Re: [Libmesh-users] Bugfixes in .xda/.xdr code

[Roy S. <roy...@ic...>]

On Fri, 9 Feb 2007, Tim Kr=F6ger wrote:

> Dear Roy and all,
>
> On Thu, 8 Feb 2007, Roy Stogner wrote:
>
>> If you need high precision .xda output, or if you do restarts on
>> problems with lots of mesh refinement, you might want to CVS update.
>
> That's what I did, but unfortunately, for my case, this seems to make t=
hings=20
> worse instead of better.

You may just be the first to test the new code in parallel instead of
serial.  :-P  The bugfix should have been parallel-safe, but I'll see
if I can reproduce (and hopefully fix) your error.

> When reading in a rather complex EquationSystems=20
> .xdr file, I get:
>
> [0]PETSC ERROR: PetscSplitOwnership() line 94 in src/sys/src/utils/pspl=
it.c
> [0]PETSC ERROR: Nonconforming object sizes!
> [0]PETSC ERROR: Sum of local lengths 13892 does not equal global length=
=20
> 13784, my local length 877!
>
> I never got this messaage before.
>
> If necessary, I will try to create a concrete example for this, next we=
ek.

I'd appreciate that.  Even just a set of stack traces might help.
---
Roy