On Fri, 2 Nov 2007, John Peterson wrote:
> Don't we want a private, static RemoteElem object in the
> Elem class? (Assume RemoteElem is a concrete subclass of Elem)
>
> private:
> static RemoteElem;
>
> Then, in an Elem member function when you'd originally have NULL'd a
> pointer, you'd now set
On Sat, 3 Nov 2007, Roy Stogner wrote:
> I've also just svn committed the parallelized DofMap code, but
> apparently prematurely; it's working on some adaptive cases but still
> tripping over unpartitioned nodes on others. I'll try to figure out
> the remaining bug this afternoon, but don't anybo
On Sat, 3 Nov 2007, John Peterson wrote:
> That sounds right. I think I checked in that bug to fix the
> "why do all the nodes have processor ID 65536?" bug (they used
> to not be set at all.) I suspect this will start to matter more
> for a parallel mesh...
For the parallel mesh I managed to h
Roy Stogner writes:
>
> Shouldn't _set_node_processor_ids be looping over all the active
> elements, not all the elements? The way it is, we can have nodes
> owned by a processor which aren't owned by any active elements on that
> processor.
That sounds right. I think I checked in that bug
