[cgma-dev] imprint mesh bug

[Boyd Tidwell]

The following commands:

     brick width 10
     brick width 6
     body 2 move x 8
     volume 1 2 size 1
     mesh volume 1 2
     imprint mesh body 1 2

cause a crash.  The same commands without the meshing works fine.


The actual failure occurs in

 void PartitionSurface::get_parents_virt( DLIList<TopologyBridge*>&
result_list )

On the assert in the following code, tmp_list.size() is zero.

    for( i = real_surf_shells.size(); i--; )
    {
      tmp_list.clean_out();
      real_surf_shells.get_and_step()->get_parents_virt( tmp_list );
      assert(tmp_list.size() == 1);
      real_surf_vols.append(tmp_list.get());
    }


A clue to the real problem may be revealed in the method called right
before the assert:


void OCCShell::get_parents_virt( DLIList<TopologyBridge*>& parents )

 for(int i = 0; i <  bodies->size(); i++)
  {
    body = bodies->get_and_step();
    TopoDS_Compound* shape = body->get_TopoDS_Shape();             <----
added line
    if (shape ==
NULL)
<---- added line

continue;
<---- added line
    TopExp::MapShapesAndAncestors(*(body->get_TopoDS_Shape()),
                                     TopAbs_SHELL, TopAbs_SOLID, M);
    if (!M.Contains(*(get_TopoDS_Shell())))
         continue;


Previously I added the lines to check if "body->get_TopoDS_Shape()" was
returning a NULL pointer.  It seemingly fixed a number of problems. I wish
I'd written down the actual things that the change fixed.  This might just
be a band-aid that hides a deeper problem but my tests run much better with
the new code in place.  Because the existing code doesn't check for a null
pointer, there seems to be an assumption that it will never be null so the
null pointer could be the real problem.  I don't have enough understanding
of the code to know what is correct.

Good Luck and thanks in advance,

  - Boyd
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/cgma-dev/attachments/20111228/4047bde1/attachment.htm>