18:59:01 <elbrus> #startmeeting 18:59:01 <MeetBot> Meeting started Wed Jun 24 18:59:01 2020 UTC. The chair is elbrus. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:59:01 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic. 18:59:05 <elbrus> pochu, adsb, jcristau, kibi, ivodd, jmw, ginggs, Sebastinas, noahm ^ 18:59:12 <Sebastinas> o/ 18:59:15 <ginggs> o/ 18:59:20 <pochu> o/ 18:59:37 <elbrus> #topic Admin 18:59:52 <elbrus> #info Previous minutes: http://meetbot.debian.net/debian-release/2020/debian-release.2020-04-22-19.00.html 19:00:12 <elbrus> #info ginggs had an action for a patch to britney 19:00:35 <elbrus> (don't think that happened yet) 19:00:40 <ginggs> no progress 19:00:46 <elbrus> no worries 19:00:57 <elbrus> #info elbrus had an action for release architecture qualification; e-mail sent 19:01:11 <elbrus> and on the agenda for later 19:01:26 <elbrus> #info elbrus had an action to remove the copy disclaimer of the rc_policy; done 19:01:43 <elbrus> #info noahm had an action to review rc_policy 19:01:49 <elbrus> ;) 19:02:38 <elbrus> nothing yet I guess? 19:02:40 <noahm> I read it, anyway. 19:02:49 <noahm> no feedback atm 19:02:54 <elbrus> so, done 19:03:05 <elbrus> #topic Transitions 19:03:30 <elbrus> how's stuff going (I haven't followed much) 19:03:33 <elbrus> ? 19:03:33 <pochu> Sebastinas: you're up :) 19:03:40 <Sebastinas> boost is about to finish. 19:03:44 <Sebastinas> kig should migrate tonight 19:04:09 <Sebastinas> icu is blocked on chromium 19:04:31 <adsb> o/ 19:04:37 <adsb> (was eating) 19:04:45 <Sebastinas> there was an upload recently but that is broken with current ffmpeg 19:04:55 <pochu> what's kig? 19:05:09 <Sebastinas> Last reverse depenency of boost1.67 19:05:22 <pochu> ah cool 19:05:45 <elbrus> the one with the upstream not ready yet? 19:05:50 <pochu> there's haskell ongoing too, but those need sourceful uploads due to the doc abi bump 19:06:01 <Sebastinas> elbrus: Yes, python support was temporarily disabled. 19:06:09 <elbrus> o, good 19:06:30 <Sebastinas> There are some smaller ones without issues. 19:06:33 <elbrus> is the haskell team picking it up? 19:06:48 <elbrus> I pinged some people on #d-devel a while back 19:07:04 <Sebastinas> Except for flint (which wasn't coordinated) 19:07:22 <elbrus> seems much better than several weeks ago (haskell) 19:07:55 <pochu> elbrus: yeah I think they are. it looks in a better shape than since I poked them last week 19:08:28 <elbrus> they asked me for wb rights 19:08:46 <elbrus> which I don't think I can/should be handing out 19:08:57 <pochu> no 19:08:58 <elbrus> I pointed them at wanna-build team 19:09:05 <pochu> that comes up every now and then 19:09:33 <ginggs> can i ask why it could be a problem? 19:09:57 <pochu> iirc it's been agreed that an automated solution would be accepted, for haskell, ocaml, rust, go... but nobody has worked on it yet 19:10:01 <elbrus> I guess: with power comes responsibility 19:10:21 <pochu> (well we the list of packages needing binnmus is already automated, but not the wb side) 19:11:14 <elbrus> pochu: I have always assumed you already had scripts in place to be able to do all those transitions in the right order 19:11:31 <pochu> nah :( 19:11:45 <elbrus> Sebastinas: also doing things by hand? 19:12:03 * elbrus started on some script based on the ben output (web scraping) 19:12:33 <elbrus> but it has been a while since Sebastinas and ginggs became active on that front 19:12:54 <Sebastinas> For ordering I rely mostly on ben. 19:13:07 <Sebastinas> And then feed a dumb script with source package names. 