DRI2 committed

[Michel Dänzer]

On Wed, 2008-02-20 at 14:09 -0500, Kristian Høgsberg wrote:
> On Feb 20, 2008 5:20 AM, Michel Dänzer <michel at tungstengraphics.com> wrote:
> >
> > On Fri, 2008-02-15 at 09:57 -0500, Kristian Høgsberg wrote:
> > >
> > > Anyway, the code is out there, shouldn't break anything unless you
> > > enable it, [...]
> >
> > Actually, it looks like you broke at least direct rendering for every
> > driver that doesn't provide __dri2DriverInitScreen because that symbol
> > is unresolved when libGL tries to dlopen them.
> 
> That doesn't sound right, I haven't touched libGL yet, it doesn't know
> anything about DRI2 at this point:
> 
>   [krh at hinata mesa]$ grep -i dri2 src/glx/x11/*.[ch]
> 
> Could you provide a few more details?

dri_util.c references __dri2DriverInitScreen and gets linked with the
driver. So if the driver doesn't provide __dri2DriverInitScreen, it's
unresolved at dlopen time.

A possible solution should be to move the DRI2 code to a separate file
like dri2_util.c and only link that with DRI2ified drivers.

On a possibly related note, someone else (Chris Rankin IIRC) reported
that even non-TTM drivers now reference TTM related libdrm symbols.


-- 
Earthling Michel Dänzer           |          http://tungstengraphics.com
Libre software enthusiast         |          Debian, X and DRI developer