Unpin cibuildwheel container images so the release can build 3.14#70
Open
yokofly wants to merge 1 commit into
Open
Unpin cibuildwheel container images so the release can build 3.14#70yokofly wants to merge 1 commit into
yokofly wants to merge 1 commit into
Conversation
Coverage Report for CI Build 27424402958Warning No base build found for commit Coverage: 94.387%Details
Uncovered ChangesNo uncovered changes found. Coverage RegressionsRequires a base build to compare against. How to fix this → Coverage Stats
💛 - Coveralls |
cibuildwheel 3.x rejects the musllinux_1_1 alias outright (it killed release run 27423060149 at startup), and the manylinux2014 images froze at CentOS 7 EOL without Python 3.14, so neither pin can produce cp314/cp314t wheels. The 3.4.1 defaults (manylinux_2_28 / musllinux_1_2) carry every targeted interpreter. Verified locally: the config now yields the full identifier set (cp38-cp314t + pp311, x86_64 + aarch64) and a real cp314t wheel builds through cibuildwheel with the smoke test passing; auditwheel grades it to manylinux2014/manylinux_2_17, so old-glibc compatibility survives the newer build image. This failure class was invisible to PR CI because only the release workflow runs cibuildwheel — so the test workflow now builds one real wheel (cp314t) through cibuildwheel on every PR, which resolves all image options upfront and fails in seconds on config rot. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
yokofly
force-pushed
the
yokofly/fix-cibuildwheel-images
branch
from
dee02e2 to
eda774d
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Release run 27423060149 for v0.3.0 failed at cibuildwheel startup:
Two stale image pins in
pyproject.tomlblock the release:musllinux_1_1— removed by cibuildwheel 3.x, hard error.manylinux2014— the images froze at CentOS 7 EOL (June 2024) and do not contain Python 3.14, so they could not build the cp314/cp314t wheels this release exists for.Fix: drop the pins and use the cibuildwheel 3.4.1 defaults (
manylinux_2_28/musllinux_1_2).Why no PR check caught this: the release workflow is the only place cibuildwheel runs, so its config can rot invisibly — PR #69 was fully green while the release config was broken. This PR closes that gap: the test workflow now builds one real wheel (cp314t) through cibuildwheel on every PR. cibuildwheel resolves all image options upfront, so any config rot fails the job in seconds instead of surfacing at release time.
Verified locally before opening this PR:
cibuildwheel --print-build-identifiersyields the full set: cp38–cp314t + pp311, x86_64 + aarch64, manylinux + musllinux. (Note: this command alone does not validate images — verified empirically — which is why the guard builds a real wheel.)manylinux2014/manylinux_2_17— the extensions only use old glibc symbols, so compatibility with old distros is preserved despite the newer build image.musllinux_1_1pin makes the new guard job fail in ~2 seconds, proving it catches this failure class.🤖 Generated with Claude Code