-
-
Notifications
You must be signed in to change notification settings - Fork 273
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
Testsuite fails on python 3.7 #593
Comments
I just released astroid 2.0 and was working on Pylint 2.0 as well. Not exactly sure why you're getting these errors though. |
From a quick look it seems most of these are related to |
Hmm, I missed this suddenly being necessary. It fixes the python 3.6 failures, but not the python 3.7 error. |
It should be installed by default though, wasn't that the case? (at least for Python 3.6, not for 3.7 where a compatible version has not been released yet). Taking a look at the other test failure. |
I see 6 tests are marked skipif(not HAS_TYPED_AST, reason="requires typed_ast") but this still errored... |
Not when you're manually assembling test dependencies without test_requires in order to run |
Yeah, that makes sense though. Curious why you're not using Regarding the other test failure, I cannot reproduce it locally nor on CI. What Python 3.7 build do you have? If you do the following in a shell, does it still reproduces?
|
Using tox means we only test your code, which you already do. Not using tox, means we check the package we're creating, specifically that it works okay with the distro packages we've added to the dependency tree. So for distro packaging purposes, tox is actively unhelpful and downright bad/counterproductive. :p |
Yes, systemd-nspawn into an Arch Linux container running the staging repositories (which is where our python 3.7 package is currently sitting while we try to rebuild everything), cd to an astroid git checkout, and run:
|
What I don't understand is what exactly is raising the Thanks for the |
I can import xml.etree.ElementTree just fine, yes. Also running tox just in case, fails too (it's only testing the actual python interpreter while grabbing fresh packages from PyPI etc. but indeed, this is failing anyway so something about the interpreter itself is obviously failing to work as expected by astroid.modutils). The python interpreter is built using this buildscript: https://git.archlinux.org/svntogit/packages.git/tree/repos/staging-x86_64/PKGBUILD?h=packages/python We shouldn't be doing anything controversial I don't think... |
I cloned the CPython repo from GitHub and build the tag |
Oh, a just fixed CPython bug (after 3.7.0 release) has the same traceback as this issue - https://bugs.python.org/issue34056. Let me try that patch. |
So basically the bug is "why does this still use imp"? :p |
Thanks for finding that bug @yan12125 ! Ideally we'd move away from |
In which case I imagine you'd like to backport that CPython patch back to Arch, until we completely remove |
Yes, that would (both) indeed be a good idea. Closing in favor of the new issue. |
Trying to rebuild astroid master (due to the lack of a release and multiple failures with the stable release -- I was trying to see if the impending 2.0.0 release would fix it) for python 3.7 on Arch Linux, I hit the following testsuite error:
I actually get more errors trying to rebuild master on python 3.6:
The text was updated successfully, but these errors were encountered: