Re: [Celestia-developers] OpenGL 2.0

[Chris L. <cl...@ww...>]

> I just ran Celestia with the OpenGL 2.0 render path, and I got errors on
> the GLSL shaders. I don't recall ever getting errors there, just it not
> working. The errors seem to have line numbers associated with them.
> Hopefully this will be useful in correcting these finally, now that
> there is feedback! For the big error, it looks like the ATI glsl parser
> doesn't like multiplying vectors by scalars? Maybe all it wants is
> typecasting?

The problem is that compliant GLSL is extremely strict with types, but
NVIDIA's GLSL compiler doesn't enforce all of GLSL's rules. Unfortunately,
this means that errors tend to slip through . . .  The NVIDIA compiler
allows multiplication of vectors by scalars, and it automatically promotes
integers to floats as appropriate.  I wish that the NVIDIA compiler was
stricter.  Actually, what I really wish is that the authors of the GLSL
standard didn't insist on such strict type rules to begin with. It makes
programs harder to read--if I want to multiply a vec3 v by 2, I have to
write this:

vec3 v2 = v * vec3(2.0, 2.0, 2.0);

Whereas the NVIDIA compiler permits this:

vec3 v2 = v * 2;

But whatever . . .  I've modified shadermanager.cpp to emit compliant (and
thus more verbose) GLSL.  I'm not quite ready to check it in, however, as
I've been doing a lot of modifications to that file in order to support
improved mesh rendering.

--Chris