Admconstraints used the old interface (and is used eg in the TOVSolver tests). I attach a patch to fix admconstraints. Note that this makes ADMConstraints inherit from TmunuBase so that all runs using ADMConstraints (even vacuum ones!) need to activate the thorn. I thus retract my previous comment about changing the default of tmunu_storage.
I agree with the change to tmunu_storage. The issue with ADMConstraints is the usage of ADMMacros. I do have a patch for that around to make it use the Tmunu grid functions and just need to dig it out if there is interest.
ADMConstraints seems to use TmunuBase directly as far as I can tell. I sets up the T[txyz][txyz] variables that ADMMacros apparently relies on. Applying only a patch to admconstraints (new_tmunu_interface.patch above) makes the TOVSolver tests (and all other tests that I could find, once I added TmunuBase to the test parfile see tmunubase_for_admconstraints.patch) pass. Apparently it does not use STRESSENERGY_guts.h which is the file in ADMMacros that uses the old interface.
According to Roland, when this ticket is resolved, the 16 currently-failing ET regression tests (http://build.barrywardell.net:8080/job/EinsteinToolkit/113/testReport/) will pass again. The longer we have failing tests, the more chance there is that another problem will appear and be difficult to disentangle. Can the patch new_tmunu_interface.patch be committed?