| opendevreview | Ke Niu proposed openstack/nova master: remove unicode prefix from code https://review.opendev.org/c/openstack/nova/+/853112 | 04:11 |
|---|---|---|
| opendevreview | Ke Niu proposed openstack/nova master: remove unicode prefix from code https://review.opendev.org/c/openstack/nova/+/853112 | 04:12 |
| opendevreview | Ke Niu proposed openstack/nova master: Cleanup py27 support https://review.opendev.org/c/openstack/nova/+/853113 | 04:14 |
| *** arne_wiebalck_ is now known as arne_wiebalck | 06:47 | |
| zigo | I'm getting failed unit tests in Nova (which is RC bug in Debian since that's failed to build...): | 13:57 |
| zigo | https://bugs.debian.org/1017217 | 13:57 |
| zigo | Can anyone help? | 13:57 |
| zigo | ricolin: I'm not sure, when reading the specs, but what is vIOMMU for? (note: I know what a PMMU is...) | 14:00 |
| *** Guest167 is now known as dasm | 14:15 | |
| artom | zigo, only thing I can think of is the recently merged double-mocking patch | 14:35 |
| zigo | artom: That's Stable / Yoga that I'm talking about, does this match? | 14:36 |
| zigo | Was it backported? | 14:36 |
| zigo | The thing is, it built fine 10 days ago in Debian Unstable. I'm currently trying to use snapshot.debian.org to see what package changed since ... | 14:37 |
| artom | zigo, hrmm, no, it's https://review.opendev.org/c/openstack/nova/+/852464 and https://review.opendev.org/c/openstack/nova/+/851445 that I'm thinking of | 14:38 |
| zigo | None of these seem to have been backported to Yoga, so it's unlikely the issue here. | 14:39 |
| zigo | In a minute, I'll see what package has changed... | 14:39 |
| artom | zigo, super bizarre... something is causing the unit test to apprently redirect | 14:44 |
| zigo | Yeah, looks scary too ... | 14:45 |
| artom | Which makes no sense, because it's a unit test, literally nothing is running except Python code | 14:46 |
| artom | zigo, so the redirect appears to be coming from websockify itself... | 14:56 |
| artom | I guess I'd look into its changelog for the version you're packaging? | 14:56 |
| zigo | artom: I haven't uploaded websockify since 2021 ... | 14:56 |
| zigo | I'm currently trying to use snapshot.debian.org to see what package has changed. | 14:57 |
| * artom is o_O | 14:57 | |
| zigo | Result: python3 python3-all python3-attr python3-distutils python3-keyring python3-lib2to3 python3-minimal python3-psycopg2 python3-typing-extensions python3.10 python3.10-minimal | 14:58 |
| zigo | I'll try them one by one... | 14:58 |
| zigo | Gosh ... upgrding the Python 3.10 interpreter triggered the bug !!! :( | 15:07 |
| zigo | python3.10 (3.10.6-1) over (3.10.5-1) | 15:08 |
| artom | o_O | 15:12 |
| zigo | Maybe related to this? https://github.com/python/cpython/issues/87389 | 15:15 |
| zigo | The patch has: | 15:18 |
| zigo | if self.path.startswith('//'): | 15:18 |
| zigo | self.path = '/' + self.path.lstrip('/') # Reduce to a single / | 15:18 |
| * artom tries in "vanilla" functional tests | 15:18 | |
| zigo | So, Python itself would strip the added /, rendering the unit test and protection useless ... | 15:18 |
| * artom is dumb for not trying before | 15:18 | |
| artom | Or rather, trying with py39 for some reason | 15:18 |
| artom | zigo, yep, fails as well | 15:19 |
| artom | OK, you've just found a gate failure :) | 15:19 |
| zigo | :D | 15:19 |
| zigo | I'm trying to add the patch to the 3.10.5 standard library, to see if I reproduce it ... | 15:20 |
| zigo | I'll then file a bug against nova. | 15:20 |
| artom | Well... I say gate failure, and yet: https://zuul.opendev.org/t/openstack/build/e2ae266c798c411784f449f6f5e3d35a | 15:21 |
| artom | I guess we're not running a new enough py10? | 15:21 |
| artom | Yeah, 10.3.4 | 15:21 |
| artom | 3.10.4, sorry | 15:22 |
| artom | It's still coming though | 15:22 |
| artom | And I hit it locally because Fedora does have the affected Python version | 15:22 |
| zigo | Ok, now I'm sure about what I'm saying ! :) | 15:24 |
| zigo | Running Python 3.10.5 interpreter, if I ran the unit test without any additional change, it passes. Now, if I just add the 2 lines of the hunk here: https://github.com/python/cpython/commit/defaa2b19a9a01c79c1d5641a8aa179bb10ead3f (which is part of Python 3.10.6), then I get the failure ... | 15:25 |
| * zigo reports the bug against nova now ... | 15:25 | |
| zigo | What's nice, is that I now know there's no security issue... it's just the python interpreter standard lib having additional protections. | 15:25 |
| zigo | https://bugs.launchpad.net/nova/+bug/1986545 | 15:32 |
| *** dasm is now known as dasm|off | 22:28 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!