14:33:56 <dgilmore> #startmeeting RELENG (2017-06-26)
14:33:56 <zodbot> Meeting started Mon Jun 26 14:33:56 2017 UTC.  The chair is dgilmore. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:33:56 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
14:33:56 <zodbot> The meeting name has been set to 'releng_(2017-06-26)'
14:33:56 <dgilmore> #meetingname releng
14:33:56 <zodbot> The meeting name has been set to 'releng'
14:33:56 <dgilmore> #chair dgilmore nirik tyll sharkcz masta pbrobinson pingou puiterwijk maxamillion mboddu Kellin
14:33:56 <zodbot> Current chairs: Kellin dgilmore masta maxamillion mboddu nirik pbrobinson pingou puiterwijk sharkcz tyll
14:33:59 <dgilmore> #topic init process
14:34:17 <nirik> morning
14:34:24 * masta is here
14:34:56 * dustymabe too
14:35:09 <dgilmore> #topic #6847 Splitting fedora-repos into edition-specific variants (F27 timeframe)
14:35:15 <dgilmore> https://pagure.io/releng/issue/6847
14:35:17 * sharkcz is here
14:35:19 <dgilmore> lets get started
14:35:42 <dgilmore> I am very against splitting the package into multiple.
14:35:53 <dustymabe> dgilmore: +1
14:36:02 <dgilmore> as we would have to update them all and in the past we failed at that which is why we have the packaging we now have
14:36:24 <dustymabe> i like the 'inject a variable' approach
14:36:26 <dgilmore> but we should be able to figure out a way to give mattdm what he wants
14:37:06 <dgilmore> it will also need coordinating with mirrormanager to make sure its all understood on that side
14:38:48 <dgilmore> does anyone have any opinion?
14:39:35 <tyll> I agree that injection a variable sounds like the best way forward
14:39:53 <dgilmore> hey tyll :) welcome
14:40:01 <nirik> but dnf folks didn't want to do that? or ?
14:40:42 <nirik> ah they just don't have it implemented... they are not against it.
14:40:44 <dgilmore> #agreed releng feels that the most appropriate way forward is with the injection of a variable into the url
14:40:57 <dgilmore> nirik: it just may take longer
14:41:01 <nirik> yeah
14:41:07 <dgilmore> but I think its the best way forward
14:41:19 <maxamillion> .hello maxamillion
14:41:20 <zodbot> maxamillion: maxamillion 'Adam Miller' <maxamillion@gmail.com>
14:41:27 <maxamillion> sorry I'm late, attempting to multi-task
14:42:39 <dgilmore> #topic #6822 using include statements in pungi-fedora
14:42:43 <dgilmore> bad maxamillion
14:42:50 <dgilmore> https://pagure.io/releng/issue/6822
14:43:08 <dgilmore> dustymabe: did you do up a PR?
14:43:31 <dgilmore> not sure why the meeting keyword was not removed after we had discussed
14:43:51 <dustymabe> dgilmore: not yet - koji hadn't been synced to stg when I was originally looking at this
14:43:58 <dustymabe> so none of my test pungi runs would work
14:44:10 <dgilmore> okay
14:44:11 <dustymabe> it is now synced to stg so I'm unblocked but haven't got back around to it yet
14:44:30 <maxamillion> dustymabe: indeed
14:44:31 <dgilmore> #info waiting on a PR from dustymabe to show how he would like to see things changed
14:44:36 <maxamillion> dustymabe: bah, tab-fail ... sorry
14:44:39 <maxamillion> dgilmore: indeed :)
14:44:51 <dgilmore> i believer no one was agianst it
14:45:03 <dgilmore> we were just not totally sure what was being proposed
14:45:09 <dgilmore> believe even
14:46:04 <dgilmore> #topic #6799 [koji] Add configuration for module build service content generator
14:46:11 <dgilmore> https://pagure.io/releng/issue/6799
14:47:29 <dgilmore> I think the only change I would like to propose here is that mbs also uploads teh logs on teh json generation
14:47:36 <dgilmore> not sure how invasive that would be
14:47:52 <dgilmore> threebean: any idea?
