Bug with AdaptiveDiscreteModels

[janmodderman]

I ran into an AssertionError which I believe to be a bug when an AppendedTriangulation is constructed using a SubCellTriangulation and an AdaptedTriangulation.

The following piece of code reproduces the error:

module debug_embedded_adaptive
using Gridap, Gridap.Adaptivity
using GridapEmbedded

model = CartesianDiscreteModel((-8,8,-8,8,-8,0),(2,2,2))                                                         
model = refine(model,2)
geo = tube(2.0,4.0,x0=VectorValue(0,0,-4.0),v=VectorValue(0,0,1))
cutgeo = cut(model, geo)
Ωᵢ = Interior(cutgeo, PHYSICAL)
end # module

The failing assertion (located in https://github.com/gridap/Gridap.jl/blob/0f2ef80ff010c558514166b03c3b276e4c4c945b/src/Geometry/AppendedTriangulations.jl#L110) requires get_background_model from both triangulations to return the same object, for the SubCellTriangulation this is correctly called and returns an AdaptedDiscreteModel, yet for the AdaptedTriangulation an UnstructuredDiscreteModel or CartesianDiscreteModel is returned, which corresponds to the background model of the AdaptedDiscreteModel (see https://github.com/gridap/Gridap.jl/blob/0f2ef80ff010c558514166b03c3b276e4c4c945b/src/Adaptivity/AdaptedTriangulations.jl#L49).

I'm not sure what the best solution would be here.

Currently, I have a workaround that changes the background model of the AdaptedDiscreteModel in the cut function, which seems to fix things.