-
Notifications
You must be signed in to change notification settings - Fork 15
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
Fix windows 3.13t wheels #50
Conversation
.github/workflows/dists.yml
Outdated
|
||
- name: Add python lib to PATH | ||
if: matrix.os == 'windows-2025' | ||
run: Add-Content $env:GITHUB_PATH "${{ env.pythonLocation }}\lib" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
something broke upstream in the past two weeks.
this is now necessary to get python313.lib discovered by maturin.
without this, we again hit PyO3/maturin-action#292:
fatal error LNK1181: cannot open input file 'python313.lib'
we encountered and fixed in #47 using generate-import-lib
now it manages to pass the 3.13 build again, but 3.13t is still red with LNK1181: ce73691 in this PR
both 3.13 and 3.13t were still working 2 weeks ago with all the same versions (maturin version, maturin-action version, pyo3 version, python 3.13.1 patch version): fc75159...4f172e2
tried both windows-2022 (no setup-python cache hit) and windows-2025 (python 3.13.1 cache hit), no difference.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
alright, we're green with 3.13t 🎉
turns out we don't need this Add python lib to PATH
step. we need an explicit 3.13t distribution installed instead.
Haven't looked carefully yet but Merry Christmas all :) |
ok, windows 3.13t is green again 🎉 and now without we're now simply setup-python'ing (near instant if py version is cached in the runner image) all python versions, including I wonder if 3.13t was really working 2 weeks ago... now we'be become robust to external changes that github is making to the runner images. |
@oconnor663 lets make a squash&merge to avoid all these debug commits on master? or even enable I deleted the broken 1.0.1 Github Release and the git tag by running |
@@ -3,7 +3,7 @@ name: dists | |||
on: | |||
pull_request: | |||
push: | |||
branches: '**' # not tags | |||
branches: [master] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I do like being able to run CI on my branches without creating a PR. Does the '**'
setting cause any problems?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
the setting causes duplicate jobs for every commit pushed to a PR. i think what you're looking for is on: workflow_dispatch
. we already added it in a previous PR: it adds a Run Workflow button on the right hand side here, where you can run CI manually on any branch without creating a PR
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Got it.
Thanks! Merged. |
with: | ||
target: ${{ matrix.target }} | ||
manylinux: ${{ matrix.manylinux || 'auto' }} | ||
# Keep in sync with tests.yml | ||
manylinux: ${{ matrix.manylinux }} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just so I follow, what's the reason for this change?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ah I was experimenting with furthet breaking up the matrix, one Python version at a time. To keep that simple, I wrote out the full matrix, including the auto
string. Ended up reverting, but decided to keep this one.
Btw, releasing 1.0.1 twice cause this issue downstream: https://aur.archlinux.org/packages/python-blake3#comment-1005277. If we need to do something similar again, maybe better to just increment the patch version number? |
ah great that you mention, sorry about that. makes total sense! |
it seems we have CI rot coming from external/upstream changes somewhere in the past two weeks: just tried releasing 1.0.1 fc75159...4f172e2 and we're no longer green