Fix FaceVariable.globalValue method

[wd15]

This ticket was prompted by

https://www.mail-archive.com/fipy@nist.gov/msg02801.html
http://thread.gmane.org/gmane.comp.python.fipy/3228

There is no easy way to obtain a FaceVariable's global value so the methods concerning global values should be moved out of FaceVariable and MeshVariable into CellVariable. Global value returns nonsense currently for FaceVariable's in parallel. Moreover, "_localNonOverlappingFaceIDs" doesn't do the right thing in parallel other than for 1D as far as I can tell.

Imported from trac ticket #657, created by wd15 on 10-17-2013 at 17:08, last modified: 10-18-2013 at 12:38

[wd15]

A number of tests rely on FaceVariable's globalValue to work in 1D. In fact the globalValue does work in 1D since the 1D mesh is the only mesh that has the correct face topology in 1D. This is quite dangerous. I think that a better solution is to only check the faceValue tests when the number of processors is 1 and remove globalValue from FaceVariable.

Trac comment by wd15 on 10-17-2013 at 17:57

[wd15]

I've taken a look at fixing this issue, but it requires some fairly major changes. There are a lot of tests in 1D and in 2D that use global face centers and values, which apparently work. Ideally, all reference to non local face quantities would be take out of FiPy including in the mesh.

Trac comment by wd15 on 10-18-2013 at 12:38

[guyer]

It turned out that it was possible to have a proper globalValue on FaceVariable.