15:03 <slangasek> #startmeeting
15:03 <meetingology> Meeting started Thu Jun 18 15:03:22 2015 UTC.  The chair is slangasek. Information about MeetBot at http://wiki.ubuntu.com/meetingology.
15:03 <meetingology> 
15:03 <meetingology> Available commands: action commands idea info link nick
15:03 <slangasek> [TOPIC] Lightning round
15:03 <slangasek> $ echo $(shuf -e barry doko stgraber jodh bdmurray slangasek caribou infinity sil2100 robru cyphermox pitti)
15:03 <slangasek> doko jodh caribou robru bdmurray pitti barry cyphermox sil2100 slangasek stgraber infinity
15:04 <slangasek> doko: hi, around?
15:05 <pitti> is jodh even still officially in the team?
15:05 <slangasek> hmm, and we have a change to this list since last time...
15:05 <barry> slangasek: i think you need to update your shuf :)
15:05 <slangasek> next time should be:
15:06 <slangasek> echo $(shuf -e barry doko stgraber bdmurray slangasek caribou infinity sil2100 robru cyphermox pitti
15:06 <slangasek> yep
15:06 <slangasek> nd no caribou here today
15:06 <slangasek> robru:
15:07 <robru> one sec
15:07 <robru> * CI Train:
15:07 <robru> - added script for scanning overlay PPA versions and comparing to wily archive; ensure that matching version numbers have matching binary hashes
15:07 <robru> * CI Train Charm:
15:07 <robru> - stop clobbering creds with empty files if creds misconfigured
15:07 <robru> - add relation for bileto
15:07 <robru> * Bileto:
15:07 <robru> - fixed some postgres-specific bugs
15:07 <robru> * Bileto Charm:
15:07 <robru> - add basenode support
15:07 <robru> - add relation for ci-train
15:07 <robru> * Train mojo Spec:
15:07 <robru> - added relation between ci-train and bileto
15:07 <robru> - various iterations on postgres configuration
15:07 <robru> (done)
15:08 <sil2100> robru the magician, charming with his mojo!
15:08 <robru> heh
15:09 <slangasek> bdmurray:
15:09 <slangasek> robru: clobbering creds with empty files> I don't suppose that helps the postgres issues at all?
15:10 <bdmurray> continued to work on improvements to apport-retrace to check launchpad for packages not available in archive or on ddebs.u.c
15:10 <bdmurray> worked with pitti to get my apport use-lp branch merged (done!)
15:10 <bdmurray> worked on hacking in ppa support to apport-retrace
15:10 <bdmurray> pushed an apport branch (with a PPA hack) for the retracers to use
15:10 <bdmurray> updated daisy config for armhf 15.04 retracers for overlay ppa
15:10 <bdmurray> submitted RT to have daisy / apport updated in staging
15:10 <robru> slangasek: no that's unrelated, that's on the ci-train side. #is doesn't have creds stored the same way we do in staging and so the charm was clobbering live creds with empty files.
15:10 <bdmurray> investigation into the right way to add ppa support to apport
15:10 <bdmurray> SRU verifications of python-pip bug fixing LP: #1324391 (fail, then pass)
15:10 <ubottu> Launchpad bug 1324391 in python-pip (Ubuntu Trusty) "pip 1.5.4 import an invalid dependencies " [High,Fix committed] https://launchpad.net/bugs/1324391
15:10 <bdmurray> discussed with mvo python-apt origin information and PPAs
15:10 <bdmurray> discussion with slangasek regarding overlay-ppa and old packages
15:10 <bdmurray> investigation into a phased-updater OOPS
15:10 <bdmurray> short week as I was in Rochester!
15:10 * doko is a bit late ...
15:10 <bdmurray> ✔ done
15:11 <slangasek> doko: hi, your turn
15:11 <sil2100> slangasek: that thing was probably related to the train outage we had on Friday when the cred files went missing
15:11 <slangasek> ok
15:12 <doko> five me 5min
15:12 <slangasek> alright then. pitti:
15:12 <pitti> systemd:
15:12 <pitti> - create script to build debs for current systemd upstream trunk (into PPA, and running all autopkgtests locally), for a more proper trunk CI
15:12 <pitti> - fix various build system and udev regressions in trunk
15:12 <pitti> - enable net.ifnames persistant network names by default
15:12 <pitti> autopkgtest:
15:12 <pitti> - set up ProdStack/Canonistack with IS for cloud-based autopkgtest running, and create autopkgtest cloud runner PoC
15:12 <pitti> - various small improvements to adt-run
15:12 <pitti> - revive armhf/ppc64el worker nodes after outage
15:12 <pitti> misc:
15:12 <pitti> - review and land bdmurray's apport branch for downloading older sources/debs from LP
15:12 <pitti> - build and test fresh vivid langpacks
15:12 <pitti> - postgresql: prepare new upstream security/bug fix releases for all supported Ubuntu releases
15:12 <pitti> - reproduce and analyze misconfigured/broken encrypted swap in LVM install mode (#1453738); in progress
15:12 <pitti> - various bug fixes in open-scsi, udisks
15:12 <pitti> plan:
15:12 <pitti> - systemd 221 is around the corner; unlike the previous two versions this should now be a breeze to land thanks to all the new CI
15:12 <pitti> - continue working on cloud-based autopkgtest runner
15:12 <pitti> - find some time to look into networkd integration with resolvconf and ifup.d/
15:12 <pitti> ^D
15:13 <barry> doko or me?
15:13 <slangasek> barry: go ahead, it hasn't been 5 minutes yet :)
15:13 <pitti> barry: 5 mins aren't over yet
15:13 <barry> sounds like a monty python sketch...
15:13 <barry> short week due to pto
15:13 <barry> debuntu: debian bug #776026; python-nose-exclude 0.2.0-4; zope.security 4.0.3-1; virtualenv 13.0.3-1 (ongoing), wheel 0.24.0-2; python-pip 1.5.4-1ubuntu3 for trusty;
15:13 <ubottu> Debian bug 776026 in src:wheel "wheel: please make whl files reproducible" [Wishlist,Fixed] http://bugs.debian.org/776026
15:13 <barry> si: 2.5.1 to the overlay; struggling with landing 3.0.1 in wily via train (and getting derailed)
15:14 <barry> started working on the python3.5 transition plan
15:14 <barry> other: upgrade bisecting and other debugging for LP: #1464365
15:14 <ubottu> Launchpad bug 1464365 in xorg (Ubuntu) "Resolution is incorrect; screen mostly occluded" [Medium,New] https://launchpad.net/bugs/1464365
15:14 <barry> --done--
15:14 <doko> - openjdk-9 fixes for ppc64el and arm64
15:14 <doko> - updated GCC 5 transition plan
15:14 <doko> - openjdk maintainer interviews
15:14 <doko> - preparing uploads for gcc 4.8, 4.9 and 5.x release candidates
15:14 <doko> - started doing some merges
15:14 <doko> - finally gcc-4.8 and binutils SRU's in trusty, uploaded cross packages
15:14 <doko> - uploaded python3.4 SRU's
15:14 <doko> (done)
15:15 <sil2100> cyphermox: ping
15:15 <cyphermox> sil2100: I'm on vacation, not actually there :)
15:15 <slangasek> cyphermox is still on vacation
15:15 <sil2100> Ah, ok :)
15:15 <slangasek> sil2100:
15:15 <sil2100> - Landing team work, silo coordination, preparing landing e-mails
15:15 <sil2100> - Prepare OTA-4 release notes
15:15 <sil2100> - OTA-4 release:
15:15 <sil2100> * Copying images and setting auto-phasing for the period of approx one day
15:15 <sil2100> * Preparing community OTA-4 images, promoting those after basic QA
15:15 <sil2100> * Promoting ubuntu-developer images
15:15 <sil2100> * Preparing non-delta images for arale
15:15 <sil2100> - Identifying the system-image situation, preparing landing
15:15 <sil2100> - Leading the RTM status meeting
15:15 <sil2100> - Updating documentation regarding package twins
15:15 <sil2100> - Further work on commitlog generation
15:15 <sil2100> - Coordinating the Unity8 big landing
15:15 <sil2100> - Lack of sleep and overall tiredness for the whole week
15:16 <sil2100> Holiday tomorrow (Friday the 19th)
15:16 <sil2100> (done)
15:16 <infinity> Do you celebrate every Friday the 19th?
