-
-
Notifications
You must be signed in to change notification settings - Fork 7.9k
[TST]: Appveyor Qt tests failing #24394
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
Comments
Neither Qt or PyQt had releases on conda-forge in the last few days which was my first guess at what is going wrong. Sorry I miss-read these logs on #24380 . |
https://www.appveyor.com/updates/2022/11/07/ Probably related to the appveyor platform updates that rolled out today |
Possibly also related: Lines 57 to 59 in 9f96205
Could be that the issue wasn't "pip vs conda", but rather some other factor... not sure what, though |
https://ci.appveyor.com/project/ksunden/appveyor-test/build/job/akqg7r4vfncf862d Replicated with a stripped down version of the tests (didn't install mpl, but installed all requirements still) |
Okay, I know what is happening now: Pywin32 released yesterday and so is on version 305 on pypi, but version 304 on conda forge We attempt to install via conda because of the reasons indicated by the comment above. However, since the pypi version is higher, the conda install gets overridden with the PyPI version since pywin32 is specified in The pip version is broken for the conda install. Our options are:
|
Reordering works with the caveat that it still shows up as the pypi version in |
The mixing of conda and pip on CI is not great. Is it plausible to go pure conda on appveyor? Another option is "does appveyor actually add coverage we do not get from azure + gha?" If not we should just drop the whole CI platform.... |
We do get windows from azure (GHA is not currently configured for windows, could be added, but haven't looked at usage caps/simultaneous run caps or anything). We do not have any other tests which use conda, if that is important to us. (in particular, conda can be a little slow when it comes to updating qt dependencies... I know over at pyqtgraph we made explicitly sure to work with the version of qt available via conda, and CI as much) As for going full conda, no reason why it couldn't be done, just would be an additional place to update deps as it can't use the requirements txt files since package names are sometimes different. |
We have an environment.yml. I think that could be used for appveyor. |
Closes matplotlib#24394 Adjust environment.yml skip fontconfig private method test on windows skip fontconfig private method test on windows revert adding fontconfig explicitly to environment Add wxpython back Skip some wx tests Upgrade to gtk4 in environment.yml ignore private modules in getattr test _backend_gtk would fail because it expects one of gtk4 or gtk3 to be imported first Clean up appveyor.yml comments skip memleak test for wxagg conda activate does not, in fact, work revert to simply activate
Closes matplotlib#24394 Adjust environment.yml skip fontconfig private method test on windows skip fontconfig private method test on windows revert adding fontconfig explicitly to environment Add wxpython back Skip some wx tests Upgrade to gtk4 in environment.yml ignore private modules in getattr test _backend_gtk would fail because it expects one of gtk4 or gtk3 to be imported first Clean up appveyor.yml comments skip memleak test for wxagg conda activate does not, in fact, work revert to simply activate
Summary
Appveyor Qt sigint tests appear to be consistently failing due DLL loading
Relevant log portion
Proposed fix
No response
The text was updated successfully, but these errors were encountered: