Update CI

[trungleduc]

Code changes

• Reorganize test workflow

• Drop python 3.8
• Remove uneeded steps in the pre-commit job
• Run UI test with packages generated from current PR, not from PYPI.
• Relax Playwright UI test condition.
• Add dev install script: calling python scripts/dev_install.py or yarn dev is enough to set up the dev enviroment.
• Add update snapshot workflow

[github-actions[bot]]

:point_left: Launch a Binder on branch trungleduc/jupyter-collaboration/update-ci

[trungleduc]

@jtpio I'm getting this error in the Check Release job: ValueError: Tag v3.0.0 already exists. Do you know what happened?

[jtpio]

Looks like it may be related to the changes to the yarn.lock file?

It seems it's not able to bump the version to 3.0.1 in the workflow, which could explain the existing tag issue:

[jtpio]

Probably not related but something that should also be fixed on CI: https://github.com/jupyterlab/jupyter-collaboration/issues/380

[trungleduc]

Looks like it may be related to the changes to the yarn.lock file?

It seems it's not able to bump the version to 3.0.1 in the workflow, which could explain the existing tag issue:

Thanks!

[trungleduc]

Probably not related but something that should also be fixed on CI: #380

Ok i will take care of this

[trungleduc]

Failed UI test looks weird but i'm not sure it relates to this PR

• Actual

• Expected

[trungleduc]

do you want to update the dependency in pyproject.toml too?

isn't it a breaking change? I thought that main is targeting 3.1.x

[davidbrochart]

I don't think we should make a major release for that. Python 3.8 has reached end of life, I think all Jupyter projects should drop support for it too.

[jtpio]

Right it's probably fine to drop support for a Python version in a minor release. For example JupyterLab dropped support for Python 3.6 in JupyterLab 3.3: https://jupyterlab.readthedocs.io/en/latest/getting_started/changelog.html#id298

As long as it's visible in the changelog and/or the docs.

[jtpio]

So the UI tests seem to be failing consistently after a restart. It would be worth looking into it as otherwise we would have to live with a failing CI, which is going to be annoying very quickly.

[trungleduc]

So the UI tests seem to be failing consistently after a restart. It would be worth looking into it as otherwise we would have to live with a failing CI, which is going to be annoying very quickly.

This is an issue of JupyterLab, I opened the issue here https://github.com/jupyterlab/jupyterlab/issues/16934

[trungleduc]

@jtpio, how should we proceed with this PR? Should we wait for a new JupyterLab patch version or update the snapshot?

[jtpio]

Does this behavior happen for real users of jupyter-collaboration, not just in the UI tests?

If so, it will probably make sense to update to the "wrong" snapshot for now to indicate this is an expected issue. And then fix it when a new JupyterLab release with the fix is available?

[trungleduc]

Does this behavior happen for real users of jupyter-collaboration, not just in the UI tests?

If so, it will probably make sense to update to the "wrong" snapshot for now to indicate this is an expected issue. And then fix it when a new JupyterLab release with the fix is available?

I can reproduce it locally https://github.com/jupyterlab/jupyterlab/issues/16934

it will probably make sense to update to the "wrong" snapshot for now to indicate this is an expected issue. And then fix it when a new JupyterLab release with the fix is available?

I agree

[jtpio]

OK then let's update the snapshot so "CI passes" and the PR can be merged?

Thanks @trungleduc for looking into this!

[trungleduc]

it's green!