[compiz] LIBGL_ALWAYS_INDIRECT=1 patch fix - any reason why it's not included?

[Colin Guthrie]

Michel Dänzer wrote:
> On Wed, 2007-07-04 at 13:29 +0100, Colin Guthrie wrote:
>> I was just struggling with the problem of starting compiz while having
>> to use LIBGL_ALWAYS_INDIRECT environment variable.
>>
>> In our packages we have a patch that means we do not need to set this
>> variable as it searches also the glxServerExtensions.
> 
> This is becoming a FAQ...
> 
> An extension can only be assumed to work if it's in the result of
> glXQueryExtensionsString(). If you don't believe me, try using
> GLX_EXT_texture_from_pixmap with a direct rendering context (and free
> drivers) - doesn't work, even though it's listed in both server and
> client strings. Moreover, when forcing an indirect context via the last
> parameter to glXCreateContext(), the GLX API provides no means to find
> out that the extension would indeed work (glXQueryExtensionsString is
> context independent), so the LIBGL_ALWAYS_INDIRECT workaround is needed.
> See https://bugs.freedesktop.org/show_bug.cgi?id=8160 and some recent
> discussion on the mesa3d-dev or dri-devel list.

Right, I thought it was Hanno who wrote the patch I linked and indeed
this bug report confirms it.

I'm still unclear as to why compiz cannot use use patch attached to 8160
(and rediffed by me for currenct compiz git as of a few days ago)?

It works fine for me. I'm running compiz git right now without
LIBGL_ALWAYS_INDIRECT thanks to this patch.

I don't really know enough about how all this works to comment on it, so
are you saying that without the ALWAYS_INDIRECT workaround that
applications other than compiz will be affected? Or is it that the patch
will not work with *all* pure AIGLX implementations (and thus free
drivers as you mention above)?

Sorry if this is a stupid question, but so far I'm not finding any need
to use LIBGL_ALWAYS_INDIRECT on my intel setup (with the patch).

Col