| ani | s there somewhere that talks about how to install all the necessary import modules? Like a pip requirements ? | 04:28 |
|---|---|---|
| ani | I am getting this error | 04:29 |
| ani | E ModuleNotFoundError: No module named 'configobj' | 04:29 |
| ani | ERROR: InvocationError for command /home/cavery/build/cloud-init-centos/ani-cloud-init/cloud-init/.tox/py3/bin/python -m pytest -vvvv --showlocals --durations 10 -m 'not hypothesis_slow' --cov=cloudinit --cov-branch tests/unittests (exited with code 4) | 04:29 |
| ani | i see https://github.com/canonical/cloud-init/blob/main/test-requirements.txt and https://github.com/canonical/cloud-init/blob/main/requirements.txt | 04:49 |
| holmanb | Hi ani o/ | 04:59 |
| holmanb | ani: tox should automatically handle that | 05:01 |
| holmanb | But yes, the files you mentioned should provide the required dependencies for unit tests | 05:03 |
| holmanb | i don't know that we document it explicitly for users - test deps should be handled by tox automatically and packaging should already work, so i don't think there's really an expectation for users to need to know which external module dependencies to provide | 05:05 |
| holmanb | Also I just tested with tox from epel on centos 9 stream, looks like tox installed deps correctly with `tox -e py3` | 05:17 |
| holmanb | Although two tests just failed... | 05:19 |
| holmanb | Heh, okay tar is a unit test dependency I guess | 05:20 |
| ani | OK | 05:45 |
| === kilian is now known as holzingk | ||
| holmanb_ | weeeeeee | 16:56 |
| holmanb_ | all the bsd PRs | 16:56 |
| holmanb_ | meena: when you get a chance could you peak at this one: https://github.com/canonical/cloud-init/pull/4683 | 16:56 |
| -ubottu:#cloud-init- Pull 4683 in canonical/cloud-init "fix: IscDhclient discover DHCP leases on FreeBSD" [Open] | 16:56 | |
| holmanb_ | oh hi ubottu, ears burning? | 16:56 |
| holmanb_ | meena: it's pretty short: +7/-1 for LOC changes, plus tests | 16:58 |
| meena | holmanb: looking | 17:22 |
| meena | holmanb: what does "This change adds allows discovering such lease files." this mean? | 17:24 |
| holmanb | "such" could be substituted with "those" in that sentance | 17:25 |
| holmanb | It's referring to the leases at that path | 17:26 |
| meena | "adds" vs "allows" feels like an either/or thing | 17:29 |
| Code_Bleu | meena: holmanb: I thought I tested this on OpenBSD before and it worked, but now I'm not sure. Now when I run 'cloud-init analyze show' I get this - https://sprunge.us/begTkI | 18:59 |
| meena | Code_Bleu: did you delete the appropriate .pyc files before applying the patch? | 19:00 |
| Code_Bleu | meena: pretty sure before I created this image I'm working with I manually cleared out everything and even ran pip uninstall cloud-init , rm -rf /run, cloud-init clean --logs...etc | 19:02 |
| meena | Code_Bleu: so you're testing this patch, or what exactly are you testing right now? | 19:06 |
| Code_Bleu | meena: I'm not really testing anything. I'm on an instance of OpenBSD that I created several days ago ...right have my PR was merged. I just so happened to connect to it and run the 'cloud-init analyze show' and got that error | 19:07 |
| meena | aaaah | 19:07 |
| meena | okay, so we need to fix that then. | 19:07 |
| meena | that works somewhat betterer on FreeBSD, so we want at least that level of support in allBSD | 19:08 |
| Code_Bleu | are you able to confirm on your end that it is an issue and it's just not me 😉 | 19:08 |
| meena | Code_Bleu: can you open a bug for that? | 19:08 |
| meena | need to put kiddo to bed now, so I'll take care of it after | 19:09 |
| Code_Bleu | 👍 | 19:09 |
| Code_Bleu | meena: disregard. I'm an ID10T. My stupid logs were 0 bytes in size, so I must have cleared it or something. So, obviously the analyze wont work without data to analyze :facepalm: | 19:15 |
| meena | why not tho? | 19:22 |
| meena | I would expect more sensible error message than a cryptic exception | 19:25 |
| holmanb_ | thanks for reviewing that meena | 19:31 |
| holmanb_ | we've talked about making that function distro-specific before, I think (in a review of some different code, iirc) | 19:32 |
| holmanb_ | or maybe that was with minimal? | 19:32 |
| holmanb_ | while back, can't remember | 19:33 |
| meena | the getter functions at least | 19:33 |
| holmanb_ | yeah exactly | 19:33 |
| holmanb_ | the filename check I don't care about so much, that's just checking string values in memory | 19:34 |
| Code_Bleu | meena: good point. have it check if logs are not 0 bytes for starters 😄 | 19:35 |
| holmanb_ | wait | 19:35 |
| holmanb_ | oh nvm | 19:35 |
| holmanb_ | Code_Bleu: agreed, if we can report better errors, we should | 19:36 |
| holmanb_ | meena: and moreso to make it easier to refactor in the future into methods in distro/ | 19:36 |
