https://bugs.kde.org/show_bug.cgi?id=498628

Steven <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|FIXED                       |---
                 CC|                            |[email protected]
             Status|RESOLVED                    |REOPENED

--- Comment #6 from Steven <[email protected]> ---
This appears to have regressed and is reproducible for me on KWin 6.7.4. I have
two relevant monitors configured: the primary running at 180Hz and the mirror
running at 60Hz, and I am observing steam games running at 240Hz.

An LLM analysis pointed to this Plasma 6.6 mirroring change, where multiple
`BackendOutput` render loops share one `LogicalOutput`:

https://invent.kde.org/plasma/kwin/-/commit/8548b5fdc2540b4c093756176e97fcbb2ee92080

I have not performed an actual bisect, so that is just a source code driven
guess.

LLM summary: "The current code calls `SurfaceItemWayland::handleFramePainted()`
for each relevant `SceneView`. Each call can invoke `frameRendered()`, consume
presentation feedback, and call `clearFifoBarrier()`. Since both mirrored views
reference the same logical output, both can progress the same surface."

It looks like MR 9082 might address this problem by selecting one `SceneView`
to send callbacks and progress the FIFO:

https://invent.kde.org/plasma/kwin/-/merge_requests/9082

But it seems to have stalled out after hitting a test failure.

Disclaimer: An LLM was used to analyze my local behavior and discover this bug
and MR, however all the words in this comment, except for the quoted section,
are my own.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to