[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kwin
Subject:    Re: Support for _NET_WM_DONT_BYPASS_COMPOSITOR hints
From:       drago01 <drago01 () gmail ! com>
Date:       2012-08-30 10:52:05
Message-ID: CAMqY-Fc8VF9mCWDZ2v1A8uRCHsyt6wMJW8B8KP_zkCwptEoeRA () mail ! gmail ! com
[Download RAW message or body]

On Thu, Aug 30, 2012 at 12:39 PM, Martin Gr=E4=DFlin <mgraesslin@kde.org> w=
rote:
> Hi,

Hi,

> sorry for having not answered your mail on the wm-spec mailing list, I so=
mehow
> missed it.

No problem.

> KDE has supported such a hint since our July release last year:
> _KDE_NET_WM_BLOCK_COMPOSITING

I know, I have noted that on the initial mail on wm-spec:
https://mail.gnome.org/archives/wm-spec-list/2012-February/msg00009.html

> We implement it in a different way as we do not use unredirection for it,=
 but
> stop the compositing completely when there is a window with such a hint.

OK, we can reword the spec so that the unredirection is an
implementation detail. (Even though technically disabling compositing
means that all other windows aren't redirected as well).

> There
> are multiple reasons to do so:
> * performance is still largely affected if the compositor still has an Op=
enGL
> context
> * unredirection mostly makes only sense for fullscreen windows which does=
n't
> take into account e.g. a video player you want to interact with (opening a
> menu would redirect it again) or do not have on fullscreen
> * unredirection is causing quite some pain with certain combinations of
> Distributions and drivers - that bad that we have considered removing
> unredirection support overall.
>
> When we added the hint we did not propose it for the NET_WM spec as we ar=
e (to
> my knowledge) the only window manager which still is able to be run witho=
ut
> compositing.
>
> If such a hint is added to the spec, I would like it to be written in a w=
ay
> that it would also cover our blocks compositing usecase. That is that the
> compositor can decide whether to unredirect or stop compositing completel=
y.

Yeah makes sense.

> Another suggested would be to add a further hint to the _NET_SUPPORTED se=
ction
> and adjust the text to say that a compositor MUST support the hint if it =
is
> announced in _NET_SUPPORTED. That way a Client can actually know whether =
it
> will have full resources or not.

I intentionally didn't do that because I didn't want to the client to
expect that (the compositor should be free to redirect the window
regardless of the hint when it feels the need to do so).
Hence why the spec says "the compositor may ignore this hint".

Thanks
_______________________________________________
kwin mailing list
kwin@kde.org
https://mail.kde.org/mailman/listinfo/kwin
[prev in list] [next in list] [prev in thread] [next in thread]