| stephenfin | M0weng[m]: You can only use other classmethods and staticmethod (plus external functions, obviously) from classmethods, so if 'is_extension_enabled' is a classmethod then you can use it from another classmethod | 10:21 |
|---|---|---|
| stephenfin | M0weng[m]: I should rephrase that: classmethods are only allowed to call other classmethods and staticmethods. They can't use instance methods. | 10:22 |
| opendevreview | Merged openstack/os-service-types master: Add Python 3.8 and 3.9 to supported runtimes https://review.opendev.org/c/openstack/os-service-types/+/843266 | 16:13 |
| opendevreview | Artem Goncharov proposed openstack/openstacksdk master: WIP: Implement auto-code generation https://review.opendev.org/c/openstack/openstacksdk/+/882326 | 18:26 |
| M0weng[m] | stephenfin: Thanks! I didn't realize there were two different `is_extension_enabled()`s; I was looking at the non-classmethod one by mistake, no wonder things didn't seem to make sense haha | 18:27 |
| frickler | gtema: regarding the sdk cache issue, looking at the implementation it seems that there is no file locking or atomic updates happening, so the errors that yatin found seems very plausible to me https://github.com/jaraco/keyrings.alt/blob/main/keyrings/alt/file_base.py | 19:38 |
| frickler | not sure if other backends would work better or maybe the whole library isn't suited for this usecase | 19:38 |
| gtema | frickler: from the failed jobs I haven't seen a single one where cache seemed like an issue, there were always unrelated issues. So I would be interested to see whether after disabling it failure rate decrease | 19:40 |
| frickler | gtema: did you look at the cache file contents that were posted in the bug report? those looked pretty convincing to me | 19:42 |
| gtema | Even corrupting Cache content SDK will renew it. On the other side in most of failed jobs keystone log was full of db issues | 19:43 |
| gtema | And that is not related to osc token. It may be exactly causing that | 19:44 |
| gtema | I haven't looked into updates to the bug, LP doesn't send me notifications (need to look where to enable that) | 19:45 |
| frickler | the library wants to read the cache file for updates, so if the read fails, it cannot update or delete the cache entry. at least that's my reading of the code above | 19:45 |
| gtema | Pls try to use osc with broken cache state. Historically for me it was working fine | 19:46 |
| gtema | Haven't checked recently though | 19:46 |
| frickler | ok, I can try that tomorrow | 19:46 |
| M0weng[m] | I'm trying to submit a patch using `git review` but it says I have the wrong email set. I changed the email to the one registered with gerrit using `git config user.email <email>` and re-ran `git review -s`, but it's still trying to use the incorrect email. | 19:51 |
| M0weng[m] | * <email>` and `git config --global gitreview.email <email>` and re-ran `git | 19:52 |
| gtema | 0weng, use sure it tries to use wrong email or it is just a mismatch to the account you setup in Gerrit? | 19:53 |
| gtema | s/Use/you | 19:53 |
| frickler | M0weng[m]: please show the full error on paste.opendev.org | 19:53 |
| frickler | gtema: if I mangle the cache file to contain two entries with identical key like in https://bugs.launchpad.net/devstack/+bug/2042943/comments/11 , I'm getting the same error locally | 19:59 |
| frickler | I also get a different error, but similarly corrupted cache file, pretty fast when running just "openstack token issue" for two clouds in multiple parallel invocations? | 20:00 |
| gtema | Hmm, ok, will think about it. This "file" backend is the only thing which work in real detached mode. Maybe we should drop that completely, I never was really convinced in that lib | 20:05 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!