Closed adamh-oai closed 6 months ago
Hi @adamh-oai, the tests pass, the current issue is unrelated, a dummy site is down.
All modified and coverable lines are covered by tests :white_check_mark:
Comparison is base (
4088710
) 99.07% compared to head (b8cc607
) 99.07%.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
Now it works for Python 3.12 but not for Python 3.8 on MacOS.
I ran the tests locally and everything works well with the same version of python on macos (m2 architecture) Can this test not works because of the content of the website during the run ? Did you try to rerun them ?
Edit: I just saw the tests are run using a x86 architecture, maybe the difference come from there
Yes, I updated the tests. I'm not sure why tests fail specifically on Python 3.8 but you're probably right in assuming different architectures lead to different outcomes.
We could try adding other versions of Python on MacOS, so far other architectures are not easily available on Github Actions as far as I know.
@adamh-oai I believe I found the right combination of settings and tests, what do you think?
lgtm!
Could you put up a new release on pypi with this change? Thanks!
Yes, I plan to do it this week for htmldate and a bit later for trafilatura.
Per https://github.com/adbar/trafilatura/issues/449, see if unpinning lxml works on MacOS.