15:17 <slangasek> it's a very famous movie in Poland
15:17 <infinity> slangasek: (you're up)
15:18 <slangasek> * helping unblock the libnettle transition; found blockage due to a mir-related API change that was causing gst-plugins-bad1.0 to FTBFS as of vivid release, working to unwind that through bug #1465958, now waiting for a platform-api landing
15:18 <ubottu> bug 1465958 in platform-api (Ubuntu) "gst-plugins-bad1.0 FTBFS: mirclient headers needed" [High,Confirmed] https://launchpad.net/bugs/1465958
15:18 <slangasek> * post mortem for the various issues we've run into over the past two months with the overlay ppa not being as featureful as the derived distribution out of the box
15:18 <slangasek> * closing in on a decision for the Java maintainer role
15:18 <slangasek> * new job opening in the process of being posted, should be visible tomorrow
15:18 <slangasek> (I know)
15:18 <slangasek> (done)
15:18 <slangasek> stgraber:
15:18 <stgraber> - Was out on Friday. Now back in Montreal.
15:18 <stgraber> - LXC/LXD
15:18 <stgraber> - Working on LXD socket activation systemd units.
15:18 <stgraber> - Made a first version of the LXD snap.
15:18 <stgraber> - Working on the CI infrastructure for LXC-related projects a bit.
15:18 <stgraber> - Spent quite a lot of time debugging a weird race condition in LXD.
15:18 <stgraber> - Code reviews, bug triaging, other admin paperwork
15:18 <stgraber> - Alpha-1
15:18 <stgraber> - Some prepwork for alpha-1, mostly sorting out nusakan <-> tracker issues
15:18 <stgraber> (done)
15:18 <infinity> - Dealing with several transitions in wily
15:18 <infinity> - Not one, but *two* emergency kernel security updates
15:18 <infinity> - Fixed kernel autopkgtesting
15:18 <infinity> - General AA/SRU stuff
15:18 <infinity> - Looking into kernel changes we might need to build armhf on arm64
15:19 <infinity> - Sobbing uncontrollably over the loss of mvo
15:19 <infinity> - Mitigating the above with the 'yay pitti' happy dance
15:19 <infinity> - debian-installer and HWE work
15:19 <infinity> (done)
15:19 <stgraber> :)
15:19 <pitti> ♩ \o/ ♫ ♪
15:19 <slangasek> infinity: that's to the tune of 'hey mickey', right?
15:19 <barry> infinity: where's the youtube video of that dance?
15:20 <pitti> (happy dance for fixing kernel tests)
15:20 <cyphermox> I think pitti wants to take over UTF-8 summary ending lead. :)
15:20 <infinity> barry: It's an easter egg in d-i, go find it.
15:20 <slangasek> heh
15:20 <slangasek> any questions over status?
15:20 <barry> infinity: do you have to fix 5 bugs before it reveals itself?
15:21 <slangasek> oh, and doko's binutis/gcc SRUs got released in trusty
15:21 <infinity> Yeah, which means we can probably undo a hack in linux-lts-vivid.
15:21 <slangasek> (only ignoring 3 and 6 autopkgtest regressions respectively)
15:21 * infinity makes a note.
15:22 <doko> autopkgtest regressions for trusty?
15:22 <slangasek> doko: yes; not caused by either of your SRUs, just caused by letting autopkgtests bitrot for a year in trusty before we started looking at them again
15:22 <slangasek> some of them being network-related failures
15:22 <sil2100> slangasek: were you able to make that document with the side-by-side comparison of overlay and derived distro for olli?
15:23 <slangasek> some being mysql tests that fail because time doesn't stand still and SSL certificates expired
15:23 <doko> ahh, I see. where can you see these?
15:23 <slangasek> sil2100: in progress; will be done this morning
15:23 <pitti> I sent a summary to u-devel@; mostly because of tests relying on quirks/bugs of the null runner which we used back then
15:23 <sil2100> Thanks!
15:23 <slangasek> doko: you probably can't get a link to the specific set anymore now that we've released the SRUs; there's nothing that you should need to look at however
15:24 <slangasek> the most interesting one was a lintian testsuite failure claiming that a binary was now a corrupted elf object - but that error started happening a week *before* binutils was uploaded to trusty-proposed
15:24 <infinity> pitti: Speaking of, did you want back on ~ubuntu-sru and ~ubuntu-release so you can have a bit more direct control over this madness?
15:24 <pitti> infinity: I need ~release for that, right?
