15:01:00 <zbyszek> #startmeeting FESCO (2018-12-10)
15:01:00 <zodbot> Meeting started Mon Dec 10 15:01:00 2018 UTC.
15:01:00 <zodbot> This meeting is logged and archived in a public location.
15:01:00 <zodbot> The chair is zbyszek. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:01:00 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:01:00 <zodbot> The meeting name has been set to 'fesco_(2018-12-10)'
15:01:00 <zbyszek> #meetingname fesco
15:01:00 <zodbot> The meeting name has been set to 'fesco'
15:01:00 <zbyszek> #chair nirik, maxamillion, jsmith, jforbes, zbyszek, tyll, sgallagh, contyk, bowlofeggs
15:01:00 <zodbot> Current chairs: bowlofeggs contyk jforbes jsmith maxamillion nirik sgallagh tyll zbyszek
15:01:03 <zbyszek> #topic init process
15:01:05 <bowlofeggs> .hello2
15:01:05 <zodbot> bowlofeggs: bowlofeggs 'Randy Barlow' <rbarlow@redhat.com>
15:01:10 <zbyszek> .hello2
15:01:11 <zodbot> zbyszek: zbyszek 'Zbigniew Jędrzejewski-Szmek' <zbyszek@in.waw.pl>
15:01:17 <nirik> morning
15:01:46 <zbyszek> afternoon ;)
15:01:51 <sgallagh> .hello2
15:01:52 <zodbot> sgallagh: sgallagh 'Stephen Gallagher' <sgallagh@redhat.com>
15:01:54 <jforbes> .hello2
15:01:55 <zodbot> jforbes: jforbes 'Justin M. Forbes' <jforbes@redhat.com>
15:02:01 <contyk> .hello psabata
15:02:02 <zodbot> contyk: psabata 'Petr Šabata' <psabata@redhat.com>
15:02:17 <zbyszek> Tha's 6, quorum
15:02:22 * contyk will be very distracted for a bit
15:02:47 <zbyszek> #topic #2023 Allow untagging of module builds from rawhide
15:02:50 <zbyszek> .fesco 2023
15:02:51 <zodbot> zbyszek: Issue #2023: Allow untagging of module builds from rawhide - fesco - Pagure.io - https://pagure.io/fesco/issue/2023
15:02:53 <zbyszek> https://pagure.io/fesco/issue/2023
15:03:26 <maxamillion> .hello2
15:03:27 <zodbot> maxamillion: maxamillion 'Adam Miller' <maxamillion@gmail.com>
15:03:45 <zbyszek> I admit I have no idea about this topic, so I'll just follow what others say
15:03:58 <contyk> I think this happens because of a suboptimal tagging behavior in MBS
15:04:02 <bowlofeggs> yeah i also don't really get what's going on here
15:04:10 <nirik> I'm +1, but would like to get the base cause fixed
15:04:26 <contyk> there's a plan to make it more flexible and put it in an external service (in dev)
15:04:39 <maxamillion> so if what Igor said in ticket is true, then yes I'm +1 because if it's an uninstallable artifact, it shouldn't matter
15:04:47 <contyk> yes
15:04:57 <contyk> +1 as well, in those cases
15:05:00 <zbyszek> Right, +1 then too
15:05:18 <bowlofeggs> i'm +1 to removing uninstallable things
15:05:24 <nirik> this particular case was caused by our relase scripts I think... not mbs... but yeah
15:05:25 <jforbes> I am +1 as well
15:05:29 <sgallagh> I think the short-term is that we want rel-eng to have permission to do this as they see fit until we have a better technical solution
15:05:31 <sgallagh> I'm +1
15:05:35 <nirik> s/release/branching/
15:05:55 <contyk> hmm
15:06:34 <maxamillion> wait ... releng doesn't have permissions to take action on build artifacts that need to be released?
15:06:35 <contyk> would love to learn more about that, nirik
15:06:36 <bowlofeggs> i'm with maxamillion, i'd think releng already has this authority
15:06:59 <nirik> there's a rule that we never untag artifacts after they have shipped out in rawhide.
15:08:09 <zbyszek> #agreed: Untagging of the builds is allowed until a longer-term solution is implemented (+6, 0, 0)
15:08:11 <zbyszek> #info: There's a plan to make it more flexible and put it in an external service (in development)
15:08:12 <nirik> contyk: when branching we clone the tag so the branch has everything... right now that includes modules...
15:08:15 <zbyszek> OK?
15:08:41 <sgallagh> zbyszek: LGTM
15:08:43 <zbyszek> #agree Untagging of the builds is allowed until a longer-term solution is implemented (+6, 0, 0)
15:08:47 <zbyszek> #info There's a plan to make it more flexible and put it in an external service (in development)
15:08:50 <zbyszek> Did I get the syntax right?
15:08:53 <contyk> nirik: yeah, let's look into that together latet
15:09:19 <zbyszek> I hope so, I'll just fixup the meeting notes ;)
15:09:20 <nirik> contyk: sure. I think sgallagh and mboddu were going to poke at it, but yeah, it needs likely some more logic or something.
15:09:21 <bowlofeggs> zbyszek: i think the #agreed one works
15:09:44 <zbyszek> There's no feedback unfortunately.
15:09:52 <zbyszek> Anyway, let's move on.
15:09:53 <zbyszek> #topic #2020 Firefox is switching from gcc to clang/2020
15:09:53 <zbyszek> .fesco 2020
15:09:53 <zbyszek> https://pagure.io/fesco/issue/2020
15:09:53 <sgallagh> nirik: Yeah, I'm going to look into that with mboddu after the New Year
15:09:54 <zodbot> zbyszek: Issue #2020: Firefox is switching from gcc to clang/llvm - fesco - Pagure.io - https://pagure.io/fesco/issue/2020
15:09:58 <sgallagh> my dance card is full until then
15:10:17 <sgallagh> zbyszek: It's agreed, not agree
15:10:46 <mboddu> I already commented on the ticket that sgallagh and I will look into it after Dec 31
15:11:11 <contyk> great
15:11:27 <bowlofeggs> sgallagh: i still kinda feel like FPC should get to decide on this before escalating to fesco
15:12:08 <jforbes> I would disagree. The policy doesn't need to change, they are simply asking for an exception.  Though I don't really have a problem with them chiming in first
15:12:11 <bowlofeggs> i'm fine with clang personally, but FPC oversees the guidelines and i think they should have a shot at considering this issue first
15:12:37 <nirik> here is the old ticket: https://pagure.io/fesco/issue/847
15:12:38 <mboddu> bowlofeggs: The ticket with FESCo is just about untagging the modules
15:12:40 <sgallagh> We don't really need to have a custody battle over this, do we?
15:13:27 <sgallagh> bowlofeggs: FPC doesn't *set* policy, they *implement* policy.
15:13:34 <nirik> the thought was that lots of things would switch and llvm would have poor/no support for many things and we would be in a bind
15:13:35 <mboddu> How to handle the the issue is a different case
15:13:39 <bowlofeggs> ah so it look slike this was a fesco policy to begin with, as linked by nirik
15:14:12 <zbyszek> bowlofeggs: where was it linked?
15:14:21 <bowlofeggs> zbyszek: https://pagure.io/fesco/issue/847
15:14:35 <bowlofeggs> apparently this policy was a fesco decision 6 years ago
15:14:37 <sgallagh> nirik: Right, and there's some argument to be made that, six years and endless corporate sponsors later, we might even consider revisiting this.
15:14:55 <nirik> perhaps.
15:15:05 <sgallagh> But I'm not proposing that today :)
15:15:18 <bowlofeggs> we could say something like "packages can use compilers supported by upstream"
15:15:40 <zbyszek> I'd be inclined to just allow this for firefox. If this starts coming up more, we can always ask FPC to change the guidelines.
15:15:41 <sgallagh> bowlofeggs: I actually disagree with that, but it's not the topic on the table today.
15:16:02 <zbyszek> bowlofeggs: I concur with sgallagh here
15:16:04 <sgallagh> Today we just have a specific exception to consider. As I noted in the ticket, I'm +1
15:16:11 <zbyszek> +1 too
15:16:14 <bowlofeggs> i can be +1 to the exception
15:16:22 <jforbes> +1 here
15:16:22 * nirik is +1 to allow firefox and if folks want to revisit, we should/can do that with another ticket.
15:16:31 <sgallagh> nirik++
15:16:31 <zodbot> sgallagh: Karma for kevin changed to 20 (for the current release cycle):  https://badges.fedoraproject.org/tags/cookie/any
15:17:07 <maxamillion> +1 here
15:17:26 <zbyszek> contyk?
15:17:53 <maxamillion> oh ... BZ got an upgrade
15:17:54 <contyk> I abstain
15:17:55 <maxamillion> huh
15:18:37 <zbyszek> #agreed The exception for firefox is granted (+6, +1, 0)
15:18:54 <zbyszek> #topic #2003 Ursa Major (modules in buildroot) enablement
15:18:54 <zbyszek> .fesco 2003
15:18:54 <zbyszek> https://pagure.io/fesco/issue/2003
15:18:56 <zodbot> zbyszek: Issue #2003: Ursa Major (modules in buildroot) enablement - fesco - Pagure.io - https://pagure.io/fesco/issue/2003
15:19:42 <zbyszek> I think we should go with the proposal as stated by contyk in the ticket
15:20:16 <zbyszek> When an implementation is ready, there'll probably be more details to figure out
15:20:23 <zbyszek> https://pagure.io/fesco/issue/2003#comment-544193
15:20:42 * sgallagh agrees
15:21:21 * nirik doesnt like telling everyone to hit koji
15:21:37 <contyk> yeah, a compose would be better
15:21:53 <maxamillion> I don't like it, I'm inclined to propose to reject ursa-major entirely until it can be implemented correctly
15:21:55 <contyk> or just mirror the Koji repos
15:22:01 <maxamillion> contyk: not an option
15:22:01 <zbyszek> Wait, "I think the only workable option here are buildroot composes."
15:22:03 <nirik> but would be vast
15:22:23 <contyk> maxamillion: can you define correctly?
15:22:30 <maxamillion> contyk: which statement?
15:22:48 <contyk> maxamillion: the one just now
15:23:24 <zbyszek> contyk, nirik: I don't understand nirik's comment. My reading of contyk's proposal was that a compose would be created, and koji would not be hit.
15:23:41 <bowlofeggs> zbyszek: the compose wouldn't be mirrored aiui
15:23:47 <nirik> if thats the case... I think our mirrors might be unhappy.
15:23:54 <contyk> zbyszek: that would indeed be my proposal
15:24:01 <nirik> I suppose everything would be hardlinkable...
15:24:05 <zbyszek> contyk: please clarify "that"
15:24:17 <contyk> zbyszek: creating a buildroot compose
15:24:30 <nirik> also it would increase compose time
15:24:31 <contyk> apologies, just getting off a plane
15:24:31 <maxamillion> contyk: this needs to be handled in a way that doesn't introduce undue burden on packagers, releng, or the infra team ... right now it's just passing the buck because the solution doesn't satisfy the requirements
15:25:47 <zbyszek> maxamillion: Do you have a (general) proposal how this would be handled better?
15:26:13 <contyk> maxamillion: so you refuse a, I think, a workable proposal as incorrect without having anything specific in mind as a replacement
15:27:05 <maxamillion> zbyszek: I don't know enough about the implementation to really offer technical architecture change suggestions, I only know the impact that's being outlined in the proposal and I don't consider them acceptable
15:27:09 <jforbes> I agree that dumping more on releng isn't a workable proposal
15:27:11 <bowlofeggs> i also think having packagers use the koji server could be a problem, fwiw
15:27:38 <maxamillion> contyk: I don't agree that creating more work for others is a "workable proposal"
15:27:44 <maxamillion> bowlofeggs: +1
15:27:46 <maxamillion> jforbes: +1
15:28:25 <nirik> I'm not sure this is much more work for releng... the aspect I am not sure about is the buildroot compose and other resources it would consume... time, disk, mirror space, etc.
15:28:32 * contyk doesn't consider setting up one more variant as a huge burden in terms of work
15:28:40 <bowlofeggs> nirik: and you think that mirroring these buildroot repos would not be "popular" among mirror admins?
15:28:57 <nirik> well, it's Everything + (some) modules
15:29:12 <contyk> I think nirik suggested we could point to artifacts in the other variants last time
15:29:20 <contyk> but maybe I misunderstood
15:29:34 <nirik> yeah, if we can do that I am ok with this. I thought that wasn't workable?
15:29:36 <zbyszek> nirik: is "Everything" required? I thought it'd be just the "+ some modules" part.
15:30:00 <contyk> nirik: you just need to make it look like a flat repo
15:30:04 <nirik> zbyszek: perhaps. would need pungi work though?
15:30:23 <zbyszek> So the question is who would be implementing this?
15:31:08 <nirik> we could ask pungi developers...
15:32:02 <nirik> if we can solve the resource issues around the buildroot repo(s), I'd be in...
15:32:16 <nirik> proposal: talk to pungi devs, revisit next week?
15:32:39 <zbyszek> Also, can we get an impact assessment from infra?
15:32:44 <contyk> if we can just point at the right rpms in the other variants (most in everything, some in modular), it would be mostly empty a provide an environment similar to koji
15:32:51 <maxamillion> zbyszek: +1
15:32:58 <nirik> zbyszek: in what sense?
15:33:34 <zbyszek> nirik: in the sense if the additional repo would be a burder to distribute
15:33:35 <nirik> It depends on how it's implemented how much disk/etc it would use
15:33:59 <zbyszek> nirik: well, it'd be just a bunch of symlinks, so probably not that much actual disk space
15:34:08 <nirik> if it was just repodata, thats nothing... if it was just modules that are in the buildroot, thats not much more, if it's everything... thats a burden
15:34:19 <nirik> we cannot use symlinks.
