19:00:05 <thaumos> #startmeeting ansible core
19:00:06 <zodbot> Meeting started Tue Oct 31 19:00:05 2017 UTC.  The chair is thaumos. Information about MeetBot at http://wiki.debian.org/MeetBot.
19:00:06 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
19:00:06 <zodbot> The meeting name has been set to 'ansible_core'
19:01:05 * gundalow waves
19:02:19 <thaumos> #chair gundalow
19:02:19 <zodbot> Current chairs: gundalow thaumos
19:02:25 * thaumos waves back at gundalow
19:02:52 * Qalthos 🌊🌊
19:03:17 <alikins> blorp
19:03:49 <jtanner> hello
19:04:16 <funzo> o/
19:04:31 <thaumos> #chair Qalthos alikins jtanner funzo
19:04:31 <zodbot> Current chairs: Qalthos alikins funzo gundalow jtanner thaumos
19:05:54 <nitzmahone> o/
19:06:30 <thaumos> #chair nitzmahone
19:06:30 <zodbot> Current chairs: Qalthos alikins funzo gundalow jtanner nitzmahone thaumos
19:07:15 <nitzmahone> hrm, this looks like #core_internal West ATM
19:07:23 <thaumos> pretty close
19:08:52 <cognifloyd> I'm here :)
19:08:58 <thaumos> hi there
19:09:01 <thaumos> #chair cognifloyd
19:09:01 <zodbot> Current chairs: Qalthos alikins cognifloyd funzo gundalow jtanner nitzmahone thaumos
19:09:12 <thaumos> I am not sure if bcoca is joining or not
19:09:24 <abadger1999> ola
19:09:25 <jtanner> what's on the agenda?
19:09:28 <thaumos> #chair abadger1999
19:09:28 <zodbot> Current chairs: Qalthos abadger1999 alikins cognifloyd funzo gundalow jtanner nitzmahone thaumos
19:09:54 <thaumos> #topic Add adhoc callbacks to mirror playbook start/stats
19:09:58 <thaumos> #link https://github.com/ansible/ansible/pull/32304
19:10:07 <thaumos> cognifloyd's PR is the only topic for today.
19:10:17 <thaumos> bcoca and him have a discussion in the PR about it.
19:10:23 <thaumos> sivel is for this as well.
19:10:37 <sivel> whoa sorry, hey
19:10:38 <thaumos> @cognifloyd have anything you'd like to add in here as a part of the discussion in the PR?
19:10:39 <sivel> +1
19:10:45 <thaumos> #chair sivel
19:10:45 <zodbot> Current chairs: Qalthos abadger1999 alikins cognifloyd funzo gundalow jtanner nitzmahone sivel thaumos
19:10:54 <thaumos> lol sivel
19:10:55 <sivel> I am in a priority planning meeting, so kinda here-ish, not here-ish
19:10:58 <thaumos> no worries
19:11:03 <nitzmahone> +1
19:11:04 <thaumos> I didn't intend for you to ping in
19:11:09 <nitzmahone> I've needed this for things in the past too
19:11:16 <thaumos> just wanted to make sure you were somewhat a part of the convo still bud
19:11:30 <thaumos> I am +1 for this as well, are we cool with it being in 2.5?
19:11:32 <alikins> I like the idea of 32304
19:11:59 <alikins> haven't tested/tried the impl
19:12:12 <cognifloyd> First: My computer is very slow until this blasted VM finishes shutting down ... :P
19:12:21 <abadger1999> +1
19:13:11 <cognifloyd> Anyway. I think it makes sense to make callbacks work the same for adhoc as for playbooks. Even if we decide not to add v2_playbook_on_start, I hope at least v2_playbook_on_stats makes it in.
19:14:00 <alikins> pr might be more accurately called something like 'Use a implicit playbook for adhoc instead of just a play'. Maybe could go a step or two further in that regard even.
19:14:51 <cognifloyd> At first I had separate callbacks, v2_adhoc_* (I did that so that it'd be completely backward compatible and callback plugins would have to opt in to supporting adhoc callbacks), but bcoca said it'd be better to reuse the v2_playbook callbacks instead of adding new ones.
19:17:31 <cognifloyd> Ultimately I did this because, as @armab pointed out, I'm trying to reuse adhoc commands in another program. For example: `ansible -m setup some.host.com | jq '.some[query].here'`
19:18:57 <abadger1999> So looks like we have 4 +1 and 0 objections from committers.
19:19:07 <jtanner> +1
19:19:21 <jtanner> swoop voting
19:19:33 <thaumos> back to my question though, are we cool with including it in 2.5?
19:19:49 * cognifloyd is, but that's probably obvious.
19:19:50 <thaumos> I assume yes due to all the +1's, but I like to trust but verify
19:19:58 <bcoca> +1 to the events, just not sure the playbook object makes sense as is (no real feedback, just a feeling ... must thingk aobut it)
19:19:59 <jtanner> if it passes tests before then, sure
19:20:13 <cognifloyd> Yeah, the playbook object felt weird.
