Fixes to get shadowing working ... enjoy!

Guys,

I've noticed a few people struggling to get shadowing working correctly but being as I've already fixed it but just not had the time to get it into the CVS, I thought I'd at least give you the fixes here. Sorry I didn't do this earlier, it's just that our game is taking almost every hour of my freetime.

Anyway, here's how to do it and I will endeavour to get it into the [HEAD] soon. Note that you can scroll down to the bottom if you're not interested in the reasoning and copy the whole functions into the SDK.

First, in "shapebase.h" around line 916, change the levels to something like...

#define StaticShape_GenericShadowLevel 0.4f
#define StaticShape_NoShadowLevel 0.01f

The next problem you will encounter is that shapes with translucency get skipped. You can force it to render these shapes such as the stock trees but you may get artifacts. If you want to stop this then you can amend the line in "void TSStatic::renderShadow(F32 dist, F32 fogAmount)" around line 311 to...

if (mShapeInstance->getShape()->subShapeFirstTranslucentObject.empty()/* || mShapeInstance->getShape()->subShapeFirstTranslucentObject[0]==0*/)
return;

The next two amendments allow you to change the behaviour of the shadows. Since statics don't move or animate, the shadows are set to non-moving/animating. This produces a higher-performance shadow but doesn't animate when you move it in the editor. You may also want to amend this behaviour dynamically as mentioned by Robert Blanchet here. It's a total choice for you but you can amend this functionality by editing two lines in "void TSStatic::renderShadow(F32 dist, F32 fogAmount)", namely...

mShadow->setMoving(true);
mShadow->setAnimating(true);

The next problem is the incorrect reference to the Obj->World transform. Brett Fattori mentions one of these here but there is another on line 331 where the position is transformed into world-space.

Because there are numerous problems in "void TSStatic::renderShadow(F32 dist, F32 fogAmount)", here is the whole function with dynamic shadows turned-on. Be careful with dynamic shadows if you've got lots of statics that don't need it. It could be made to be a $Prefs option that you can turn-on/off at will I guess.

void TSStatic::renderShadow(F32 dist, F32 fogAmount)
{
   if (Shadow::getGlobalShadowDetailLevel()<StaticShape_NoShadowLevel)
      return;
   if (mShapeInstance->getShape()->subShapeFirstTranslucentObject.empty()/* || mShapeInstance->getShape()->subShapeFirstTranslucentObject[0]==0*/)
      return;

   // for safety, since I can't yet guarantee when this gets called.
   dglNPatchEnd();

   if (!mShadow)
      mShadow = new Shadow();
   mShadow->setGeneric(Shadow::getGlobalShadowDetailLevel() < StaticShape_GenericShadowLevel);
   // The following set to TRUE slows things down but you can also see the changes to shadows
   // immediately within the editor.
   mShadow->setMoving(true); // MM: Set this to false to improve performance.
   mShadow->setAnimating(true); // MM: Set this to false to improve performance.
   
   Point3F lightDir(0.57f,0.57f,-0.57f);
   F32 shadowLen = 3.0f * mShapeInstance->getShape()->radius;
   Point3F pos = mShapeInstance->getShape()->center;
   // this is a bit of a hack...move generic shadows towards feet/base of shape
   if (Shadow::getGlobalShadowDetailLevel() < StaticShape_GenericShadowLevel)
      pos *= 0.5f;

   pos.convolve(mObjScale);
   getRenderTransform().mulP(pos);

   // pos is where shadow will be centered (in world space)
   mShadow->setRadius(mShapeInstance,mObjScale);
   if (!mShadow->prepare(pos,lightDir,shadowLen,mObjScale,dist,fogAmount,mShapeInstance))
      return;

   F32 maxScale = getMax(mObjScale.x,getMax(mObjScale.y,mObjScale.z));

   if (mShadow->needBitmap())
   {
      mShadow->beginRenderToBitmap();
      mShadow->selectShapeDetail(mShapeInstance,dist,maxScale);
      mShadow->renderToBitmap(mShapeInstance,getRenderTransform(),pos,mObjScale);
      mShadow->endRenderToBitmap();
   }
   
   mShadow->render();
}

Next up is the function "void TSStatic::renderObject(SceneState* state, SceneRenderImage* image)". This is so wrong! I'll let you work out what I changed so here's the whole thing...

