XWayland on Xorg server 1.15

[Tiago Vignatti]

On 06/14/2013 11:20 AM, Quentin Glidic wrote:
> Hello guys,
>
> I rebased the xwayland-1.12 branch of xorg-server on top of the current
> git master (which is the 1.15 dev version IIUC).
>
> You can find it here:
> http://git.sardemff7.net/wayland/xorg-server/log/?h=xwayland-1.15
>
> I also updated the Intel and wlshm DDXs to build and run against it.
> http://git.sardemff7.net/wayland/xf86-video-intel/log/?h=xwayland
> http://git.sardemff7.net/wayland/xf86-video-wlshm/
>
> They are just *rebased* branches, I am no X11 expert, and I simply hope
> I understood the parts right. It was not a big deal so there is no harm
> is just throwing them away if they are not relevant any more.
>
>
> The old segfault of XWM is still there though.
> https://bugs.freedesktop.org/show_bug.cgi?id=59983
> The proposed semi-patch from Tiago was pushed here for testers:
> http://git.sardemff7.net/wayland/weston/log/?h=wip/xwayland
>
>
> The current state in my setup is:
> Intel DDX works with Weston X11 backend. It does not with the DRM
> backend (“Failed to auth drm fd”). SNA is not supported.
> wlshm DDX just render a black area for X clients.
>
>
>
> Now, what is needed to get the whole thing right?
> Tiago had some work on XWayland/XWM to make it better, and I do not
> remember which parts of the stack it touched, so I will consider it
> touches the Xorg server and the Weston XWM.
>
>
> As a short-term goal, I think pushing the temporary fix is a good idea.
> That would let some more people testing it, and hopefully finding bugs.
> Another thing to do is to rebase other DDXs, it should be quite easy and
> would make everyone up-to-date for the X server part.
> Regarding the Intel DDX, SNA support should be fairly easy. I am not
> sure it is worth it, but it deserve a quick test, IMO.
>
>
> As a long-term goal, we should rebase Tiago’s work on top on these new
> branches, if relevant. I can help a bit on this topic, but I am no X
> expert, so I may be useless here.
>
>
> Tiago, can you point me to your XWM work (git branch, patches on the ML)
> and possibly tell me what is needed to do on that, or if we should start
> for scratch (again? :-) ).

In short, this work is out of question now. No need to tackle that anymore.

The long version is that, a few months ago (maybe a year), we wanted to 
clean up the XWayland architecture and also fix a nasty deadlock on it 
[0]. Back there my approach was to separate XWM but Kristian later 
preferred to go with something different (avoiding one of the processes 
doing blocking calls when the other was also doing it [1]). I never 
liked what he suggested to be honest, but I worked on it anyways and 
then stopped my contributions on core Wayland afterwards.

Now, if you want to bring your rebase to upstream you basically have to 
coordinate together with Kristian only. I think it's a nice idea overall.


Thanks,

[0] 
http://cgit.freedesktop.org/wayland/weston/commit/?id=19dadf26171c18d9ad1e96c523a0e0dabd673d20
[1] 
http://cgit.freedesktop.org/xorg/xserver/commit/?h=xwayland-1.12&id=e4566c049c7b385e2307fef3c72ef61762729f42

Tiago