15:03:48 #startmeeting Fedora QA meeting 15:03:48 Meeting started Mon Aug 17 15:03:48 2015 UTC. The chair is adamw. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:03:48 Useful Commands: #action #agreed #halp #info #idea #link #topic. 15:03:53 #meetingname fedora-qa 15:03:53 The meeting name has been set to 'fedora-qa' 15:03:56 #topic Roll Call 15:04:02 morning folks! who all else is here? 15:04:18 * tflink is around 15:04:21 * satellit listening 15:04:30 hi there 15:05:10 * kparal pokes pschindl 15:05:44 try the whip 15:06:41 welp, that should be enough to get started 15:06:47 #topic Previous meeting follow-up 15:06:48 * pschindl is here 15:07:14 I'm seeing no actions from previous meeting, so that's OK 15:07:18 #info no actions from previous meeting 15:07:30 #topic Flock outcomes 15:08:13 I figured those of us who were both at Flock and here can give those of us who weren't at Flock but are here (and those who weren't at Flock and aren't here, via the minutes...) an idea of what went on and what future directions we got out of it 15:08:25 sound cool? any specific questions before we start? 15:08:53 * kparal reads the sentence for the third time 15:09:17 * adamw continues his campaign to write for The Times in 1863 15:09:28 me and tim are gonna tell y'all what happened at flock 15:09:29 how's that? 15:09:49 great 15:09:53 =) 15:10:08 so anything in particular anyone wanted to ask? 15:10:37 are we changing to the BSD kernel? 15:11:23 no, we decided on the hurd. 15:11:30 yay 15:11:46 also, we'll be rewriting all tools in PHP with a lesstif GUI 15:12:21 I'm afraid any further news can't beat that 15:12:23 * tflink waits for phoronix article saying "fedora cannot tolerate more insults, moves to hurd and starts to rewriting tooling in php" 15:12:28 =) 15:13:15 any other big change decisions made on flock? 15:13:35 not really 15:13:46 this wasn't one where there was a 'big' idea like Fedora.next 15:14:18 there was some discussion on moving forward the rings concept in some sense, and we all bailed on that in 20 minutes because oh god life's too short, but from what I heard after there weren't really any big decisions made 15:15:11 anything QA specific? 15:15:38 bodhi 2.0 is coming on Wednesday 15:15:43 right 15:15:46 changes to both blockerbugs and libtaskotron are needed 15:16:00 well, allegedly. we have our suspicions as to whether they'll make it. but definitely soon 15:16:13 the client was just released this morning 15:16:20 i can confirm that blockerbugs noworky with the new bits 15:16:28 there was an amusing point where lmacken was like "oh no, blockerbugs won't need any changes, it's totally cool...(looks at code for 20 minutes)...okay, so blockerbugs needs changes" 15:16:33 I'll play with it tomorrow. is that an update to python-fedora, or a new package? 15:16:51 the git repo is updated, threebean did a scratch build for el7 and rawhide 15:16:57 * tflink looks for links 15:17:21 rawhide scratch build for new python-fedora: http://koji.fedoraproject.org/koji/taskinfo?taskID=10731280 15:17:38 F22 build would be great 15:17:41 el7 scratch build for new python-fedora: http://koji.fedoraproject.org/koji/taskinfo?taskID=10731255 15:17:53 it's easy enough to grab the .src.rpm and fire an f22 scratch build... 15:18:02 * tflink is going to be busy working on blockerbugs today, will be asking for testers before long 15:18:16 * adamw has to spend some family time but will try to be around for testing at some points 15:18:22 I'll try to rebuild, hopefully the deps will be satisfied 15:19:05 there's also koji 2.0 coming at some point which might affect some things (it'll certainly affect fedfind), but the time scale on that is longer 15:19:21 there's koji 2.0 in development? cool 15:19:33 is it going to integrate taskotron as bodhi 2 does? 15:19:36 http://sched.co/3rPn 15:19:58 kparal: probably not, it's an update that mostly comes from the RH fork of koji AFAIK 15:20:08 sure, i volunteered you to write it =) I don't know, that wasn't th focus of the talk. and yeah, it's coming from RH's fork, brew 15:20:19 a bit off topic, are there recordings of flock talks somewhere? 15:20:30 recordings were made, but i don't know if they're up anywhere yet 15:20:31 I didn't see any blogpost about it on fedora planet 15:20:34 ok 15:20:41 #info many Flock talks were recorded, but not sure if they were uploaded yet 15:20:58 #info Bodhi 2.0 is planned to land this week and will require changes to blockerbugs and libtaskotron 15:21:11 #action tflink to work on Bodhi 2.0 related tooling changes 15:21:22 the patch to libtaskotron has been submitted, haven't looked at it in detail yet 15:21:58 adamw: blockerbugs needs changes because even though it uses the BodhiClient from python-fedora, it bypasses the query() method and just uses the underlying ProxyClient.send_request to hit the bodhi1 API endpoint directly ;( 15:22:10 * lmacken will submit a patch 15:22:46 thanks 15:22:51 tflink: submitted where? I haven't noticed anything in Phab 15:23:01 kparal: pull request in bitbucket 15:23:05 ah 15:23:13 too...many...mechanisms... 15:23:15 I didn't see it until lmacken told me about it, either 15:23:41 yeah, too many mechanisms 15:23:55 +1 for moving git repos into phab - no pull requests :) 15:23:57 we'll need to do something about it in the future. thanks for mentioning it 15:24:09 :) 15:25:09 OK, so aside from that, the big thing i wound up talking about a lot was improving integration with releng particularly in the compose process 15:25:17 tflink: that patch seems to affect only unreleased code - not even in dev yet 15:25:34 kparal: move to #fedora-qa? 15:25:38 ok 15:25:45 thanks 15:27:18 it definitely seems like everyone's thinking in the same CI-ish direction: i'd say the executive summary of the basic idea is we'll have nightly composes that look the same as TC/RC composes, better metadata and fedmsg messages from the compose process, and that'll allow some form of gating using our existing automated test stuff 15:27:38 there was similar talk about improving the loop between releng and the taskotron depcheck test, for repo composes 15:28:23 yeah, that seems like some relatively low-hanging fruit that'll help a lot 15:28:43 * tflink has a lot of tickets to file from Flock 15:29:03 so far as the compose stuff goes, my basic read was that in a lot of ways we're waiting on the new pungi, which implements the 'nightlies look like real composes' and 'better metadata' stuff 15:29:17 the idea is to run depcheck on the rawhide buildroot's repodata every time that changes (~ every 6 minutes) and check for borked repotrees 15:29:38 * Corey84 peeks in 15:29:42 hi corey! 15:29:48 o/ 15:29:53 don't let all the tooling talk scare anyone off, we'll have other topics in a bit :) 15:30:16 tflink: yeah, i think that's something we can sensibly do right now and should really try 15:30:21 adamw, I need to learn that too its all good 15:30:45 6 minutes that seems quick 15:30:54 adamw: yep, agreed. jskladan and I talked a bit about how to easily do that before we left Rochester 15:31:19 Corey84: people do Rawhide builds *all the time*. so every few minutes, koji runs a task that updates the buildroot with the latest finished build 15:31:24 the current plan is to do it a bit hacky for now to see if it's useful. if it is indeed helpful, we'll work on making that change less hacky 15:31:37 Corey84: so if you want to build a new library version then rebuild all its deps, you don't have to wait a day for the new library build to land first 15:32:04 Corey84: the idea is that instead of only doing depcheck daily when the actual repo compose happens, we do it for every run of the buildroot generation instead 15:32:05 makes sense but lordy 15:32:29 like a --ref on dnf update on a "stable" install 15:32:30 * tflink shrugs - that's what automation is for :) 15:32:35 tflink: and then ten years from now we'll read the comment that says # we're going to make this less hacky next week if it's useful 15:32:39 lol 15:32:44 adamw: pretty much, yeah 15:32:47 = 15:32:49 =) 15:33:04 but it's either the hacky bits or wait until disposable clients are done 15:33:09 hacky bits it is! 15:33:24 disposable clients ? the buildroots? 