# NME generates invalid GLSL **URL:** \ **Category:** Bugs\ **Tags:** material, shader, nodematerial\ **Created:** [April 3, 2023, 1:27pm UTC](https://forum.babylonjs.com/t/nme-generates-invalid-glsl/39585 "2023-04-03T13:27:22Z")\ **Posts on this page:** 8\ **Page:** 1
**Author:** ![vinhui](https://sea2.discourse-cdn.com/flex024/user_avatar/forum.babylonjs.com/vinhui/32/41914_2.png) [@vinhui](https://forum.babylonjs.com/u/vinhui)\ **Post date:** [April 3, 2023, 1:27pm UTC](https://forum.babylonjs.com/t/nme-generates-invalid-glsl/39585/1 "2023-04-03T13:27:22Z")
Hi, I have a nme material that generates invalid GLSL shader code. This is the material: [Babylon.js Node Material Editor](https://nme.babylonjs.com/#8GBF2X) It calculates the reflection in the vertex shader but it wants to do it before the worldnormal variable is created. This results in a `normalize()` call without any arguments. I also have this variant of the shader, which should functionally be the same: [https://nme.babylonjs.com/#8GBF2X#1](https://nme.babylonjs.com/#8GBF2X#1) This however results in it wanting to use an `output5` variable before it’s declared. This is a GLSL output snipped of the incorrect code: ```plaintext v_output6 = output6; #ifdef REFLECTIONMAP_SKYBOX0 positionUVW = position.xyz; #endif #if defined(REFLECTIONMAP_EQUIRECTANGULAR_FIXED0) || defined(REFLECTIONMAP_MIRROREDEQUIRECTANGULAR_FIXED0) directionW = normalize(vec3(u_World * vec4(position.xyz, 0.0))); #endif #if defined(USESPHERICALFROMREFLECTIONMAP) && defined(USESPHERICALINVERTEX) vec3 reflectionVector = vec3(reflectionMatrix * vec4(normalize(output5).xyz, 0)).xyz; #ifdef REFLECTIONMAP_OPPOSITEZ0 reflectionVector.z *= -1.0; #endif vEnvironmentIrradiance = computeEnvironmentIrradiance(reflectionVector); #endif vec4 worldPos = output6; mat4 view = u_view; #include[0..maxSimultaneousLights] //World position vec4 output3 = u_World1 * vec4(position, 1.0); //Reflection texture //World normal mat3 u_World_NUS = mat3(u_World2); #ifdef NONUNIFORMSCALING u_World_NUS = transposeMat3(inverseMat3(u_World_NUS)); #endif vec4 output5 = vec4(u_World_NUS * normal, 0.0); ``` Obviously, it also doesn’t work if I try to use the exported `NodeMaterial` code. Is there a workaround I could use for now? Manually fixing the glsl and using that doesn’t seem like a great option since I’m not sure what all the uniform should be set to. ---
**Author:** ![vinhui](https://sea2.discourse-cdn.com/flex024/user_avatar/forum.babylonjs.com/vinhui/32/41914_2.png) [@vinhui](https://forum.babylonjs.com/u/vinhui)\ **Post date:** [April 3, 2023, 1:52pm UTC](https://forum.babylonjs.com/t/nme-generates-invalid-glsl/39585/2 "2023-04-03T13:52:02Z")
To narrow it down a bit, the following also does not work: [https://nme.babylonjs.com/#8GBF2X#2](https://nme.babylonjs.com/#8GBF2X#2) So it seems that the `ReflectionBlock` generates the invalid code. This leads me to believe that this code is not working correctly: [Babylon.js/pbrMetallicRoughnessBlock.ts at master · BabylonJS/Babylon.js · GitHub](https://github.com/BabylonJS/Babylon.js/blob/master/packages/dev/core/src/Materials/Node/Blocks/PBR/pbrMetallicRoughnessBlock.ts#L1026) ---
