Control: tags -1 pending

Niels Thykier:
[...]
Hi Nicolas,


First off, thanks for the update in bd250c063ce2b68d7a71c342550360c1ad035d1d. I squashed it into cf4b4ce36925c84339a78415822ac7c54e5e5209 and have merged the code. I also added a minor tweak to the migration code such that it detects `dh-builtusing` patterns and emits a warning about having to migrate the manually.

Ideally, that migration would be automated but the migration infrastructure does not support patching `debian/control` yet, so that was a "better than nothing" fix.

Since metadata detectors require a integration level of `full` or `dh-sequence-zz-debputy`, I also added that restriction to the new keywords. I failed to mention this as a restriction when I decided to port this feature to use a metadata detector. Let me know if you had plans to use this feature specifically with `dh-sequence-zz-debputy-rrr`.

I have gotten a bit further. I pushed some WIP code to the bug-1121339- built-using branch. It fails because I ported some code wrong/ incompletely and I did not have energy to fix it. I will look at it later (some features are not available where they should be, I got a type error, not dealt with the `_w`/`_d` functions, etc.).


Obviously fixed these. I tweaked the logging so it now uses the exact path to the pattern in the manifest when logging (such as `packages.NAME.built-using[0].sources`)

However, the tests are ported to the general layout I would want and are present in the `tests/test_built_using.py` file. It can be run with `py.test` (but will fail due to the above).

I have one TODO in it about the test running the detection code even though the package is not supposed to be built (real run would skip the `built-using` code here). We are talking here about the `01basic` test with the `package-disabled-by-arch` + `package-disabled-by-profile` code.

As I read the definitions, a real run of `debputy` (with the tweaked code) would never run the built-using code for these packages (in this port), since these packages are not "active". In other words, the test should really be looping over `manifest.active_packages` rather than `test_definition.packages` to match reality.

Best regards,
Niels


I tweaked the tests to skip running the metadata detector when the package is not active. I am happy to accept patches or merge requests for additional tests to cover anything you might feel would be lost by this change.

I plan to do an upload soon (possibly later today) with these changes, so you can start using this feature. :)

Best regards,
Niels


Attachment: OpenPGP_signature.asc
Description: OpenPGP digital signature

Reply via email to