19:13:23 <elbrus> right, but I guess with a bit more integration, we can generate the wb commands from ben output 19:13:32 <pochu> would be better to have ben write some machine readable files. it already makes one for tracker.debian.org, we'd just need some other information 19:13:44 <elbrus> exactly 19:14:14 <elbrus> I don't mean ben creating the wb commands, but just better output than the web site 19:15:02 <elbrus> mehdi_ said he would take bugs filed against ben and try to improve it 19:15:22 <elbrus> so, I guess we could just ask on that front 19:16:08 <elbrus> anybody willing to take an action on this front? 19:16:53 <Sebastinas> I can take a look at it. 19:17:38 <elbrus> #action Sebastinas see if an improvement bug for ben can be created to provide better machine parsable output 19:17:54 <elbrus> something like that? 19:18:21 <elbrus> anything more on transitions? 19:18:28 <pochu> not from me 19:18:43 <elbrus> nodejs? 19:19:03 <elbrus> blocked on loads of autopkgtest 19:19:08 <Sebastinas> elbrus: Yes, sounds good. Some yaml file with the data could already be enough to help on the front 19:19:10 <elbrus> not seeing much happening 19:20:07 * elbrus will file some more regression bugs for that transition 19:20:23 <elbrus> next topic? 19:20:38 <elbrus> #topic Release architecture qualification follow-up 19:20:52 <elbrus> thread started at https://lists.debian.org/debian-release/2020/05/msg00089.html 19:21:35 <elbrus> as said, I can mostly reuse the mail from nthykier from last release 19:21:46 <elbrus> does anybody have more to add? 19:22:28 <pochu[m]> I wouldn't like to remove architectures for the sake of removing them 19:22:48 <elbrus> me neither 19:23:15 <elbrus> and YunQiang Su promised faster mips* 19:23:33 <elbrus> mipsel 19:23:39 <pochu[m]> If they have hw and users and we can reasonably support them then I'd keep what we can. Of course if there are valid concerns from the relevant teams then we'd hear them 19:23:54 <pochu[m]> So yeah that call makes sense 19:24:31 <elbrus> the point of course is that "reasonably" depends on who you ask 19:25:04 <elbrus> e.g. bunk raised an issue for s390x as the last big endian release architecture 19:25:05 <pochu[m]> We had a discussion about automating the process. Defining some criteria that would need to be met, and if it wasn't for some amount of time the arch would be removed. Tfheen volunteered to help but then buster got released and it probably became low prio 19:25:53 <elbrus> jcristau: said last time we discussed this something along the lines of "difficult decisions are difficult" 19:26:19 <pochu[m]> That itself doesn't sound like an issue 19:26:33 <elbrus> a bit of automation helps for making it more objective, but the criteria remain ... 19:27:34 <pochu[m]> In the end we may need to make an executive decision if we don't have a better way 19:27:43 <pochu[m]> Just like it happened in Vancouver :p 19:28:19 <kibi> :) 19:28:47 * elbrus has heard the term vancouvered, not as a great name 19:28:55 <pochu[m]> Whatever we do, some people will be unhappy 19:28:59 <elbrus> that 19:29:29 <pochu[m]> So I'd say let's send that call to relevant teams, and consider and weigh the pros and cons 19:29:38 <elbrus> #action elbrus send release architecture qualification call to the related teams 19:29:42 <elbrus> there 19:30:01 <pochu[m]> thanks! 19:30:06 <noahm> What are the relevant teams in this case? 19:30:18 <ansgar> The mail refernce said mipsel has the lowest ram limit? What is it out of curiosity? 19:30:31 <pochu[m]> Wanna build, dsa, security, kernel 19:30:44 <elbrus> DSA, security, gcc team, glibc team 19:30:52 <noahm> makes sense, thanks 19:31:07 <elbrus> let's join those 19:31:13 <pochu[m]> and toolchain yeah 19:31:48 <elbrus> #topic Should autopkgtest regressions be RC for all release architectures? 19:31:52 <elbrus> ginggs ^ 19:32:08 <elbrus> https://salsa.debian.org/release-team/release.debian.org/-/merge_requests/7 19:33:00 <elbrus> no ginggs anymore? 