Switching shader causes single frame glitch

[bbeaumont]

This is an issue when changing a shader on a material at runtime with a directional light. The first frame after the switch the object appears black. It's something that @lidev fixed on our fork but I don't think it's made it into the main fork, he can probably advise a bit better on what the problem is, but here is the commit that fixed it.

https://github.com/mindshapes/away3d-core-fp11/commit/37f8603c9c20a425f15e96ae9327e381a2e2824a

not sure if thats going to be visible outside the organisation, so here the patch

From 37f8603c9c20a425f15e96ae9327e381a2e2824a Mon Sep 17 00:00:00 2001
From: Alejandro Santander <palebluedot@gmail.com>
Date: Fri, 18 May 2012 12:12:46 -0300
Subject: [PATCH] fixed the dreaded mouse over/out highlight shader bug    

---
 src/away3d/lights/DirectionalLight.as              |    4 ++++
 src/away3d/materials/methods/BasicAmbientMethod.as |    5 +++--
 src/away3d/materials/passes/DefaultScreenPass.as   |    2 ++
 3 files changed, 9 insertions(+), 2 deletions(-)
diff --git a/src/away3d/lights/DirectionalLight.as b/src/away3d/lights/DirectionalLight.as
index deeac4c..edea32e 100644
--- a/src/away3d/lights/DirectionalLight.as
+++ b/src/away3d/lights/DirectionalLight.as
@@ -1,5 +1,6 @@
 package away3d.lights
 {
+
    import away3d.arcane;
    import away3d.bounds.BoundingVolumeBase;
    import away3d.bounds.NullBounds;
@@ -47,6 +48,9 @@
         */
        public function get sceneDirection() : Vector3D
        {
+           if( _sceneTransformDirty ) {
+               updateSceneTransform();
+           }
        return _sceneDirection;
        }
diff --git a/src/away3d/materials/methods/BasicAmbientMethod.as       b/src/away3d/materials/methods/BasicAmbientMethod.as
index 1a0f4aa..1de9e14 100644
--- a/src/away3d/materials/methods/BasicAmbientMethod.as
+++ b/src/away3d/materials/methods/BasicAmbientMethod.as
@@ -1,11 +1,12 @@
 package away3d.materials.methods
 {
+
    import away3d.arcane;
    import away3d.core.managers.Stage3DProxy;
    import away3d.materials.utils.ShaderRegisterCache;
    import away3d.materials.utils.ShaderRegisterElement;
    import away3d.textures.Texture2DBase;
-   
+
    import flash.display3D.Context3D;
    import flash.display3D.Context3DProgramType;

@@ -165,7 +166,7 @@ package away3d.materials.methods
        /**
         * Updates the ambient color data used by the render state.
         */
-       private function updateAmbient() : void
+       public function updateAmbient() : void
        {
            _ambientData[uint(0)] = _ambientR = ((_ambientColor >> 16) & 0xff) / 0xff * _ambient * _lightAmbientR;
            _ambientData[uint(1)] = _ambientG = ((_ambientColor >> 8) & 0xff) / 0xff * _ambient * _lightAmbientG;
diff --git a/src/away3d/materials/passes/DefaultScreenPass.as b/src/away3d/materials/passes/DefaultScreenPass.as
index f1769fd..071cbcb 100644
--- a/src/away3d/materials/passes/DefaultScreenPass.as
+++ b/src/away3d/materials/passes/DefaultScreenPass.as
@@ -1,5 +1,6 @@
 package away3d.materials.passes
 {
+
    import away3d.arcane;
    import away3d.cameras.Camera3D;
    import away3d.core.base.IRenderable;
@@ -552,6 +553,7 @@ package away3d.materials.passes
            _ambientMethod._lightAmbientR = _ambientLightR;
            _ambientMethod._lightAmbientG = _ambientLightG;
            _ambientMethod._lightAmbientB = _ambientLightB;
+           _ambientMethod.activate( stage3DProxy );
            if (_shadowMethod) _shadowMethod.setRenderState(renderable, stage3DProxy, camera);
            _diffuseMethod.setRenderState(renderable, stage3DProxy, camera);
            if (_usingSpecularMethod) _specularMethod.setRenderState(renderable, stage3DProxy, camera);
-- 
1.7.3.2

[richardolsson]

Thanks for reporting. I took the liberty of formatting the patch a little as you are indeed right that it's not visible outside your organization.

@lidev Maybe you can explain the fix? To me it looks a little bit weird that updateAmbient() would need to be public. Did you make any further changes in client code required for this fix to work?

@DerSchmale What do you think?

[eternauta1337]

I've re-run the experiment on the latest release branch and was unable to reproduce shader glitches when adding/removing a method from a material. Considering it fixed then.

[bbeaumont]

I've just been in the process of upgrading to the Gold release, and I'm still seeing this issue occuring. However, reimplementing the above fix does still work.

[ontheronix]

@lidev can you explain the fix? Is it a fix or a bypass?