[Snap] Ship the native Wayland backend by no longer unsetting GDK_BACKEND=wayland
Categories
(Firefox Build System :: Third Party Packaging, task)
Tracking
(Not tracked)
People
(Reporter: jan, Unassigned)
References
(Depends on 1 open bug, Blocks 2 open bugs)
Details
(Keywords: nightly-community)
Remove https://github.com/canonical/firefox-snap/blob/1804379e439f2ab46d9fb628000e4f7e75478d4c/firefox.launcher#L55-L58 to ship MOZ_ENABLE_WAYLAND for Snap on Wayland.
If desired, $ DISABLE_WAYLAND=1 snap run firefox
allows going back to Xwayland.
Comment 1•1 year ago
|
||
I'm not sure what is the purpose of this bug, the problem is not unknown, but canonical was waiting on our CI have enough coverage of Wayland for us to be confident to give them the greenlight on this. We already have bug 1725245 blocking snap-wayland meta.
Reporter | ||
Comment 2•1 year ago
|
||
CI team bug 1725245 tracks testing EGL/Linux with newer packages because the CI is testing an explicitly unsupported configuration, it force-enables WebRender on blocked Intel DDX (bug 1710400). CI neither tests what most X11/Xwayland Ubuntu users are actually using.
- It doesn't block shipping Snap Wayland (this bug), as Snap Xwayland is worse than Snap Wayland.
- There are so many comments that MOZ_ENABLE_WAYLAND=1 fixed this and that.
- Xwayland/Nvidia should have never been shipped at all, it has been more fragile than Wayland/Nvidia. Wayland should be force-enabled on Nvidia.
- Could you take bug 1826330? Search for
if SNAP_NAME env var is not empty
. That's what's needed to fix Snap Wayland non-Gnome window icons.
Offtopic, but shouldn't be forgotten: From my perspective, re-introduced bug 1834589 (previously bug 1726923) unnecessarily delays shipping Wayland with highest confidence. It can be fixed by partially going back to previous code (bug 1834589 comment 14).
How few/many users who have enabled experimental fractional scaling on their desktop (instead of changing display resolution) get a Wayland protocol error? Why should sudden exiting be better than generating a crash report, knowing crash volume and trying to fix or work around it? Most cases have been fixed.
Comment 3•1 year ago
|
||
(In reply to Darkspirit from comment #2)
CI team bug 1725245 tracks testing EGL/Linux with newer packages because the CI is testing an explicitly unsupported configuration, it force-enables WebRender on blocked Intel DDX (bug 1710400). CI neither tests what most X11/Xwayland Ubuntu users are actually using.
- It doesn't block shipping Snap Wayland (this bug), as Snap Xwayland is worse than Snap Wayland.
- There are so many comments that MOZ_ENABLE_WAYLAND=1 fixed this and that.
the discussion on that matter has been done, and the outcome was and is still that we are not confident in moving firefox stable out of XWayland because we lack completely CI on wayland. This is not my call, but this is the current status.
- Xwayland/Nvidia should have never been shipped at all, it has been more fragile than Wayland/Nvidia. Wayland should be force-enabled on Nvidia.
- Could you take bug 1826330? Search for
if SNAP_NAME env var is not empty
. That's what's needed to fix Snap Wayland non-Gnome window icons.
I dont understand the bug and the required fix, and I am busy on more urgent matters
Offtopic, but shouldn't be forgotten: From my perspective, re-introduced bug 1834589 (previously bug 1726923) unnecessarily delays shipping Wayland with highest confidence. It can be fixed by partially going back to previous code (bug 1834589 comment 14).
How few/many users who have enabled experimental fractional scaling on their desktop (instead of changing display resolution) get a Wayland protocol error? Why should sudden exiting be better than generating a crash report, knowing crash volume and trying to fix or work around it? Most cases have been fixed.
I'm sorry I was just asking why there was this bug, please dont be offended.
Comment 4•1 year ago
|
||
Please don't fiddle with GDK_BACKEND env variable. It's used for testing only and breaks various components. Correct way how to enable/disable Firefox/Wayland is to set MOZ_ENABLE_WAYLAND variable to 1/0.
Reporter | ||
Comment 5•1 year ago
|
||
Snap sets GDK_BACKEND=wayland by default on Wayland. Firefox' launch script currently unset
s it.
$ sudo snap install gnome-terminal --edge
$ printenv | grep GDK_BACKEND
(=nothing)
$ snap run gnome-terminal -- bash -c "printenv | grep GDK_BACKEND; sleep 5"
GDK_BACKEND=wayland
Edit:
So far I'm only aware of bug 1522780. Which child processes that don't support Wayland should Firefox try to launch inside the Firefox Snap?
Description
•