14:48:55 * threebean reads up
14:50:40 <threebean> eh, should be possible.
14:50:43 <threebean> but will take a code change.
14:51:02 <dgilmore> threebean: cool
14:51:09 <threebean> will file an issue to track it.
14:51:12 <threebean> can we add it in later?
14:51:21 <dgilmore> I think it wou,ld be usefult to have as part of the auditability
14:51:29 <dgilmore> I think so
14:51:29 <threebean> (for now at least, the logs are accessible on mbs-backend01 in the journal)
14:51:53 <dgilmore> threebean: thats not public and open and transparent :D
14:52:21 <threebean> yup, yup.
14:52:25 <threebean> but it is auditable!
14:53:15 <dgilmore> #agreed that under the condition we add the logs creating the json to the upload we can add the content generator
14:53:56 <threebean> oh - wait.  we need to implement this first before we can turn on the CG?
14:55:28 <threebean> not having the CG is effectively blocking (or slowing) bodhi dev work atm.
14:55:59 <dgilmore> threebean: no, you need to agree to prioritize doing it and agree to doing it as quickly as possible
14:56:05 <threebean> gotcha.
14:56:22 <threebean> we can queue it up for next sprint (starting wednesday).
14:56:48 <dgilmore> cool thanks
14:57:32 <dgilmore> #topic broken dependencies
14:57:53 <dgilmore> hey tyll so you had some things to raise over cleaning up of broken deps?
14:59:00 <tyll> dgilmore: thx, yes, usually we retire all pkgs with broken deps for the final freeze
14:59:41 <tyll> however, there are still a lot of pkgs with broken deps (about 87)
15:00:29 <tyll> So I was wondering if we have also another option to maybe just block the packages in a way that they will not be in the Everything repo but could still be added as an update later
15:01:37 <nirik> well, we could untag all builds of that package so there are none in the tag...
15:01:38 <dgilmore> the only way to do that would be to block them in the f26-compose tag
15:01:41 <tyll> here is a list of all pkgs and the archs they have broken deps in: https://paste.fedoraproject.org/paste/AtaI4nsPqBDMvw7jm1z5IA
15:01:43 <nirik> but then they would get GCed
15:02:28 <dgilmore> tyll: what is the arches?
15:03:39 <tyll> dgilmore: in the repo for that arch there is a broken dependency
15:04:03 <dgilmore> tyll: so does that mean they are not broken on other arches?
15:04:15 <tyll> the pkgs where it is just some arches should be "easily" fixable by adding some excludearch/exlusivearch and making them non-noarch
15:04:26 <dustymabe> maxamillion: origin is in that list
15:04:32 <tyll> dgilmore: yes, some of them for example require docker which is not available everywhere
15:04:51 <maxamillion> dustymabe: yeah, it shouldn't be
15:04:59 <maxamillion> dustymabe: I don't know why it's getting listed there
15:05:05 <dgilmore> tyll: asterisk: x86_64, armhfp, ppc64le, aarch64, ppc64, i386
15:05:15 <dgilmore> I thought jsmsaid he was working to fix that
15:05:26 <dgilmore> maxamillion: origin: ppc64
15:05:36 <tyll> the info is from this report: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org/thread/NR6YEE577F5N22S6BII53LC5LIWFQMJZ/
15:05:38 <dgilmore> maxamillion: it apparently has broken deps on ppc64
15:05:45 <dgilmore> maxamillion: there is no docker on ppc64
15:05:51 <maxamillion> dgilmore: ah
15:06:00 <maxamillion> dgilmore: ppc64 isn't a primary arch, why is this an issue?
