16:00:30 #startmeeting Ansible Core Public IRC Meeting 16:00:30 Meeting started Thu Jan 12 16:00:30 2023 UTC. 16:00:30 This meeting is logged and archived in a public location. 16:00:30 The chair is shertel. Information about MeetBot at https://fedoraproject.org/wiki/Zodbot#Meeting_Functions. 16:00:30 Useful Commands: #action #agreed #halp #info #idea #link #topic. 16:00:30 The meeting name has been set to 'ansible_core_public_irc_meeting' 16:00:35 #info agenda https://github.com/ansible/community/issues/661 16:00:45 o/ 16:00:59 hello! hope you are feeling better 16:01:26 #topic https://github.com/ansible/ansible/pull/79370 16:01:27 yes, I'm pretty fine, I had a cold, so nothing too bad... 16:01:43 not good enough to go outside, but good enough to work and live :) 16:02:01 I'm glad it's not bad 16:02:35 I was just looking over this one again, looks like all my requested changes were made. 16:03:16 I'm curious whether this has a chance of getting merged soon :) 16:03:20 --metadata-dump still includes the private plugins, but no longer listed otherwise 16:03:49 I just hadn't gotten around to reviewing it yet, it's been on my to-do list for a while 16:04:10 LGTM, I will merge later unless anyone objects 16:04:13 that's intentional, as the plugins still exist, so tools showing the documentation (which use --metadata-dump) should have access to it 16:04:30 cool, thanks a lot! once its merged I'll merge an antsibull-docs PR which uses that data ;) 16:04:54 I didn't want to merge it until its clear that ansible-doc will also use this flag 16:05:39 bcoca, do you want a chance to review again too^ ? 16:08:47 #action shertel to review/merge https://github.com/ansible/ansible/pull/79370 16:09:09 thanks! 16:09:10 #topic https://github.com/ansible/ansible/pull/77737 16:09:13 im still not convinced we need this 16:09:14 oh, hey, forgot about this thing 16:09:34 i think documenting the description as 'internal use only' is a good convention and there is no need to 'codify' it 16:09:59 sivel: the meeting, or the PR? 16:10:03 meeting :) 16:10:29 I am +0 on it, have mostly just been relaying other people's reviews 16:10:54 (on codifying 'internal use only') 16:11:05 shertel: maybe bring it up in next weeks core team meeting just to ensure we do want to move forward with it, before merging :) 16:11:20 I think it's important to make sure that users are not unnecessarily exposed to internals, and I did collect some other opinions (see the community topic) 16:11:23 ok, this was the first complaint I heard 16:11:29 there is also the scenario .. don't you want options documented/readable for the 'internal people'? don't assume author is always around (or even remembers) what the plugin does/how it works 16:11:48 I am basically +0 too. There have been a fair number of requests for it 16:11:57 bcoca: you can still see the docs, but not everyone will see it right away 16:12:00 felixfontein: 'community topioc'? 16:12:20 bcoca: https://github.com/ansible-community/community-topics/issues/154 16:13:43 for one, using the file name has strong opposition in core 16:14:04 _ is also overloaded with 'deprecated' .. though that does not apply to collections anymore since 2.14 16:14:08 yeah, I don't mind not using the filename, that was just the easiest thing I came up with :) 16:14:24 he, why _deprecated was imp;lemented in 1.8 ... 16:14:38 _ deprecation never really worked for collections 16:14:52 I think we can move on to 77737 16:15:01 #action shertel to bring up https://github.com/ansible/ansible/pull/79370 at next internal core meeting for more discussion 16:15:27 +1 on that 16:15:27 felix is this one just waiting on a review too? 16:15:39 shertel: yes it is... 16:16:15 I think this one shouldn't be controversial, it's just doing for test and filter plugins what the tests do for all other plugin types, namely check that ansible-doc doesn't choke on the docs 16:16:51 I can add it to my list, I might not get to it this week so if anyone else wants to merge sooner please do 16:17:03 (having a schema test for the docs would be better, but that's more work...) 16:18:26 I'll keep that in mind when I review it 16:18:44 thanks again :) 16:18:55 np 16:18:57 #topic https://github.com/ansible/ansible/pull/79471 16:19:23 this one is controversial I think :) 16:20:00 I think somewhere (I don't remember anymore where) it was mentioned that core team never wanted the env variable fallback for modules in the first place 16:20:05 cause it is confusing about env vars, most people expect to set them on controller and then the modules pick them up (which they do if they are localhost or 'fake remotes' like networking) 16:20:39 that's definitely true, but the same problem is also true for modules that read files... 16:20:45 and the current implementation which is via 'fallback to generic function, just happens to be the only one implemented is lookup env var ..' 16:21:31 I still think it's good to be able to document such fallbacks in a uniform way; in particular this allows the docs generation to add a hint that these vars must be defined in the correct environment 16:21:47 (which you don't want to do manually in every description where you have to manually write which env var is used as a fallback...) 16:22:07 i would just rewrite the callbacks and allow for direct env= declaration in argspec .. but that is just me 16:22:21 making clear they are low precedence 16:22:39 but that's orthogonal to them being allowed in the docs 16:22:40 and env being a list 16:23:18 felixfontein: not really https://github.com/ansible/ansible/pull/79720 since i hope to merge this for 2.15 16:23:32 docs/arsgpec will be much tighter 16:23:59 oooh, hadn't seen that yet 16:24:23 hmm, I'll have to look into that in more detail 16:24:42 it still will take a long time for collections to actually use that 16:24:55 its mostly adding 'required_X' to docs, but if we add env= .. .that also should be added there since it also converts 'docs to argspec' and viceversa 16:25:12 so action plugin will now be able to use 'argspec' directly w/o hardcoding/duplicating 16:26:05 felixfontein: not really, some are already using their 'home grown' version https://github.com/ansible/ansible/pull/79713 <= this si attempt to have validate-modules ignore their hack 16:27:20 ah, I haven't seen that one yet either... I really like having this information in DOCUMENTATION, but mainly so that antsibull-docs can show it 16:27:36 exactly! 16:27:54 didn't they have a similar PR some time ago (by another person)? 16:28:07 `we want a format not only to equate to the argspec, but be readable by users` <= my response to that pr 16:28:27 yeah, that would be nice 16:28:40 felixfontein: yes, also diff implementation (both that started from some code i wrote around 2.2 to do conversion from doc to argspec and back) 16:28:55 well, the version from argspec can be made readable by extra code 16:29:04 felixfontein: also why 'diff ' and 'check mode' made it into attributes 16:29:11 I think russoz will also be very much interested in your PR 16:29:26 I guess in that case 79471 is tabled for now 16:29:34 okay 16:29:38 #topic https://github.com/ansible/ansible/issues/79695 16:29:40 last one 16:29:53 yet another 'allow to hide stuff' PR ;) 16:29:58 i've been posting it and asking for feedback since novembre 16:30:12 bcoca: I must have missed that... 16:30:51 i used to be a branch that iwas posting ... mostly ignored, so i created draft pr as it is easier to show diffs 16:31:13 79695 is important for sanity testing modules that also have an action plugin and uses some parameters to communicate between action plugin and module that shouldn't be used by users 16:31:23 felixfontein: iirc you are only one that actually gave feecback when i did ... 16:31:44 this one seems like a reasonable request, we should still be able to validate private options 16:31:46 +1 16:32:08 also more importantly not having to disable certain tests ensures that these still run for new public options 16:32:18 exactly 16:32:18 my issue with that, users can still pass the argument themselves 16:32:30 it will depend on action plugin if that is passed through or overwritten 16:32:31 otherwise PRs might add new options without docs and they get accidentally merged because the sanity tests don't complain 16:32:52 bcoca: true, but that's already the problem right now 16:33:26 not saying it isnt, saying that we should implement filter and not give the option of passthrough if the argument is really private 16:33:35 the proposal won't change anything with regard to that, it will only make sanity testing more reliable for such module/action plugin pairs 16:33:46 the problem with filter ... it requires my PR first ... 16:34:07 since right now controller is oblivious to argspec 16:35:31 is your main objection that this might be a breaking change when this is implemented before your PR? 16:35:33 that said, i don't see it as blocker 16:36:08 no, just sayign its an issue we'll need to address ALSO .. but requires more infra (which my PR provides) to be able to do so 16:37:04 is the general idea (having a `undocumented` boolean flag in argspec) ok? if yes I can start with a PR for that (to make validate-modules aware of that) 16:37:35 bikeshed moment: internal=yes 16:37:42 :D 16:37:51 since we also want filtering, undocumented might be 'incomplete name' 16:38:04 I agree `internal` is better when there is also filtering involved 16:38:07 if we flag as internal, we can justify removign 'user input' later on 16:38:16 I mainly chose `undocumented` because I didn't expect filtering to happen anytime soon 16:39:00 'soon' is realtive ... 