Bug found, L2 error calculation working

The problem was an temporary r-value, for which the memory got "stolen" by mfem. So the reference was reset which hindered the mfem VectorFunctionCoefficiente Evaluation process..

tests/time_dependent_gridfunction.cpp

@@ -119,6 +119,25 @@ TEST_CASE("Time Dependent Gridfunction", "[time_dependent_gridfunction]") {
                 double const eoc = std::log2(error/error2);
                 REQUIRE_THAT(eoc, Catch::Matchers::WithinAbs(order+1., 0.1));
             }
+            SECTION("Second Order"){
+                int order = 2;
+                mfem::H1_FECollection fec_x(order, mesh_x.Dimension());
+                mfem::H1_FECollection fec_t(order, mesh_t.Dimension());
+                mfem::FiniteElementSpace X(&mesh_x, &fec_x, vdim);
+                mfem::FiniteElementSpace T(&mesh_t, &fec_t);
+                vtint::time_dependent_gridfunction_t u_gf1(T, X, u_analytical_vectorial);
+                double const error = u_gf1.l2_error(u_analytical_vectorial);
+                //write to file to see if it is reasonable
+                //u_gf1.save_as_vtk("first_order_vectorial", 0.05);
+                mesh_t.UniformRefinement();
+                mesh_x.UniformRefinement();
+                mfem::FiniteElementSpace X2(&mesh_x, &fec_x, vdim);
+                mfem::FiniteElementSpace T2(&mesh_t, &fec_t);
+                vtint::time_dependent_gridfunction_t u_gf2(T2, X2, u_analytical_vectorial);
+                double const error2 = u_gf2.l2_error(u_analytical_vectorial);
+                double const eoc = std::log2(error/error2);
+                REQUIRE_THAT(eoc, Catch::Matchers::WithinAbs(order+1., 0.1));
+            }
         }
     }SECTION("H1-L2 seminorm error (i.e. in H^1(0,T;L^2(\\Omega))") {
         SECTION("First order") {

vtint/time_dependent_gridfunction.cpp

@@ -250,7 +250,8 @@ namespace vtint {
             mfem::GridFunction spatial_gridfunction = this->spatial_gridfunction_for_time(t);
             auto u_analytical_for_time = [u_analytical, t, vdim](const mfem::Vector &x, mfem::Vector & res){
                 res.SetSize(vdim);
-                res = u_analytical(t, x);
+                mfem::Vector const u_tmp = u_analytical(t, x); //we have to do that, otherwise it will be a "stealing copy" from an r-value, which leads to mfem problems...
+                res = u_tmp;
                 std::cout << "t: " << t << "\n";
                 std::cout << "x: " << "\n";
                 x.Print(std::cout);