| cardoe | tkajinam: actually I'm looking... save_and_reraise_exception's entire functionality has been supported since Python 3.3. It causes type hints throughout the code base to be entirely wrong. It's making everyone that's using Claude/Codex submit these patches full of getattr() slop. I'm now thinking we mark it as @deprecated and tell people to just use native exception handling | 02:06 |
|---|---|---|
| cardoe | It causes a lot of warnings about possible unset variables as well. | 02:07 |
| frickler | so this "cannot import name 'md5' from 'oslo_utils.secretutils'" issue is affecting a couple of projects and it makes me wonder whether we should actually revert this for this cycle. rolling out such a change 5 minutes before we go into library freeze seems like very bad timing, IMO it needs to be done early in the cycle | 07:17 |
| frickler | in particular since we've struggled with enough other external issues over the last weeks already. see e.g. https://zuul.opendev.org/t/openstack/build/0aa9da1022ab4345b967461d93286cb3 | 07:19 |
| frickler | tkajinam: ^^ iiuc that was your change? | 07:19 |
| zigo | Hi. According to my Debian package build, oslo.service needs latest cotyledon, but it's not explicit in the current requirements. I'd suggest fixing this (which I expressed in debian dependency restrictions). | 08:58 |
| zigo | frickler: Agreed! | 08:59 |
| frickler | fyi I was told that the pending tempest release https://review.opendev.org/c/openstack/releases/+/977624 should fix all known issues with the md5 removal. but bad timing anyhow | 09:01 |
| opendevreview | Thomas Goirand proposed openstack/oslo.concurrency master: fix multiprocessing spawn failure in FileBasedLockingTestCase https://review.opendev.org/c/openstack/oslo.concurrency/+/971765 | 11:51 |
| tkajinam | zigo, it's in optional dependencies | 13:02 |
| tkajinam | zigo, or are you saying that the min version defined in the optional dependencies should be bumped ? | 13:03 |
| tkajinam | frickler, yeah that (tempest new release) is what we actually need | 13:03 |
| tkajinam | AttributeError: 'PythonSpec' object has no attribute 'machine' | 13:39 |
| tkajinam | hmm I'm wondering if this is known ... | 13:39 |
| zigo | tkajinam: Yeah, the min version is wrong. | 15:09 |
| stephenfin | cardoe: I took a look at that (I added the hints originally). Have +2'd and will +W if tkajinam can't get to it today. Have you an example of the type issues you're seeing though, just for my own reference? | 15:34 |
| stephenfin | or even one of these slop PRs you're referring to | 15:38 |
| cardoe | stephenfin: https://review.opendev.org/c/openstack/ironic/+/977569 | 15:58 |
| cardoe | the whole _is_post_bios_reboot_complete | 15:59 |
| cardoe | TheJulia's also done a bunch of networking stuff with Claude recently and it really wants to just catch Exception. | 15:59 |
| stephenfin | cardoe: I don't see use of save_and_reraise_exception in any patchset? | 16:01 |
| TheJulia | One can train it to be better at that through being very specific with feedback to influence weighting, but its first crutch is "Exception" | 16:01 |
| cardoe | stephenfin: https://opendev.org/openstack/ironic/src/commit/526174567b2e26b7c8ea9d083ac1bb478a44cd26/ironic/drivers/modules/redfish/utils.py#L389 https://opendev.org/openstack/ironic/src/commit/526174567b2e26b7c8ea9d083ac1bb478a44cd26/ironic/drivers/modules/redfish/utils.py#L511 | 16:02 |
| stephenfin | ack | 16:04 |
| cardoe | So if you use pyright or anything on the code base for example. It's quite convinced that the get_system() on line 389 can return None. The reason for that is line 511 which is the save_and_reraise_exception() | 16:04 |
| cardoe | So I burned my whole context window last night on Opus 4.6 analyzing why Claude Code keeps coming to these conclusions and it pointed the finger at that. | 16:04 |
| cardoe | The other thing is since it can raise RuntimeException, Claude really believes that the best approach is to use Exception and nothing more specific | 16:05 |
| tkajinam | cardoe, stephenfin I voted +2+A but am still hoping that we can somehow refactor it to reduce redundant code ... | 16:33 |
| tkajinam | hmm https://review.opendev.org/c/openstack/oslo.utils/+/977692 failed at tempest test. rechecking it... | 16:33 |
| cardoe | Yeah when I didn't have it redundant it was failing for me as well in any form I tried. So I threw in the towel to at least get something going by the release cause I've got a bunch on my plate. | 16:34 |
| cardoe | I still wanna work with stephenfin to get the websso auth into keystoneauth and gophercloud but 2026.1 is staring me down realllll hard right now. | 16:34 |
| opendevreview | Merged openstack/oslo.utils master: fix force_reraise to be annotated as NoReturn https://review.opendev.org/c/openstack/oslo.utils/+/977568 | 17:31 |
Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!