[Spice-devel] [RFC spice-gtk v3 1/1] Gstreamer: Use GstVideoOverlay if possible

[Frediano Ziglio]

> 
> Hi,
> 
> 
> On 05/09/2018 06:39 PM, Frediano Ziglio wrote:
> >> Currently when gstreamer is used to decode a full-screen
> >> stream sent from the server, the decoding frames are being
> >> forced to RBGA format and pushed using appsink to be scaled
> >> and rendered to screen.
> >>
> >> Today most of the gstreamer sinks supports the GstVideoOverlay
> >> interface which allows to render directly from the pipeline to
> >> a window by a given windowing system id (i.e. xid). This patch
> >> makes playbin to use this feature if possible.
> >>
> >> Setting the DISABLE_GSTVIDEOOVERLAY environment variable will
> >> make gstreamer to avoid of using the gstvideooverlay interface.
> > Acked-by: Frediano Ziglio <fziglio at redhat.com>
> >
> > Took some time to properly test, tested combination of:
> > - using normal drawing and not;
> > - using mjpeg or h264;
> > - using h264 full screen;
> > - xorg or wayland;
> > - using vaapi (gstreamer-vaapi and intel drivers) and software.
> >
> > Had some issues on Wayland with RHEL 7, on client connection graphics
> > on VM fails, but happens also without this patch (checked multiple
> > times, is consistent), seems related to fast resolution changes due
> > to some issue with Wayland.
> 
> Thanks for your tests, currently in wayland environment
> old rendering path is used, so any issues with wayland
> are likely to exist also without this patch.
> 
> 
> >
> > Maybe there were some minor missing point but can be follows up,
> > the patch is a good step ahead, thanks for it.
> 
> There is one patch regarding the signal's description comment (as we
> talked on irc)
> so it won't be documented, do we want to apply it? (see attached)
> 

Yes, I remember, just I think is more a minor thing.
I would extend a bit the commit message, something like:

"This signal is not intended to be part of the public interface
but for internal spice-gtk usage so avoid to document to make
clear it should not be used.
This allows us to remove or change the signal at any time."

I would send an updated, separate, patch on the ML.

> 
> Snir.
> 
...

Frediano