[Getfem-users] Problem with vtk_export of the solution of a 2D laplacian problem embedded in 3D

[Andres Peratta]

Hi
I am solving a laplacian problem and bumped into troubles with the vtk_export 
and the interpolation function. While trying to export the solution with 
write_point_data(mf_u, U, "Potential") the following error on "dimensions 
mismatch" pops out:


============================================
|      An error has been detected !!!      |
============================================
Error in /usr/local/include/getfem/getfem_interpolation.h, line 466 :
Dimensions mismatch

* The error appears when running an executable compiled with -O3 flag, 
but everything works fine in the debug version of the code.

* The corresponding piece of code is: 
getfem::vtk_export exp("output.vtk"); // runs okay
exp.exporting(mf_u);  // runs okay
exp.write_point_data(mf_u, U, "Potential"); // The problem seems to start 
here; and appears when vtk_export::write_point_data attempts to use the 
interpolation function from getfem_interpolation.h.


Below are more details of the problem. 
Many thanks for your help/comments, 

Andres


--------

The integration domain consists of a 2D surface in 3D space. (i.e. the surface 
is arbitrarily warped in 3D space). The mesh, imported directly from GMSH,  
consists of flat triangular elements and the line segments of the boundary. 

I modified "laplacian.cc" in order to read the mesh from gmsh as follows:

getfem::import_mesh("gmshv2:input.msh",mesh);   // Import from gmsh
N = mesh.dim();    // dimensionality

The mesh represents a 2D surface warped in 3D space, however N becomes 3; I am 
not sure if this is still okay; however I get good solutions when running the 
debug version. 

// I use the following parameters:

MESH_TYPE = "GT_PK(2,1)" 
FEM_TYPE = "FEM_PK(2,2)"
INTEGRATION = "IM_TRIANGLE(6)"

// The postprocessing part of the code is as follows:

getfem::vtk_export exp("output.vtk");
exp.exporting(mf_u); 
exp.write_point_data(mf_u, U, "Potential"); // This line triggers the error

// The compiler is gcc version 4.2.1; but most certainly it's me doing 
// something wrong rather than a compiler issue.