[06:35] <pitti> jibel: FYI, disabling wazn; it needs a reboot
[07:14] <elfy> morning DanChapman
[07:22] <DanChapman> morning elfy :-)
[07:28] <KI7MT> GM all ..
[07:53] <Noskcaj> evening KI7MT
[07:56] <KI7MT> Noskcaj, Hows goes the battles  .. saw your not about the up coming meetings, I'm really interested in is autotest and autopkg stuff, was very interesting reading .. but me a python don't see eye-to-eye on things :)
[07:56] <KI7MT> *note
[07:56] <jibel> Good morning
[07:56] <KI7MT> GM
[07:57] <jibel> pitti, why does wazn need a reboot?
[07:57] <Noskcaj> KI7MT, for autopilot you'll want to talk to DanChapman. For autopkg, i'm guessing jibel and/or pitti
[07:57] <jibel> pitti, nm, I read your comment on other channels
[07:58] <Noskcaj> KI7MT, Timezones make it difficult for me to make meetings
[07:58] <pitti> jibel: trying to run VMs just hangs; there is a kswapd0 process eating CPU, and causing high load
[07:58] <pitti> jibel: and it seems there is a stalled apt-get update or so
[07:58] <pitti> jibel: I don't have root on the box, do you?
[07:59] <KI7MT> Noskcaj, Yeah TZ  can be a problem for allot of us .. was just good reading about it, interesting stuff.
[07:59] <pitti> jibel: btw, I'm currently working on adt-run understanding a simple package name
[07:59] <pitti> jibel: i. e. "adt-run libpng --- adt-virt-*"
[07:59] <pitti> jibel: this will use apt-get source in the testbed
[08:00] <pitti> jibel: that entirely avoids copying the source, so this should help with linux
[08:00] <pitti> jibel: and also, it will make our ARM/lxc runners more convenient to use as we don't need to download the source to the host in jenkins
[08:00] <pitti> actually curious that nobody requested this yet :)
[08:00] <pitti> in fact I just got it running, need to write tests now
[08:01] <pitti> OMG I understand autopkgtest's code, I'm d00med
[08:01] <jibel> pitti, I don't *officially* have root on that box, so I'd rather wait someone from CI
[08:04] <jibel> pitti, calling adt directly with a package name is definitely interesting
[08:04] <jibel> pitti, I've been thinking about enabling arm nodes, and we don't need all the overload of the wrapper script around adt-run
[08:05] <jibel> pitti, just a jenkins slave that calls adt-run, with the results copied into the workspace should be enough
[08:05] <jibel> and call the ARM job as a non-blocking post-build action of x86 jobs
[08:06] <pitti> jibel: right, and I fixed --output-dir to be working and actually useful (and cleaning up cruft)
[08:06] <pitti> jibel: that was my idea, that we can reduce the wrapping of adt-run to a minimum
[08:07] <pitti> it should just be adt-run --output-dir=... <pkgname> --- adt-virt-lxc --ephemeral trusty-cloud
[08:07] <pitti> jibel: we'll still need some maintenance jobs for updating the container etc., and you probably don't want to hardcode "trusty" here ^ either, but by and large that ought to work?
[08:12] <jibel> pitti, yes, but we can keep the setup phase to a minimum
[08:16] <pitti> jibel: we still need some option to either run "apt-get update" after opening the testbed, or a more generic "run this script before doing anything", right?
[08:17] <pitti> or should we do that as a separate step?
[08:17] <pitti> i. e. call lxc-* to update the container, and then call adt-run?
[08:18] <pitti> the update would then happen in the actual container, not the ephemeral one; although this sounds much less robust
[08:18] <pitti> probably best if I add an --apt-update option to the lxc runner
[08:27] <pitti> jibel: hm, hang on, that's not enough -- we also want to enable -proposed
[08:30] <pitti> jibel: LXC alreay has a concept of hooks; would it work to pass a script to adt-virt-lxc which it could then pass on to lxc for running, or should it rather do that by itself with just the usual autopkgtest way of running commands?
