https://bugs.kde.org/show_bug.cgi?id=522988

--- Comment #5 from Bjorn <[email protected]> ---
(In reply to Zamundaaa from comment #4)
> The root cause is the same.
> 
> > GL context loss is still there: "Could not delete render time query because 
> > no context is current"
> That's not a context loss, it's a normal thing that happens when a QtQuick
> effect closes.
> 
> > KWin restarts compositor with new pid
> What do you mean "with new pid"?
> 
> > and PipeWire still negotiates Modifier: 0x0 (no dmabuf)
> Modifier 0 doesn't mean "no dmabuf". What are you basing this stuff on?

Thanks for the corrections.
Regarding the new PID, after the GPU hang/engine reset, kwin_wayland restarted
with a different PID (from 3237 to 2964 in the journal), suggesting a full
compositor restart.
Regarding Modifier 0x0, I was assuming that meant linear format with no dmabuf
zero-copy, but I may be wrong about that. Could you clarify what Modifier 0x0
means in the PipeWire negotiation context, and whether the current behavior is
actually incorrect? The main symptom I'm seeing is OBS screen capture causing
near-100% CPU usage on all 20 cores at any resolution (only tested 1080p and
4K)

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

Reply via email to