19:21:02 <bcoca> cognifloyd: as for the json output, you might be able to once my 'globalize callbcks' PR gets in
19:21:10 <bcoca> though you might need |head in verbose mode
19:21:33 <alikins> cognifloyd: what does using adhoc that way gain you over 'ansible-playbook -i some.host.com -e "some_param=foo" | jq ...' ?
19:21:41 <abadger1999> thaumos: yah, +1 from me for 2.5
19:21:58 <cognifloyd> Almost. Right now, the json callback (with globalize callbaccks) would still only work if there was an on_stats or something similar at the end.
19:22:26 <thaumos> so, for the topic, it looks like that we're good to merge.
19:22:35 <thaumos> should we continue the rest of the discussion in #ansible-devel?
19:22:37 <bcoca> alikins: json callback does most display work on stats, rest of methods just gather data for final output
19:22:38 <thaumos> or here?
19:23:11 <thaumos> We don't have anything else on the agenda, unless dag shows up for his item
19:23:37 <bcoca> open floor?
19:23:51 <thaumos> #topic Open floor
19:23:53 <thaumos> yeah
19:23:54 <thaumos> it is
19:23:57 <bcoca> https://github.com/ansible/ansible/pull/32275
19:24:20 <thaumos> #topic playbook dir option
19:24:28 <thaumos> #link https://github.com/ansible/ansible/pull/32275
19:24:36 <thaumos> +1 from me on this
19:24:44 <bcoca> ^ for console/adhoc since neither of them have a playbook dir and cannot pick up extra group/host/vars plugins vars
19:25:06 <bcoca> also 1st step for allowing include_role to work in adhoc
19:25:42 <bcoca> and includes in general, for when they need templates/ etc
19:30:04 <abadger1999> CLI Option?  sure, looks nice.  Could you write an integration test too?  Should be an easy script to write.
19:30:13 <abadger1999> (but not blocking the initial PR for the test)
19:30:26 <sivel> I just had someone complaining that ansible-inventory didn't pick up play relative group_vars, so I'd say +1 it oculd be useful
19:30:44 <bcoca> sivel: yep, aslo added to that
19:31:39 <jtanner> +1 ... "inventory" is an amalgamation of vars from various places and all should be considered
19:32:36 <jtanner> include_role in adhoc is an interesting topic though
19:32:41 <bcoca> that was part of it, its basically a stopgap to a 'project' concept
19:33:04 <cognifloyd> project concept?
19:33:16 <bcoca> jtanner: include_tasks also ... import_playbook ... i'm tempted to give specific .. just use asible-playbook warning
19:33:42 <bcoca> cognifloyd: most people treat their ansible layout as a 'project' and expect the top level dirs to be meaningful to playbooks all the way down
19:33:49 <bcoca> test/playbooks/x.yml
19:33:56 <bcoca> look in group_vars/stuff
19:33:59 <bcoca> or roles/stuff
19:34:54 <cognifloyd> ok
19:37:31 <thaumos> ok so vote
19:37:48 <thaumos> merge playbook dir option +1, -1
19:38:00 <thaumos> +1
19:38:37 <abadger1999> +1
19:39:41 <cognifloyd> +1 (from a non-committer *shrugs*)
19:40:50 <alikins> +1
19:41:20 * thaumos captured jtanner's +1
19:41:32 * thaumos and sivels
19:41:35 <bcoca> merged
19:41:44 <alikins> sorta rebirth of Ansiblefile
19:41:57 <bcoca> alikins: ?
19:41:57 <jtanner> lulz ... 'while you were out voting, i was merging'
19:42:04 <bcoca> https://github.com/ansible/ansible/pull/32280
19:42:19 <thaumos> pssash whatever
19:42:20 <bcoca> jtanner: nah, was waiting for any negative, pressed merge after YOUR +1
19:42:22 <thaumos> lol
19:42:30 <thaumos> bcoca ninja merged it anyways
19:42:36 <thaumos> after the first +1 came in
19:42:40 <thaumos> I saw it happen
19:42:43 <bcoca> bcoca merged commit 95b31a3 into ansible:devel  38 seconds from now
19:42:49 <bcoca> ^ i merged it in my tardis
19:42:55 <thaumos> exactly
19:43:02 <thaumos> k, next open floor
19:43:06 <jtanner> so by "globalize", you mean bring it up to the cli layer and pass it down?
19:43:07 <thaumos> #topic Open Floor
19:43:08 <bcoca> ^ new PR above
19:43:21 <thaumos> missed that, thanks bcoca
19:43:24 <bcoca> jtanner: in part, also means 'hijacking' most of display methods and allow callbacks to controll them
19:43:34 <thaumos> #topic Globalize callback
19:43:35 <bcoca> i also update json callback as example
19:43:41 <thaumos> #link https://github.com/ansible/ansible/pull/32280
19:44:00 <bcoca> not ready for merge, jsut want to make sure everyone is OK with my approach
19:44:04 <bcoca> its really hackish
19:44:22 <sivel> I haven't really reviewed the approach, just a couple of quick things I noticed
19:44:30 <bcoca> but its only way i saw to do this in w/o rewriting every call to display
19:45:17 <bcoca> also preserves current behaviour unless user/callback specifically wants to override it
19:46:10 <bcoca> https://github.com/ansible/ansible/pull/32280/files#diff-b433171851ee6e9ab6c4f98f233678b5R100 <= the 'hackish' part
19:46:13 <abadger1999> bcoca: I'd rather implement that using a more standard oop pattern
19:46:20 <abadger1999> but the idea is okay.
