https://bugs.kde.org/show_bug.cgi?id=513785

--- Comment #4 from [email protected] ---
(In reply to David Edmundson from comment #3)
> I agree it's important, I could close this so quickly because we have
> previously discussed it all. 

I was not able to find anything abou this. I searched for any mention of the
protocol here, on the forum, and using multiple search engines and wasn't able
to finf anything. I also briefly looked for a general statement of KDEs policy
regarding wayland protocols, and didn't find anything, but there i wasn't as
thorough and definitely could have missed something.

> Me too, but ultimately we need portals for the stuff that's not wayland
> related (file open etc.) so have to do them anyway. Us supporting anything
> else is increasing an ecosystem split not fixing it. 

I can understand that for a view limited to KDE, but as many other compositors
have or are working on implementing the protocols, i don't think anything will
improve. Are we cursed to forever having a split between Portals-using DEs and
Protocol-using DEs?
Would it not make more sense to offer both from the Compositor/DE side, to
avoid application/library developers needing to implement both systems and
decide which to use at runtime, which can lead to issues? (issues like this
one, in a library i was considering
https://github.com/nashaofu/xcap/issues/227)

> This is something we are fixing both in KDE and upstream which benefits
> everyone.
> Out of curiosity, do you maintain one of these relevant apps?

Not really, i was working on a side-project which needs rare (every few
minutes) but reasonably low latency screenshots/image buffers of the  display
when i came across the fact that its been quite a while since the protocol was
merged into wayland-protocols.

As for some more on my pain points:
Since i opened this i switched from using a screencapture library to rolling it
myself using a portals library and pipewire bindings, which led me to lern
about session restore tokensz which alleviate the "constant popups" pain point.

The main functionality pain-point (besides the surrounding ones like being
forced to depend on pipewire) is the lack of a way to take screenshots to
memory/fd/drm-buf. The screenshot portal always saves to a file on disk, which
is nice if thats what you want, but horrible for latency when you need to do
anything to the screenshot beyond saving it.
To solve this in my application, i start a full screencast through pipewire
which has a callback that puts the raw frame data into a channel to send it to
a thread which only exists overwrite the last stored frame with the new one.
This is quite a hassle and a giant waste of CPU cycles just to be able to get a
reasonably up-to-date frame with low latency. Heres the project, if you care:
https://github.com/laundmo/wf_overlay/

But i think i've talked in a closed issue for long enough, and will at some
point take the time to write a new one about the screenshot API lacking support
for *not* writing to disk. I'm frustrated sith KDEs decision on this, but at
this point it seems pretty set in stone.

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

Reply via email to