Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Problem with release on conda-forge #350

Closed
pbranson opened this issue Aug 18, 2022 · 8 comments
Closed

Problem with release on conda-forge #350

pbranson opened this issue Aug 18, 2022 · 8 comments

Comments

@pbranson
Copy link
Contributor

There appears to have been a problem with the latest release, see ocean-eddy-cpt/gcm-filters#156

The shuffle codec hasn't compiled correctly it seems.

@MSanKeys963
Copy link
Member

Cc: @jakirkham

@jakirkham
Copy link
Member

jakirkham commented Aug 19, 2022

@pbranson, can someone from your team please share a reproducer? Specifically can someone please fill out this issue template? Unfortunately there is not enough to go on here.

FWIW pulling a random Conda package (Linux x86_64 for Python 3.9) from conda-forge shows the _shuffle filter included, but that could be a different package than the one used.

We also run the full test suite on the feedstock before each release; so, would think we would have caught this, but maybe we are missing something.

@jakirkham
Copy link
Member

Ah missed the user is on PyPy. This is currently not supported in conda-forge ( conda-forge/numcodecs-feedstock#83 ). Are they getting the package from somewhere else?

@pbranson
Copy link
Contributor Author

I think the test environment for gcm-filters is using mamba and the conda-forge channel. @NoraLoose are you able to provide the information required in the comment above?

Apologies for raising the issue without sufficient information!

@jakirkham
Copy link
Member

Right. Given there are no PyPy packages for numcodecs on conda-forge, the big question is what packages are being referred to and how were they installed?

@pbranson
Copy link
Contributor Author

pbranson commented Aug 26, 2022

OK I have looked through the github actions and I think I have found where the test environment is setup and the version of numcodecs that was failing:

https://github.com/ocean-eddy-cpt/gcm-filters/runs/7838966081?check_suite_focus=true#step:5:179
This results in
https://github.com/ocean-eddy-cpt/gcm-filters/runs/7838966081?check_suite_focus=true#step:7:48

E   ModuleNotFoundError: No module named 'numcodecs._shuffle'

And the same is for Python 3.6 -> 3.9
For example this is the version installed by mamba for 3.9:

  + numcodecs                            0.9.1  py39h295c915_0       pkgs/main/linux-64          1 MB

But on closer inspection, it seems that even though the tests are meant for different Python versions, they are all installing/using python 3.9! @jbusecke I think there is something wrong with the test setup!

HTH, I am only new to this project.

@jbusecke
Copy link

jbusecke commented Sep 8, 2022

@pbranson thanks for the ping. I believe you are right. I have submitted a fix here.
It seems that now python 3.6->3.10 pass without a pinned numcodec version.

@jakirkham
Copy link
Member

Thanks for the update! 🙏 Going to close as resolved

Side note: It is worth pointing out that pkgs/main/linux-64 means the package is coming from defaults and not conda-forge. So would raise issues about packages from defaults with Anaconda

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

No branches or pull requests

4 participants