A simple accelerated GL compositor

[Felix Bellaby]

Tomasz Torcz wrote:
> > > You might still find that my code will work without transparency on your
> > > platform. Try setting GLX_ALPHA_SIZE to 0 where it appears.
> 
>   Tried that. Your composer runs, but screen goes black, only mouse
> cursor is displayed and I can move it. By pressinc ^C I can return to
> normal drawing.
>
> > Could you try changing GLX_DRAWABLE_TYPE (line 1154) to GLX_PIXMAP_BIT
> > (i.e. drop the redundant GLX_WINDOW_BIT) ? This _might_ get you past the
> > ChooseFBConfig even with GLX_ALPHA_SIZE set to 1 (line 1152).
> 
>   Tried that, too. It doesn't work in way similar to very first failure.

Okay, a blank screen may not seem like much, but I may know enough now
about the FBConfigs on your card to try reproducing its capabilities on
my own hardware. I need to think through how the capabilities of your
card are likely to effect the code before we can get much further. I
hardly expected that code written to use radically different hardware
would draw properly and a blank screen was always a probably outcome. 

> Today I haven't patience to sprinkle glcompmgr.c with printf(), but I
> suspect the same failure. On the side note, slightly more debugging
> output (or any output) from glcompmgr could help.

Setting DEBUG to 1 (line 65) should list every non-damage X event and
error. Try running it like that and press alt-tab a few times before
killing the program from a VT. If you get a list of Xevents with no
errors then there is probably little that could be learned from any
further debugging output. If you get some Xerrors then I would like to
see the output. If you get nothing then things are badly wrong.

I will add explanatory output into the code to detail any failures
during program initialisation. 

> > If you find that setting GLX_ALPHA_SIZE to 0 (line 1152) remains the
> > only way to get a valid FBConfig then either the mga drivers do not
> > provide access to the alpha buffers or there aren't any.
> 
>   I will try with Matrox closed drivers. But I fear that they are to
> very old X release and may now work with 6.9.0.

It would be very interesting if they did, but your doubts are probably
justified.

Felix