void TSStatic::renderObject(SceneState* state, SceneRenderImage* image)
{
   AssertFatal(dglIsInCanonicalState(), "Error, GL not in canonical state on entry");

   if (!DetailManager::selectCurrentDetail(mShapeInstance))
      // we were detailed out
      return;

   RectI viewport;
   glMatrixMode(GL_PROJECTION);
   glPushMatrix();
   dglGetViewport(&viewport);

   installLights();

   // Uncomment this if this is a "simple" (non-zone managing) object
   state->setupObjectProjection(this);

   // This is something of a hack, but since the 3space objects don't have a
   //  clear conception of texels/meter like the interiors do, we're sorta
   //  stuck.  I can't even claim this is anything more scientific than eyeball
   //  work.  DMM
   F32 axis = (getObjBox().len_x() + getObjBox().len_y() + getObjBox().len_z()) / 3.0;
   F32 dist = (getRenderWorldBox().getClosestPoint(state->getCameraPosition()) - state->getCameraPosition()).len();
   if (dist != 0)
   {
      F32 projected = dglProjectRadius(dist, axis) / 350;
      if (projected < (1.0 / 16.0))
      {
         TextureManager::setSmallTexturesActive(true);
      }
   }
   
   glMatrixMode(GL_MODELVIEW);
   glPushMatrix();
   dglMultMatrix(&mObjToWorld);
   glScalef(mObjScale.x, mObjScale.y, mObjScale.z);

   dglNPatchBegin();

   // RENDER CODE HERE
   mShapeInstance->setEnvironmentMap(state->getEnvironmentMap());
   mShapeInstance->setEnvironmentMapOn(true,1);
   mShapeInstance->setAlphaAlways(1.0);

   Point3F cameraOffset;
   mObjToWorld.getColumn(3,&cameraOffset);
   cameraOffset -= state->getCameraPosition();
   dist = cameraOffset.len();
   F32 fogAmount = state->getHazeAndFog(dist,cameraOffset.z);

   if (image->isTranslucent == true)
   {
      TSShapeInstance::smNoRenderNonTranslucent = true;
      TSShapeInstance::smNoRenderTranslucent    = false;
   }
   else
   {
      TSShapeInstance::smNoRenderNonTranslucent = false;
      TSShapeInstance::smNoRenderTranslucent    = true;
   }

   mShapeInstance->setupFog(fogAmount,state->getFogColor());
   mShapeInstance->animate();
   mShapeInstance->render();

   dglNPatchEnd();


   glMatrixMode(GL_MODELVIEW);
   glPopMatrix();

   renderShadow(dist,fogAmount);

   TSShapeInstance::smNoRenderNonTranslucent = false;
   TSShapeInstance::smNoRenderTranslucent    = false;
   TextureManager::setSmallTexturesActive(false);

   uninstallLights();
   dglSetCanonicalState();

   if (GameBase::gShowBoundingBox) {
      glDisable(GL_DEPTH_TEST);
      Point3F box;
      glPushMatrix();
      dglMultMatrix(&getTransform());
      box = (mObjBox.min + mObjBox.max) * 0.5;
      glTranslatef(box.x,box.y,box.z);
      box = (mObjBox.max - mObjBox.min) * 0.5;
      glScalef(box.x,box.y,box.z);
      glColor3f(1, 0, 1);
      ShapeBase::wireCube(Point3F(1,1,1),Point3F(0,0,0));
      glPopMatrix();

      glPushMatrix();
      box = (mWorldBox.min + mWorldBox.max) * 0.5;
      glTranslatef(box.x,box.y,box.z);
      box = (mWorldBox.max - mWorldBox.min) * 0.5;
      glScalef(box.x,box.y,box.z);
      glColor3f(0, 1, 1);
      ShapeBase::wireCube(Point3F(1,1,1),Point3F(0,0,0));
      glPopMatrix();
      glEnable(GL_DEPTH_TEST);
   }

   glMatrixMode(GL_PROJECTION);
   glPopMatrix();
   glMatrixMode(GL_MODELVIEW);
   dglSetViewport(viewport);

   AssertFatal(dglIsInCanonicalState(), "Error, GL not in canonical state on exit");
}

Enjoy.

- Melv.