19:26:03 #startmeeting ansible core irc public meeting 19:26:03 Meeting started Tue May 28 19:26:03 2019 UTC. 19:26:03 This meeting is logged and archived in a public location. 19:26:03 The chair is bcoca. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:26:03 Useful Commands: #action #agreed #halp #info #idea #link #topic. 19:26:03 The meeting name has been set to 'ansible_core_irc_public_meeting' 19:26:17 #topic https://github.com/ansible/ansible/pull/56817 19:26:17 hi! 19:26:23 * bcoca waves 19:26:30 i appoligize, we had runover for post mortem 19:26:35 Hi 19:26:37 MOAR MEETINGS 19:26:38 o/ 19:26:44 our grave overfloweth 19:26:44 sdoran: ssssh!~ 19:27:00 jtyr? 19:27:35 ^ pr above adds loop_control toggel to how we notify handlers in loops, right now we notify all handlers if 'task' is changed, toggle allows for doing it 'per item' 19:27:43 brb 19:27:46 i.e notify: {{ handlers[item] }} 19:28:05 i dislike name, but im good with feature 19:28:08 any comments? 19:28:51 it seems useful 19:28:59 +1 feature 19:29:53 yeah, looks ok to me- +1 for having it under loop_control instead of a new top-level "knob" 19:30:16 +1 for the feature 19:30:28 k, i have some notes on implementation, but as long as we are all happy with feature, will give jtyr green light 19:30:45 #topic https://github.com/ansible/community/issues/466#issuecomment-496615437 19:31:15 +1 for the feature 19:31:19 shipit 19:31:28 Adding to my original comment on the agenda, abadger1999 has suggested that we allow unit tests to allow satisfy the test requirement. 19:31:39 (also +1) 19:31:40 + 1to test requirements 19:31:57 +1 to moar tests 19:32:30 do I understand 3. correctly that if you even create a really small bugfix PR (maybe just a typo in the docs) for a module which has no tests, that sanity tests will fail? 19:32:41 felixfontein: Yes 19:32:56 that will be fun ... people will avoid fixing those ? 19:33:23 Now's the time to change that if that is considered undesirable behavior. 19:33:30 i would make exception to #3 on 'small patch' 19:33:34 without tests, in many cases, they're just breaking things they didn't know about 19:33:41 this is probably discouraging some people to do (or complete) little bugfixes / typo corrections 19:33:51 for x in (y) changed to for x in (y,) is obvious fix 19:33:53 What's the reasoning for that? That seems a bit confusing to invoke a failure if a module is on the ignore list. 19:33:55 that's one where I wish we had a non-blocking result from sanity that could say "you really oughtta add some tests" but not prevent it from being automerged 19:34:02 also +1 for requiring tests. AWS modules have been following those rules for a while, but have had an exception for urgent small patches. 19:34:17 nitzmahone: I can do that with ansible-test -- we just need to get the bot to pick up those messages. 19:34:35 Yeah, we'd talked about it previously, just assumed it was probably never gonna happen 19:35:19 So two variations on what I originally proposed: 1) allow unit tests to satisfy the requirement 2) make changes to ignored modules a warning instead of an error 19:35:33 works4me 19:35:38 me too 19:35:57 updated comment to reflect 19:36:00 sounds good 19:36:00 I just worry that item three could have the effect of discouraging obvious and maybe good bugfixes from getting in if we throw up the "please write tests" barrier. 19:36:12 sdoran: reread now 19:36:51 k, lets vote on 'ammended' text 19:36:52 +1 19:37:01 +1 19:37:04 +1 19:37:18 +1 19:37:20 +1 19:37:44 * bcoca should ask Bercow on palamentary proceedures 19:37:52 k, seems like it passes 19:38:05 #topic https://github.com/ansible/community/issues/466#issuecomment-494938855 19:38:09 piou? 19:38:13 pilou? 19:38:41 ah +1 for previous (sorry was busy in kitchen) 19:38:44 hmm, default always in 'choices' even when 'None'? which we use as indicator of 'unset' 19:39:05 i would say 'no' , cause if user sets to None .. its not 'unset' ... 