15:06:21 <dgilmore> maxamillion: with the merge of the arches we pick up things like that now
15:06:30 <dgilmore> maxamillion: likely it should have a ExcludeArch ppc64
15:06:58 <dgilmore> #info We should try and get the deps sorted out,
15:07:21 <dgilmore> maxamillion: you should be getting daily emails that the deps are broken
15:07:49 <dgilmore> threebean: do you have ideas on how we could make the broken deps notifications more effective
15:08:17 <tyll> dgilmore: asterisk-dahdi seems to be a broken subpackage on all archs
15:08:18 <maxamillion> dgilmore: alright, will do
15:08:39 <dgilmore> tyll: but the dep was getting added back
15:08:51 <maxamillion> dgilmore: I have to excludearch for armv7hl right because of a golang compiler bug also :/
15:08:52 <tyll> dgilmore: maybe it is not yet in stable?
15:09:18 <dgilmore> tyll: https://koji.fedoraproject.org/koji/buildinfo?buildID=897526 its in updates-testing still
15:10:07 <dgilmore> java-1.8.0-openjdk: x86_64, i386
15:10:10 <dgilmore> that seems bad
15:10:23 <tyll> dgilmore: java will be fixed with the next push
15:10:44 <tyll> dgilmore: at least there is an update for the missing dep
15:10:54 <dgilmore> tyll: okay
15:11:08 <dgilmore> I think we should try push people to fix things
15:11:09 <tyll> https://bodhi.fedoraproject.org/updates/FEDORA-2017-e898ddd8ca
15:11:40 <dgilmore> contyk: modularity-testing-framework: ppc64
15:12:09 * contyk reads the backlog
15:12:20 <tyll> the upcoming notification of the pkgs being retired got at least some pkgs being fixed
15:12:31 <tyll> (at least in testing)
15:13:13 <dgilmore> contyk: that package has broken deps on ppc64 and is at risk of being kicked out
15:13:35 <contyk> dgilmore: because of golang? it's not mine, I don't know much about it
15:13:36 <dgilmore> tyll: so an option would be to block in the -compose tag
15:13:42 <contyk> but I can prod its owners
15:13:52 <dgilmore> but that would not make the package be removed from the nightly composes
15:15:44 <dgilmore> #info We need to figure out how to deal with broken deps better, from notifications to cleaning up
15:16:21 <dgilmore> a bunch of them seem to be packages that should add "ExcludeArch: ppc64"
15:16:38 <dgilmore> because of docker not being supported on ppc64
15:16:52 <tyll> I have some time later to fix at least the pkgs that only require exludearch updates
15:17:32 <nirik> but we will need to get the fixed packages through freeze.
15:17:44 <dgilmore> nirik: indeed
15:17:59 <dgilmore> they can be proposed as FE's
15:18:21 <dgilmore> adamw: how would you feel about 80 odd updates to fix broken deps as FE's
15:18:24 <dgilmore> ?
15:19:41 <tyll> when we just block the pkgs in the compose we should have a deadline until they get properly retired
15:21:00 <dgilmore> sure
15:22:07 <dgilmore> do we have a proposal?
15:23:40 <tyll> 6 weeks is the usual deadline we give for orphaned pkgs so it sounds like a good candidate
15:25:08 <dgilmore> #agreed we will block all the packages in f26-compose, and give 6 weeks for things to be cleaned up. If the updates get stable or FE's for final we will unblock to get them included in F26
15:25:13 <dgilmore> sound good?
15:25:21 <tyll> yes
15:25:25 <tyll> +1
15:25:39 <nirik> +1
15:26:04 <dgilmore> sounds like a plan
15:26:22 <dgilmore> #action to make sure mboddu and Kellin are aware of the agreed upon process
15:26:31 <dgilmore> #topic open floor
15:26:39 <dgilmore> does anyone have anything to bring up?
15:27:17 <dgilmore> if not will wrap up?
15:27:37 <dgilmore> #endmeeting