How to get rid of segfault when using assemble()

[dokken]

If you assemble on a subdomain using the full mesh, but with a MeshFunction as input to the integration measure:

dx_ = Measure("dx", domain=mesh, subdomain_data=mesh_function)
a = ...*dx_(cell_marker)
A = assemble(a)
A.ident_zeros()

as shown in for instance:

MixedElement vs MixedFunctionSpace

Hi @Thomas2, I don’t know how to thank you, I was FINALLY able to run the simulation thanks to your suggestions! For clarity, I needed to change your pseudo code a bit, this is the final version: F = # my weak form a, L = lhs(F), rhs(F) A = assemble(a, keep_diagonal=True) b = assemble(L) for bc in bcs: bc.apply(A, b) A.ident_zeros() sol = PETScLUSolver() x = Function(V) sol.solve(A, x.vector(), b) This solution does not need the use of MixedFunctionSpace but just of MixedElement, I also run t…

or

Obtain velocity normal to boundary

As the FacetNormal as implied by its name, is defined on each Facet, and not for each degree of freedom, you have to approximate the normal at the degrees of freedom (for instance at corners). For your particular case, the problem is that you are trying to interpolate an ComponentTensor, that is an expression using ufl syntax. To work around this, I would suggest using an UserExpression as follows: n = FacetNormal(mesh) # Interpolate mesh values into boundary function u_boundary = Function(D) …