19:39:41 sounds like omit? 19:39:42 users already have 'omit' magic to 'unset' options 19:39:48 cyberpear: jinx! 19:39:49 Yeah, seems like some undefined behavior with the intersection of `required: true` among other things 19:40:25 im going to table, i really dont understand the question, as phrased it seems really wrong 19:40:36 since pilou is not here, pushing till next meeting 19:40:50 is some module using required:no and default:None with choices? 19:41:03 probably plenty 19:41:03 default: None is implied 19:41:09 default: none is 'the default' 19:41:16 I think this was related to the PR we saw in triage https://github.com/ansible/ansible/pull/56892 19:41:26 so most all with choices and required: no are effectively doing that today 19:42:53 I don't like modules that specify string 'null' as a choice but it seems like this could alter modules doing that 19:43:00 yeah, that is wrong 19:44:00 k, i have another item i added, but didnt notify interested party, for botmeta, so going to skip and 'open floor' 19:44:04 #topic open floor 19:44:11 (the limitation I've seen with omit is that you can't set_fact the value) 19:44:27 cyberpear: its for use on the option itself 19:44:33 which you can always do 19:44:58 botmeta? 19:45:09 jtanner: skipped since i didnt notify user yet 19:45:30 is it on the issue? i might be able to resolve out of band 19:45:48 https://github.com/ansible/ansible/pull/56858 19:46:09 oh okay 19:46:22 https://github.com/ansible/ansible/pull/55473 <= also wanted to tell him tha tupdating 40 modules at a time is not good idea 19:47:15 https://github.com/QijunPan is a ghost 19:47:43 probably huawei employee that created account just to add the modules 19:48:14 we've not been picky about setting folks as maintainers so far, so honestly i'd just merge it 19:48:24 i can do so if desired 19:48:42 go4it then 19:48:50 and multi-module PRs are sufficiently controlled by the bot with a message and needs_revision 19:49:07 50 is normal thershold 19:49:13 40 .. too close for my comfort 19:49:19 heh 19:49:30 looks like just lots of unrelated bugfixes 19:49:53 which is why i paused, if it was 'same fix and repeat' .. diff issue 19:50:12 modifying 40 files whlie replacing same code, is one thing ... doing tons of unrelated fixes ... 19:50:25 ah, i see. not 40+ new modules 19:50:31 just touching 40+ modules 19:50:47 yep 19:50:57 it won't get automerged anyway =) 19:53:37 hello bcoca and the rest, could we see https://github.com/ansible/ansible/pull/49432 about check mode markers 19:53:40 i noted the issue in the PR 19:53:44 and merged it 19:54:40 * nitzmahone not opposed to such a thing, but would need to be (at least) per-task, since tasks can opt in/out 19:54:53 it is, thought i reviewed this morning 19:55:45 ah, didnt check in comment 19:56:16 zer0_her0: anything else on that or you were just looking for review? 19:56:44 actually i see the support:core label and I am wondering if this means that we should discuss this here 19:57:02 you can, but really means core team member must review/merge 19:57:35 I think reviews are accepted for this PR. Is anything else to be done for merge ? 19:57:38 anyone against idea of 'dry run' captions in check mode? 19:57:48 zer0_her0: i had changes 19:57:50 Tests? 19:58:05 Since this is adding a Core feature. 19:58:11 callbacks/default.py 19:58:31 bcoca, I did the changes (version_added) 19:58:46 ^ you just got asked for tests also 19:59:04 and i also said changelog was missing 19:59:05 what should we add for this ? 19:59:30 ... i lost the other comments ... 19:59:40 (changelog is ok), I mean for test, what kind of tests ? 19:59:59 that toggle works as intended 20:00:29 current output unchagned and 'expected' results in diff scenarios x) --check-mode x) check_mode:true on task/block/role 20:00:46 zer0_her0: Look at test/integration/targets/callback_default/runme.sh 20:00:55 new folder like tests/integration/targets/dryrun, with a runme.sh file in there that executes the playbook with the new feature and checks that it worked 20:01:06 ok, I'll c heck this 20:01:18 That should give you an idea of what to do. You can ask for further help in #ansible-devel if you get stuck. 20:01:36 and on the the ansible-devel mailing list 20:01:40 #endmeeting