Dear All,
we regularly use a self compiled ParaView version (5.7) that uses Mesa as rendering backend for in-situ visualization and processing. So far, that works well and the generated images look flawless. The same version we now want to use in batch mode using pvbatch on the compute nodes. We use Python state files and traces for this, and it works as well, yet on the images that are produced, we see speckles, black and white noise, see image excerpt below. I scaled it up using nearest neighbor. 
We tried both jpeg and png with varying compression level, this is not the issue. I also looked at the data, and it is of course correct. In the image, you can see the triangular cells, the noise is well below the grid resolution. OpenSWR is loaded correctly, as is indicated by the message:
“SWR detected AVX2 instruction support (using: libswrAVX2.so).”
Running the same script using the same ParaView version that is compiled with OpenGL and running on GPUs does NOT produce those speckles. Is there a compression somewhere in Mesa?
Other than that, I wish everyone a very merry Christmas, some time to relax, and all the best for 2020.
Thanks and Cheers, Niklas