[08:32] <jibel> pitti, we could do that in a mount hook, not sure how it works with ephemeral containers
[08:32] <pitti> jibel: I just don't want to fiddle with the container configuration too much; for starters, I wouldn't even know where that lives, and I didn't see an lxc-start optoin for that
[08:35] <jibel> pitti, another option would be to bind mount a specific directory for each run and execute any executable script in a specific location
[08:35] <jibel> lxc-start-ephemeral has an option to specify a directory to bind mount
[08:36] <pitti> jibel: well, adt-virt-* already know how to run a shell script in the testbed
[08:36] <pitti> jibel: so we could just re-use that
[08:37] <pitti> then it wouldn't even be limited to adt-virt-lxc, but work for any runner
[08:37] <pitti> it would then need to be a shell script, but that doesn't seem like a too bad restriction?
[08:43] <jibel> pitti, how do you do that (make adt-virt-* run a shell script)?
[08:44] <pitti> jibel: well, you'd read it and call Testbed.execute(), adt-run does that all the time
[08:44] <jibel> ah, ok
[08:45] <pitti> that's using the auxverb or shstring, whatever the testbed provides
[09:36] <elfy> balloons: https://code.launchpad.net/~elfy/ubuntu-manual-tests/1249362/+merge/202809
[09:36] <elfy> all that I thought needed doing are done
[10:22] <slickymaster> morning all
[10:36] <pitti> rbasak: FYI, what we talked about yesterday: http://anonscm.debian.org/gitweb/?p=autopkgtest/autopkgtest.git;a=commitdiff;h=a615e39eb92b
[10:37] <rbasak> pitti: that looks great. Thanks!
[10:38] <rbasak> I hate that whole global variable thing, btw. But I guess the whole virtsubproc interface needs to be changed to fix that.
[10:38] <pitti> rbasak: in case you do this often, http://anonscm.debian.org/gitweb/?p=autopkgtest/autopkgtest.git;a=commitdiff;h=2262ca0 should make things quite a bit simpler, too; we'll use that for running archive autopkgtests on ARM, to avoid unnecessary copying and extra wrappers around adt-run
[10:40] <rbasak> That looks handy.
[11:01] <davmor2> Morning all
[11:17] <pitti> jibel:
[11:17] <pitti> ./run-from-checkout --setup "xapt-get update" libpng --- adt-virt-schroot trusty
[11:17] <pitti> jibel: err, ignore the "x"apt-get, that was for testing
[11:17] <pitti> jibel: --setup-commands can either get a file name, or the actual commands
[11:24] <jibel> pitti, setup will be executed every time the testbed is started? I'm thinking about the revert capability of lxc based containers where the ephemeral container is destroyed between tests
[11:26] <pitti> jibel: hmm, I need to check that; but that would only happen if you use the binaries from the built source, no?
[11:26] <pitti> oh, you mean for multiple tests
[11:26] <jibel> yes for multiple tests
[11:26] <jibel> run test1, revert, run test2, ...
[11:28] <pitti> I'll add a test case for that (haven't written tests yet)
[11:28] <pitti> ATM I just run it once, so that's wrong
[12:24] <pitti> jibel: ok, now works with lxc/revert and multiple tests
[13:22] <pitti> jibel: oh, debian bug 736416 just came in, interesting
[13:28] <jibel> pitti, oh, that's interesting
[13:32] <pitti> jibel: I'm currently writing a response
[13:52] <pitti> jibel: rebuilding VMs on wazn, the current ones seem broken (connecting hangs at 1/18 tries)
[14:01] <jibel> pitti, ack
[14:27] <senan> balloons, danchapman : Good Eve
[14:32] <DanChapman> senan, hey
[15:10] <pitti> jibel, rbasak: ok, I now also reduced the number of test bed resets (with LXC) from 3 to 1, there were several unnecessary ones
[15:10] <rbasak> Nice!
