warnings when adding DerivativeTwoPhaseMaterial to KKS example

[jessecarterMOOSE]

If I add a DerivativeTwoPhaseMaterial to a KKS example problem using this block:

  [./two_phase]
    type = DerivativeTwoPhaseMaterial
    f_name = Fsum
    args = 'cl cs'
    eta = eta
    derivative_order = 2
    fa_name = fl
    fb_name = fs
  [../]    

I get some warnings from the kernels when running the problem:

*** Warning ***
Missing coupled variables {eta} (add them to args parameter of ChemPotSolute)


*** Warning ***
Missing coupled variables {eta} (add them to args parameter of ChemPotSolute)


*** Warning ***
Missing coupled variables {eta, cs} (add them to args parameter of CHBulk)

I'm curious why this is. I see now that the kernels mentioned in the warnings take the single-phase chemical energy materials and do not need eta, so the DerivativeTwoPhaseMaterial is adding a global dependency on eta? Is this something I need to worry about? I wouldn't imagine that adding extra variable dependences would matter (except maybe in terms of performance) but wanted to check that this is expected behavior.

[laagesen]

Hey Jesse, these warnings are triggered by the validateNonlinearCoupling function in initialSetup() for both these kernels. Are you supplying Fsum to these kernels after you added it? I'm guessing not but just wanted to make sure.

[jessecarterMOOSE]

No I left everything else the same. I'm only making some extra materials for validation/postprocessing reasons.

[jessecarterMOOSE]

@laagesen the results come out the same either way, so is it safe to ignore the warnings? Would that always be the case?

[laagesen]

Yes it should be safe to ignore it. Although the message is a bit confusing and we will try to look into why it's getting triggered in this situation, which is not desired. In general this warning is only telling you that you are missing coupled variables which might be needed for a calculation of Jacobians. So it should not affect the results you get even in a situation where you are actually missing something. It could potentially make convergence worse, but should not affect the answer you get (assuming convergences tolerances are set to reasonable values).

[jessecarterMOOSE]

Makes sense. Thanks!