-
-
Notifications
You must be signed in to change notification settings - Fork 10.8k
BUG: aarch64 linux wheel builds failing on highway_qsort
SVE error
#25445
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 occasiona 8000 lly send you account related emails.
Already on GitHub? Sign in to your account
Comments
This should include google/highway#1884 which may fix numpy#25445
@rgommers, a cursory glance over the git history for Highway reveals google/highway#1884, so first thing I think we should try is bumping it (#25446). Can you kick off the wheel build in that PR? |
This should include google/highway#1884 which may fix numpy#25445 [wheel build]
This should include google/highway#1884 which may fix numpy#25445
Sorry about the build error. We recommend using releases instead of arbitrary revisions/versions, those are more thoroughly tested. |
This should include google/highway#1884 which may fix numpygh-25445. [skip azp] [skip actions] [skip circle]
Thanks @jan-wassenberg. Comment from @Mousius on the linked PR:
That sounds reasonable to me indeed. For now, whatever works to get the aarch64 wheel builds fixed so nightlies start uploading again (they're outdated by a month, and we've broken ABI since then). |
Got it, thanks. Let's update. |
The Cron jobs have been running, it's just that the aarch64 entry has been failing for that amount of time. |
The earliest anyone has talked about doing a numpy 2.0 release candidate is midway through January, so no rush on getting that done before new years. Appreciate you bumping the priority on that for us. |
We had to pull in some git commits after the initial Highway merge, but it makes way more sense to use their well tested versions 😸 Hopefully fixes numpy#25445
We had to pull in some git commits after the initial Highway merge, but it makes way more sense to use their well tested versions 😸 Hopefully fixes numpy#25445
Thanks @jan-wassenberg. I grabbed a machine today and reviewed my assumption. It turns out I was wrong; I think I just started with git hashes and ended up with git hashes, which may have introduced compounding issues 🤔 #25446 should roll back to |
Oh, thanks for double-checking. Good that we can use 1.0.7 after all. |
Issue observed in gh-25440, and is present on
main
too. Example from this CI log:A copy of this comment from the PR:
A search for that error comes up empty, however we haven't run these wheel builds in 1 months and 8 days I think (that was the last nightly upload for aarch64 to https://anaconda.org/scientific-python-nightly-wheels/numpy/files).
That last successful build was before the initial merge of
highway_qsort
(gh-24018), so this probably never worked in the wheel builds. @Mousius @seiko2plus does that error look familiar perhaps?The text was updated successfully, but these errors were encountered: