| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2025-02-27 | |||
| 17:16:42 | opendevreview | Artem Goncharov proposed openstack/codegenerator master: Add required to compute service responses https://review.opendev.org/c/openstack/codegenerator/+/942934 | |
| 22:14:33 | M0weng[m] | <stephenfin> "0weng: mea culpa: looks like I..." <- Thank you!! 🙏 | |
| 22:17:39 | opendevreview | Oria Weng proposed openstack/python-openstackclient master: [DNM] Identity: Migrate 'registered limit' commands to SDK https://review.opendev.org/c/openstack/python-openstackclient/+/942735 | |
| 23:06:48 | M0weng[m] | Uh oh, I got an internal server error when trying to update a federation protocol's remote_id_attribute only:... (full message at <https://matrix.org/oftc/media/v1/media/download/AcXQUdcnE8LuNEEjXHuiKeEWrFA55TiLOD29oGhp3ZbkicDx4zxI1G4im0Zt-fQMtYG0Le1RpPpuGMKdswrfz29CeVkcKpUwAG1hdHJpeC5vcmcvTFJuTVRqbGpSTkNybnZQQ2VBTXZtb211>) | |
| 23:10:09 | M0weng[m] | I have also been able to successfully edit just the mapping_id or both the mapping_id and the remote_id_attribute together, just not remote_id_attribute by itself. | |
| 23:49:53 | opendevreview | Merged openstack/codegenerator master: Hardcode barbican quota schemas https://review.opendev.org/c/openstack/codegenerator/+/941543 | |
| #openstack-sdks - 2025-02-28 | |||
| 01:43:28 | opendevreview | Merged openstack/python-openstackclient master: Add four new network agent types to the list command filter https://review.opendev.org/c/openstack/python-openstackclient/+/942175 | |
| 01:43:29 | opendevreview | Merged openstack/openstacksdk master: exceptions: Deprecate http_status, request_id params https://review.opendev.org/c/openstack/openstacksdk/+/929017 | |
| 02:14:02 | opendevreview | Merged openstack/openstacksdk master: tests: Rename cloud name variables https://review.opendev.org/c/openstack/openstacksdk/+/942712 | |
| 02:14:04 | opendevreview | Merged openstack/openstacksdk master: identity: Fix limit, registered limit creation https://review.opendev.org/c/openstack/openstacksdk/+/942818 | |
| 08:08:49 | opendevreview | Merged openstack/codegenerator master: BS volume bootable parameter is a string https://review.opendev.org/c/openstack/codegenerator/+/942920 | |
| 08:09:33 | opendevreview | Merged openstack/codegenerator master: Add required to compute service responses https://review.opendev.org/c/openstack/codegenerator/+/942934 | |
| 08:40:13 | opendevreview | Rajesh Tailor proposed openstack/python-openstackclient master: Add support for image properties in server show https://review.opendev.org/c/openstack/python-openstackclient/+/940799 | |
| 08:41:26 | opendevreview | Rajesh Tailor proposed openstack/openstacksdk master: Add support for image properties in server show https://review.opendev.org/c/openstack/openstacksdk/+/940798 | |
| 08:46:26 | priteau | Hello stephenfin. The kayobe CLI broke yesterday following the release of cliff 4.9.0. I have opened a bug with more details here: https://bugs.launchpad.net/kayobe/+bug/2100579. I don't know yet if this is a regression in cliff or if kayobe is using it wrongly. | |
| 10:50:13 | stephenfin | priteau: Thanks. I'll take a look now | |
| 11:13:30 | stephenfin | priteau: So two things. First, it's a definite logic mistake in cliff: the filter and list comprehensions are not the same thing. I'll fix that now. | |
| 11:14:46 | stephenfin | Secondly: the CommandHook class stated that get_parser method should return an ArgumentParser instance, while get_epilog should return a string. Kayobe's 'HookDispatcher' is returning None for both. | |
| 11:15:34 | stephenfin | But, seeing as that's been tolerated before and there was the filter for the epilog, I'm inclined to think that cliff's own docstrings were wrong | |
| 11:38:36 | opendevreview | Stephen Finucane proposed openstack/cliff master: command: Filter on empty epilog, not empty hooks https://review.opendev.org/c/openstack/cliff/+/942992 | |