16:39:19 more precisely, I didn't expect anything to happen for at least a year or so 16:39:36 i hope you are wrong ... i fear you might not be 16:39:40 but then I didn't knew about your PR / branch (or if I did I forgot about it again...) 16:39:52 fair enough 16:40:08 using `internal` would keep our options open for the future 16:41:21 when we are in bikeshedding mode: should https://github.com/ansible/ansible/pull/79370 also use 'internal' instead of 'private'? 16:41:30 one thing we could also do is change the ignore to be specific `validate-modules:undocumented-parameter:is_file` 16:41:59 that might require a lot more code 16:42:12 felixfontein: yes, but again, think we dont need to codify this .. private also implies more 'secret/secure' than internal does 16:42:26 or maybe not... might be a good Q for mattclay 16:42:34 felixfontein: yes, but the advantage of changing the test, you avoid changing runtime code 16:42:49 and having to change moduels/argspec/docs/doc printing 16:43:23 the docs printing is not affected since it doesn't have access to the argspec (except with your PR, but it could filter that out) 16:43:29 which allows us more time to add my PR and then open the floodgates to changes that requried controller to have argspec knowledge (i.e validate module params at action plugin time!) 16:44:50 ok for me, assuming that mattclay doesn't object 16:45:21 not saying that 'is the way' .. but that it might be an alternative 16:46:11 it's generally a good thing, though, not just for this application 16:46:47 sometimes you have screwed up things in the docs and have to keep it for some time for compatibility purposes, then at least you can make sure that sanity tests will catch the same error in another option 16:47:38 (though in most cases the error is 'syntax error' and very unspecific... so it won't help in these cases) 16:48:07 (syntax error because schema validation failed, not because invalid YAML) 16:50:23 felixfontein: do you need anything else with this one? Not sure what the implementation looks like, but seems reasonable to me. 16:51:15 shertel: you mean 'extended'/hierarchical error codes? I'd mainly like a quick opinion from mattclay what he thinks, but besides that I think I'm happy 16:51:37 it might be a little early for him, I'm sure he'll respond when he sees it 16:51:47 but this isn't urgent at all, it's mainly that this discussion came up again recently and I decided to create a proposal for it :) 16:52:02 yeah, that's totally fine. thanks! 16:52:13 sorry there is not more of a quorum here! 16:52:22 we are running out of time, so im not going to bring up the issue of 'meeting format change' .. but it might be something to discuss, specially with timezones limitations 16:52:59 somethine more async will make meetings slower, but increase participation, visibility and quorum 16:53:16 waiting up to a month for the next meeting isn't exactly fast either :) 16:53:34 so in average it will make things go faster, I think 16:53:44 agreed, but the byweekly was clearly too often and mostly cancled 16:54:03 monthly wasn't that busy either 16:54:09 idk what best solution is, just seeing the current one is not working 16:54:21 in general I'd prefer just to get pings directly in #ansible-devel/github when someone wants my opinion/a review 16:54:49 part of me wants to go back to ansible/proposals ... but that is also mixed since there is no onus to make people comment nor timeframes attached 16:55:06 I'd be in favor of trying ansible/proposals again 16:55:08 ^ direct pings work for 'when SME is online and available' 16:55:31 shertel: that's fine for uncontroversial stuff / 'simple' PRs, but I think for proposals / controversial PRs I'm not sure direct pings is a good idea 16:55:53 ansible/proposals SGTM 16:55:58 quorum is hard since we are distributed across many timezones, why 'direct ping' cannot scale 16:56:20 I don't even know who is in core team right now :) 16:56:55 well I know (or well, mostly assume) some folks are, but I don't know if there are more 16:57:00 i have rough idea .. but not 100% sure myself ... 16:57:05 lol 16:57:10 :) 16:57:17 ^ not kidding, 16:57:26 reminds me a bit of $JOB.... 16:57:50 some people have a 'nebulous status' like nitz .. he is both core and not core, he is heisen-core 16:58:59 yeah, I guess I do not count him exactly as core but part of a superset 16:59:00 sounds like core team is a bit of a probability cloud 16:59:41 fyi, im out next week, but might still pop in 16:59:58 and every other week till mid march 17:00:12 enjoy! 17:00:12 note to self: try to get all the PRs merged that bcoca doesn't like 17:00:32 thanks all for coming to discuss :) 17:00:34 #endmeeting