15:33:47 Corey84: disposable clients for taskotron 15:33:49 #info For Rawhide, we will aim to have depcheck run on every buildroot generation task instead of daily on the repo composes, so we can see when things break faster and more precisely identify the culprits 15:34:22 I'm hoping to have that in place in a few weeks - will depend on the fallout from bodhi2 and a few other things, though 15:34:30 tflink++ 15:34:30 adamw: Karma for tflink changed to 5: https://badges.fedoraproject.org/tags/cookie/any 15:34:50 yay! cookies! 15:35:07 even if I suspect they were unintentional, never look a gift cookie in the mouth ... 15:35:10 :-D 15:35:21 =) no, not unintentional 15:35:56 so, on the compose side of things, there's a lot of stuff i/we *could* do basically through fedfind 15:36:31 fedfind already kinda exists as a fig leaf for all the inconsistencies in the compose process and I can add various bits to it to effectively figure out all the info we *ought* to be getting from releng 15:36:52 * tflink would love to see the output from a pungi4 run 15:37:05 tflink: I've seen it somewhere and now i can't find the damn link 15:37:17 * tflink is asking releng 15:37:28 dgilmore should have one though 15:37:41 iirc there's a compose.json file with lots of useful info about what images the compose generated and where they are 15:37:58 right now it's kinda incomplete because it doesn't include a lot of the images that are built in koji, but aiui they plan to fix that. 15:38:16 basically, it does the stuff fedfind does, only from the correct end and hence far less stupidly. 15:38:55 the other bit i'd like to have is composedb, at least as I understand it, which would cover the bit of fedfind that finds the actual composes. 15:39:35 anyhow, yeah - tl;dr is we *could* do a bunch of stuff right now, but it'd all be somewhat hacky and probably need to be redone fairly soon anyway, so i'm not inclined to go *too* far down that path 15:40:20 there are bits it makes more sense to work on, that maybe aren't as immediately visible but are more likely to be needed in the long term 15:40:46 the specific thing i'm thinking of there is taskotron/openQA integration 15:41:19 #info we could do various work around improving compose test integration already, but it'd all be hacky and need to be replaced with the new pungi, so we're inclined to wait on that 15:42:02 on the taskotron front - folks seem excited about our plans around dist-git style tasks 15:42:17 oh yeah, want to cover that before i talk taskotron/openQA integration? 15:42:35 and indeed, people seemed to like the idea 15:42:36 adamw: either way 15:42:56 most of the other taskotron bits are somewhat low level 15:43:10 OK 15:43:15 is there anything really new-from-flock there? 15:43:22 i got the impression flock business there was mostly evangelism 15:43:38 yeah, there's a low level change that we'll be making soon 15:43:51 it'll be significant to devs but other folks shouldn't notice 15:43:57 to taskotron devs, rather 15:44:16 we're planning ot go forward with ditching TAP 15:44:35 ok 15:44:40 maybe best for a qa-devel meeting, then? 15:44:43 jskladan and I spent some time working out the new yaml-based format and writing out examples 15:44:44 yep 15:44:58 cool 15:45:24 * kparal noticed some of that already 15:45:40 in fact openQA/taskotron integration is kinda similar, and we need to get to test days, so maybe i'll try and show up for a qa-devel meeting too or mail the list, if that's OK 15:46:01 * randomuser` talked to tflink about buildbot, has started epel7 packaging and identified a handful of deps that need packaging 15:46:06 WFM 15:46:07 very short version: we need an improvement to openQA before taskotron could schedule tasks in it (or a bad workaround), and i'm gonna try and work on that. 15:46:25 in fact i already did, but didn't manage it yet. :) 15:47:47 i'm also going to do some low-hanging work on compose stuff using the current process: basically i'm writing a thing that should generate a daily 'compose report' for Rawhide and Branched composes, listing significant missing images and images that have gone missing (or shown up) since the previous compose, and maybe some other interesting bits. it's not too hard or time-consuming so i figured it was worthwhile. 15:48:24 oh! and there was one other flock thing I can think of - the talk from someone outside our team, that we weren't sure about 15:48:30 'mastering the art of open source testing', or something 15:48:53 it turned out to be by a guy who does Mozilla testing, and he was basically outlining his personal approach to organizing exploratory testing 15:49:16 * Corey84 & 15:49:50 he had some interesting ideas, and it's not an area we really focus on heavily, so it might be worth watching a recording (if there was one and it goes up), and/or his slides 15:50:01 i'll try and find his slide deck and post it to the list later 15:50:19 #action adamw to find slide deck from Flock 'exploratory testing' talk and post it to the list 15:50:39 ok, anything else from flock before we move along? 15:52:32 nothing I can think of, no 15:53:43 rg 15:53:44 rgr 15:53:58 #topic l10n and NetworkManager test days 15:54:17 so as a quick reminder - we have l10n (that's localization, basically translation testing) Test Day tomorrow, and NetworkManager Test Day on Thursday 15:54:36 #info l10n Test Day is tomorrow 2015-08-18: https://fedoraproject.org/wiki/Test_Day:2015-08-18_L10N 15:54:48 #info NetworkManager Test Day is Thursday 2015-08-20: https://fedoraproject.org/wiki/Test_Day:2015-08-20_NetworkManager 15:55:14 I wanted to remind folks to come along and check if anything needed doing, and there's a more specific issue for the l10n Test Day too 15:55:43 the l10n team have set up to use the 'testdays' webapp, but apparently that might be going away (or have gone away already) with the new cloud changes 15:55:52 tflink? 15:56:11 the old app is on the old cloud which was scheduled to go away a week ago or so 15:56:22 if it's still up, I'm not sure how much longer that will be true 15:56:35 we have a new testday app written but haven't gotten to deploying it yet 15:57:16 it's still up at least *right now* 15:57:21 if it makes it to tomorrow we'll be OK 15:57:23 i think so, yes 15:57:36 * tflink doesn't know what the plans for killing off the old cloud are 15:57:45 since it's still up right now, what i'll do as a backup plan is 'export' the wikicode right now so we can stick the empty tables back into the page and update the instructions if necessary 15:57:53 last I heard, some bits of jenkins were blocking the kill-off process 15:58:01 go jenkins! 15:58:05 who would be involved in that process? 15:58:29 #info l10n test day is set up to use the testdays webapp, which is living on an old cloud that will be going away soon (should have gone away already) 15:58:31 * tflink checks 15:58:48 #action adamw to store the empty results wikicode for the l10n test day so we can fall back on manual result entry if necessary 15:58:58 * adamw has improved wikitcms to parse test day results, but not enter them yet :/ 15:59:28 i think nirik and pingou would know what's going on with jenkins and the old cloud 16:01:16 OK, I stored the empty tables. 16:01:33 i'll see if we can contact nirik/pingou later to see whether we'll get lucky 16:02:13 i suspect that even if the decommissioning was planned for tomorrow, if we asked nicely it could be put off for a couple days 16:02:28 but honestly, i think that people are pretty focused on getting bodhi2 ready for wednesday 16:02:34 right 16:02:53 so, other than that, any action items for the test days? 16:03:14 l10n needed a live image, but it sounds like that's been built, might just need the page updating to link to if it hasn't been alreadyu 16:04:16 at some point, we need to get the instance migrated to the new cloud 16:04:18 pschindl built it, I uploaded it, apater said he will take it from there 16:04:34 sorry, still thinking testdays app 16:05:18 kparal: awesome 16:05:28 tflink: what's the hold up with deploying the new one? 16:06:39 thanks pschindl kparal :) 16:06:50 adamw: spare cycles 16:06:55 =) 16:06:58 the usual answer 16:07:27 i suspect that I'll be busy with blockerbugs and libtaskotron updating/testing for the next couple days 16:07:46 we might be able to get it done later this week, not sure 16:07:53 * tflink didn't think it was a high priority 16:08:42 no, not really, i was just curious 16:08:57 i'd want to have *some* testdays app up, but having the old one up would be good enough. 16:09:16 welp, we're nearly 10 mins over, so let's do a very quick: 16:09:18 #topic Open floor 16:09:29 any other business, folks? any Alpha stuff crop up while we were at Flock, or anything? 16:09:48 I should probably not ask, but is there going to be a blocker bug meeting today? 16:10:07 oh - for the record, Beta TC1 is scheduled for 08-25 16:10:45 I haven't noticed any Alpha fires 16:10:46 um. we do have blockers to review, but roshi's not around and I might suffer domestic violence if I sit here running meetings all day 16:10:50 it sounded like roshi was planning on a meeting today 16:10:55 oh, he is? funsies 16:11:01 * adamw may have to be absent for some/all 16:11:09 * satellit I have lockups with intel graphics in f23 workstation 16:11:16 otherwise we could do it tomorrow or later in the week 16:11:25 satellit: yep, we saw - sounds like the thing to do there is report an X or kernel bug. 16:11:26 * kparal is fine with that 16:11:33 k 16:11:58 we have 6 proposed beta blockers and 7 proposed final, so we should definitely do one. 16:12:07 tflink: can you kick 'current' on blockerbugs over to beta? 16:12:16 adamw: sure 16:12:23 adamw: usual reminder that you have perms to do that, too 16:12:39 usual reminder that i forgot :P 16:12:51 so much forgetting 16:12:57 we're obviously not elephants 16:13:12 i'm sure it was nothing to do with that bottle of freeze exception 16:13:32 because we're dwarfs digging for gems in a mountain 16:13:51 oh god yes, the dwarf mountain 16:13:54 wwoods needs to write a book 16:14:20 alrighty, we're over time, so setting a very short quantum fuse 16:14:23 thanks for coming, folks! 16:14:33 am sorry I just saw the l10n testdayapp discussion (late for the meeting :(). So is it okay to go with http://209.132.184.193/testdays/show_event?event_id=28 ? 16:15:06 blockerbugs/current is beta now 16:15:32 apeter: basically, we don't know :) 16:15:39 apeter: but i stored the empty wikicode in case it goes away 16:15:55 apeter: if it goes away between now and tomorrow we can fall back to manual result entry 16:16:08 apeter: i'll try and ask the relevant folks later today whether it's likely to stick around until tomorrow 16:16:21 adamw, okay :) 16:16:28 if we do wind up being able to use the app we'll definitely want to save the data from it regularly during the day in case it suddenly disappears 16:16:42 you know how to export the data? it's really easy 16:16:52 No I dont.. need to check 16:16:54 okay 16:17:10 go to 'admin interface', click 'export testday', and paste in the metadata URL (that's https://fedoraproject.org/wiki/L10N/TestdayApp/F23 ) and click Submit 16:17:18 and you get the wikicode with whatever results are currently submitted 16:17:31 it doesn't throw anything away afaik so it's safe to do it as often as you like 16:17:57 i'll try and be around to do it too, but best if two of us are :) 16:18:09 ah okay. got it.. thats pretty easy :) 16:18:12 yup :) 16:18:34 adamw, alright noted. I will be around too :) 16:18:37 yay 16:18:40 alrighty, thanks again everyone 16:18:40 :D 16:18:52 #endmeeting