| 11:38:37 | opendevreview | Stephen Finucane proposed openstack/cliff master: hooks: Update type hints to reflect reality https://review.opendev.org/c/openstack/cliff/+/942993 | |
| 11:38:46 | stephenfin | gtema, priteau: ^ | |
| 11:50:48 | priteau | stephenfin: I have just tested 942992, it resolves the issue in kayobe | |
| 11:51:15 | priteau | Would you be able to push a new release soon? | |
| 11:51:48 | stephenfin | Yes, I'd just like gtema to sanity check it, then we can merge and cut a 4.9.1 release | |
| 11:52:06 | gtema | sure | |
| 11:53:34 | stephenfin | https://review.opendev.org/c/openstack/releases/+/942996 There's a release patch. We'll just recheck it once those patches are merged. | |
| 12:02:05 | priteau | stephenfin: Would you still recommend we fix the HookDispatcher class? I can look into it once our CI is green again. | |
| 12:04:25 | stephenfin | priteau: You can fix your CI now by returning '' from get_epilog https://paste.opendev.org/show/bgLM8MsGNB0sWVVtjzsS/ | |
| 12:04:51 | stephenfin | As an alternative to capping and uncapping | |
| 12:06:26 | priteau | Any CI fix we do ourselves involves fixing the oldest branches first, because we have upgrade jobs that are failing. We don't use upper constraints in kayobe so a new release would magically fix all our branches :) | |
| 12:07:47 | priteau | Then I can fix get_epilog in master and backport the usual way | |
| 12:18:07 | stephenfin | priteau: Okay, then maybe hold tight until this evening and we'll hopefully be able to get that fix out sharpish | |
| 12:18:52 | stephenfin | otherwise, IMO you might as well apply the fix to all branches rather than a cap followed by an uncap. It's literally half the work :) | |
| 12:19:01 | gtema | stephefin - I assume we may want to add explicit test like I posted in releases channel for capturing regressions | |
| 12:19:33 | stephenfin | sorry, what explicit test? | |
| 12:19:39 | gtema | https://paste.openstack.org/show/byZpxP2y2ujhv4TCKehT/ | |
| 12:19:43 | gtema | something like that | |
| 12:19:51 | gtema | just convert print to assert | |
| 12:19:54 | stephenfin | ah, yes | |
| 12:20:52 | gtema | I assume you have idea how to fix the regression assuming you introduced this change? You should have had reasons for that | |
| 12:21:48 | stephenfin | Yeah, I'll come up with something. I first trying to figure out why it has that impact though | |
| 12:22:16 | gtema | great, I am afk for now, but feel free to ping me | |
| 12:23:18 | stephenfin | As for why I changed that, we're asserting that if you give e.g. data_type=str, the output will a str | |
| 12:23:57 | stephenfin | If we just returned the value the user provided without doing anything to it, you do e.g.: | |
| 12:24:14 | stephenfin | _conver_type(123, data_type=list) | |
| 12:24:26 | stephenfin | but the result will be the int 123 | |
| 12:24:41 | stephenfin | which is obviously wrong, and mypy complained as such | |
| 12:25:00 | stephenfin | hence the change. and the horrific type hints :) | |
| 12:36:50 | stephenfin | gtema: Okay, there's a bigger issue with _convert_type here. It popped up in older releases if you passed list_type also. For example https://paste.opendev.org/show/bZOYPdh75lIJLal2P65T/ | |
| 12:37:23 | stephenfin | idk if we should/can support this, at least while continuing to offer type conversion :-\ | |
| 12:41:48 | stephenfin | tbc, the issue is that we're using this inside '__getattribute__', and '_convert_type' is occasionally returning a _copy_ of a field, rather than the field itself | |
| 12:42:13 | stephenfin | Thus, attempts to modify a field in-place modify the copy rather than the "source" | |
| 12:42:40 | stephenfin | To fix this properly, I think we need | |
| 12:42:47 | stephenfin | ...to kill _convert_type | |
| 12:44:12 | gtema | We should rather redesign everything from OpenApi anyway | |
| 13:22:19 | stephenfin | https://review.opendev.org/c/openstack/horizon/+/943001 | |
