Skip to content
This repository was archived by the owner on Apr 16, 2021. It is now read-only.

Change upload process and upload to anaconda #14

Merged
merged 1 commit into from
Nov 24, 2020
Merged

Change upload process and upload to anaconda #14

merged 1 commit into from
Nov 24, 2020

Conversation

janaknat
Copy link
Contributor

@janaknat janaknat commented Nov 5, 2020

Change from Rackspace to Anaconda upload.

@janaknat
Copy link
Contributor Author

@tacaswell Any comments on this PR?

@janaknat
Copy link
Contributor Author

@mattip Any comments on this PR?

@tacaswell
Copy link
Contributor

I am a bit concerned about why the tests are failing (which look like real failures not just from the upload logic).

@janaknat
Copy link
Contributor Author

test_subfigure_ss() was added after the arm64 wheel PR was merged. Not sure why it was not caught.

test_custom_transform() has a change where the tolerance was modified. Patch merged on June 29. Arm64 PR merged on July 24.

7deb36dcdebdf310993b1455043cb6eea2aa1b1a - Increase tolerance for ppc64le and s390x

aarch64 had specific tolerance before. After this patch, aarch64, ppc64le and s390x have the same tolerance of 0.034.
Failure log shows RMS (0.039) in test_custom_transform(). Not sure if my understanding of this is correct.

@tacaswell
Copy link
Contributor

As this is the nightly wheels I'm going to go ahead and merge this so it is ready to go when we fix those regressions upstream (earlier today I incorrectly had in my head this was building a tag).

@tacaswell tacaswell merged commit 490759c into MacPython:master Nov 24, 2020
@mattip
Copy link

mattip commented Nov 24, 2020

While the change to anaconda.org is good, this will need to be redone soon to move off travis since they have severely limited their free offering. Note that the upstream builds stopped running on travis three days ago, I imagine these will also reach their credit limit soon.

@mattip
Copy link

mattip commented Nov 24, 2020

See also numpy/numpy#17824

@janaknat
Copy link
Contributor Author

janaknat commented Dec 2, 2020

@tacaswell Is there a timeline for when the aarch64 wheels show up in PyPI?

@janaknat
Copy link
Contributor Author

@tacaswell @mattip Will there be an aarch64 wheel released for matplotlib?

@janaknat
Copy link
Contributor Author

@tacaswell @mattip Ping. Any timeline for when aarch64 wheels will be available for matplotlib?

@mattip
Copy link

mattip commented Dec 17, 2020

My guess is with the next release of matplotlib. You may want to reach out in other forums like the mailing list https://mail.python.org/mailman/listinfo/matplotlib-devel

@tacaswell
Copy link
Contributor

Are there aarch64 wheels someplace to upload? I am not seeing them at https://anaconda.org/multibuild-wheels-staging/repo?type=pypi to upload.

@mattip
Copy link

mattip commented Dec 18, 2020

It turns out there was a problem with the .travis.yml file - there were missing semicolons at the end of the lines in this PR. I am working on a follow up to move the linux, macos jobs to github actions and fix the travis builds for arm64.

It seems there are no CI-built wheels for windows. Is there a recipe for building them somewhere?

@mattip
Copy link

mattip commented Dec 18, 2020

@tacaswell Since you are using cibuildwheel in the main repo, what is the added value of this repo? Can't you upload to anaconda.org directly from the main repo?

@QuLogic
Copy link
Contributor

QuLogic commented Dec 19, 2020

That was to spread out the builds across CI; with Travis being all but dead, those could be switched on for non-tagged commits too.

@mattip
Copy link

mattip commented Dec 19, 2020

So it would seem to me to make sense to continue with cibuildwheel as a preferred solution and consolitate activity there. I commented on matplotlib/matplotlib#19147.

@janaknat
Copy link
Contributor Author

@mattip Will there be aarch64 wheels in the next release?

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants