra work for us but then what’s the impact? is the new version stable enough? GNOME isn’t likely to transition fully over a cycle, does it bring risks?
  • yaru and the new GTK, how ready is it?

Those are topics we are going to spend resources on and ideally we would be helping to move things forward, but it’s already mid-cycle, we didn’t account for any of those and the team is already overworked.

We should be careful bringing any GTK4 update in the default installation until the theming question is sorted out at least. It plays nicely that Debian starts its freeze around now so it’s not likely that autosyncs will create issues for us.

Hopefully we manage to resolve enough of those questions in the remaining of the cycle to be in a better position to include the new versions when I -serie opens.


Hi there :wave:

My/our plan is like always, to wait until gnome-shell master has cooled down, then rebase Yaru on gnome-shell upstream theme and re-apply our design/color changes. So I think, as long as they are in the time of their cycle it’s doable.

Have a gtk4 yaru theme ready, only need to get this into our build. You can test it here:

I wouldn’t lose hope, as long as the build is set up, and I personally would need help there because I am a meson noob, things should work out

Edit: one thing though, I would re-consider the dock position on the left, I think this won’t work anymore with the gnome-shell40 as a dock on the left will hide parts of the workspace thumbnail



I know usual traditions about LTS but will Focal be upgraded to 3.38 ?
Groovy & Hirsute & 20.04 snap GNOME stack will use 3.38.

I know usual traditions about LTS but will Focal be upgraded to 3.38 ?

The question is offtopic there, please don’t sidetrack the discussion

Perhaps a PPA could be provided, like when GNOME 3 was new, for testing?

1 Like

I think I saw a developer of GNOME 40 working on a setting to allow dock to be placed on the left exactly because many distros use dock positioned on the left. But I don’t know if it actually got implemented in the end, in has to be checked first.

Yes, we are probably going to stage the changes we have in ppa, unsure what is going to be done this cycle though since one of the problem is the lack of resources

A post was split to a new topic: Ubuntu should make it’s own QT desktop

Maybe we will have an Ubuntu GNOME again? With pure GNOME 40 without changes?

Maybe we will have an Ubuntu GNOME again? With pure GNOME 40 without changes?

The unmodified upstream GNOME experience is available in Ubuntu itself since the desktop switched back from Unity to GNOME, no need of a Ubuntu GNOME respin, install gnome-session and select GNOME on the login screen selector.


Hey Frederik, thanks for the reply and details! We should be able to help with the build system changes, I will add a card to our backlogs to make sure it’s being handled.

It’s likely that some applications are going to transition to the new GTK version so having a compatible theme in the distribution would be useful (even if it’s not used by anything on the default desktop)

The dock position is indeed one of the design questions we are going to need to resolve before updating to the new shell version.


Something of a bummer, but ok, I understand. Regarding extensions, and this was discussed recently, lack of timely maintenance is a problem now, there is no need for GNOME 40 or 400 for that. I can think of other reasons for delaying the update, but if the ones that weight the most in the decision were dash-to-dock and desktop icons, it’s my opinion that the time is ripe for dropping them and track upstream more closely. Both extensions are sources of bugs that renew with each release and accumulate in Launchpad, sometimes embarrassing bugs. I can see a value in having a fixed dock but not when it makes windows in the overview show partially out of screen. I can see a value in having desktop icons but not when they are the culprit of performance regressions, look distorted, don’t play well with my auto-hiding dock and are lacking many essential features. Yaru, OTOH, is a great theme that’s well maintained and has never been cause of headaches, at least for me.


Well, I wouldn’t bet on it that the GNOME restructuring is ready in time for GNOME 40.

I can’t believe dropping the dock and desktop icons is even being discussed.


I think that the Ubuntu team should just abandon the dock extension and the desktop extension. Those two extensions are bug-magnets and have a maintenance burden on the team, (just look at their Launchpad bug reports), and they actually degrade the performance of gnome-shell in an Ubuntu session compared to a vanilla gnome session, especially the desktop extension.


I think that sticking to 3.38 would be a perfect moment to upgrade to desktop-icons-NG and move to Pipewire/Wayland and get screencasting working.


Not upgrading to gnome-shell 40 and GTK4 in 21.04 means that the new shell design and the new GTK4 toolkit are going to be widely tested in only one interim release, 21.10, before the next LTS release, 22.04.

So I suggest that the Ubuntu team just treat the next two interim releases as experiments to pave the way to the new shell and toolkit, and a chance for their users to adapt to the new shell, and switch to Wayland by default in order to shake as many bugs as possible by the time the next LTS is released.

There will be some not-very-happy users because of this, but that’s the way things happen in a limited resources team, and there is 20.04 LTS for however needs the old shell and a default X11 session.


Well, my guess is that we will see GNOME 40 early in the H+1 cycle instead of late in the Hirsute cycle. Thus the time to be tested will not be THAT different. And there is still a chance that the whole redesign will be postponed to autumn by GNOME itself.


Yes and yes. desktop-icons-ng instead of desktop-icons seems to be recommended by a lot of users. @seb128, that’s probably something we should look into(?)


it’s the backlog, it would be useful to ask upstream if they have recommendation on which variant is the preferred one though

1 Like