This only works for cc at the moment right?

I am referring to the output module...

yeah because maybe we want to do something more fancy for box involving svcs. Then the OutputModule has to become a property.

you mean for the saturation output in case of using an interface solver? Maybe for now we can write the output such that it works for box and cc and then modify it when we have the interface solver. I suppose the output should easily work for the two methods, right? Using scv.dofIndex() to get the privars etc...

ok I'll check

Then we could merge it and do the transition to the new output module step by step... The current implementation would simply fail for a model using the box method right? There is no check whether cc is used here to skip this output no!?

The problem is not only the interface solver. But also spatial params. The current version now continues the "bug" in the case of element parameters that randomly the last scv determines the vertex value if the parameter is not defined on the box. For solution dependent spatial params the only option is to output per scv.

Of course, but what happens right now in the output module when running a box model? It should be clear and well defined before merging the branch into next...

I guess the same happens as on master. Although the iteration sequence might be different, but I think we use the same loop of elements and local vertices.

So on master it's not clearly defined. Do you want to improve it before or after merging? Improving it, we might need to exchange some reference solutions too.

I guess I wasn't clear :). To me it seemed that the implementation of the output module only works for cc.. but is it called also for box models? You did commit calls to this writer, e.g. for the 1p model right? So what happens if I want to use the 1pbox model now.. I could also have tested it, that would've been faster I guess :). I just wanted the box output the way it was implemented here or something that makes sure we don't get a segfault when running a box model...

I guess you didn't see the latest commit?

Ah great, that's what I wanted from the beginning ;)