Re: [Libmesh-devel] Mesh Corruption In XDA Format

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

On Thu, 17 Mar 2016, Derek Gaston wrote:

> Everyone: I could really use some help on this one.
> I have created a repository with the simplest code I can that shows the issue: https://github.com/friedmud/dinas_problem
> 
> (it should compile easily as long as you have LIBMESH_DIR set to point to your libMesh)
> 
> It is definitely an off-by-one error that happens (in this particular case) after 2 uniform refinements... then writing out to XDA and reading
> back in.  The code will print this:

In devel mode I get an in->good() assertion failure.

This looks a lot easier to replicate than the previous triggering case
I was trying to hunt down; thanks.  I don't suppose you can get it to
failure on a smaller mesh?

My suspicion was that something was going wrong with the new flag
indicating presence or absence of unique ids in the file, but that
flag gets read and written *after* nodal coordinates, so I don't see
how it...

Damn it!

I've got unique id data being written at the end of *each* chunk of
nodes.  Of course this is passing all our tests on small meshes and
breaking horribly on large ones.  Let me see if I can fix it quickly.

Sorry,
---
Roy