On Thu, Aug 2, 2012 at 9:24 AM, Roy Stogner <roystgnr@ices.utexas.edu> wrote:

On Thu, 2 Aug 2012, Cody Permann wrote:

Hey, I'm glad you're looking into threading issues regarding
periodic bcs.  That's long been an issue with some of our codes for
sure.  However, your most recent changes to fe_base.C have cause all
of our Periodic Tests to fail (changeset 5895).  It appears that
many of the constraints aren't working now at all.  All of these
results were run serially without threading.  I've attached a bunch
of pictures that illustrate the
problem.

Hopefully the attachments get through the list :)

They did; thanks!


This is big changeset so I'm not immediately sure where to start
looking for the issue.

No worries; you shouldn't have to reverse-engineer my stuff.


If you have any insight or need any more information from us, please
let me know.

Could you get that rectangle_pic.png case into a code short enough to
send me?  Or better yet - into a new example code?  I'll tweak
adaptivity_ex5 to see if I can get any problems to crop up there, but
it's not the best case for debugging.

I'll see what I can do.  That rectangle example is actually the most complicated one in the bunch!  it is using subdomain restricted terms, but I think the problem shows up even in simpler examples.  It'll be nice when MOOSE goes open-source ;)
 
(pre-r5894 it was actually an atrocious case for debugging... I assume
your other codes don't have the same "trying to impose both periodic
and penalty BCs on the same boundary" error, though)

Oh, that is bad, and yes, I don't believe we have anything like that in our suite.

Thanks,
Cody
  
---
Roy