[MOAB-dev] iMesh_setAdjTable() causes error in iMesh_load()

[Jed Brown]

On Thu, Jun 9, 2011 at 17:20, Iulian Grindeanu <iulian at mcs.anl.gov> wrote:

> he6.h5m is a file that should contain only hexas and the skin organized in
> 3 neumann sets (bottom, top and lateral quads)
> Are you loading it with iMesh, and then saving it again?
>
> Can you do an
> mbconvert he6.h5m newhe6.h5m ?
>

Yes, this works fine.


> The new one will be stripped down to bare essentials (the sets should be
> preserved)
>
> Maybe it is something wrong in he6.h5m, although I can't say what.
>

I doubt it because I was doing simulations on it and saving the state just
fine a few weeks ago. Now I'm solving different equations (and I updated
MOAB in the mean-time) and it has stopped working. Loading he6.h5m is still
working fine as long as I use a month-old MOAB.

I committed my code that calls iMesh_setAdjTable() on May 1 and it
definitely worked then, but my MOAB may have been a week or two old at that
point. I can bisect the history, but it's mildly time consuming because I
have to cherry-pick my own patches back onto that state (so the build works,
this is a couple git commands), rebuild, and relink my library, and run. I
haven't written a condensed test case yet, but I can. Then I would be able
to automate the bisection process to find which commit broke it.

But note that the iMesh API changed around that time. A minimal example
probably wouldn't need to exercise those functions, but if I test using my
library, I have to revert my updates for the new API.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/moab-dev/attachments/20110609/903a2fe0/attachment.htm>