Fix floating point precision bug handling alpha channel (#332) (#333)

Fixes #332
Fixes #108

Signed-off-by: Matias N. Goldberg <[email protected]>

ogre2/src/media/materials/programs/depth_camera_final_fs.glsl

@@ -31,10 +31,18 @@ uniform float far;
 uniform float min;
 uniform float max;
 
+uniform vec4 texResolution;
+
 void main()
 {
   float tolerance = 1e-6;
-  vec4 p = texture(inputTexture, inPs.uv0);
+
+  // Note: We use texelFetch because p.a contains an uint32 and sampling
+  // (even w/ point filtering) causes p.a to loss information (e.g.
+  // values close to 0 get rounded to 0)
+  //
+  // See https://github.com/ignitionrobotics/ign-rendering/issues/332
+  vec4 p = texelFetch(inputTexture, ivec2(inPs.uv0 *texResolution.xy), 0);
 
   vec3 point = p.xyz;
 

ogre2/src/media/materials/scripts/depth_camera.material

@@ -74,6 +74,8 @@ fragment_program DepthCameraFinalFS glsl
   default_params
   {
     param_named inputTexture int 0
+
+    param_named_auto texResolution texture_size 0
   }
 }
 

test/integration/depth_camera.cc

@@ -249,6 +249,8 @@ void DepthCameraTest::DepthCameraBoxes(
       unsigned int ma = *mrgba >> 0 & 0xFF;
       EXPECT_EQ(0u, mr);
       EXPECT_EQ(0u, mg);
+      // Note: If it fails here, it may be this problem again:
+      // https://github.com/ignitionrobotics/ign-rendering/issues/332
       EXPECT_GT(mb, 0u);
       EXPECT_EQ(255u, ma);
 
@@ -441,6 +443,8 @@ void DepthCameraTest::DepthCameraBoxes(
           unsigned int a = *rgba >> 0 & 0xFF;
           EXPECT_EQ(0u, r);
           EXPECT_EQ(0u, g);
+          // Note: If it fails here, it may be this problem again:
+          // https://github.com/ignitionrobotics/ign-rendering/issues/332
           EXPECT_GT(b, 0u);
           EXPECT_EQ(255u, a);
         }