dolfin team mailing list archive

[kent-and@xxxxxxxxx]

> kent-and@xxxxxxxxx wrote:
>>> One or more new changesets pushed to the primary dolfin repository.
>>> A short summary of the last three changesets is included below.
>>>
>>> changeset:   4730:6548c25c33352492c9279c035509a139caab323b
>>> tag:         tip
>>> user:        "Garth N. Wells <gnw20@xxxxxxxxx>"
>>> date:        Tue Sep 09 13:08:58 2008 +0100
>>> files:       dolfin/fem/Assembler.cpp dolfin/pde/LinearPDE.cpp
>>> description:
>>> Revert to old assembly in LinearPDE due to bug in symmetric assembly.
>>>
>>> There is a problem with exterior facets in the symmetric assembly. The
>>> code needs to be broken up to make debugging easier.
>>>
>>
>> Agree that it should be broken up. But it will take some effort.
>> Do you have an example where the bug is apparent ?
>>
>
> I was using a RT0 element (not one of the demos).  It should be
> reproducible by adding a non-zero Dirichlet bc (given be a Function) to
> the mixed-poisson demo.
>
> I had a quick look, but I couldn't find the problem, so I thought it
> better to wait until the code is broken up.
>
> Garth
>

Strange, I see no reason for this not to work.
(since the code involving only cell integrals is pretty clean, but ...)

Anyway, I noticed that with my setup, CG is used. CG can not be used
in this case since the problem is not positive. Did you use CG ? Did
it converge (and does this really mean convergence in the usual sense) ?

Kent