Wednesday, 2025-07-23

opendevreviewRiccardo Pittau proposed openstack/requirements master: Updated from generate-constraints  https://review.opendev.org/c/openstack/requirements/+/95542107:35
fricklerrpittau: ^^ that's not enough, see https://review.opendev.org/c/openstack/requirements/+/907665/108 . you can squash the stack to get a working mergeable patch07:45
rpittaufrickler: yeah, I realized after uploading that, I'm testing it locally07:45
*** ykarel_ is now known as ykarel07:52
opendevreviewRiccardo Pittau proposed openstack/requirements master: Updated from generate-constraints  https://review.opendev.org/c/openstack/requirements/+/95542108:01
rpittauooook this ^ should work08:01
*** bauzas8 is now known as bauzas08:32
stephenfinralonsoh: frickler: Sorry, I forgot to open hexchat yesterday. The reason it can't find the package is because the last version of pkg_resources supported on Python 3.8 still has a bug where it is not normalizing packages names, and since January we are correctly naming wheels as e.g. oslo_foo instead of oslo.foo.09:38
ralonsohstephenfin, should be use an older pkg_resources version there? only of py3809:39
stephenfinralonsoh: I'm under the impression it was always broken. We just never noticed because we weren't correctly normalizing wheels before that change in January (fungi has the specifics)09:40
stephenfinso I don't think there is an older version we can switch to09:41
stephenfinhttps://github.com/pypa/setuptools/issues/505109:41
stephenfinThe thing I still haven't figured out is why pkg_resources is getting invoked in the first place. That should never happen. On newer versions of pip, pip itself generates console scripts and doesn't use pkg_resources. On older versions of pip, pbr overrides the distutils InstallScripts command to generate a version of the script that doesn't use pkg_resources, overriding the setuptools default09:43
stephenfinI'm guessing either my patch isn't actually getting used, or we're doing something like an editable install and the PEP-517 backend is getting invoked (so all the distutils logic from pbr is bypassed) but pip is somehow not generating console scripts for us as you'd expect09:45
stephenfinYeah, it doesn't look like my patch is being used? Looking at log lines with timestamps 2025-07-22 08:42:53.321531 and 2025-07-22 08:48:52.426214 from the build log https://storage.gra.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_d8e/openstack/d8e3003ef09a4ae9bd504a96a6ee1dc2/job-output.txt09:48
fricklerstephenfin: ah, your patch is for 2025.1, but the test is running against 2024.2 where the issue appears. so I think we need to backport your change to 2024.2 and test with that? it doesn't look like 2025.1 is affected, not sure why10:38
stephenfinAha, good spot. I couldn't figure that out. Had assumed required_projects was unset but it was10:39
stephenfinI can propose the backports10:40
stephenfinfrickler: ralonsoh: Done https://review.opendev.org/c/openstack/neutron/+/951494 Let's wait and see now10:45
ralonsohstephenfin, thanks10:46
fricklerstephenfin: why did you drop the dep on https://review.opendev.org/c/openstack/requirements/+/951735? without that we won't see the issue, or am I missing something?10:58
stephenfinfrickler: That was a mistake, sorry. I thought the issue was reproducing in its current state. Re-added11:17
fricklerthx. I'll set up an autohold for that job so that we can take a closer look if needed11:24
stephenfinfrickler: Looks like that passed https://review.opendev.org/c/openstack/neutron/+/95149412:38
stephenfinralonsoh too12:38
ralonsohstephenfin, so we need the revert^2 and the designate patch then12:39
stephenfinyes, but designate first12:40
opendevreviewMerged openstack/requirements master: Updated from generate-constraints  https://review.opendev.org/c/openstack/requirements/+/95542116:23

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!