| 13:52:06 | opendevreview | Stephen Finucane proposed openstack/openstacksdk master: Support server unshelve to specific availability zone https://review.opendev.org/c/openstack/openstacksdk/+/942877 | |
| 15:02:13 | opendevreview | Stephen Finucane proposed openstack/openstacksdk master: fields: Save converted attributes before returning https://review.opendev.org/c/openstack/openstacksdk/+/943009 | |
| 15:02:25 | stephenfin | gtema: ^ | |
| 15:19:58 | opendevreview | Merged openstack/cliff master: command: Filter on empty epilog, not empty hooks https://review.opendev.org/c/openstack/cliff/+/942992 | |
| 15:19:59 | opendevreview | Merged openstack/cliff master: hooks: Update type hints to reflect reality https://review.opendev.org/c/openstack/cliff/+/942993 | |
| 16:26:18 | opendevreview | Douglas Viroel proposed openstack/openstacksdk master: Bump compute max microversion to 2.100 https://review.opendev.org/c/openstack/openstacksdk/+/938833 | |
| 16:54:52 | priteau | stephenfin: Thanks a lot for the quick fix. It looks like our CI is green again. | |
| 17:34:39 | opendevreview | Artem Goncharov proposed openstack/codegenerator master: Fix network external_gateway_info schema https://review.opendev.org/c/openstack/codegenerator/+/943036 | |
| 20:03:12 | opendevreview | Artem Goncharov proposed openstack/codegenerator master: Bootstrap magnum OpenAPI build https://review.opendev.org/c/openstack/codegenerator/+/943058 | |
| 20:11:58 | opendevreview | Artem Goncharov proposed openstack/codegenerator master: Bootstrap magnum OpenAPI build https://review.opendev.org/c/openstack/codegenerator/+/943058 | |
| 20:13:51 | opendevreview | Artem Goncharov proposed openstack/codegenerator master: Bootstrap magnum OpenAPI build https://review.opendev.org/c/openstack/codegenerator/+/943058 | |
| 20:25:15 | opendevreview | Manuel Osorio proposed openstack/openstacksdk master: Add share transfer to shared file system https://review.opendev.org/c/openstack/openstacksdk/+/925653 | |
| 20:43:34 | opendevreview | Artem Goncharov proposed openstack/codegenerator master: Bootstrap magnum OpenAPI build https://review.opendev.org/c/openstack/codegenerator/+/943058 | |
| 21:52:14 | opendevreview | Manuel Osorio proposed openstack/openstacksdk master: Add share transfer to shared file system https://review.opendev.org/c/openstack/openstacksdk/+/925653 | |
| 21:56:16 | opendevreview | Oria Weng proposed openstack/python-openstackclient master: [DNM] Identity: Migrate 'limit' commands to SDK https://review.opendev.org/c/openstack/python-openstackclient/+/936279 | |
| 22:03:00 | opendevreview | Oria Weng proposed openstack/python-openstackclient master: [DNM] Identity: Migrate 'limit' commands to SDK https://review.opendev.org/c/openstack/python-openstackclient/+/936279 | |
| 22:03:13 | opendevreview | Oria Weng proposed openstack/python-openstackclient master: [DNM] Identity: Migrate 'limit' commands to SDK https://review.opendev.org/c/openstack/python-openstackclient/+/936279 | |
| #openstack-sdks - 2025-03-01 | |||
| 05:23:19 | frickler | gtema: stephenfin: after https://review.opendev.org/c/openstack/python-openstackclient/+/929712 the "openstack role show" command is case-sensitive, which it wasn't before. arguably this could be considered a regression, if you call it a bug fix I think at least a reno might be needed | |
| 05:24:00 | frickler | amongst others this is breaking keystone jobs as in https://zuul.opendev.org/t/openstack/build/46f9132008204e3b9910b9d826af20ad as M0weng[m] discovered | |
| 05:28:40 | frickler | https://paste.opendev.org/show/bsDrVMhANQmJomtHRGms/ | |
| 09:02:43 | gtema | frickler - I noticed that also last night. But I need to object - OSC was and is case sensitive. I also checked that creation of the role with osc 7.2.0 and master work the same. I assume something has changed in the osc wrapper or wherelse. To not to spend too much time I proposed https://review.opendev.org/c/openstack/keystone/+/943091 fixing the test which is not correct, but was working before | |
