/**
 @file shadowVolume.h
 
 @maintainer Morgan McGuire, morgan@graphics3d.com
 
 @created 2001-12-16
 @edited  2006-02-31
 */

#ifndef G3D_SHADOWVOLUME_H
#define G3D_SHADOWVOLUME_H

#include "G3D/platform.h"
#include "GLG3D/PosedModel.h"

namespace G3D {

/** See G3D::markShadows */
void beginMarkShadows(RenderDevice* renderDevice);

/** See G3D::markShadows */
void endMarkShadows(RenderDevice* renderDevice);

/**
 Given a posed model, renders a shadow volume using the algorithm
 described in:
 
   C. Everitt, M. Kilgard, "Practical and Robust Shadow Volumes", 
   NVIDIA Technical Report, 2002
   http://developer.nvidia.com/object/robust_shadow_volumes.html

 This algorithm <B>requires</B> models with no broken edges.  Shadows
 will not be rendered for models with broken edges.

 Note that faster methods of rendering shadow volumes have since been
 proposed, such as: 

   M. McGuire, J. F. Hughes, K. T. Egan, M. Kilgard, C. Everitt,
   "Fast, Practical, and Robust Shadow Volumes", NVIDIA Technical
   Report 2003
   http://developer.nvidia.com/object/fast_shadow_volumes.html

 Set up your rendering code as follows:

  <PRE>
    renderDevice->clear(true, true, true);
    renderDevice->pushState();
        renderDevice->enableLighting()
        // Configure ambient light
        renderDevice->setAmbientLightColor(Color3::white() * .25);

        for (int m = 0; m < model.size(); ++m) {
            model[m]->render(renderDevice);
        }

        renderDevice->disableDepthWrite();    
          for (int L = 0; L < light.size(); ++L) {
            beginMarkShadows(renderDevice);
                for (int m = 0; m < model.size(); ++m) {
                    markShadows(renderDevice, model[m], light[L].position);
                }
            endMarkShadows(renderDevice);

            renderDevice->setBlendFunc(RenderDevice::BLEND_ONE, RenderDevice::BLEND_ONE);
            renderDevice->setStencilTest(RenderDevice::STENCIL_EQUAL);
            renderDevice->setStencilConstant(0);
            renderDevice->setAmbientLightColor(Color3::black());
            renderDevice->setDepthTest(RenderDevice::DEPTH_LEQUAL);

            renderDevice->setLight(0, light[L]);
            for (int m = 0; m < model.size(); ++m) {
                model[m]->render(renderDevice);
            }
        }
    renderDevice->popState();
  </PRE>

  <B>Limitations</B>

    This function requires the far clipping plane at z = -inf (the
    default for GCamera).

    Assumes that render calls do
    not set the blend func or stencil funcs, or change the
    camera matrix or projection matrix.

    Transparent objects should not be used
    with shadow marking.
 */
void markShadows(
    RenderDevice*           renderDevice, 
    const PosedModel::Ref&    model,
    const Vector4&          light);

}

#endif