| === gschanuel6 is now known as gschanuel | ||
| blackboxsw | falcojr: I think I've finally addressed all the complexities of the user/groups schema. Too many options please recheck docs I rewrote for typo/grammar misses etc. | 17:37 |
|---|---|---|
| minimal | blackboxsw: side-equestion - do the schemas in general take into account a requirement for some values to be quoted (so they *are* treated as strings, not numbers)? | 17:52 |
| minimal | thinking of values like dates where (from memory) they are treated as numeric equations (i.e. "-" as minus) unless quoted | 17:52 |
| blackboxsw | minimal: the JSON schema looks at the YAML type as it is interpreted. For dates specifically, we'd have to declare something like {"type": "string" "format": "date"} for our schema to allow without error. Ultimately when YAML loads the date, it can accept things like `date: 2022-04-15` and yaml.safe_load turns it into the python dict `{"date": datetime.date(2032, 4, 15)}` | 18:02 |
| blackboxsw | it may have been an older version of YAML that bungled the date formatted strings | 18:03 |
| blackboxsw | but to answer your underlying question, JSON schema validation is on the content which has been interpreted by YAML during load I believe | 18:04 |
| minimal | blackboxsw: I'm thinking back to a MR I had merged maybe 12-18 months ago to fix some of the user-data examples including things like expiry date as they didn't work without quotes - so you think the quotes are no longer needed? | 18:07 |
| minimal | ah, here's the MR I'm thinking of: https://github.com/canonical/cloud-init/pull/618 | 18:10 |
| ubottu | Pull 618 in canonical/cloud-init "Correct documentation and testcase data for some user-data YAML" [Merged] | 18:10 |
| blackboxsw | minimal: and examples that have landed on every cloudinit/config/cc_* module are automatically run through validation by CI in our unittests per https://github.com/canonical/cloud-init/blob/main/tests/unittests/config/test_schema.py#L401 | 18:10 |
| blackboxsw | minimal: ahh right, so I just made the schema in this case specifically require a type: string https://github.com/canonical/cloud-init/pull/1379/files#diff-c18cfd88ec0bd5f1077576fee00a0b2e87cf0b30aa547ad1eaacfd5742e4a757R27 | 18:15 |
| ubottu | Pull 1379 in canonical/cloud-init "Schema users" [Open] | 18:15 |
| blackboxsw | but hrm, maybe worth properly supporting a date type. Reviews comments welcome on my open PR here. | 18:15 |
| blackboxsw | minimal: though I might want to pull in that functionality in a separate PR as this one has gotten a lot bigger than I'd hoped | 18:16 |
| minimal | blackboxsw: no issue with it defining string, was just wanting to clarify if single quotes are therefore still needed in the YAML for it to be treated as a string | 18:16 |
| blackboxsw | so at the moment, schema will tell you you have to have a string | 18:16 |
| minimal | I guess my question is regarding implicit versus explicit strings in YAML | 18:18 |
| minimal | as I thought the underlying problem was the Python YAML lib used could misinterpret implicit strings like dates | 18:19 |
| falcojr | minimal: I think that's just a yaml problem more broadly | 18:21 |
| minimal | in other words will "expiredate: 2022-04-16" pass both the schema validation and be treated by the code as a date, or will it pass the schema validation but be treated by the code as a number, in which case the schema needs to enforce "expiredate: '2022-04-16'" instead | 18:22 |
| blackboxsw | minimal: sorry just dropped the laptop trying to paste that example. Schema will warn about non string I. That case | 18:23 |
| minimal | blackboxsw: warn about non-string? or non-explicit-string? | 18:24 |
| blackboxsw | minimal https://paste.ubuntu.com/p/WgscmSjkxg/ | 18:24 |
| blackboxsw | yet quoted `expiredate: "2024-10-04" ` will pass without warnings | 18:25 |
| minimal | for that paste, is login for a Launchpad acct? | 18:26 |
| blackboxsw | bah | 18:26 |
| blackboxsw | pasting elsewhere | 18:26 |
| blackboxsw | minimal: 2nd try https://paste.opendev.org/show/bgROMPxKtqtahai5C6w0/ | 18:27 |
| === holman is now known as holmanb | ||
| minimal | ok, that seems to answer my question. Thanks | 18:28 |
| blackboxsw | I do wonder if JSON schema can handle date discrete date types anyway, but as falcojr mentioned it's probably a JSON schema/YAML general prob to sort. | 18:29 |
| blackboxsw | I forsee that I'll spend some googling time today :) | 18:29 |
| holmanb | started in on ephemeraldhcp for v6 and one bugfix led to another and I guess I just refactored the networking code a bit: https://github.com/canonical/cloud-init/pull/1383 | 18:30 |
| ubottu | Pull 1383 in canonical/cloud-init "Network functions refactor and bugfixes" [Open] | 18:30 |
