[Paraview] Clean to Grid in parallel

[Matthew Mosby]

Hello all,

I am having some trouble with the clean to grid filter (paraview 3.10.1)
in client/server mode with the server having np > 1 on a pvtu dataset.

pvtu dataset:
www.nd.edu/~mmosby1/figs/paraview_clean_to_grid_filter/data.tgz

www.nd.edu/~mmosby1/figs/paraview_clean_to_grid_filter/fig1.png
Figure 1 shows the domain visualized on 1CPU with opacity = 0.5. The
boundaries of the (8) different pieces are visible.

www.nd.edu/~mmosby1/figs/paraview_clean_to_grid_filter/fig2.png
Figure 2 shows the domain visualized on 1CPU with opacity = 0.5 after
applying the "Clean to Grid" filter. The boundaries of the pieces are
correctly removed.

www.nd.edu/~mmosby1/figs/paraview_clean_to_grid_filter/fig3.png
Figure 3 shows the domain visualized on 4CPU with opacity = 0.5 after
applying the "Clean to Grid" filter. The boundaries of the pieces are
NOT correctly removed. It seems that the boundary is only removed if the
processor owns both of the coincident nodes.

I thought this was an obvious feature to have, so maybe I am doing it
wrong... I also tried to use the filter after the D3 communication
filter thinking that this might fix my problem, but I get what is shown
in Figure 4.

www.nd.edu/~mmosby1/figs/paraview_clean_to_grid_filter/fig4.png
Figure 4 shows the domain visualized on 4CPU with opacity = 0.5 after
applying the "D3" filter then the "Clean to Grid" filter. All of the
horizontal (xy-plane) boundaries are gone, but the new processor
boundaries are still visible.

Have any of you seen this behavior? Am I doing something wrong? Is this
a bug/missing feature? If it is a bug/missing feature, is it fixed in
newer versions?

Thanks,
Matt Mosby