SkeletonPairs and transient problems #1125
Description
Hello to all.
Wanting to learn about the discontinous Galerkin method (and later move on to HDG and HHO), I tried to apply the approach to transient problems. To that end, I combined tutorials 6 and 18 to test the DG formulation on a transient nonlinear problem. The thing almost works, as I can set up the test space with :L2 conformity and a transient trial FESpace and then solve the system using solvers from DifferentialEquations.jl (see issue #962).
Unfortunately, I can only make it work if I drop the weak form terms corresponding to the skeleton triangulation, and so the global solutions are of course not continuous. As soon as I try to include the additional terms defined on the skeleton, I typically get an error saying that the function (involving jump and mean operators) belongs to the interface definition and cannot be used. I am attaching a MWE, note that the error persists even if I replace the external solver with the one proposed in Gridap tutorials for transient problems.
In the MWE, I have excluded the skeleton contribution to the weak form, but the error can be easily reproduced
if the last term in line 58 is uncommented:
a(t, u, v) = ∫(∇(v) ⋅ (α(t, u) * ∇(u)))dΩ + a_Γ(t, u, v) #+ a_Λ(u, v)Is this a bug or are skeleton pairs simply not intended to work with transient problems?
Best,
Jan