Bug#1002651: Please cherry-pick the commits after the 3.38.6 release from upstream to bullseye-updates

[Amr Ibrahim]

Am Sonntag, dem 26.12.2021 um 20:18 +0000 schrieb Simon McVittie:
> 
> Which commits do you have in mind and what bugs do they fix? The more
> we backport, the higher the risk of regressions is, particularly if
> those commits aren't in a formal release yet.

Yes, they aren't in a formal release yet, however, since they come from
the gnome-3-38 branch, the risk is minimized I believe.


> 0db4ac73 "wayland/subsurface: Check for parent actor before
> unparenting",
> 60afd63e "wayland/subsurface: Handle node unlinking on parent
> destroyed",
> 58343f43 "window-actor/wayland: Remove subsurface actors on dispose"
> seem to be fixing crashes with native Wayland apps that use
> subsurfaces, and in particular Firefox. Is that correct?

Yes, that's correct. They fix this issue in particular:
https://gitlab.gnome.org/GNOME/mutter/-/issues/1892



> bb85f21d "xwayland: Fix condition for queuing a window to
> META_QUEUE_CALC_SHOWING" seems to be preventing unnecessary event
> messages from being sent to X11 apps in Wayland, so it's fixing a
> performance problem. Is that correct?

Correct. That's actually a regression starting from 3.36 due to a typo:
https://gitlab.gnome.org/GNOME/mutter/-/issues/1955



> 5a6b3c4b "wayland/data-offer: Accept final preferrence of drop
> destination" seems to be fixing a protocol error that would result in
> Firefox being terminated during Alt + drag-and-drop. Is that correct?

Yes, correct. It fixes this bug:
https://gitlab.gnome.org/GNOME/mutter/-/issues/1952



> 15ac84da "surface-actor: Fix unobscurred_fraction calculation" seems
> to be fixing an incorrect refresh rate when fullscreen apps in
> multi-monitor setups use Vsync. Is that correct?

Correct. It fixes this bug:
https://gitlab.gnome.org/GNOME/mutter/-/issues/1971



> For a9e60bad "wayland/surface: Switch order for calculating surface
> damage",
> aeda2cbc "wayland/surface: Use correct default viewport for surface
> damage",
> 0b26ad63 "wayland/surface: Do not uncoditionally process surface
> damage"
> I don't understand what bug they are fixing, so I am going to find it
> hard to "sell" this change to the release team: we can't weigh up
> benefit vs. regression risk if we don't understand the benefit. If
> you need these changes, please could you explain the impact to
> bullseye users of not having them?

Unfortunately I also don't have an explanation.


> I've also asked for clarification on the upstream MR
> <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2168>.

You were also mentioned in the backport MR:
https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2178