Missing decency on pywin32

[vidartf]

In jupyter/jupyter_client#469, a platform dependent dependency was added on pywin32. This dependency does not seem to have made it into the conda forge package, as reported here: jupyterlab/jupyterlab#7259

[jasongrout]

CC @owen-synergy, who filed the original issue in JuptyerLab.

[msbuckas]

Hi - I'm still getting the error when starting JupyterLab (as per the original post) for WIndows x64.

I have done a conda update --all and have the ffg. in my env:

Name Version Build Channel
jupyter_client 5.3.3 py_0
jupyter_core 4.5.0 py_0
jupyterlab 1.1.3 pyhf63ae98_0
jupyterlab_server 1.0.6 py_0

Do I have to wait for a new jupyter_client version to get this fix?

[jasongrout]

The fix is in the most current jupyter_client on conda-forge. It did change from a noarch package to an arch-specific package which may complicate things?

Can you try explicitly updating jupyter_client? The correct packages have build 1: https://anaconda.org/conda-forge/jupyter_client/files

[msbuckas]

@jasongrout I got the direct download to work.

Yes, I think the -noarch to -win64 was a complication, therefore:

1. conda remove jupyter_client (this also removed jupyterlab)
2. conda install --offline "C:<mydir>\jupyter_client-5.3.3-py37_1.tar.bz2"
3. conda install jupyterlab

NB! I first tried just step2 - but it did not fix the prob. So it looks like the -noarch pkg must be removed before the -win64 pkg gets installed? And step3 may also subsequently install other dependencies because of this hack?

Thanks for your quick help!

[jasongrout]

You should be able to remove just the jupyter_client package with conda remove jupyter_client --force or something.

[msbuckas]

This issue is 'Closed' so I assume no further work is being done to resolve the arising -noarch vs -win64 packaging complication? Surely, the workaround is a temporary hack and users are not indefinately required to remove -noarch jupyter_client and then download and install the -win64 jupyter_client?

[jasongrout]

As a follow-up, pywin32 was removed as a dependency in jupyter_client 6.0, so I think this is no longer an issue: https://github.com/jupyter/jupyter_client/blob/1a25d7ddb1e68aef3ae3aeef87058c0a7a2daf0d/CHANGELOG.md#600