| minimal | yeah YAML is a mess with its "magic detection" of types - remember the issue I'd raised again 18months or so ago about MAC address treatment lol | 18:30 |
| blackboxsw | heh true. I had forgotten about that. | 18:30 |
| blackboxsw | -> relocating for lunch | 18:30 |
| blackboxsw | nice fast work holmanb | 18:31 |
| minimal | blackboxsw: if you ever doing any future work on expiredate (and/or "inactive"), there's no testing in place for those (validating the numbers put into /etc/shadow fields) and I've never found the time to have a go at it myself | 18:31 |
| holmanb | @blackboxsw: don't get too excited, that's only bugfixes + refactorization ;) - no ephemeral work in that | 18:32 |
| holmanb | blackboxsw: I was just like "alright let's make is_ipv6_addr not so awful real quick", and after fixing it suddenly 125 tests were failing :/ | 18:37 |
| blackboxsw | Thanks falcojr https://github.com/canonical/cloud-init/pull/1379 landed. | 19:53 |
| ubottu | Pull 1379 in canonical/cloud-init "Schema users" [Merged] | 19:53 |
| falcojr | nice! | 19:53 |
| blackboxsw | knocked out a number of JIRA tickets with that given the 'enum' doc handling and 'hidden' key thank! | 19:53 |
| blackboxsw | thank you for the correct assessment of that schema crud in review | 19:53 |
| blackboxsw | falcojr: one minor PR for migrating legacy schema coming | 19:56 |
| blackboxsw | $ fgrep -r "schema =" cloudinit/config/cc_* | 19:56 |
| blackboxsw | cloudinit/config/cc_ntp.py:schema = { | 19:56 |
| holmanb | @falcojr: Thanks for the review, just landed https://github.com/canonical/cloud-init/pull/1383 (tests passed after the spelling nit commit) | 20:01 |
| blackboxsw | minor followup to migrate ntp legacy schema https://github.com/canonical/cloud-init/pull/1384 | 20:47 |
| ubottu | Pull 1384 in canonical/cloud-init "cc_ntp: migrated legacy schema to cloud-init-schema.json" [Open] | 20:48 |
| blackboxsw | ok hitting James' review now. | 20:48 |
| blackboxsw | falcojr: cc_* meta looks good I think minimally we want to filter docs out of `cloud-init devel schema --docs all` for those modules without JSON schema. | 21:47 |
| blackboxsw | otherwise big +1 on cleanup\ | 21:47 |
| falcojr | TIL 'cloud-init devel schema --docs all' :D | 21:47 |
| falcojr | never realized you could get doc info on the cli | 21:48 |
| holmanb | :) | 21:48 |
| blackboxsw | yeah developer docs | 21:48 |
| blackboxsw | yeah it's been my saving grace while going through and reviewing and schema changes | 21:48 |
| blackboxsw | much faster than iterating through `tox -e doc` | 21:48 |
| blackboxsw | and visually comparing | 21:48 |
| falcojr | so I DIDN'T need to 'tox -e doc' a million times this whole time? | 21:48 |
| falcojr | ;) | 21:49 |
| blackboxsw | ha! well next time we rewrite the whole schema we can use it :) | 21:49 |
| falcojr | heh, just give it a couple more years | 21:49 |
| blackboxsw | #toosoon | 21:49 |
| holmanb | if we throw half of it away first it won't take so long on the rewrite :) | 22:00 |
| blackboxsw | getting to EOW holmanb but minor thought on ssh-import-id https://github.com/canonical/cloud-init/pull/1334/files | 22:07 |
| ubottu | Pull 1334 in canonical/cloud-init "declare dependency on ssh-import-id (SC-864)" [Open] | 22:07 |
| holmanb | @blackboxsw: looking now | 22:09 |
| blackboxsw | sorry might want to refresh on latest comment | 22:10 |
| holmanb | +1: I assumed that's what you meant | 22:10 |
| holmanb | oh last comment, not updated suggested change, I see now | 22:14 |
| blackboxsw | ahh sorry will redact my suggested change in that first comment. as I think the latter is more appropriate given multiple entrypoints to ssh-import-id | 22:16 |
| holmanb | blackboxsw: np at all - at the rate that my own EOW is approaching, however, I think this is going to have to wait until monday to implement the second comment | 22:18 |
| blackboxsw | holmanb: totally get that. I'm about turning the lights off over her | 22:19 |
| blackboxsw | *here | 22:19 |
| blackboxsw | hey CloudCIX is going to look at implementing EphermealDHCP4 support for local timeframe IMDS detection... so that'll be good. that PR will likely idle to stale while they do that. | 22:20 |
| holmanb | I saw that, sounds good to me | 22:21 |
| holmanb | making some progress on EphemeralDHCP6 today, I grok it quite a bit more at least | 22:22 |
| holmanb | *made | 22:22 |
| blackboxsw | awesome, that'll help a few folks trying to spin up systems ipv6/dhcp6 only envs. | 22:31 |
| blackboxsw | for that ds-identify branch detected NoCloud in `/etc/cloud/cloud.cfg(.d/*)` I've got to run it through the paces with live-server installer ephemeral environment and make sure we don't get into a pickle with current implementation | 22:34 |
| blackboxsw | so, probably won't finish that review today | 22:34 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!