19:46:37 <bcoca> abadger1999: he, expected that, im open to suggestions, just got this working in 5mins (to my suprise)
19:46:43 <abadger1999> <nod>
19:47:11 <abadger1999> bcoca: <nod> I'll look and make a suggestion and how to change it
19:47:19 <bcoca> i had never had an object 'copy and paste' methods from another
19:47:34 <alikins> I like the idea of #32280 but not really the current impl
19:48:15 <bcoca> im not even looking at this as permanent solution, but quick way to enable the functionality while we work on a more thorugh design
19:48:37 <bcoca> i just feel it is currently very hackish, want to see what other alternatives we can work with
19:49:06 <alikins> abadger1999/bcoca: 'I'd rather implement that using a more standard oop pattern'  - yup
19:49:46 <bcoca> this breaks every pattern i ever though possible i'm 'side classing' from 2 objects w/o having inheritance
19:51:06 <bcoca> im both horrified and proud of said hack ...
19:51:30 <abadger1999> heh.  Yeah, I think this is a has-a pattern
19:51:49 <abadger1999> Just instead of monkey patching, call callback.v() if it exists.
19:51:54 <bcoca> abadger1999: but reversed, display has a callback and then steals it's method for tiself
19:52:35 <abadger1999> Might want to allow callback to have v undefined too.
19:52:38 <bcoca> abadger1999: tried that but got very slow during debug
19:52:42 <abadger1999> for some definition of undefined
19:52:45 <abadger1999> Hmm...
19:52:53 <bcoca> abadger1999: if not defined, it is not overriden
19:53:00 <bcoca> also avoids expenisve hasattr calls
19:53:20 <abadger1999> bcoca: yeah but I mean... it would be nice for the callback baseclass to define all of the methods that are special to callback plugins
19:53:24 <bcoca> once 'set' its all direct calls, all the expense is on initial setup
19:53:36 <abadger1999> So it would be nice for it to have a v() method that does not override display.
19:53:45 <bcoca> abadger1999: agreed, again, this was quick fix, and callbcaks used to NOT have all display functions
19:53:53 <alikins> re-implement display as callbacks[1] seems like a reasonable goal. [1] callback in the general non-ansible specific meaning
19:53:57 <bcoca> abadger1999: ???
19:54:10 <bcoca> doesnt it having a v method mean it SHOULD override display?
19:54:11 <abadger1999> alikins: yeah... I was thinking that might also work
19:54:21 <abadger1999> bcoca: not if v() does nothing.
19:54:34 <bcoca> abadger1999: then you basicaly break existing v?
19:54:44 <abadger1999> bcoca: I'll show you some code so you can see what I'm saying
19:55:11 <bcoca> alikins: this is basically a 'callback' by overriding teh function ref directly vs a 2 step one
19:57:20 <bcoca> abadger1999: the other thing is that there are  callbacks that dont use callbackbase, so i was remiss on making this depend on that
19:57:22 <abadger1999> bcoca: https://paste.fedoraproject.org/paste/7Wxt2xDJR-SZCd5OYqzteQ
19:57:43 <abadger1999> bcoca: but then, of course, we have to change how display decides to use the callback versus its own thing
19:57:57 <bcoca> abadger1999: i see the code, still dont understand how that does not break display
19:58:09 <abadger1999> bcoca: we're coming up on the end of the hour... shall we move this to #ansible-devel
19:58:09 <bcoca> cause method exists, so display would use the 'noop' vs existing
19:58:14 <thaumos> let's continue this in ansible-devel?
19:58:18 * dag is back
19:58:20 <bcoca> k
19:58:21 <thaumos> we have to wrap up for our windows friends
19:58:22 <thaumos> dag...
19:58:24 <bcoca> dag is here, lets leave
19:58:25 <thaumos> wth man
19:58:28 <dag> not sure if there's still time :)
19:58:30 <abadger1999> bcoca: (and the answer is... yes, it breaks display... that's why I say we have to change display as well)
19:58:34 <thaumos> you always come back at the end
19:58:43 <thaumos> not in two minutes dude
19:58:49 <bcoca> abadger1999: eventually, yes, but did not want to go that far at this point
19:58:54 <thaumos> can you be here on thursday?
19:59:01 <dag> yeah, it's halloween :-)
19:59:07 <dag> thaumos: sure, no urgency
19:59:25 <thaumos> ok
19:59:27 <alikins> I have to mention that is starting to make callback/display look more and more like an accidental 'logging' re-implementation
19:59:51 <thaumos> alright folks, closing out for today!
19:59:51 <bcoca> alikins: part  of this was to make 'display.log' load your 'python logging' callback
19:59:54 <thaumos> thanks for the discussions!
20:00:01 <thaumos> #endmeeting