[cairo] glitz WGL backend

David Reveman c99drn at cs.umu.se
Mon Sep 20 18:24:39 PDT 2004

On Mon, 2004-09-20 at 22:37 +0000, Tor Lillqvist wrote:
>  > PS: could you mail me your code, so i can try out some things?
> Did you have any luck with the code I sent you?
> For me, the day after I sent you that, my code suddenly stopped
> working (after I had catched up with the CVS updates)... Just a black
> window.
> To make matters worse, I started getting crashes from nvoglnt.dll when
> calling glTexImage2D... (I have an ageing GeForce 2.) I dunno, could
> there be a bug in NVIDIA's code? It *seems* as if it tries to access
> textures past their limits (512 bytes form the start of the texture
> even if it is smaller). Sounds incredible that there would be such a
> bug in NVIDIA's code which surely has gone through lots of QA. Oh
> well, I put in a workaround, I always malloc() at least 512 bytes for
> the texture data ;-) Also preventing my code from finding the
> *_texture_rectangle extension(s) seems to help...
> Still, the black window problem prevailed. No matter how much
> debugging printout I added I couldn't figure out what's going on.
> Then suddenly today, after catching up with the (not that relevant to
> my problem, it seemed?) changes in CVS from today, it works
> again. Ugh... Although rendertest reports quite many of the cases "not
> supported". I could have sworn I earlier did see the projective
> transform tests work, for instance, but not any longer.

That's probably correct results. Your GeForce2 card doesn't support
GL_ARB_texture_border_clamp, right?

Without border clamp glitz can't do projective transformations, fill
type repeat, fill type reflect or fill type nearest. But glitz can do
fill type transparent without border clamp and that's much more
important as it's used for 95% of all rendering. Earlier versions of
glitz couldn't do fill type transparent without border clamp but could
instead do projective transformations, fill type nearest... which mean
more cases of "success" when running rendertest but not for the
important cases.


More information about the cairo mailing list