[Bf-blender-cvs] [817e975dee2] master: Fix T50849: Transparent background produces artifacts in this compositing setup

[Sergey Sharybin]

Commit: 817e975dee27640947bf7d083db39d1d70120385
Author: Sergey Sharybin
Date:   Wed Mar 8 15:56:50 2017 +0100
Branches: master
https://developer.blender.org/rB817e975dee27640947bf7d083db39d1d70120385

Fix T50849: Transparent background produces artifacts in this compositing setup

The issue was caused by sometimes negative color returned by the filter node.

Seems to be caused by precision issues. Don't see any reason why we would want
negative colors in output. Those only causing issues later on.

===================================================================

M	source/blender/compositor/operations/COM_ConvolutionEdgeFilterOperation.cpp
M	source/blender/compositor/operations/COM_ConvolutionFilterOperation.cpp

===================================================================

diff --git a/source/blender/compositor/operations/COM_ConvolutionEdgeFilterOperation.cpp b/source/blender/compositor/operations/COM_ConvolutionEdgeFilterOperation.cpp
index e1ada9a8c39..5f78067220a 100644
--- a/source/blender/compositor/operations/COM_ConvolutionEdgeFilterOperation.cpp
+++ b/source/blender/compositor/operations/COM_ConvolutionEdgeFilterOperation.cpp
@@ -94,4 +94,10 @@ void ConvolutionEdgeFilterOperation::executePixel(float output[4], int x, int y,
 	output[2] = output[2] * value[0] + in2[2] * mval;
 	
 	output[3] = in2[3];
+
+	/* Make sure we don't return negative color. */
+	output[0] = max(output[0], 0.0f);
+	output[1] = max(output[1], 0.0f);
+	output[2] = max(output[2], 0.0f);
+	output[3] = max(output[3], 0.0f);
 }
diff --git a/source/blender/compositor/operations/COM_ConvolutionFilterOperation.cpp b/source/blender/compositor/operations/COM_ConvolutionFilterOperation.cpp
index 68ec2be5ebd..6ac1ff9a1eb 100644
--- a/source/blender/compositor/operations/COM_ConvolutionFilterOperation.cpp
+++ b/source/blender/compositor/operations/COM_ConvolutionFilterOperation.cpp
@@ -107,6 +107,12 @@ void ConvolutionFilterOperation::executePixel(float output[4], int x, int y, voi
 	output[1] = output[1] * value[0] + in2[1] * mval;
 	output[2] = output[2] * value[0] + in2[2] * mval;
 	output[3] = output[3] * value[0] + in2[3] * mval;
+
+	/* Make sure we don't return negative color. */
+	output[0] = max(output[0], 0.0f);
+	output[1] = max(output[1], 0.0f);
+	output[2] = max(output[2], 0.0f);
+	output[3] = max(output[3], 0.0f);
 }
 
 bool ConvolutionFilterOperation::determineDependingAreaOfInterest(rcti *input, ReadBufferOperation *readOperation, rcti *output)