SSRPass: throws err for a simple scene.

6 thoughts on “SSRPass: throws err for a simple scene.”

1. Probably a duplicate of #18533.

2. Yeah, I also found in this non-composer scene https://jsfiddle.net/gonnavis/hjgdrfv5/2/ , the instancedMeshes (two spheres) which probably caused this problem are rendered, but the ordinary mesh (the cube) not rendered. So it should be the error throwed in instancedMesh blocked the subsequent ordinary mesh render.

   After swap the creation in this non-composer scene https://jsfiddle.net/gonnavis/hjgdrfv5/5/ , I see one ordinary mesh and one instance rendered ( should be the first instance rendered ). So the normal infos still not all correct, this is why in https://jsfiddle.net/trch4517/ just one sphere has reflection.

3. Okay, in this case I mark this issue as a duplicate of #18533 since it’s the same cause (and the same generic solution is required).

4. @ycw @Mugen87 The main problem may because of instancedMesh and regular mesh can’t share same material now.

   One obvious low-performance workaround is, at here

   three.js/examples/jsm/postprocessing/SSRPass.js

   Line 256
   in
   c21b290

   insert

   	this.scene.traverseVisible( child => {
   
   		child._normalMaterial = new MeshNormalMaterial();
   
   	});
   

   and replace these codes

   three.js/examples/jsm/postprocessing/SSRPass.js

   Lines 576 to 578
   in
   c21b290

   	this.scene.overrideMaterial = overrideMaterial;
   	renderer.render( this.scene, this.camera );
   	this.scene.overrideMaterial = null;

   with

   		this.scene.traverseVisible( child => {
   
   			child._SSRPassMaterialBack = child.material;
   
   			child.material = child._normalMaterial;
   
   		} );
   		renderer.render( this.scene, this.camera );
   		this.scene.traverseVisible( child => {
   
   			child.material = child._SSRPassMaterialBack;
   
   		} );
   

   then it works well. If do some type check, just create two MeshNormalMaterials for instanced or regular mesh seperately, would be better.
   

   But the fundamental problem seems too in-depth for me to solve~~

5. If it’s only about material sharing, #20135 should fix it.

   In any event, we should not start to add fixes per pass. The renderer should handle this internally.

6. @gonnavis @Mugen87 @mrdoob i tried to merge #20135 ( w/ few bugfixes ) to r126 dev and .. it works!

   https://jsfiddle.net/m7bcex6s/1/

   thanks.