19:33:09 <ginggs> still here :) 19:33:20 <pochu[m]> Where are we running autopkgtests these days? 19:33:21 <elbrus> take it away 19:33:29 <elbrus> arm64 and amd64 19:33:29 <pochu[m]> Or is that theoretical? 19:33:43 <pochu[m]> So it's a no op atm? 19:33:43 <ginggs> it's theoretical 19:33:48 <elbrus> we have some ppc64el workers, but not enough for unstable-to-testing 19:33:58 <ginggs> as in the test are not automated 19:34:14 <elbrus> you can run it on porterboxes 19:34:23 <ginggs> but if someone finds a regression (or it comes from ubuntu) 19:34:26 <elbrus> and Ubuntu has more archs of course 19:34:31 <pochu[m]> I wouldn't change that right now. I'd rather get them running first, then add them as blocking when reasonably ready 19:34:40 <ginggs> and it can be reproduced e.g. on a porterbox 19:34:55 <ginggs> then can a bug be filed, and should it be RC? 19:35:08 <elbrus> bug can be filed 19:35:15 <pochu[m]> E.g. if I don't think it would be a great idea if mipsel gets added and lots of packages become insta RC 19:36:06 <ginggs> well, if someone finds a mipsel FTBFS, it is RC 19:36:09 * elbrus likes the idea of adding archs first and *then* declaring regressions RC 19:36:28 <elbrus> but that's because the package has to be able to build 19:36:41 <elbrus> to be in the release 19:36:46 <ginggs> on the flipside, often an autopkgtest regresses, and then we find out later it's actually a FTBFS 19:37:06 <ginggs> i.e. the autopkgtest was an early warning 19:37:10 <elbrus> I am very much not against fiing bugs 19:37:24 <elbrus> maybe even important 19:37:30 <elbrus> but not yet RC 19:37:54 <elbrus> but if other people like the idea of ginggs, I'm fine either way 19:38:28 <ginggs> just to reiterate, i'm only talking about regressions 19:38:54 <elbrus> as opposed to what we state about arm64 and amd64 19:38:58 <pochu[m]> I'm fine with that 19:39:00 <ginggs> whereas, if i understand correctly, any failure (old or new) is RC on arm64 and amd64 19:39:00 <elbrus> those must pass 19:39:36 <pochu[m]> They block testing migration atm right? So that effectively makes them RC bugs 19:40:20 <elbrus> regressions, yes 19:40:28 <elbrus> for amd64 and arm64 19:40:47 <ginggs> right, but policy says all failures are RC? 19:40:53 <elbrus> new tests that fail are also "regressions" 19:41:30 <elbrus> yes, we agreed on that (all failures are RC) for amd64 and arm64 19:41:33 <ginggs> what about old tests that always failed? 19:41:44 <elbrus> on amd64 and arm64: RC 19:41:57 * elbrus has been filing several bugs like that 19:42:08 <elbrus> just not a whole lot 19:42:09 <ginggs> understood, but those don't block migration? 19:42:18 <elbrus> no 19:42:41 <elbrus> just like RC bugs that are not regressions also don't block migration 19:43:25 <ginggs> ooo, i think i get it now 19:44:43 <elbrus> so, you want to change your proposal, or it stands? 19:45:00 <ginggs> i'm clearer now on amd64 and arm64, but still think regressions on other release architectures should be RC 19:45:28 <pochu[m]> Hmm yeah regressions would be fine 19:45:53 <elbrus> as said, I can live with it 19:46:02 <elbrus> anybody else? 19:47:36 <elbrus> #agreed autopkgtest *regressions* are RC on release archs, merrily failing on anything but arm64 and amd64 is not 19:48:07 <elbrus> ivodd: are you in? 19:48:16 <kibi> /C 19:48:20 <kibi> gah. :) 19:48:32 * elbrus though about discussing "point release process" 19:49:06 <elbrus> if not, we'll leave it till next time 19:49:22 <elbrus> #topic AOB 19:49:52 <elbrus> anybody? 19:50:35 <elbrus> #topic Next meeting 19:50:46 <elbrus> #info Next meeting is 22nd July at 19:00 UTC (import into your calendar via https://release.debian.org/release-calendar.ics) 19:51:20 <elbrus> #endmeeting