Re: view/renderer handling of surface shaders

[James W. <ja...@fr...>]

Roger Holmes wrote:
> On 1 Feb, 2007, at 17:27, James W. Walker wrote:

>> It still seems to me that the view could have a single concept of
>> current surface shader (as the IR does), which is the most recently
>> seen surface shader, subject to pushing and popping when you enter/
>> leave a group and when you start/finish a geometry.
> 
> But pushing and popping the graphics state is a slow operation, you  
> would
> not want to do it for every geometry, and certainly not for every  
> part of
> a geometry (you can have an attribute set on every vertex, edge and
> triangle for some geometries, though what a surface shader does for
> a point I wouldn't care to guess, but I think I have seen them on edges
> when we were first developing 3DWorld).

We're already pushing/popping for retained geometries that have 
attribute sets containing surface shaders.  See 
E3Renderer_Method_SubmitGeometry.

Pushing/popping the view state does not update everything, the view does 
make an effort to update only aspects that have changed, using the 
stackState mask.

The test file that I was getting incorrect results with, 
texture-interference.3dmf, contains an attribute set containing a 
surface shader.  I added another test file, texture-interference-2.3dmf, 
that has a free-floating texture shader.  The second file was working 
correctly.  However...

In IRGeometry_Attribute_Handler, the shader contained in an attribute 
set attached to the geometry was being submitted using

	Q3Object_Submit( *theShader, theView );

As a test, I changed that to say

	TQ3AttributeSet	tempAtts = Q3AttributeSet_New();
	Q3AttributeSet_Add( tempAtts, kQ3AttributeTypeSurfaceShader, theShader );
	Q3Object_Submit( tempAtts, theView );
	Q3Object_Dispose( tempAtts );

With that change, texture-interference.3dmf worked correctly, but 
texture-interference-2.3dmf failed.  This makes me more convinced that 
it's just causing confusion having two surface shaders in the view state.

I have removed the second surface shader from E3View.c.  Now both of my 
test files work, and I haven't seen any other problems.  Of course, let 
me know if you see any bugs.
-- 
   James W. Walker, Innoventive Software LLC
   <http://www.frameforge3d.com/>