Global skip does not work in multi-output recipe

[h-vetinari]

Since this issue was reopened: the issue is that adding skip: true # [something] in a multi-output recipe does not translate to a skip for all outputs (which is why conda-smithy will keep generating jobs).

There's an argument to be made that this is somewhat correct based on the fact that the "global" output is not really all that special from the POV of conda. For example, if an output uses a script: ... variable, then it's conceivable that an output could still be produced even without running build.sh.

I don't know of any multi-output recipes though that are this clearly separated. In most cases, multi-output recipes are chosen at all because there's some substantial overlap in what needs to happen between the different outputs, and then the "global" build step becomes a hard requirement, without which trying to build makes no sense (hence the expectation that skipping it should be a clear enough signal to give up).

The - somewhat annoying - workaround is to add skip: true # [something] to every single output.

---

Previously:

In conda-forge/modin-feedstock#13, I updated skip: true # [py<37], since upstream isn't compatible with 3.6 anymore.

While trying to turn this recipe into a multi-output version, rerendering cause the 3.6 CI jobs to reappear (even though the build is skipped, so CI should stay green).

At first I thought this was a conda smithy issue, but was advised there that this is actually coming from conda render under the hood, and to raise an issue here.

[github-actions]

Hi there, thank you for your contribution!

This issue has been automatically marked as stale because it has not had recent activity. It will be closed automatically if no further activity occurs.

If you would like this issue to remain open please:

1. Verify that you can still reproduce the issue at hand
2. Comment that the issue is still reproducible and include:
   - What OS and version you reproduced the issue on
   - What steps you followed to reproduce the issue

NOTE: If this issue was closed prematurely, please leave a comment.

Thanks!

[leofang]

Can we please reopen this issue? We hit this issue in skipping PY38 in a conda-forge feedstock. I think it is a waste of the CI resources to launch pipelines and yet do nothing.

cc: @kenodegard

[h-vetinari]

Given that this was reopened, I updated the OP and the title. Relevant snippet:

The - somewhat annoying - workaround is to add skip: true # [something] to every single output.

[leofang]

Even adding skip to every single output does not work.

[h-vetinari]

Every single output, including the "global" one. I've never seen CI jobs created once every possible output was skipped.

[jakirkham]

Yes that is what Leo had tried earlier. Reference meta.yaml for discussion

[jakirkham]

Think an important point to note is once we added python in the top-level requirements/host, the skip (and conda smithy rerender in our case) worked

So think it is some combination of top-level requirements and skip that make the difference. We could probably poke at that example more to learn what is going on

[github-actions]

Hi there, thank you for your contribution!

This issue has been automatically marked as stale because it has not had recent activity. It will be closed automatically if no further activity occurs.

If you would like this issue to remain open please:

1. Verify that you can still reproduce the issue at hand
2. Comment that the issue is still reproducible and include:
   - What OS and version you reproduced the issue on
   - What steps you followed to reproduce the issue

NOTE: If this issue was closed prematurely, please leave a comment.

Thanks!

[h-vetinari]

Even though I know how to work around this now, it still feels pointless. A package-level skip should also skip all the outputs IMO.