Github Actions: test newer python versions, take 2 #507
Conversation
nosetests no longer works with python 3.10 onwards, but we do not actually need it
Try to fix node.js 12 deprecation warning when building jobs by using newer versions of actions/checkout, actions/setup-python, and slackapi/slack-github-action Assuming usage stays the same...
| @@ -8,7 +8,7 @@ jobs: | |||
| strategy: | |||
| matrix: | |||
| os: [ubuntu-latest] | |||
| python-version: [ '3.7', '3.8', '3.9'] | |||
| python-version: [ '3.7', '3.8', '3.9', '3.10', '3.11', '3.12-dev' ] | |||
There was a problem hiding this comment.
do we really need to test so early version like 3.12-dev? Those tests need to pass for all the following commits, right?
There was a problem hiding this comment.
I think we can always ignore build failures?
It's better to know it'll fail so we know we have something to fix for the next python version, than to not know and have someone tell us a few weeks after the release.
There was a problem hiding this comment.
The purpose of the tests is to validate the code changes are working for all the supported version. Having the green tick is very helpful to know this is working fine. Having a always red tick and us needing to go check it to see if this is the always failing test due to some changes in 3.12 is burdensome.
dev is dev... As long as the code is not stable and some changes could be make some failures only transient... i feel like this is adding more effort than helping us. Except if we agree to work on the failure quickly, and none of us are likely to have time for that.
There was a problem hiding this comment.
Well, disabling the check is trivial -- can we just leave it on for now and disable it if that ever is a problem (e.g. a check keeps failing for more than a handful of PRs) ?
I was personally annoyed with the 3.10 breakage we had a couple of years ago, so if we had such a check in place at the time I'd have been happy to fix it before things broke on my laptop. fedora normally ships the next version right with its release (synchronized release of python and new release of fedora) so I'll always be one step ahead and really think I'd benefit from this.
Tests themselves look fine, just the running with nosetests that is apparently a problem, so adding an ugly case seems to work.
Coverage isn't easy to make work directly so I've skipped it for now, we didn't actually use the coverage results as far as I can see?
Also updated the actions we use to fix deprecation warnings on every test run while I was here.