#2071 F30 Change: Firefox Wayland By Default On Gnome
Closed: Accepted 5 months ago by bowlofeggs. Opened 7 months ago by bcotton.

Firefox is going to run natively on Gnome Wayland session and won't use XWayland/X11 Gtk+ backend. This change affects Gnome only and won't be enabled for other Wayland compositors (KDE Plasma, Sway).


How exactly is this going to work? Is there just one launcher that uses Wayland if run on GNOME on Wayland, but uses X11 otherwise? It is not clear for my from the discussion on devel.

I'd love to see this, but based on personal experience, I have doubts that Firefox on wayland is ready for prime time. I have been running firefox-wayland almost exclusively for the last few months, and while @stransky is very responsive, there are still multiple visual glitches, mostly related to sizing of windows and subwindows on a HiDPI display. Some of those are serious enough to make the browser hard to use, for example some menus are only 25% visible. As a beta tester, those are annoyances that I can deal with, but I don't think we should make this the default version of firefox.

My sample has n=1, so it's possible that I'm just unlucky. Before the decision is made, I'd like some declaration on what the expectations are and what is the plan for dealing with regressions like this:
- what about usability-impacting regressions (compared to firefox on X11)
- what about visual regressions, e.g. noise when resizing, that do not impact usability
- how frequent the issues have to be before we treat them as important enough to delay this change

(And also: should I report the problems I see in Fedora bugzilla, or upstream? All or just the most important ones?)

I'd love to see this, but based on personal experience, I have doubts that Firefox on wayland is ready for prime time. I have been running firefox-wayland almost exclusively for the last few months, and while @stransky is very responsive, there are still multiple visual glitches, mostly related to sizing of windows and subwindows on a HiDPI display. Some of those are serious enough to make the browser hard to use, for example some menus are only 25% visible.

That's on my TODO list for next FF cycle.

As a beta tester, those are annoyances that I can deal with, but I don't think we should make this the default version of firefox.

Sure, we can switch back to X11 any time. Let's aim the goal and be ready for backup plan. I think the beta freeze is a reasonable check point for it. There's no intention to ship a broken browser, that won't serve to anybody.

My sample has n=1, so it's possible that I'm just unlucky. Before the decision is made, I'd like some declaration on what the expectations are and what is the plan for dealing with regressions like this:
- what about usability-impacting regressions (compared to firefox on X11)
- what about visual regressions, e.g. noise when resizing, that do not impact usability
- how frequent the issues have to be before we treat them as important enough to delay this change
(And also: should I report the problems I see in Fedora bugzilla, or upstream? All or just the most important ones?)

Please download latest mozilla nightly (https://www.mozilla.org/en-US/firefox/channel/desktop/) and launch with MOZ_ENABLE_WAYLAND=1. Report all issues upstream unless they're already tracked at https://bugzilla.mozilla.org/show_bug.cgi?id=635134

+1 here. I too have been using this for a while. There are a few glitches left here, but nothing that I would consider major (it's improved a ton from the first versions that had a lot of input issues).

After 1 week, I see a vote of +1,0,-0. Can this be added to the agenda for Monday's meeting?

Metadata Update from @bowlofeggs:
- Issue tagged with: meeting

7 months ago

I think the delay is mainly my fault. Let's vote in the ticket.

+1, but I think we need to be ready to revert back before F30 final.

I'm +1; I trust that @stransky will make reasonable calls as to whether this is ready for the bulk of our users or not.

So we're at +4, 0, 0 after 10 days. Do we need to discuss this during the meeting?

+1 sorry, thought I voted on this last week

Approved in today''s FESCo meeting. Processing the Change proposal as accepted.

Indeed!

  * AGREED: F30 Change: Firefox Wayland By Default On Gnome is approved
    but might get reverted if the glitches are not fixed in time for F30
    final (+8, 0, -0)  (contyk, 16:26:53)

Metadata Update from @psabata:
- Issue untagged with: meeting
- Issue close_status updated to: Accepted
- Issue status updated to: Closed (was: Open)

7 months ago

So, this feature did not make the 'code complete deadline' or the Beta freeze. A Beta FE has been requested.

