test_tv_run_inversion broken in parallel

[dngoldberg]

The test_tv_run_inversion test throws a runtime error when run with MPI (n>1):

fenics_ice/tests/test_runs.py

Line 71 in 38a72f8

def test_tv_run_inversion(persistent_temp_model, monkeypatch):

Note that an assertion error is thrown but NOT the one tested by pytest, but rather a line 259 of solver.py.

The toml has both alpha and beta active, and a dual inversion is run; the solver object therefore expects a tuple of parameter functions. im not sure the best way to handle it. One way would be to modify the toml (at runtime or by changing the file) and set beta_active to False. But then the inversion would not be a dual one, and the gradient of beta would not be tested. i think alpha is the more important to test due to the grounding line though.

The same likely applies to test_tv_run_forward but i have not tested it.

[bearecinos]

Weird I see what you mean I'll have a look at this after I finish sending oggm runs ... maybe these tests will benefit from a independent toml