libMesh: A C++ Finite Element Library

On Thu, 24 Sep 2009, Kirk, Benjamin (JSC-EG311) wrote:

> Is there anything outstanding or more broken now than it was in 0.6.3?

That Hilbert hashing collision in the I/O global renumbering.  We've
verified that that's a real bug, not just a glitch in Vikram's
application code, right?  On the other hand, that code existed already
in 0.6.3, so we're probably not talking about a new bug or an easy to
trigger bug...

There's something misbehaving with the new adjoint stuff in parallel.
That may be a bug with the goal-oriented refinement code alone, but
I'd like to take a deeper look at it soon; the last bug I caught this
way turned out to be a more fundamental send_list problem.  I'll try
to have this one diagnosed or hopefully fixed by this weekend.

> I'd like to get 0.6.4 out and hold us to that elusive goal of creating more
> releases more frequently...

Agreed.
---
Roy

Let me do just a bit more testing with ghosted vectors and JFNK.  I am  
still seeing some oddness... and I want to make sure it's me before we  
push out a new version.

Derek

On Sep 24, 2009, at 1:54 PM, Kirk, Benjamin (JSC-EG311) wrote:

> Is there anything outstanding or more broken now than it was in 0.6.3?
>
> I'd like to get 0.6.4 out and hold us to that elusive goal of  
> creating more
> releases more frequently...
>
> -Ben
>
>
> ------------------------------------------------------------------------------
> Come build with us! The BlackBerry® Developer Conference in SF, CA
> is the only developer event you need to attend this year. Jumpstart  
> your
> developing skills, take BlackBerry mobile applications to market and  
> stay
> ahead of the curve. Join us from November 9-12, 2009. Register  
> now!
> http://p.sf.net/sfu/devconf
> _______________________________________________
> Libmesh-devel mailing list
> Libmesh-devel@...
> https://lists.sourceforge.net/lists/listinfo/libmesh-devel

>> Is there anything outstanding or more broken now than it was in 0.6.3?
> 
> That Hilbert hashing collision in the I/O global renumbering.  We've
> verified that that's a real bug, not just a glitch in Vikram's
> application code, right?  On the other hand, that code existed already
> in 0.6.3, so we're probably not talking about a new bug or an easy to
> trigger bug...

Yes, that is something that was in 0.6.3 and I still need to fix.

> There's something misbehaving with the new adjoint stuff in parallel.
> That may be a bug with the goal-oriented refinement code alone, but
> I'd like to take a deeper look at it soon; the last bug I caught this
> way turned out to be a more fundamental send_list problem.  I'll try
> to have this one diagnosed or hopefully fixed by this weekend.

Have you tried the brute-force workaround - fill the send_list with the
identity map so you over-communicate but eliminate the possibility of an
incomplete send list?  it is instructive for possibly eliminating the
send_list as a source of the problem.

-Ben