Getting egl clients to work

[Pekka Paalanen]

On Mon, 02 Jun 2014 10:46:29 -0700
Bill Spitzak <spitzak at gmail.com> wrote:

> Actually all I am trying to do is get software EGL to work. I did
> not think there was a chance at all that the nVidia drivers could
> be used. But maybe it is a possibility...
> 
> > Sounds like you are using Mesa (software renderer) on your
> > system otherwise using Nvidia proprietary stack (right?).
> 
> Yes
> 
> > At some point I thought you wanted to run Weston using Nvidia's
> > EGL to get hardware acceleration in Weston itself, not Mesa EGL
> > which on Nvidia proprietary drivers can only ever offer software
> > rendering.
> 
> Actually all I am trying to do is get mesa's software EGL to
> work, and only in the X11 backend. I did not think there was a
> chance at all that the nVidia drivers could be used...
> 
> My assumption was that both weston and any clients would load my 
> locally-compiled mesa, which would then figure out that only
> software rendering would work, and use that into SHM buffers.
> 
> It now looks like weston successfully does this, but clients fail?

Indeed yes. You can confirm it by looking at Weston's log (stderr),
where it reports the GL version string etc.

> > You should not install Mesa at all, if you wanted to use
> > Nvidia's libEGL, or otherwise make sure the correct libs get
> > loaded.
> >
> > Nvidia's libEGL can in theory work only with the x11-backend of
> > Weston.
> 
> That would be very interesting if it works...
> 
> > egl_dri2 is a built-in, not a file like egl_gallium.so. A
> > "side-effect" of egl_dri2 is that it does not implement the
> > generic software rendering path on the Wayland platform (i.e.
> > for Wayland apps).
> >
> > Again, for Wayland apps to be able to use a software renderer
> > from Mesa, Mesa has to install egl_gallium.so and the apps need
> > to use it. When this is done properly, it does not matter to
> > apps what EGL Weston itself is using.
> 
> Okay, this seems to be the problem I am having. For some reason
> the clients fail to use egl_gallium. Is this a bug? Is Mesa
> compiled incorrectly? Should I set an environment variable? This
> is all a mystery to me.

It is again the --enable-gallium-egl option to Mesa's configure,
that lets it build egl_gallium.so.


Thanks,
pq