tests/test_CLI.py: create $GNUPGHOME on the fly

[josch]

• avoid expiration of keys by re-creating them
• prevent gnupg version on developer's system being incompatible with $GNUPGHOME in git

Storing binary data is bad because:

• git is not good at handling binary data
• binary data is harder to inspect (remember the xz incident)

With this commit, test_clearsign fails with "TypeError: 'NamedFile' object is not iterable" but I read that this issue is addressed by #1 so I'm not adding this here.

[josch]

@jo-so-nx I rebased your commits from #1 on top of mine, except for 385f4d6 as the expiration extension by 50 years is not necessary if the key is generated on-the-fly.

What do you think?

[jo-so-nx]

@jo-so-nx I rebased your commits from #1 on top of mine, except for 385f4d6 as the expiration extension by 50 years is not necessary if the key is generated on-the-fly.

What do you think?

I'm fine with this. You can pick and modify them as you like.

I must admit we have abandoned bmaptool in our project and have no longer the infrastructure to test anything.

[josch]

I'm fine with this. You can pick and modify them as you like.

Thank you!

I must admit we have abandoned bmaptool in our project and have no longer the infrastructure to test anything.

I'm a bit worried about the same having happened to Yocto as there was very little activity in this project since March 2024. What is the status @JPEWdev @twoerner?

[zeha]

ping @JPEWdev @twoerner please, could you check/apply this? thanks :)

[JPEWdev]

I'm looking at it, but I've lost the ability to trigger workflows, so until that's done I can't run the CI to test this

[JPEWdev]

@josch Sorry, I cannot figure out why I can't retrigger the CI on this. Can you please rebase your branch on main and push it again? That should trigger it.

[josch]

Can you please rebase your branch on main and push it again? That should trigger it.

Done, everything seems to pass, thank you! 😃

[JPEWdev]

The signing tests where skipped for some reason, so I'm not sure the test run is valid. Can you please look into that?

[josch]

Thank you, nice catch! I wonder though what is missing

I tried adding python3-gpg into .github/workflows/ci.yml apt-get installation line to no avail. The file pyproject.toml lists dependencies = [ "gpg >= 1.10.0" ] and the log says:

2025-01-13T18:59:32.8524042Z   Building wheel for gpg (pyproject.toml): started
2025-01-13T18:59:34.2209346Z   Building wheel for gpg (pyproject.toml): finished with status 'done'
2025-01-13T18:59:34.2213954Z   Created wheel for gpg: filename=gpg-1.10.0-cp312-cp312-linux_x86_64.whl size=40395 sha256=ed66083c743d0e3f2505316b7750464de7bfb979cef656163111a5e6503799fa
2025-01-13T18:59:34.2216864Z   Stored in directory: /home/runner/.cache/pip/wheels/7b/3e/c4/50fc15a2e3f7f5264dc8df81e6dec60449bf3f252808415713
2025-01-13T18:59:34.2236498Z Successfully built bmaptool gpg
2025-01-13T18:59:34.2461959Z Installing collected packages: gpg, six, platformdirs, pathspec, mypy-extensions, click, bmaptool, black
2025-01-13T18:59:34.5483116Z Successfully installed black-24.10.0 bmaptool-3.8.0 click-8.1.8 gpg-1.10.0 mypy-extensions-1.0.0

Still, in tests/test_CLI.py it fails to import the gpg module... I'll keep investigating...

[JPEWdev]

Maybe temporarily remove the try...except around the import and see if it gives a more useful error

[JPEWdev]

Ah right.... I attempt to fix this here: f61172a but it still doesn't work because it can't seem to import gpg after the change.

It's an unfortunate combination of the gpg and setuptools version.... and I'm not actually sure if there is a way to fix it cleanly since we are kinda stuck with version of gpg that comes from the distro.

[JPEWdev]

I was looking at your changes, and I finally figured out what the problem with the CI is. Basically, the pre-canned python versions provided by GitHub Actions can't possibly provide the gpg module correctly, because it must come from the host package. To facilitate this, I made a patch that extends the CI testing to test with "native" python, which can use the host python3-gpg, and thus run the tests. This should cherry-pick on your branch in place of the wip patch you have: JPEWdev@eb978ae and I'm pretty sure this will fix the CI

Thanks so much for looking into this, it was really helpful!

[josch]

Ah, sorry, I panic-removed my last comment. I must've clicked the wrong test because I still saw those skip messages. It looks fine for the native case now. Thanks! :)

[JPEWdev]

Had to squash and merge because this repo only allows rebase or squash, and rebase wasn't an option because the branch was not up to date. Doesn't matter too much :)

Thanks!