Bug 448839

Is anyone working on fixing this bug? Is this issue complex or easy to fix?

Created attachment 146739 [details]
(Wrong bug)

(In reply to Marcelo Bossoni from comment #2)
> Created attachment 146739 [details]
> heaptrack (400+ MB being used after closing all windows/apps)

Wrong bug.. how can I delete the attachment?

> Wrong bug.. how can I delete the attachment?

Done :)

The issue still occurs in latest stable plasma release 5.24.1 with latest NVIDIA driver, 510.54.

My issue looks somewhat different, as you can see from the attachment. I'm on AMDGPU (Ryzen 4500U).

Created attachment 147215 [details]
Blurry thumbnails

Restarting plasmashell fixes the issue for me.

(In reply to Michael D from comment #6)
> My issue looks somewhat different, as you can see from the attachment. I'm
> on AMDGPU (Ryzen 4500U).

Is this issue affecting all Plasma users then?

Seems it may not be nvidia-specific, if that's what you mean. Maybe Intel's unaffected.

I notice it is rare that I get black, blurry thumbnails--it's only happened to me once so far. Are yours always like that?

(In reply to Michael D from comment #10)

> to me once so far. Are yours always like that?

Yup. Its always like that, no matter what application I open or how many times I restart plasmashell.

Same thing for screen recording, which uses the same codepath under the hood

*** Bug 448944 has been marked as a duplicate of this bug. ***

Is the screen recording problem definitely the same as the thumbnails problem? I've never had blurry or black thumbnails but screen recording has never worked for me under Wayland.

(In reply to Eli T. Drumm from comment #14)
> Is the screen recording problem definitely the same as the thumbnails problem? 

Sort of. Window thumbnails and *window* recording are the same. This is fixed by https://invent.kde.org/plasma/kwin/-/merge_requests/2387

*screen* recording is still broken with nvidia

(In reply to Nicolas Fella from comment #15)
> (In reply to Eli T. Drumm from comment #14)

>This is fixed by https://invent.kde.org/plasma/kwin/-/merge_requests/2387
So, will this be back-ported to 5.24 or will we have to wait for Plasma 5.25?

> *screen* recording is still broken with nvidia
Any extra info on this? Any open merge requests for fixing screen recording in NVIDIA?

Window thumbnails, window capture and screen capture all now work with no issues on Plasma 5.25.90 using OBS Studio 28.0.1.
However with Firefox only window capture works, screen capture shows a black screen with a cursor (tested from https://mozilla.github.io/webrtc-landing/gum_test.html).

(In reply to sporif from comment #17)
> Window thumbnails, window capture and screen capture all now work with no
> issues on Plasma 5.25.90 using OBS Studio 28.0.1.
> However with Firefox only window capture works, screen capture shows a black
> screen with a cursor (tested from
> https://mozilla.github.io/webrtc-landing/gum_test.html).

I can confirm that output screencasting doesn't work with nvidia driver, but I'm not sure that it's a kwin bug. At the moment, kwin renders a frame in a gbm_bo, then imports that gbm_bo and copies its contents in pipewire's dmabuf. It works with amdgpu, but that doesn't work for some reason with nvidia driver

I cannot share neither windows nor screens on Firefox or Chromium. I only get a black screen (although with a cursor).

OBS Recording *does* work as well as thumbnails.

System:
Plasma 5.27.2
Frameworks 5.104.0
Fedora RPMs
Nvidia 1070
Nvidia Drivers: 525.89.02

There's a MR, which somehow didn't get linked here automatically: https://invent.kde.org/plasma/kwin/-/merge_requests/3827

(In reply to postix from comment #20)
> There's a MR, which somehow didn't get linked here automatically:
> https://invent.kde.org/plasma/kwin/-/merge_requests/3827

Yes, I would appreciate some testing. 

Note: This change above also needs https://invent.kde.org/plasma/kwin/-/merge_requests/3821 to be complete and work properly.

A possibly relevant merge request was started @ https://invent.kde.org/plasma/kpipewire/-/merge_requests/41

Git commit c325f8269a36694867938394af57af5500e87c71 by Jan Grulich.
Committed on 16/03/2023 at 11:28.
Pushed by grulich into branch 'master'.

Use a different API call to make importing DmaBufs work on Nvidia

Use glReadPixels() instead of glGetTexImage() to import DmaBufs into
memory. This makes Nvidia happy and it no longer gives us just black
screen.

M  +14   -1    src/dmabufhandler.cpp

https://invent.kde.org/plasma/kpipewire/commit/c325f8269a36694867938394af57af5500e87c71

Git commit 6a6438e681ac60d8ec4670b075ce77de0fdd2e11 by Aleix Pol Gonzalez, on behalf of Jan Grulich.
Committed on 16/03/2023 at 17:22.
Pushed by apol into branch 'Plasma/5.27'.

Use a different API call to make importing DmaBufs work on Nvidia

Use glReadPixels() instead of glGetTexImage() to import DmaBufs into
memory. This makes Nvidia happy and it no longer gives us just black
screen.


(cherry picked from commit c325f8269a36694867938394af57af5500e87c71)

M  +14   -1    src/dmabufhandler.cpp

https://invent.kde.org/plasma/kpipewire/commit/6a6438e681ac60d8ec4670b075ce77de0fdd2e11

Git commit 75c7689f678409b594e3f4ed0161aa12f2baf1ec by Jan Grulich.
Committed on 31/03/2023 at 12:31.
Pushed by grulich into branch 'master'.

Screencast: fix renegotiation when client fails to use DMABuf modifiers

When client fails to import a DMABuf with a certain modifier, it will
remove it from offerings on his side and offer a new set of modifiers
for us to use. In case the list ends up being small enough that even we
won't be able to use any of the modifiers, we would still keep these
modifiers on our list and offering them to the client. With this we
would end up in a loop renegotiating and never actually fallback to use
MemFD buffers instead. We should also offer modifier-less DMABufs as
in the past. Using these we just need to avoid offering them as they
should be used as the last option.

M  +25   -1    src/plugins/screencast/screencaststream.cpp

https://invent.kde.org/plasma/kwin/commit/75c7689f678409b594e3f4ed0161aa12f2baf1ec