Let me check what would be the proper channel to forward this information or get the appropriate person to work with you on this matter.
As per the SO thread:
I read from an "out" variable. Once I changed the shader code to only write to the outColor variable and have a separate local vec3 for read-write access, the issue was gone.
But -- seems like that's still a GL/GLSL implementation bug (ie. to-be-fixed) in Intel HD's driver: it is legal to read from an out global variable in a shader. Now it may not be zeroed --fine-- but in the given shader code, the first thing that happens to the out var is a write access, even though it gets passed as an inout into a function first for that.
We have replicated this issue and are currently working to fix the issue. Thank you for identifying the issue and bringing it to our attention.
Please note that Intel® has released a new driver that contains several fixes. We recommend you installing the new driver and let us know if it works.
You can download the driver at the following link:
For Windows* 7/8 32 bit:
For Windows* 7/8 64 bit: