Add pybind11 to dockerfile for system test

[Fujikawas]

According to #696 (comment), the micro-manager participant in two-scale-heat-conduction case depends on pybind11.

Checklist:

• I added a summary of any user-facing changes (compared to the last release) in the changelog-entries/<PRnumber>.md.
• I will remember to squash-and-merge, providing a useful summary of the changes of this PR.

[IshaanDesai]

pybind11 is technically a dependency of micro-dumux, and not of the adapter. Could we install it right before the participant micro-dumux is compiled?

[Fujikawas]

pybind11 is technically a dependency of micro-dumux, and not of the adapter. Could we install it right before the participant micro-dumux is compiled?

You are right. As pybind11 is also not an independent component that we would test etc. like the other adapters/MM, I'd tend to merge it into the installation of Micro Manager since it's often used between C++ case and MM. Also I would like to base the dumux-adapter on preCICE instead of MM.

[IshaanDesai]

You are right. As pybind11 is also not an independent component that we would test etc. like the other adapters/MM, I'd tend to merge it into the installation of Micro Manager since it's often used between C++ case and MM. Also I would like to base the dumux-adapter on preCICE instead of MM.

Makes sense! Let's trigger the system test from this PR once the changes are done.

[MakisH]

Besides #739 (comment), the rest of the changes make sense.

[IshaanDesai]

Despite changing the DUMUX_ADAPTER_REF to a commit hash on its develop branch, the system test manual run still fails with an error that occurs because an older version of the adapter is used. This version was previously used in the system test. Could it be that the cache is not cleared?

[MakisH]

Link to the job output?

[Fujikawas]

Link to the job output?

https://github.com/precice/tutorials/actions/runs/23255829787

[IshaanDesai]

Link to the job output?

https://github.com/precice/tutorials/actions/runs/23255829787

The error that the participants two-scale-heat-conduction/macro-dumux, free-flow-dumux, and porous-media-dumux appear because the DuMuX adapter v3.0.0 gets picked. This happens despite b4d1d58, in which Jun changed the adapter reference to a commit on its develop branch. The question is why, despite doing this, the old version of the adapter got picked.

[MakisH]

from the output:

WARNING: No argument provided for needed parameter DUMUX_VERSION. Substituting with 3.7
WARNING: No argument provided for needed parameter DUMUX_ADAPTER_REF. Substituting with main
WARNING: No argument provided for needed parameter DUNE_VERSION. Substituting with 2.9

are the changes already in main?

[IshaanDesai]

Good catch! These changes are not in main. @Fujikawas, this means that the DUMUX_ADAPTER_REF is not set correctly.