Screwed up geometry of ghosted plots

[Jan Blechta]

Plots of ghosted meshes seem to have correct topology but geometry of ghosted cells is rubish.

from dolfin import *
import sys

parameters['ghost_mode'] = sys.argv[1]

mesh = UnitSquareMesh(2, 2)
rank = MPI.rank(mesh.mpi_comm())

for opt in ['regular', 'ghost']:
    for c in cells(mesh, opt):
        print rank, opt, c.index(), c.global_index()

plot(mesh, title='mesh at rank %d' % rank)
interactive()

Run mpirun -n 2 python 1.py shared_facet or mpirun -n 2 python 1.py shared_vertex and push 'c' in plots.

Next step when this is resolved is checking whether Functions are plotted correctly.

[Jan Blechta]

Maybe easy for Chris?

[Chris Richardson]

Is it really useful to use plot in parallel?

[Jan Blechta]

Well, not for serious stuff. But debugging, checking topological properties is exactly thing where it is useful. Feel free to reassign to me. I guess that the issue has a good ratio priority/difficulty.

[Chris Richardson]

I'm not into hacking around in VTK. I'd like to see more effort into getting some matplotlib plot routines working (that's what I used to debug ghost meshes).

[Jan Blechta]

Fix issue #551

Fix issue #551 - Screwed up geometry of ghosted plots

→ <<cset a736331c6228>>

[Jan Blechta]

Fix issue #551

Fix issue #551 - Screwed up geometry of ghosted plots

→ <<cset a736331c6228>>

[Chris Richardson]

OK - so that was easy.

[Jan Blechta]

The sad part is that is another of several bugs introduced by omitting MeshEntityIterator(mesh) -> MeshEntityIterator(mesh, opt) in whole DOLFIN.

[Chris Richardson]

I guess it was always a danger - where are the other problems?

[Jan Blechta]

Other was #501.

[Anders Logg (Chalmers)]

Removing milestone: 1.7 (automated comment)