Odd Results with the MESA drivers and gl_Vertex

[LinGreg]

I'm not sure whether to post this here or in the linux forum. But here it goes...

I have a varying variable that saves the raw z position in the vertex shader as such:

ZPosition = gl_Vertex.z;

Normally, this works great using Nvidia's or ATI/AMD's drivers. But when I ran this under the Mesa's drivers, the x position got saved instead. As a matter of fact, the x position got saved using any of these lines:

ZPosition = gl_Vertex.x;
ZPosition = gl_Vertex.y;
ZPosition = gl_Vertex.z;

But produced sensible results when gl_Vertex was transform via:

gl_Position = gl_ModelViewProjectionMatrix * gl_Vertex;

This is very strange, esp. since according to glxinfo, this set of Mesa drivers should support up to opengl 2.1. Has anyone else had this issue with the Mesa drivers?

[kyle_]

If this is indeed driver bug, you may be able to work around by adding some bolierplate code.
Since you know that gl_Vertex is ok and more complex operations on it are OK too, try dot'ing its with a value that will get you what you need, eg:

uniform float zero;  // not set elsewhere
ZPosition = dot(vec4(zero,zero,1,zero), gl_Vertex);

As your matrix computations are correct, this has a chance of producing good result as well - it may be just Mesa getting confused with too simple case.
The point of the excersie is to show that there is a bug in the driver - this is hardly a good workaround.

Of course, since you didnt show your shader its hard to blame the driver before more is known.