15:34:28 <nirik> it would be hardlinks or copies
15:34:40 <zbyszek> Right, hardlinks.
15:35:18 <zbyszek> But actually no new data, right? By defininition all those rpms must already be visible somewhere else.
15:35:36 <nirik> sure, but they would take up 100,000 inodes or whatever.
15:35:49 <contyk> it could be new data
15:36:29 <contyk> but very little of it if any - that's the buildroot only modules that would require FESCo ack
15:37:00 <zbyszek> OK, so who can talk to pungi devs (and whoever else should be included)?
15:37:15 <zbyszek> nirik, contyk?
15:37:17 <nirik> I can file a ticket... and everyone can watch/add to it?
15:37:40 <zbyszek> Works for me.
15:37:45 <contyk> +1
15:38:04 <maxamillion> +1
15:38:14 <jforbes> +1
15:38:26 <bowlofeggs> +1
15:38:41 <sgallagh> +1
15:39:03 <zbyszek> +1 too
15:39:07 <zbyszek> #action nirik to file a ticket with pungi
15:39:07 <zbyszek> #agreed We'll gather more information about the required changed to pungi and additional resources required for this and revisit next week (+7, 0, 0)
15:39:34 <zbyszek> #topic Next week's chair
15:39:57 <contyk> I'll be free
15:40:14 <sgallagh> This is my last FESCo meeting of 2018
15:40:41 <zbyszek> Is this the last meeting before after the elections?
15:41:02 <jforbes> next week
15:41:42 <zbyszek> So, contyk or sgallagh, who wants to chair the next meeting less?
15:41:53 <contyk> I can
15:41:59 <contyk> sgallagh can't
15:42:03 <zbyszek> #action contyk will chair next meeting
15:42:24 * zbyszek has trouble figuring out if "this" is this, or "the one beeing discussed"
15:42:26 * contyk nods
15:42:28 <sgallagh> Right, sorry.
15:42:38 <sgallagh> s/this/today/
15:42:41 <contyk> it's this
15:42:42 <zbyszek> Thanks contyk.
15:42:44 <contyk> clearly
15:42:49 <zbyszek> #topic Open Floor
15:43:09 <contyk> I suspect we won't have quorum anyway
15:43:47 <zbyszek> Anyone?
15:43:50 <jforbes> One item for open floor, the koji outage next month
15:43:59 <zbyszek> It's pretty long, no?
15:44:34 <jforbes> More of an infra thing, but the concern is, what if there is a critical security update during that long window? Is there a way to force a build/exclude s390?
15:44:44 <zbyszek> #info OUTAGE: Koji system 2019-01-11 -> 2019-01-14
15:45:16 <zbyszek> #info https://pagure.io/fedora-infrastructure/issue/7429
15:45:45 <zbyszek> jforbes: ExcludeArch ?
15:46:01 <nirik> https://pagure.io/pungi/issue/1095 <- pungi ticket
15:46:07 <jforbes> I am certainly not proposing that we just keep things going, it would be a nightmare to catch up S390, but there should be an exception process
15:46:15 <nirik> please expand on anything I got wrong. ;)
15:46:17 <jforbes> zbyszek: Yes, but the way it read, koji as a whole would be down
15:46:31 <nirik> koji will not be down, only s390x builders.
15:46:53 <jforbes> Okay, that alleviates my concern
15:46:57 <zbyszek> Hmm, but does this mean that we should discourage non-noarch builds?
15:47:09 <jforbes> zbyszek: They should fail
15:47:19 <zbyszek> Ah, OK.
15:47:19 <jforbes> unless you ExcludeArch
15:47:26 <nirik> they would queue up... waiting
15:47:37 <nirik> well, until they hit the 2 day timeout probibly.
15:48:31 <zbyszek> OK, anything else to mention here?
15:48:41 <jforbes> No, that addresses my concern
15:48:41 <zbyszek> If not, I'll close in a minute.
15:48:42 <bowlofeggs> it's a super computer so it should be able to do a super amount of builds when it comes back online
15:49:09 <jforbes> bowlofeggs: I think by legal definition, an iPhone is a super computer these days
15:49:18 <bowlofeggs> ha
15:49:39 <zbyszek> proposal: replace s390x by an iphone, call it quits
15:49:57 * sgallagh is not entirely sure if zbyszek is joking
15:50:02 <jforbes> zbyszek: It should at least be android
15:50:39 <zbyszek> #endmeeting