Note that I don't always get empty logs. In previous states of debugging, I was getting output. I think that, whenever it is building the adapter, the log gets too long or something similar. But even when I build based on the cached layer (that's why I split the meson build and the ninja install), I get no output.

Or maybe the Docker layers were cached, so nothing was printed in the stdout.log, but somehow the running did not print any logs.

It is actually failing in the build step, due to dependencies (e.g., mpi4py).
Logging continues to be a puzzle: sometimes the stdout.log has content, sometimes not (but at the moment mostly not).

New observation: This run actually completed the build step, and failed in the beginning of the run step. No stdout.log was written at all (not an empty file; just no file at all), while we got simulation logs.

The system tests now succeed: https://github.com/precice/tutorials/actions/runs/8672889794/job/23783659705

I have updated all reference results in this PR as well (we cannot only update reference results for specific tests with the current workflow).

I also needed to increase the global timeout, as the SU2-FeniCS test currently takes 8min.

@valentin-seitz could you please have a quick sanity check? Have I forgotten anything? Have I misused anything?

Regarding the logging issue, let's postpone and offload it here: #522

The system tests now succeed: https://github.com/precice/tutorials/actions/runs/8672889794/job/23783659705

I have updated all reference results in this PR as well (we cannot only update reference results for specific tests with the current workflow).

I also needed to increase the global timeout, as the SU2-FeniCS test currently takes 8min.

@valentin-seitz could you please have a quick sanity check? Have I forgotten anything? Have I misused anything?

Regarding the logging issue, let's postpone and offload it here: #522

I think we discussed most of it already via phone.
We should most definitely have a look into the spurious logging behavior and maybe document the possibilty of the docker compose --file docker-compose.tutorials.yaml up thing somewhere in the cli.
The last safeguard is the adding of the reference results. Here the pipeline should not fail. Which it did not, so 👍