Removed interface compatibility.

[alarshi]

In benchmarks/solkz

[bangerth]

/run-tests

[tjhei]

It looks correct, so I am surprised that the output changes. I guess we ignore this slight change and update the results?

[bangerth]

@tjhei: Even though this patch changes a file in benchmarks/, are there tests that reference the material model?

[bangerth]

@tjhei -- no, the output shows errors, and they are a factor of 5-10 smaller than before. We need to figure out what is different here before this can be merged. Something is substantially wrong here.

[tjhei]

@tjhei: Even though this patch changes a file in benchmarks/, are there tests that reference the material model?

Yes. We include the .cc in the test.

@tjhei -- no, the output shows errors, and they are a factor of 5-10 smaller than before.

One additional Stokes iteration can cause such a difference (errors are small to begin with).

[bangerth]

Really? This is crazy. What you're saying is that the total error is completely determined by the iteration error, not the discretization error.

[bangerth]

Addresses #1538.

[tjhei]

Well, not 100% sure...

[bangerth]

Let's leave this for some later time after the hackathon to investigate. We're not going to get to the bottom of it tonight.

[gassmoeller]

Found the problem. The particle_interpolator_bilinear_solkz test uses a material model that is derived from the SolKzMaterial class, which changed in this PR. The derived model overloads the density and viscosity functions and modifies their behavior. Because of this PR these functions are not longer called, and therefore the modified behavior is never triggered. @alarshi: Please keep this PR as is, except that the new class still needs the density and viscosity function exactly as they were before. evaluate then needs to call these functions to determine the density and viscosity.

[tjhei]

I did the fix.

[bangerth]

Yes, @tjhei's is the way to go. But can you (@tjhei) run astyle on your patch?

OK to merge if the tester is happy.

[tjhei]

Beginner mistake. ;-)