During the review meeting we agreed we're willing to give it an FE if FESCo still thinks it should land in F30 at all, but we wanted to asked FESCo to check that again given a couple of things:

  1. There is a known bug affecting Wayland-native apps that use subsurfaces which would affect Firefox if we make it Wayland-native. This apparently affects screenshots, screencasting and animations.

  2. Firefox 66 is not scheduled for release until 2019-03-19, six days before the first Beta release target.

So, we would like FESCo to consider whether it still believes we should pull this Change into F30. If so, we believe it makes sense to try and get it into Beta and so the FE will be used.

Thanks!

@kevin @bcotton can one of you re-open this, or should I file a new ticket? thanks!

Metadata Update from @churchyard:
- Issue status updated to: Open (was: Closed)

6 months ago

Metadata Update from @jforbes:
- Issue tagged with: meeting

6 months ago

@stransky Is there any chance the bug with screenshots is going to be addressed by the release? Or is it known to be postponed?

@stransky Is there any chance the bug with screenshots is going to be addressed by the release? Or is it known to be postponed?

This is a bug in mutter component, not in Firefox. So you need to ask gnome folks about it. Please ask at https://gitlab.gnome.org/GNOME/mutter/issues/146

From https://gitlab.gnome.org/GNOME/mutter/issues/146

Jep, we're on it. Unfortunately a proper fix (not a rather hacky solution like !384 (closed) and gnome-shell!341 (closed)) requires some bigger changes. The first step to look out for is !409.

This will hopefully be ready by 3.34, but will not get backported to <=3.32. So for Fedora 30 the answer is no :(

P.S.: Note that there could have been a workaround from FF, which is to make the main window the root surface (and make the window-shadow a subsurface; currently it's the other way around, which is why window-screenshots only capture the window shadow, which again gets clipped out because it's outside of the reported window size). Unfortunately that's also not possible because of #440. So overall there are several areas where we can improve concerning subsurface support. And they are not trivial.

So mutter should be fixed in time for F31. The other options of either backporting the patches or doing some workaround in Firefox itself, are not appealing at all.

I have been testing upstream FF 66 nightly, and various issues that I had with hidpi scaling are still present. I haven't tested the Fedora rpms for 66, but I expect them to be the same.

Based on those two data points, I think we should postpone the switch to F31.

I have been testing upstream FF 66 nightly, and various issues that I had with hidpi scaling are still present. I haven't tested the Fedora rpms for 66, but I expect them to be the same.

No, Fedora Firefox 66 is not the same as upstream nightly 66, extra wayland patches are included. You can expect a final package tomorrow.

btw. you can use this build for testing:

https://koji.fedoraproject.org/koji/taskinfo?taskID=33603803

Also if you have any issues with hidpi scaling please report that at #BZ (redhat) - I'm not aware of anyone.

Firefox Wayland by default on gnome is deferred to F31, but also advertise some way to do a "preview" of the feature for F30 users. (+8,0,-0) (jforbes, 15:14:46)

Metadata Update from @jforbes:
- Issue untagged with: meeting

6 months ago

Builds with disabled Wayland backend are here:
https://koji.fedoraproject.org/koji/buildinfo?buildID=1237137

I'm also going to submit them as update for Fedora 30.

Metadata Update from @zbyszek:
- Issue tagged with: pending announcement

6 months ago

Metadata Update from @zbyszek:
- Issue tagged with: next release

6 months ago

The postponement announced along today's meeting agenda.

Metadata Update from @zbyszek:
- Issue untagged with: pending announcement

6 months ago

I'm going to close this since it has been accepted for Fedora 31.

Metadata Update from @bowlofeggs:
- Issue close_status updated to: Accepted
- Issue status updated to: Closed (was: Open)

5 months ago

Login to comment on this ticket.

Metadata