Skip to content

Set ceiling for setuptools#8509

Merged
jsignell merged 3 commits intodask:mainfrom
jsignell:setuptools
Dec 21, 2021
Merged

Set ceiling for setuptools#8509
jsignell merged 3 commits intodask:mainfrom
jsignell:setuptools

Conversation

@jsignell
Copy link
Copy Markdown
Member

All the tests are failing this morning since setuptools was just released and it now warns for when you use LooseVersion. We've already made the shift in dask, but some of our environments test older versions of pandas and numpy which didn't have the change yet. So this puts a ceiling on setuptools in those envs.

@jsignell
Copy link
Copy Markdown
Member Author

@jakirkham is there anyway this could be handled on the conda-forge side? Like maybe the pip version should be pinned on python, and the setuptools version can be pinned on that? Probably not that, but something like that?

@jsignell jsignell changed the title [TEST] Set ceiling for setuptools with python < 3.9 [TEST] Set ceiling for setuptools Dec 21, 2021
@jsignell jsignell changed the title [TEST] Set ceiling for setuptools Set ceiling for setuptools Dec 21, 2021
@jsignell
Copy link
Copy Markdown
Member Author

This unsticks us for now, so I'm going to go ahead and merge.

@jsignell jsignell merged commit f0e7ab5 into dask:main Dec 21, 2021
@jsignell jsignell deleted the setuptools branch December 21, 2021 16:12
aeisenbarth pushed a commit to aeisenbarth/dask that referenced this pull request Jan 6, 2022
Fixes the tests that are failing on import since setuptools was just released and it now warns for when you use `LooseVersion`. We've already made the shift in dask, but some of our environments test older versions of pandas, numpy, and distributed which didn't have the change yet.
@zzhengnan
Copy link
Copy Markdown
Contributor

zzhengnan commented Jan 20, 2022

@jsignell Hi Julia. In conda-forge/conda-forge.github.io#1575 (comment) you mentioned that setuptools 60.0.3 caused Dask's CI to fail.

It would seem that the deprecation warning in question was introduced as early as 59.6.0. Any idea why it wasn't until 60.0.3 that Dask's CI started to fail?

Context: my colleagues and I have been investigating an issue where conda fails to create an environment based off a "frozen" environment file (i.e., one generated by conda export) that had previously been fully functional. We identified the root cause to be related to the hotfix in conda-forge/conda-forge-repodata-patches-feedstock#204, and that discussion led me to this Dask PR. I have a broader question about the hotfix which I'll post in the conda forge repo.

@jsignell
Copy link
Copy Markdown
Member Author

That had not occurred to me as a situation @zzhengnan thanks for taking the time to comment here. I had no idea about conda hotfixes either until this issue came up. I'm not sure why 59.6.0 is fine and > 60 isn't. I didn't dig through the commit history, but I assume that something changed about how or when setuptools raises the warning.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants