intake icon indicating copy to clipboard operation
intake copied to clipboard

Tests failing for Python 3.8 and Python 3.9

Open shayandoust opened this issue 5 years ago • 7 comments

Hello,

It seems like pytest is failing for the Python versions mentioned in the title. Hence, this is blocking an upload, which rectifies a RC bug.

Here is the output of the tests. Please let me know if there is anything I can try to rectify these tests failing, or if you require any further information.

I do see some errors within the log like

   def info(self, path, **kwargs):
        path = self._strip_protocol(path)
>       out = os.stat(path, follow_symlinks=False)
E       FileNotFoundError: [Errno 2] No such file or directory: '/<<PKGBUILDDIR>>/.pybuild/cpython3_3.8_intake/build/intake/source/tests//calvert_uk_filter.tar.gz'

/usr/lib/python3/dist-packages/fsspec/implementations/local.py:60: FileNotFoundError

however, nothing has been touched and I'm not too sure why these files are missing.

Thanks, Shayan

shayandoust avatar Nov 15 '20 21:11 shayandoust

Is it possible to find out the set of package versions that are used in the test, and the instal method? Intake's CI does test against python 38, so that won't be the issue.

martindurant avatar Nov 16 '20 17:11 martindurant

Did you have time to delve deeper into this?

martindurant avatar Jan 19 '21 17:01 martindurant

Hello martindurant,

Apologies for the lengthy absence. It seems like that issue was to do with what files were to be included into the autopkgtest, so that issue above now stands resolved. Now there persists a final issue, which I think started happening only for Python 3.9 for the latest release tarball of intake.

Here is the bug report opened: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=982699 If you would like a full build log from my system, please let me know. Hopefully the issue is trivial.

Hopefully this is the final issue we can get resolved as we are entering the next stage of Debian freezes.

shayandoust avatar Feb 15 '21 23:02 shayandoust

I believe that is a result of a new version of pandas, and different conventions for data type matching therein. It's probably already passing on master.

Would our releasing a new version solve things for you?

martindurant avatar Feb 16 '21 14:02 martindurant

Yes please. Once you've released a new version I will test it and let you know of any issues. Though if the issue is trivial and only affecting tests then it might be sensible to suggest a patch to ensure quick migration within testing and unstable.

shayandoust avatar Feb 16 '21 16:02 shayandoust

The new version is on pypi and conda-forge

martindurant avatar Feb 17 '21 18:02 martindurant

Thanks! I can confirm the fix for Debian

mr-c avatar Mar 15 '21 10:03 mr-c