Refactor Github CI #287
Refactor Github CI #287
Conversation
|
Hmm, only macos is failing 🤔 |
|
Strange. It fails for small distortion cases. How about following https://github.com/pypa/cibuildwheel#example-setup |
Seeing the errors raised, I don't believe it is an os version issue, but let's give it a try anyway. |
|
@lan496 do you have a |
|
In my environment (macOS-13.3.1(a), arm64), all Python tests pass. pip install ".[test]"
pytest python/test/ -v |
|
How about going with It seems |
Please test cibuildwheel explicitly. It might be specific to a Python version
With
Checked the changelog, didn't seem to have anything relevant. |
|
Looking at the tests though, I wonder if they are implemented correctly |
|
Hmm, ctests fail on macos-12 and macos-13, but work on macos-11 (using apple-clang) at the test_broken_symmetry |
|
OK, only |
|
Is it possible for non-sudo users to run this cibuildwheel in the first place? The current script tries to install libsymspg under /usr/local and it requires the sudo permission. |
That's only the ctests that fail, there are python tests failing as well. I suspect the python tests failing are related
Try reverting 5c5f40d. I wonder why it doesn't run in a container. You could spin up a manual container though? |
I agree. But what I wanted to mean was many tests failed at 73a9f8d https://github.com/spglib/spglib/actions/runs/5222954738/jobs/9429134918. |
Yes, but these tests are still failing , it's just that the CI is failing earlier now at the ctest stage now |
9459bcd to
e74f52d
Compare
e23e007 to
3705faa
Compare
|
For now this should at least build the wheels for all distros. The remaining issue is with Fedora 39/rawhide, which is an usptream issue that will be fixed with scikit-build/scikit-build-core#463 |
Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
Temporarily revert the wheel build to macos-11. Further testing framework needs to be fixed to handle macos-12 Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
Numpy cannot be built at the moment. Needs development numpy version Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
Signed-off-by: Cristian Le <cristian.le@mpsd.mpg.de>
Codecov ReportPatch coverage has no change and project coverage change:
Additional details and impacted files@@ Coverage Diff @@
## develop #287 +/- ##
===========================================
- Coverage 83.64% 83.60% -0.05%
===========================================
Files 23 23
Lines 6274 7942 +1668
===========================================
+ Hits 5248 6640 +1392
- Misses 1026 1302 +276
Flags with carried forward coverage won't be shown. Click here to find out more.
☔ View full report in Codecov by Sentry. |
For now this PR:
cp311-*variants) in default CI. Should not be necessary since the default CI runs them as well, but another sanity check doesn't hurt.One thing I've noticed. The wheel builds are not really used in uploading to
PyPI.PyPIis building the package from source on their side as well. We should get the badge and put it on theREADME.mdto keep track of itLet me know if the tests done here are a bit excessive and I can move some of them to
workflow_dispatch. E.g. we could make it run only python3.11and3.12on the normal CI.