Add basic rootless podman support

[SoMuchForSubtlety]

The implementation is mostly borrowed from the rootless docker strategy.

[SoMuchForSubtlety]

Hi, when you have a chance could you review my PR? 🙂

[eddumelendez]

Hi @SoMuchForSubtlety! Sadly, we don't have enough tests to make sure this is working. We tried adding GHA #4583 but without success. We don't feel comfortable adding this when AFAIK there are more things to check against podman.

[TheHaf]

So, what kind ofthings would be needed to check against podman in order to make this happen? I for one can't use Docker anymore and need to get Testcontainers to work with Podman properly and I feel that I am not alone in this. Seeing how any and every podman support PR is being closed makes me kinda sad.

[bsideup]

Seeing how any and every podman support PR is being closed makes me kinda sad.

We are also very sad that Podman isn't fully Docker API compatible yet! We kindly offered Podman team our help with making sure Podman is better compatible, plus recommended using our test suite to ensure that, since Testcontainers uses many Docker API and exposes any incompatibilities, but it seems that the team is okay with the way it is right now.

So far, none of the "Podman support PRs" have passed the acceptance criteria and required changing our tests and/or even worse changing the implementation. Our tests represent real world usage of Testcontainers and, if some of them aren't passing, it means that it won't work for our users too.

We will gladly accept a PR that only adds the strategy and GitHub Actions setup to test it, without touching anything else.

[TheHaf]

Thank you for the clarification, that makes the whole situation much clearer for me. It sounds like this is probably also a somewhat frustrating situation for you.

If I understand you correctly, one important step would be to take your test suite and run it against podman in order to bring any compatibility issues to light. Though the podman community would need to be willing to acknowledge any such issues and improve upon them. But in any case, It would be interesting to see where podman actually stands in regard to passing this test suite.

Regarding the handling of such podman support PRs, maybe it's not entirely clear the the creators of the PR what they were missing or what they did wrong. I have a feeling that providing more information about why a PR was rejected would help in such cases. In this regard, I would think that your last sentence is probably very helpful to the PR creator, for example.

One thing that is not clear to me is your statement about changing the implementation being not possible because it would break things for the users. I can see that things should not break for users, but shouldn't the behavior of the interface according to the specification be the benchmark and not the behavior of the implementation? And what would be the way to resolve things in cases where the observed behavior is actually proven to be wrong?