 |
Copied from external conversation: I think there should be some shadow response but specular also shouldn't cutoff right on the shadow edge. I think the correct thing to do is reduce the specular response in the shadow region rather than stopping it completely |
|
|
Yep, looking at literature specular or reflections will produce glancing illumination regardless of occlusion due to the reflection angle but the shadow will partially attenuate the reflection. |
 |
Translucent objects cannot have stencil shadows cast on them. They compute shadows in screen space once per light, so they can only compute them at solid objects. To support translucent objects, one would have to recompute all shadows individually for each translucent object. As for shadow maps, they support translucent object by some kind of bug/accident. Here is the code inside InteractionStage which disables soft shadows for translucent objects: if ( !translucent && ( vLight->globalShadows || vLight->localShadows ) && !viewDef->IsLightGem() ) { uniforms->softShadowsQuality.Set( r_softShadowsQuality.GetInteger() ); } else { uniforms->softShadowsQuality.Set( 0 ); } In case of stencil shadows, stencil test is used in non-soft case, which is basically disabled just before rendering translucents: qglStencilFunc( GL_ALWAYS, 128, 255 ); But in case of shadow maps, the shadow map still gets attached to the shader and thus takes effect (in hard code). Now the logical resolution to this situation would be to put shadows on translucents under cvar in case of shadow map, thus making default behavior consistent between two modes. |
|
|
Committed in svn rev 10647. I guess yet another reason to improve shadow maps in the future... sigh. |
- Anonymous
