Bug Build-Only per-frame GC caused by missing shaders

I'm running a pass on cleaning up per-frame GC from our game, and there's one entry left, which comes from URP with 2D lights;



This is profiling a build, built with Deep Profiling enabled, with:
Unity 2020.13.1f1
URP 8.2.0

What's happening with the first GC cost is pretty straight-forwards. RenderingUtils.errorMaterial looks like this:

Code (csharp):

1. static Material s_ErrorMaterial;
2. static Material errorMaterial
3. {
4.     get
5.     {
6.         if (s_ErrorMaterial == null)
7.         {
8.             // TODO: When importing project, AssetPreviewUpdater::CreatePreviewForAsset will be called multiple times.
9.             // This might be in a point that some resources required for the pipeline are not finished importing yet.
10.             // Proper fix is to add a fence on asset import.
11.             try
12.             {
13.                 s_ErrorMaterial = new Material(Shader.Find("Hidden/Universal Render Pipeline/FallbackError"));
14.             }
15.             catch { }
16.         }
17.  
18.         return s_ErrorMaterial;
19.     }
20. }

It seems like the shader "Hidden/Universal Render Pipeline/FallbackError" doesn't exist in builds, as if I run this code in a build:

Code (csharp):

1. var s = Shader.Find("Hidden/Universal Render Pipeline/FallbackError");
2. Debug.Log(s);
3. Debug.Log(s == null);

The output is "null" and "true". The shader does get loaded when playing in the editor.

The second GC cost, that's labeled the absurd "Object.__icall_wrapper_ves_icall_array_new_specific()" is probably just this line in Render2DLightingPass.Execute():

Code (csharp):

1. const int blendStylesCount = 4;
2. bool[] hasBeenInitialized = new bool[blendStylesCount];

So that should probably be cleaned up with a static array or by turning the pass unsafe and stackallocing the array instead.

Should I report this as a bug, or is it something you're aware about and that will be fixed on it's own? I'd really love to not have to have GC spikes due to running the 2D renderer.

 

Please report it as a bug or else it does not get fixed/resolved. I agree that it should be a bug.
But I have sort of given up on waiting for the fix as I suspect the URP team might be pretty small and limited in resources as it is not a breaking issue.
Because of this minor stupid issue, I implemented my own custom renderer to use my own custom DrawObjectPass that omits the resolving via errorMaterial. If something fails to render, I just ignore it. It might be a pain-in-the-ass to debug faulty shaders later on but still I avoid having the initial GC spike.

 

Hey there! Reporting it as a bug will always help, so please feel free to do so!

 

Perhaps the error material should have been created during the Renderer creation and not on demand.

 

For anybody running into this problem still, I worked around this by forcing the Hidden/URP/FallbackError shader to be included in the build via the "Always Included Shaders" in ProjectSettings>>Graphics.

• I can confirm that this removed the problem on my android build (I checked the profiler before and after the change)
• I did not check the change in build size or anything else yet so be aware in case that is important to you
  • Though I will note that the build time wasn't materially impacted
• Obviously this is not a substitute for the bug actually being fixed. @Shaunyowns i've seen a couple threads now mentioning this and that you say a bug is created but I cant find a link to it. Can you provide that / give an update please?