Re: [Libmesh-devel] Distributing exodiff with libmesh

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

On Mon, 21 Mar 2011, John Peterson wrote:

> I guess one feature that's pretty useful in exodiff is the ability
> to determine exactly which nodes two solutions differ at, since
> sometimes it's only a single node.

> I'm too lazy to pull out my laptop and do an svn update to see if
> your code will do that ;)

Not yet - it's designed less for regression testing (diffing the same
file *against itself* gives O(10^-17) L2 error due to the inverse
mapping done by MeshFunction) than for convergence testing (the two
solutions don't have to be on the same mesh).

> Oh and also the ability to write out a file with the diff solution
> on the mesh, both relative and absolute, would be good.

That would be pretty cool - I wrote a script to do this with gmv
output from a non-libMesh program and it's turned out to be
fantastically useful.
---
Roy