[Libmesh-devel] [BUG] Boundary Mesh -- fixed

[John P. <pet...@cf...>]

Martin L,b|(Bthi writes:
 > Hi John
 > 
 > John Peterson writes:
 >  > Thanks for reporting that bug.  Please try updating and running
 >  > your test code again -- hopefully it is fixed.
 > 
 > Thanks a lot for your very fast response. Unfortunately the example
 > code still breaks in exactly the same way. 

Doh!  For the original problem, instead of a seg-fault I hit an assert
in debug mode.  So it may be that, although there is still a
seg-fault, it is now for some other reason.

 > Just some guesses:
 > 
 > o All nodes are copied in the beginning, but they are still references
 >   to the initial mesh. Upon cleanup they get destroyed and so the
 >   original mesh has many corrupted references.

They aren't references to the original mesh... they are their own
separate memory-allocated-for entities.  One should be able to
completely delete the original mesh and still have everything function
perfectly (in theory).

 > o Is the BoundaryMesh renumbered after creation, say, when writing to
 >   a file?

The BoundaryMesh would potentially be renumbered if prepare_for_use
is ever called.

 > I'll try to find out where the problem happens, as soon as I am able
 > to build the library in debug mode (currently I'm having problems with
 > tetmesh's tetrahedralize not being defined).

Yeah, I'll need a simple, reproducible test case which demonstrates
the problem to be able to continue bug hunting.

Do you mean tetgen?  I think it is working OK in debug mode for me
anyway.

-J