| 09:06:19 | gtema | actually a keystone is throwing exception claiming that the role with this name already exists, so it doesnt look to me like a osc issue | |
| 15:38:22 | frickler | gtema: did you check my paste? that's pure OSC without a wrapper and with 7.2.1 or earlier the "role show" command is not case sensitive. both "role show Member" and "role show member" or even "mEmBeR" give the same role as result | |
| 15:40:56 | frickler | checking with debug, the keystone API request also is case-insensitive, the same "member" role is returned, but despite that result osc/sdk says "no result found" for 7.3.0 | |
| 16:13:40 | gtema | frickler - no I missed the paste. But I am speechless since that works for me (perhaps one more reason to dislike Apfel devices). Technically what has changed is that between 7.2.1 and 7.3.0 role methods are switched to sdk from old python-keystoneclient+osc-lib, but I do not see where case would be altered | |
| #openstack-sdks - 2025-03-03 | |||
| 10:10:16 | stephenfin | frickler: gtema: TIL that keystone is case-insensitive. The difference in behavior appears to be because Resource.find is case sensitive https://paste.opendev.org/show/bkHaAZHo7xaUKu2I5WwW/ | |
| 10:11:23 | stephenfin | Changing that globally is likely to be a large change that could have knock-on effects for other services that may not (a) allow filtering by name query string parameters and/or (b) are case sensitive. idk if we want to do that | |
| 10:11:46 | gtema | neah, we must stick to case sensitivity | |
| 10:12:22 | gtema | services themselves are free to be case sensitive or insensitive. SDK must respect case | |
| 10:12:26 | stephenfin | This is also presumably a change we've seen in the other identity resources that have switched to using SDK in recent releases. I'm guessing this is just the first one we've noticied | |
| 10:12:29 | stephenfin | *noticed | |
| 10:13:33 | stephenfin | If this is important enough as a regression, we probably need to temporarily reimplement Resource.find in OSC and issue warnings on case-insensitive matches | |
| 10:14:27 | stephenfin | s/as a/to be classified as a/ | |
| 10:14:28 | gtema | wrt keystone I see additional undocumented "case_sensitive" parameters support in the code | |
| 11:57:56 | frickler | stephenfin: gtema: how about defining this to be a feature instead of a regression, but being nice to consumers and add a reno with a big warning? | |
| 11:58:50 | gtema | I would also prefer feature rather than regression | |
| 13:09:11 | opendevreview | yatin proposed openstack/openstacksdk stable/2024.1: [Stable Only] Do not pin openstacksdk to master https://review.opendev.org/c/openstack/openstacksdk/+/943177 | |
| 14:39:10 | opendevreview | yatin proposed openstack/openstacksdk stable/2024.1: [Stable Only] Do not pin openstacksdk to master https://review.opendev.org/c/openstack/openstacksdk/+/943177 | |
| 14:39:36 | opendevreview | yatin proposed openstack/openstacksdk stable/2024.2: [Stable Only] Do not pin openstacksdk to master https://review.opendev.org/c/openstack/openstacksdk/+/943182 | |
| 19:44:56 | opendevreview | Artem Goncharov proposed openstack/codegenerator master: Bootstrap magnum OpenAPI build https://review.opendev.org/c/openstack/codegenerator/+/943058 | |
| #openstack-sdks - 2025-03-04 | |||
| 10:23:01 | opendevreview | Stephen Finucane proposed openstack/openstacksdk master: Do not pin openstacksdk to master https://review.opendev.org/c/openstack/openstacksdk/+/943236 | |
| 11:21:12 | opendevreview | Merged openstack/openstacksdk stable/2024.2: [Stable Only] Do not pin openstacksdk to master https://review.opendev.org/c/openstack/openstacksdk/+/943182 | |
| 12:09:29 | opendevreview | Merged openstack/openstacksdk stable/2024.1: [Stable Only] Do not pin openstacksdk to master https://review.opendev.org/c/openstack/openstacksdk/+/943177 | |