| Code_Bleu | https://github.com/canonical/cloud-init/issues/4686 | 19:42 |
| -ubottu:#cloud-init- Issue 4686 in canonical/cloud-init "cloud-init analyze on OpenBSD needs better error handling" [Open] | 19:42 | |
| Code_Bleu | meena: holmanb_ ^ | 19:43 |
| holmanb_ | Code_Bleu: appreciated | 20:02 |
| meena | Code_Bleu: i would venture a guess that this particular behaviour won't be restricted to OpenBSD | 20:19 |
| Code_Bleu | meena: I was thinking the same thing, but thought I'd be specific and not assume it did work on other OS or other BSD. If ya'll think it makes sense, I can remove 'OpenBSD' from the bug. | 20:25 |
| Code_Bleu | I'll leave the Env details of the OS I ran it on OpenBSD, but I'll update the description to just be generic. | 20:26 |
| Code_Bleu | meena: updated | 20:27 |
| meena | Code_Bleu: thanks. | 20:50 |
| Code_Bleu | meena: well 💩 We prob should have updated the 'build-on-openbsd' to remove the dmidecode with your #4654 PR ? | 20:55 |
| Code_Bleu | https://github.com/canonical/cloud-init/blob/4006c23cba5a5e65f0dba4db97491a09cd9a851e/tools/build-on-openbsd#L15 | 20:57 |
| holmanb | Ah well, we can do that in a followup Code_Bleu | 21:47 |
| holmanb | Code_Bleu: have you tried running unittests on OpenBSD? | 21:49 |
| holmanb | Curious if they pass | 21:49 |
| meena | holmanb: my guess is that more tests fail than on FreeBSD… and on FreeBSD currently only 2 tests fail with python netlink patches | 21:52 |
| meena | OpenBSD has no netlink, so those will fail for sure, and should be excluded from running / expected to fail on not (Linux / FreeBSD) | 21:52 |
| meena | I still haven't figured how to fix those damn ls output tests | 21:53 |
| meena | Code_Bleu: tox -e py3 # should do everything needed, but you'll need tox and bash as a start | 21:53 |
| meena | Code_Bleu: the next iteration should be to make dmidecode the fallback on OpenBSD | 21:59 |
| meena | holmanb: need to fix this one https://gist.github.com/igalic/d7770b21294d95bfdbae505e122d0a68 | 22:14 |
| holmanb | tar is also required if it's not already installed | 22:15 |
| meena | holmanb: tar is in the base system on BSDs | 22:18 |
| meena | on OpenBSD, perl is still in base@ | 22:18 |
| Code_Bleu | holmanb: meena: I have ran tox -e py3 before, but not sure how that relates to not removing the dmidecode from the 'build-on-openbsd'. Or maybe you were asking for other reasons? | 22:26 |
| meena | Code_Bleu: how badly does tox -e py3 fail on OpenBSD? | 22:27 |
| Code_Bleu | I had tested using meena's fix before it got merged and had dmidecode removed from my test system and it all worked. I just didn't remove it from my PR since meena's PR hadn't been merged yet and forgot to mention to have that just removed from the build-on-openbsd as a package requirement. | 22:28 |
| meena | Code_Bleu: to be fully functional, it's still a dependency, but now it's very optional | 22:29 |
| meena | given that we kinda cover 75% of all requests that our sources ask of dmidecode | 22:29 |
| Code_Bleu | meena: correction. I ran the tox -e py3 on the repo that was on my local Gentoo box. I never ran it on OpenBSD itself. I just ran it for purposes of getting tired of pushing commits up and the CI fail :P | 22:29 |
| Code_Bleu | meena: so we just plan on keeping dmidecode as a package that the build-on-openbsd is going to install? | 22:31 |
| meena | Code_Bleu: i'm torn, but my instinct is to rip it out | 22:31 |
| meena | holmanb: those ls tests already have @pytest.mark.parametrize, i wonder if i can use that to say, this is what it says on busybox / BSD | 22:36 |
| meena | minimal: how come we have no cloudinit.util.is_Alpine() ? | 22:46 |
| meena | I have committed a crime: https://gist.github.com/igalic/b4ff1fefad329331f43d1f74bfdc44c9 | 23:07 |
| minimal | holmanb: which function were you referring to? | 23:07 |
| minimal | meena: isn't that covered by is_Linux() ? | 23:12 |
| meena | minimal: no, busybox ls and gnu ls work differently | 23:12 |
| minimal | ok, but that has nothing to do with Linux / not Linux | 23:13 |
| meena | busybox and BSD agree on the output, but disagree on the return code | 23:13 |
| minimal | so where are you expecting that any is_Alpine() function may be needed? | 23:14 |
| meena | really, so far, just for this stupid test | 23:14 |
| minimal | which test? | 23:14 |
| meena | maybe i should rewrite it to something more sensible | 23:14 |
| meena | see this patch, https://gist.github.com/igalic/b4ff1fefad329331f43d1f74bfdc44c9 | 23:14 |
| minimal | Alpine can have either Busybox tar or "fall fat" tar installed, do determining if code is running on Alpine does not indicate which version of tar is installed... | 23:17 |
| minimal | the only way to know is do so "tar --version" or equivalent | 23:17 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!