**Author:** ![Evgeni\_Popov](https://sea2.discourse-cdn.com/flex024/user_avatar/forum.babylonjs.com/evgeni_popov/32/4432_2.png) [@Evgeni\_Popov](https://forum.babylonjs.com/u/Evgeni_Popov)\ **Post date:** [April 3, 2023, 3:08pm UTC](https://forum.babylonjs.com/t/nme-generates-invalid-glsl/39585/3 "2023-04-03T15:08:41Z")
It’s a bit counterintuitive, but you must use the same `World` matrix block that you use for the vertex shader (to compute `WorldPos`) when you compute the world normal! This will force the system to generate the code for the normal before the code in the `Reflection` block needs it. That’s a bit awkward but I don’t see any other solution… ---
**Author:** ![vinhui](https://sea2.discourse-cdn.com/flex024/user_avatar/forum.babylonjs.com/vinhui/32/41914_2.png) [@vinhui](https://forum.babylonjs.com/u/vinhui)\ **Post date:** [April 3, 2023, 3:59pm UTC](https://forum.babylonjs.com/t/nme-generates-invalid-glsl/39585/4 "2023-04-03T15:59:41Z")
I’m pretty sure I’ve also tried that. I’ve had all the duplicate nodes removed and that has resulted in the situation with the `normalize()` without a parameter code. ---
**Author:** ![Evgeni\_Popov](https://sea2.discourse-cdn.com/flex024/user_avatar/forum.babylonjs.com/evgeni_popov/32/4432_2.png) [@Evgeni\_Popov](https://forum.babylonjs.com/u/Evgeni_Popov)\ **Post date:** [April 3, 2023, 4:17pm UTC](https://forum.babylonjs.com/t/nme-generates-invalid-glsl/39585/5 "2023-04-03T16:17:58Z")
The fix makes [https://nme.babylonjs.com/#8GBF2X#2](https://nme.babylonjs.com/#8GBF2X#2) work, and I could not get your other NMEs to fail by setting a texture on the env block in the NME. Can you provide a link to material that fails even with the fix I mentioned above? ---
**Author:** ![vinhui](https://sea2.discourse-cdn.com/flex024/user_avatar/forum.babylonjs.com/vinhui/32/41914_2.png) [@vinhui](https://forum.babylonjs.com/u/vinhui)\ **Post date:** [April 3, 2023, 5:01pm UTC](https://forum.babylonjs.com/t/nme-generates-invalid-glsl/39585/6 "2023-04-03T17:01:17Z")
You are right, that is indeed the solution. When removing all duplicate nodes I missed one. It does make things less readable but it’ll do for now. It would be great if something like the reroute nodes from ue4 were a thing, that would help make the spaghetti a little bit more manageable. Thanks! ---
**Author:** ![Evgeni\_Popov](https://sea2.discourse-cdn.com/flex024/user_avatar/forum.babylonjs.com/evgeni_popov/32/4432_2.png) [@Evgeni\_Popov](https://forum.babylonjs.com/u/Evgeni_Popov)\ **Post date:** [April 3, 2023, 6:08pm UTC](https://forum.babylonjs.com/t/nme-generates-invalid-glsl/39585/7 "2023-04-03T18:08:42Z")
You can do ALT+left click on a link to create a node and improve the layout: ![image](https://us1.discourse-cdn.com/flex024/uploads/babylonjs/original/3X/1/9/1921ce0d59a52804c336bf68d5bd423cdb754d06.png) ![image](https://us1.discourse-cdn.com/flex024/uploads/babylonjs/original/3X/a/a/aa43c639b110707457b2684e143f87b69670788c.png) ---
**Author:** ![vinhui](https://sea2.discourse-cdn.com/flex024/user_avatar/forum.babylonjs.com/vinhui/32/41914_2.png) [@vinhui](https://forum.babylonjs.com/u/vinhui)\ **Post date:** [April 4, 2023, 7:34am UTC](https://forum.babylonjs.com/t/nme-generates-invalid-glsl/39585/8 "2023-04-04T07:34:20Z")
Wow, didn’t know that was a thing. Very helpful, thanks!