[18:26] problems with the CI? I'm seeing several python versions' unit tests failing for a new PR I just opened [18:43] minimal: FS_LABELS and ISO9660_DEVS look like intermediary values that get used to detect which cloud is used [18:43] agreed it is duplication of work to get that in both ds-identify and in cloud-init python code [18:45] definitely an opportunity for improvement - though I think we'd want to think hard about how best to share data with cloud-init's python code before doing it [18:45] minimal: tests pass locally with your PR, I'll rerun the github actions jobs [18:45] holmanb: yeah I'm not really seeing many benefits for using ds-identify vs just setting datasource_list to a small subset [18:46] can't see WHY the CI is failing as the logs are so huge my laptop's fan is going like a hurricane [18:46] minimal: agreed - if you don't mind modifying your image then setting datasource_list then there isn't much benefit as it is now [18:47] minimal: the real benefit imo is to not have to modify your image in the first place [18:47] I'm not modifying it, I'm setting datasource_list as part of *creating* the disk image ;-) [18:48] anyway I'm still testing ds-identify here [18:48] I can see it being useful for "generic" disk images but then not sure that would work well with my thought of packaging each DS as a separate subpackage [18:50] Agreed - if you're creating a disk image for a known datasource, there isn't much benefit [18:51] but also if you have a single datasource in your datasource list, ds-identify trusts the image builder and skips the checks [18:52] basically I'm currently wondering if I can create Alpine subpackages like "cloud-init-ds-aws", "cloud-init-ds-azure" etc with the relevant sources code moved out from the main package [18:53] and then have cloud-init having a dep on "cloud-init-ds-all" to pull them all in but it could be overriden by doing "apk add cloud-init cloud-init-ds-aws" [18:53] ahhh, I see [18:55] my guess is that it would be really hard to untangle tests in the subpackages [18:55] but if you don't care about that it probably wouldn't be too hard [18:55] but still a lot of work [18:55] I guess that would get you a really "minimal" build ;) [18:58] well there's approx 1 MB of DS sources + whatever size the pyc files are too, so being able to only install a single DS would shrink things a bit. Having separate DS packages would also allow me to set per-DS deps which I can't do currently [18:58] minimal: if you do that you might want a way to have ds-identify only run if cloud-init-ds-all is installed (or if more than one of the -{aws,azure,etc} subpackages are installed) [18:59] yeah I'm still thinking things through [18:59] anyway I'm still going to submit my ds-identify changes anyway [19:00] just need to get a couple of other PRs out to the way first [19:00] just re-kicked the CI for you [19:01] thanks [19:01] it's a fairly simple PR so was surprised when it had CI errors [19:03] nope, it's failing again [19:05] ah, this might be an updated dependency issue actually [19:06] I just ran again against your branch with tox -r -e py3 and I see failures [19:08] yeah jsonschema just pushed release 4.21.0 [19:08] and the failing tests are schema related [19:08] I'll try to get a fix in shortly and ping you minimal [19:09] ah right. I didn't check tox before pushing as the change was minor lol [19:09] thanks [19:40] no worries [19:41] I didn't see it the first time I checked your PR either because I normally don't run unittests with -r (because it takes longer and dependency changes are infrequent) [19:45] minimal: just merged the fix - you'll need to rebase on the tip of main and push again to get tests to run with the pinned dependency [20:06] holmanb: ok, let me do that [22:27] 2024 cloud-init release schedule is posted here: https://discourse.ubuntu.com/t/2024-cloud-init-release-schedule/41679 [23:36] ok, so more 1 month to get several more PRs raised and merged ;-) [23:42] I need to get the run vs var/run stuff fixed until then