memory leaks, bug in import 1D for cracks
M src/SMDS/SMDS_UnstructuredGrid.cxx
M src/SMESH/SMESH_MeshEditor.cxx
M src/SMESH/SMESH_MesherHelper.cxx
M src/SMESHUtils/SMESH_MeshAlgos.cxx
M src/SMESHUtils/SMESH_MeshAlgos.hxx
M src/SMESHUtils/SMESH_Octree.hxx
M src/SMESHUtils/SMESH_Tree.hxx
M src/SMESH_PY/smeshstudytools.py
M src/StdMeshers/StdMeshers_Import_1D.cxx
http://www.salome-platform.org/forum/forum_10/998544058
class SMESH_2D_Algo
{
+ /*!
+ * \brief Method in which an algorithm generating a structured mesh
+ * fixes positions of in-face nodes after there movement
+ * due to insertion of viscous layers.
+ */
+ virtual bool FixInternalNodes(const SMESH_ProxyMesh& mesh,
+ const TopoDS_Face& face);
Now UnRegister(), corresponding to the Register() called by AttributeIOR,
is invoked by an object publisher i.e. by SMESH_Gen_i::publish() but not by GUI
because GUI is unable to track publication of all objects created by the Engine:
for example groups can be created while mesh.Compute() is called.
1) call myFilterDlg->UnRegisterFilters();
2) re-create an actor after converting a group to standalone as a
CORBA object becomes different from that held by the actor
Now UnRegister(), corresponding to the Register() called by AttributeIOR,
is invoked by an object publisher i.e. by SMESH_Gen_i::publish() but not by GUI
because GUI is unable to track publication of all objects created by the Engine:
for example groups can be created while mesh.Compute() is called.
A better fix has been found by improving the getNextFree() method in ObjectPool.hxx. When there is no "hole" in the ID list we don't iterate on the _freeList to find the next free ID. We jump straight to the last occupied ID.
This fix is simpler and can benefit to other methods of SMESH_Editor like QuadTo4Tri for example.
The whole procedure performance was almost O(n^2) due to insertion of for example 5 elements in a mesh with a free ID at the beginning. The second element is then inserted with a O(n) complexity.
The hexas are now removed after all tetra insertions, which guarantees a O(n) complexity for the whole procedure at a limited memory cost (transient additional cost of 1/5 of total memory occupation at the end).
1) New general mechanism to avoid dump of functors not used later as arguments
2) Use the New general mechanism to filters and sub-meshes
3) Erase the 1st incarnation of re-created objects.
A re-created object appears e.g. as follows:
- create a mesh
- remove the mesh
- save and re-open the study
- create another mesh -> ( two meshes have same Entry ) == ( the 1st mesh is re-created )
