Description

The AnalyticLightNode should test the scene camera layers property instead of the current shadow camera layers property (light.shadow.camera.layers in _renderScene).

We could introduce a renderer._cameraLayersOverride property or a getter/setter just like _renderObjectFunction (the latter being overkill in my opinion).

The main issue I currently have is that in updateShadow, if we use frame.camera as the reference and there are multiple lights casting shadows in the scene, the next value of frame.camera won't be the actual camera of the scene anymore but the previous Orthographic shadow camera from the previous light.

renderer._cameraLayersOverride = camera.layers;

renderer.setRenderObjectFunction( ( object, ...params ) => {

	if ( object.castShadow === true ) {

		renderer.renderObject( object, ...params );

	}

} );

/cc @sunag

Reproduction steps

in webgpu_shadowmap.html add camera.layers.enableAll(), then pillar1.layers.set( 2 ); and no shadow pillar will be visible in the scene as it should.

Version

r164