15:24 <infinity> pitti: release for devel series, sru for SRUs.
15:24 <bdmurray> infinity: you are just going to let this new guy on these teams?
15:25 <infinity> pitti: We split the hints across the teams, just like we split the queues.
15:25 <infinity> bdmurray: He's emeritus on both teams. :P
15:25 <bdmurray> infinity: He might have forgotten how to do things!
15:25 <infinity> That's a chance I'm willing to take.
15:25 <pitti> I just select all and click "accept"
15:26 <infinity> Okay, not willing anymore.
15:26 <cyphermox> pitti: I think you mean select all and click Reject :)
15:26 <pitti> I'm not yearning for more work, just to manage stuff that's innocently stuck in -proposed a little better
15:26 <slangasek> heh
15:27 <slangasek> doko: fwiw if you want an overview of failing autopkgtests, in theory https://jenkins.qa.ubuntu.com/view/Trusty/view/AutoPkgTest/ is some kind of one
15:27 <infinity> pitti: Team membership doesn't imply forced duties, though it might get you pinged occasionally for help with something.  Anyhow, your call, no one's voluntold into these teams.
15:27 <doko> slangasek, thanks, not so nice as update_excuses ...
15:27 <pitti> infinity: right, please add me
15:27 <stgraber> pitti: select all and click "reject" with "Spite" as the reason, everyone will think you're infinity :)
15:28 <infinity> slangasek: Did anyone investigate the lintian regression and sort out what it really was?  The last time lintian had a testsuite explosion like that, it really was a (weird) toolchain regression.
15:28 <slangasek> doko: update_excuses gives you the overview for packages that are in progress. for SRUs that were already accepted, it no longer applies
15:28 <pitti> oh, we don't have archives for SRUs
15:28 <pitti> we do have them for devel
15:29 <infinity> pitti: Reactivated in both.
15:29 <slangasek> infinity: no.  what I did sort out was that the timing didn't line up with the binutils change, which was what had triggered the test rerun
15:29 <pitti> actually, I lie: http://people.canonical.com/~ubuntu-archive/proposed-migration/update_excuses/trusty/
15:29 <pitti> ^ should have all the old excuses?
15:29 <slangasek> infinity: (nor the gcc change; first lintian test failure was May 4, SRU accepts were May 13-14)
15:30 <slangasek> pitti: oh, it should? interesting
15:30 <infinity> slangasek: Sure, timing says it clearly wasn't *that* upload, still would be nice to figure out the cause.  Maybe I'll poke later in my copious free time.
15:30 <pitti> conveniently un-clickable, but at least there
15:30 <slangasek> doko: ^^ so there is an archive of the excuses if you really care; but with my SRU hat I already reviewed them and declared them to be false positives
15:31 <slangasek> [TOPIC] AOB
15:31 <slangasek> anything else?
15:31 <infinity> Sure.  Since we already hijacked the meeting for sru/release stuff...
15:31 <pitti> yes, right now we need to mostly ignore trusty, unless we upload a couple of SRUs to fix tests
15:31 <infinity> slangasek, stgraber, pitti : ~ubuntu-release only has one Admin (me), which is suboptimal.  Anyone want to volunteer to be a second Admin?
15:31 <slangasek> are we actually to where we're meant to be wrt not regressing the testbeds?
15:32 <slangasek> infinity: you can give me bits
15:32 <pitti> slangasek: since utopic, yes; trusty had a completely different infrastructure
15:32 <infinity> slangasek: Done.
15:33 <slangasek> pitti: my meaning is: are we at the point now where the testbed we're using to test trusty SRUs properly models what we were using at the time of trusty release
15:33 <slangasek> (per the discussion in Austin)
15:33 <infinity> slangasek: I think he's saying "close as it can get", since the runner is completely different, and we can't fix that.
15:33 <slangasek> ok
15:33 <slangasek> I did see some test failures that appeared to be regressions in the network policy
15:34 <slangasek> i.e., use of proxy + accessing Canonical network resources still returning a failure
15:34 <pitti> slangasek: they now have build-essential pre-installed; but e. g. the "tests don't have their build deps installed" behaviour is still differnet
15:34 <pitti> that too
15:34 <pitti> but there's not much I can do about that, we don't have unrestricted network there
15:34 <infinity> Network policy is about to change again, so that's an ongoing battle.
