Problem, Textures updated by plugin won't behave as expected in combination with HLSL shader

Hi

I've already posted this issue in the forums, but I still haven't got any answer.

I'm developing under Windows 7 with a GF 210M (DX 10.1) and have the following problem. I'm updating 2 textures from my c++ plugin per glTexSubImage2D. This part works, I can read those textures as GUITexture or render them with a shader to the backbuffer. They are displayed correctly. I run my program in forced-opengl mode. This part of the code works perfectly fine.

The second thing I do is render my scene, and then, in the OnRenderImage part I apply a post processing shader with Blit. I can access and display the rendered depthtexture as well as the rendered colortexture. I can also access the other 2 textures I've mentioned before. But, if I use all of those at the same time, one of the textures will display wrong data.

First off, here's the script which is attached to the camera :

 using UnityEngine;
 using System.Collections;
 
 public class MixAR : MonoBehaviour 
 {
     public Shader mixShader;
     public float kinectMaxDepth;
 
     private Texture2D colorTex;
     private Texture2D depthTex;
     private Material targetMaterial;
 
     IEnumerator Start () 
     {
         targetMaterial = new Material(mixShader);
 
         yield return StartCoroutine(KinectManager.Init(KinectManager.NUI_INITIALIZE_FLAG_USES_COLOR | KinectManager.NUI_INITIALIZE_FLAG_USES_DEPTH));
         KinectManager.InitImageFetch(ImageType.NUI_IMAGE_TYPE_COLOR, ImageResolution.NUI_IMAGE_RESOLUTION_640x480);
         KinectManager.InitImageFetch(ImageType.NUI_IMAGE_TYPE_DEPTH, ImageResolution.NUI_IMAGE_RESOLUTION_320x240);
 
         colorTex = new Texture2D(640, 480);
         depthTex = new Texture2D(320, 240, TextureFormat.RGB24, false);
 
         KinectManager.BindTexture(colorTex, ImageIndex.INDEX_RGB);
         KinectManager.BindTexture(depthTex, ImageIndex.INDEX_DEPTH);
 
         targetMaterial.SetFloat("_MaxDepth", kinectMaxDepth);
         targetMaterial.SetTexture("_KinectColorSource", colorTex);
         targetMaterial.SetTexture("_KinectDepthSource", depthTex);
     }
     
     void OnRenderImage(RenderTexture src, RenderTexture dst)
     {
         Graphics.Blit(src, dst, targetMaterial, 0);
     }
 }

Furthermore, here's the mixShader's code, which is attached to this script:

 Shader "Custom/KinectOcclusion"
 {
     Properties 
     {
         _KinectColorSource ("Kinect Color Camera", 2D) = "" {}
         _KinectDepthSource ("Kinect Depth Source", 2D) = "" {}
         _MainTex ("", 2D) = "" {}
         _CameraDepthTexture ("", 2D) = "" {}
         
         _MaxDepth ("Maximum KinectSDK Depth", Float) = 4096
     }
     SubShader 
     {
         Pass
         {
             ZTest Always Cull Off ZWrite Off
         
             CGPROGRAM
             
             #pragma vertex identityTransf
             #pragma fragment fragment2Depth
             #pragma only_renderers opengl d3d9
             //#pragma target 3.0
             #include "UnityCG.cginc"
             
             uniform sampler2D _KinectColorSource;
             uniform sampler2D _KinectDepthSource;
             uniform sampler2D _MainTex;
             uniform sampler2D _CameraDepthTexture;
             
             float _MaxDepth;
         
             struct VertexInput
             {
                 float4 vertex : POSITION;
                 float4 texCoord : TEXCOORD0;
             };
             struct VertexOutput
             {
                 float4 vertexPosition : SV_POSITION;
                 float4 texCoord;
             };
             struct FragmentOutput
             {
                 float4 color : COLOR;
             };
             
             VertexOutput identityTransf(VertexInput input)
             {
                 VertexOutput output;
                 
                 output.vertexPosition = input.vertex;
                 output.vertexPosition.xy *= 2.0f;
                 output.vertexPosition.xy -= 1.0f;
                 output.texCoord = input.texCoord;
                 
                 return output;
             }
             
             FragmentOutput fragment2Depth(VertexOutput input)
             {
                 FragmentOutput output;
                 
                 float2 depthComponents = tex2D(_KinectDepthSource, input.texCoord.xy).rg;
                 float newDepth = (depthComponents.g * 255.0f + depthComponents.r) * 255.0f / _MaxDepth;
                 float oldDepth = DECODE_EYEDEPTH(tex2D(_CameraDepthTexture, input.texCoord.xy).r) / 10.0f;
                 
                 if (newDepth < oldDepth)
                 {
                     output.color = tex2D(_KinectColorSource, input.texCoord.xy);
                 } else
                 {
                     output.color = tex2D(_MainTex, input.texCoord.xy);
                 }
                 
                 return output;
             }
                         
             ENDCG
         }
     }
 }

If I execute this code, _MainTex contents will be replaced with the contents of _KinectDepthTex, which is basically a depth texture encoded in the red and green channel. However, if I bind a normal/static texture from the assets to colorTex and depthTex, everything works fine and as expected.

I've read a lot of the unity documentation, and I didn't find anything I've forgotten. Any help would be appreciated.