Closed webknjaz closed 2 weeks ago
I am experiencing the same issue as well. Is there a way of changing the version to have it work? I do not see a dependency in my packages so I am not sure where it is coming from.
If I understood correctly, as per https://github.com/pytest-dev/pytest-xdist/issues/1071#issuecomment-2067565494 pytest==3.6.0
has been yanked, and probably new versions of pytest-xdist
, execnet
and pytest
will be coordinated to fix the problem, right?
So my understanding is that for the time being we don't have to pin any dependency as a workaround. (Also happy to add a workaround if I understood things wrong).
@abravalheri I haven't tested it but I'd expect that it's execnet
that needs to be yanked, not pytest-xdist
. So my guess is that it's still possible to get a broken version. You can restart any GHA workflow in the repo and see if it breaks..
@abravalheri I haven't tested it but I'd expect that it's
execnet
that needs to be yanked, notpytest-xdist
. So my guess is that it's still possible to get a broken version. You can restart any GHA workflow in the repo and see if it breaks..
Possibly working? (https://github.com/pypa/setuptools/actions/runs/8759517556/job/24055273503, although cygwin
seems to have trouble on finding the dependencies).
Possibly working? (https://github.com/pypa/setuptools/actions/runs/8759517556/job/24055273503
Yep
although
cygwin
seems to have trouble on finding the dependencies).
No, that's an incorrect assessment. This is a classical case of "who knows what broke in the CI step" — if you look closer, you'll see that pip install
was successful and coverage xml
is what actually failed. This is one of the reasons I always insist on having a single command per CI step (it actually enables more confusing behaviors when there's several invocations inside, like a "false success").
Any progress on fixing this issue?
@MorganWeiss I don't see any immediate action to be taken by setuptools. Maybe it is relevant for you to follow the issues in pytes-xdist
and co and possibly contributing towards the fix there?
I am not aware of those, do you have any helpful links?
Get Outlook for iOShttps://aka.ms/o0ukef
From: Anderson Bravalheri @.>
Sent: Saturday, April 20, 2024 11:33:31 PM
To: pypa/setuptools @.>
Cc: Morgan Weiss @.>; Mention @.>
Subject: Re: [pypa/setuptools] [good first issue][CI breakage] pytest-xdist == 3.6.0
makes the CI fail (Issue #4317)
@MorganWeisshttps://github.com/MorganWeiss I don't see any immediate action to be taken by setuptools. Maybe it is relevant for you to follow the issues in pytest, pytes-xdist and co?
— Reply to this email directly, view it on GitHubhttps://github.com/pypa/setuptools/issues/4317#issuecomment-2067924865, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AKH4GZL5ILPLPFJ4UG5JPSDY6NMTXAVCNFSM6AAAAABGP3UTMGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANRXHEZDIOBWGU. You are receiving this because you were mentioned.Message ID: @.***>
You can start with the link shared in the first message of this thread, and then you can follow the discussions (they point to more links there).
v3.6.1 has just been released. Hopefully, this is enough to address this problem.
Since the release of new
pytest-xdist
roughly an hour ago, all CI jobs fail as follows:(https://github.com/pypa/setuptools/actions/runs/8759251151/job/24041851385?pr=4092#step:9:114)
I also filed the upstream report here: https://github.com/pytest-dev/pytest-xdist/issues/1071.
Immediate action item
~Temporarily restrict the version of this dependency by adding
, != 3.6.0
to the dependency specifier intox.ini
and maybe elsewhere, this needs investigation.~ It looks likeexecnet
is what needs to be capped below v2 per https://github.com/pytest-dev/pytest-xdist/issues/620#issuecomment-2067333447.