15:34 <infinity> But hopefully we can do it smoothly.
15:34 <slangasek> I'm not asking for an unrestricted network
15:34 <pitti> infinity: in scalingstack you mean?
15:34 <slangasek> I'm asking for the network policy to actually match what I understood to have been agreed with CI :)
15:35 <infinity> pitti: Yeah.  Though, we can probably do better with the move to scalingstack.
15:35 <slangasek> which was "external resources on the Canonical network ok (but not yet implemented); truly external resources outside of our control, not ok"
15:37 <pitti> slangasek: ah, we still have the old "anything goes via proxy" policy ATM
15:37 <slangasek> barry: ^^ this also impacts your system-image tests of course, so if you want to know when your autopkgtests will be allowed to work... :)
15:37 <stgraber> ah, that wasn't what you last told me, if that's the plan, then I'll have to turn off a bunch of tests in the LXC testsuite with the next upload
15:37 <slangasek> pitti: oh, ok.  I saw some test failures that implied that things were not working, including using the proxy
15:37 <pitti> was that "no arbitrary net access" already decided? (I object rather strongly..)
15:37 <infinity> I object to that too, unless Debian adopts the same policy.
15:38 <stgraber> unless "external resources on the Canonical network ok" includes "squid.internal" in which case, I've got what I need
15:38 <infinity> Cause hacking every test from Debian is ungood.
15:38 <barry> slangasek: yeah.  although i haven't gotten that far yet, i did re-enable the autopkgtests that do smoketesting against s-i.u.c
15:38 <slangasek> tests relying on arbitrary external resources are unreliable tests
15:38 <pitti> well, we would just change "one out of 1000 runs fails" with "all runs fail"
15:38 <pitti> TBH, the unreliability of tests pales compared to the unreliability of our infrastructure
15:38 <pitti> to the point that I can't even remember a single case
15:39 <pitti> I mean wrt. remote net access, not in general
15:39 <barry> it's also not necessarily a bad thing if you learn that a test relying on an external resource is failing <wink>
15:40 <pitti> if I have to retry a failed net-accessing test twice a year, I'll be reeeeeeally happy :)
15:40 <slangasek> stgraber: well, at least for the moment it's possible to access via the proxy
15:40 <slangasek> (except when this randomly doesn't work)
15:40 <infinity> slangasek: We did note that there are valid uses for tests talking to external resources, but often those don't relate to uploads, but should be daily "does my IM client software still work" tests.
15:41 <stgraber> slangasek: ok. I got a bunch of e-mails about test failures and the log looked like no network connectivity but it wasn't triggered by one of my uploads so I didn't really look into the failure
15:41 <slangasek> infinity: yes
15:41 <infinity> slangasek: But if we figure out how to cater to that case, we still need to be able to talk to external resources to do it.
15:41 <slangasek> infinity: the conclusion was "this doesn't belong in an autopkgtest run as part of proposed-migration"
15:42 <infinity> slangasek: Sure, "... as part of p-m", I agree with, but we're building infra that should be flexible and reusable.
15:42 <slangasek> it can be run under a different tenant with a different network policy
15:42 * infinity shrugs.
15:42 <slangasek> using the same charm on the same cloud
15:42 <infinity> Anyhow, stgraber's use case also seems reasonable.  Unless we want to host all his blobs in our network just to make the test 100% reliable instead of 99%, which seems like overengineering for a problem we don't have.
15:43 <slangasek> stgraber: I heard something to the effect that Canonical would be doing internal hosting of lxc images; is that the plan?
15:43 <stgraber> slangasek: nope
15:43 <pitti> infinity: ^ or at least I don't think it's sensible to think about this while 20% of our tests fail due to random infrastructure problems
15:44 <stgraber> LXD will be using the cloud images at some point in the near future, but nothing's changing for LXC
15:44 <slangasek> stgraber: ah, of course
15:44 <infinity> pitti: I like the theory that we'll make the infra five-nines reliable, so other problems become more obvious.
15:44 <stgraber> and LXD's own testsuite already doesn't rely on network access at all
15:44 <infinity> pitti: But, yeah, let's try to get there first.
15:44 <slangasek> ok I think we've discussed this to death
15:44 <slangasek> shall we release our captive audience?
15:45 <infinity> +1
15:45 <slangasek> #endmeeting