Re: [Libmesh-users] Problem in example 18

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

On Wed, 21 Oct 2009, Roy Stogner wrote:

> Still not sure what the nonconforming object sizes problem is; I'm
> recompiling with a debug-mode PETSc now.

It appears that I was being overly aggressive about keeping
SNES objects around; PETSc tried to reuse the same ILU preconditioner
which led to errors when we changed the problem size out from
underneath it.

I'm now being overly defensive instead: PetscDiffSolver now does the
same as PetscNonlinearSolver, and builds a new SNES before and tears
it down after every solve.

Would you give the latest svn a try and make sure it works for you
too?

Thanks,
---
Roy