Roy Stogner <[email protected]> wrote on 06/08/2011 12:45:06 PM:

> 
> On Wed, 8 Jun 2011, David Andrs wrote:
> 
> > > Would it be sufficient to change that assert from
> > > libmesh_assert(last_local_idx < this->size());
> > > to
> > > libmesh_assert(last_local_idx+1 <= this->size());
> > > ?
> > 
> > Yes, it fixes the issue as well.
> 
> Good to hear; that's in SVN now.
> 
> Sorry about that.  r4500 was an attempt to fix a bug in the
> "processor N has 0 DoFs" corner case; I should have been looking
> closely enough to see that there's be another bug if N=0.

No problem at all.  Thanks for such a quick fix.

> 
> Mind if I ask how you cooked up a regression test that ended up
> with 0 DoFs on proc 0?  I've seen n_dofs-equal-0 cases come up when
> partitioning M elements onto N>M processors, but I thought our
> partitioners left the highest-numbered processors element-free in
> those cases.  And I've seen it come up when partitioning a fine mesh
> then doing adaptive coarsening without load balancing, but there
> shouldn't be any reason to skip load balancing on a non-ParallelMesh,
> should there?

This was actually a serial run with adaptivity and this bug poped out 
after first refinement when calling EquationSystems::reinit().

I'll try to replicate this behavior with pure libMesh calls and send it to 
you, it might be worth to include it into regression tests.

--
David

------------------------------------------------------------------------------
EditLive Enterprise is the world's most technically advanced content
authoring tool. Experience the power of Track Changes, Inline Image
Editing and ensure content is compliant with Accessibility Checking.
http://p.sf.net/sfu/ephox-dev2dev
_______________________________________________
Libmesh-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-devel

Reply via email to