Create github action to check consistency of requirements.txt and notebook_requirements.txt #297
Conversation
.github/workflows/requirements.yaml
Outdated
| jobs: | ||
| requirements: | ||
|
|
||
| runs-on: ubuntu-latest |
There was a problem hiding this comment.
Should we test other OSs? I guess it depends on whether we are checking that the environment can be installed easily, or checking whether the package versions are compatible (which is a weaker requirement than the package versions can be installed easily on all OSs).
There was a problem hiding this comment.
Although I'm a little worried we will open a can of worms, it does sound useful to check on windows and osx. Can we give it a try?
There was a problem hiding this comment.
Or maybe just add the other OSs to the pytest workflow?
|
This looks good, but i'm imagining cutting 2.1.1, so we should probably update the changelog about the xgboost requirement |
|
Given this is just a maintenance PR, do we need to cut a release? I imagined the xgboost requirement change to be a short-term change that gets reverted before we're ready to cut another release. If that's the case then I don't see a point in a changelog entry. |
|
I think we should do a patch, an updated |
|
Oh right, if it's broken for us, it's broken for everyone. Forgot about that part. Changelog added! |
__init__.py@mdeceglie suggested adding a job to verify the consistency of
requirements.txtandnotebook_requirements.txtto make it easier to review PRs from dependabot. Note that consistency ofrequirements.txtis already checked by the pytest jobs, but currently no CI job usesnotebook_requirements.txt. Also note that dependabot does suggest updates tonotebook_requirements.txt, not justrequirements.txt, e.g. #288. Maybe this job will become redundant if we ever get around to #270.I tested some inconsistent environments on my fork to see what failures look like; here are the corresponding jobs:
Note that
inconsistent 1requested a version of numpy without wheels for py 3.7, so it tried to compile it from source and it was that compilation that failed, not a consistency error, so it's not really relevant. Butinconsistent 2shows an actual consistency failure.