[15:10] <pitti> jibel, rbasak: can you think of anything else which I should add/change/fix for our arm/lxc autopkgtest endeavour?
[15:10] <rbasak> Have you considered how to keep the lxc template up-to-date?
[15:11] <pitti> rbasak: that isn't the job of adt-run itself, but we quickly discussed it
[15:11] <rbasak> I noticed that "apt-get update" doesn't get run, incidentally. Not sure if that matters if you're not building anything.
[15:11] <jibel> rbasak, that should be a separate process
[15:11] <pitti> rbasak: ideally we'd rebuild the trusty-cloud container every day from the current cloud image
[15:11] <pitti> rbasak: ah, for that I added http://anonscm.debian.org/gitweb/?p=autopkgtest/autopkgtest.git;a=commitdiff;h=6e643986c6
[15:12] <pitti> rbasak: i. e. for every test we can run a setup script to enable proposed and do apt-get update
[15:12] <rbasak> Ah. That'll be handy!
[15:12] <pitti> rbasak: and also apt-get dist-upgrade if we desire
[15:12] <pitti> rbasak: but of course we don't want dist-upgrade to run on a weeks old base image, so we need to refresh that daily
[15:12] <pitti> either by rebuilding the container, or dist-upgrading it
[15:13] <pitti> as I said yesterday, the output of lxc-create -t ubuntu-cloud is rather poor
[15:13] <pitti> it has broken apt sources and old packages for some reason
[15:13] <pitti> if we fix that, we can use that; until then, perhaps just run dist-upgrade
[15:14] <pitti> perhaps that's also an artifact of using saucy's lxc packages, I don't know
[15:15] <rbasak> The only other thing I can think of is the necessity to say --gain-root=sudo, instead of adt-virt-lxc somehow telling adt-run what the default should be. But that isn't a blocker.
[15:19] <pitti> rbasak: how do you mean?
[15:19] <pitti> for now you need to run lxc-start as root anyway, so the entire adt-run runs as root
[15:19] <rbasak> pitti: I have to say: "adt-run --gain-root=sudo ... --- adt-virt-lxc"
[15:19] <pitti> and virt-lxc cna then su to the ubuntu user for non-root tests/builds
[15:19] <pitti> aah
[15:20] <pitti> nice, I didn't try that
[15:20] <rbasak> Since adt-virt-lxc specifies suggested-user=ubuntu and gives adt-run root.
[15:20] <pitti> nope, doesn't work
[15:20] <pitti> that calls virt-lxc as user instead of root
[15:21] <rbasak> Yes
[15:21] <rbasak> That's the intention.
[15:21] <rbasak> I run adt-run as a normal user.
[15:21] <rbasak> adt-virt-lxc calls sudo as necessary.
[15:21] <rbasak> brb phone
[15:23] <balloons> mzanetti, ping
[15:23] <pitti> rbasak: ah, you use per-user containers then? I created mine system-wide
[15:23] <mzanetti> balloons: pong
[15:24] <balloons> I was going to ask your opinion on reminders app mocking, though I already know it :-)
[15:25] <balloons> I was planning on going down the server mocking route with autopilot as you know. elopio mentioned this as the best example he's found: http://bazaar.launchpad.net/~ubuntuone-hackers/unity-scope-click/trunk/files/head:/src/tests/autopilot/unityclickscope/
[15:27] <mzanetti> balloons: what does this do?
[15:28] <rbasak> pitti: (done on phone) No, I use system-wide ones.
[15:29] <mzanetti> balloons: ah... simulates a server and the app would use localhost instead of the real server
[15:29] <balloons> mzanetti, yes
[15:29] <mzanetti> balloons: not sure what evernote sends... you'd need to sniff the connection
[15:29] <rbasak> But since lxc-* (at least in the past) required root, I tried to hide that inside adt-virt-lxc, so that the interface wouldn't change in the future.
[15:29] <pitti> rbasak: hm, if I call lxc-* as user, it tries to use containers from my ~/.local/...
[15:29] <pitti> $ lxc-config lxc.lxcpath
[15:29] <pitti> /home/martin/.local/share/lxc/
[15:29] <mzanetti> balloons: I'm a bit afraid it might be a binary protocol
[15:29] <rbasak> Thus adt-virt-lxc calls sudo, but adt-run should be unaware of that root requirement.
[15:29] <pitti> $ sudo lxc-config lxc.lxcpath
[15:29] <pitti> /scratch/lxc
[15:29] <balloons> I figured before I dove in on this it might be easier to get your thoughts
[15:29] <balloons> mzanetti, ahh, so not nice json :-)
[15:29] <rbasak> From adt-run's point of view, it gets root inside the container, and can switch to ubuntu if necessary.
[15:30] <rbasak> adt-virt-lxc specifies suggested-user=ubuntu, and adt-run uses that.
[15:30] <pitti> rbasak: right, it's just that lxc now has these system/user modes
[15:30] <rbasak> But then adt-run tries to call fakeroot, and that doesn't work.
[15:30] <rbasak> (since fakeroot isn't installed)
[15:30] <pitti> rbasak: it hasn't been like that in saucy yet
[15:30] <pitti> $ lxc-config lxcpath
[15:30] <mzanetti> balloons: no... it's not json for sure. it might be xml, but no idea what libthrift does
[15:30] <pitti> /data/lxc
[15:30] <rbasak> Which is a bit silly, because adt-run *had* root inside the container, but gives it away because of suggested-user
[15:30] <pitti> rbasak: so I think that's new in trusty
[15:30] <rbasak> So I have to specify --gain-root
[15:32] <pitti> rbasak: is your host running saucy or trusty?
[15:32] <rbasak> pitti: trusty right now. I hadn't noticed anything different though. Apart from needing to "chmod 755 /var/lib/lxc", I haven't had to change how I work at all.
[15:33] <pitti> rbasak: hm, so I wonder why lxc as user now wants to look into my $HOME for containers
[15:34] <elfy> balloons: if you missed the ping - those testcase updates are done now
[15:35] <balloons> elfy, I did indeed. Shall I review then?
[15:35] <elfy> up to you - I've done my bit :D
[15:36] <elfy> I'd do the testcases - but you don't get karma for it :p
[15:36] <elopio> I'm walking with the dog.
[15:36] <elopio> bbs
[15:36]  * balloons reviews
[15:49] <pitti> ubuntu@cyclops-node02:~$ adt-run -o /tmp/adt-libpng --setup "apt-get update" libpng  --- adt-virt-lxc --ephemeral trusty-cloud
[15:49] <pitti> jibel, rbasak: ^ this now works well, and just takes a minute or so
[15:49] <pitti> of course the --setup needs to be extended a bit
[15:49] <pitti> /tmp/adt-libpng/ then has apt0t-build-stderr, apt0t-build-stdout, and log
[15:49] <pitti> IOW, the jenkins attachments that we want
[15:51] <balloons> elfy, I like your branch names :-)
[15:51] <jibel> pitti, so in jenkins basically what we need in the job is something like adt-run -o $WORKSPACE/results --setup extended_setup_script $PACKAGE --- adt-virt-lxc ...
[15:51] <pitti> jibel: correct
[15:51] <elfy> balloons: lol
[15:51] <pitti> jibel: where setup_scripts shoudl enable -proposed, run apt-get update, and perhaps dist-upgrade -y (not sure whether we currently do)
[15:51] <jibel> and cleanup the results dir before jenkins grabs the attachments
[15:52] <pitti> jibel: the only clutter that I see in it is an empty binaries/ directory
[15:52] <pitti> jibel: for build-needed they will be there, but not in our case; I'll fix that
[15:52] <pitti> jibel: all the cruft (fifos, "xout" files, etc.) that you saw in earlier versions are completely gone
[15:54] <pitti> jibel: enabling -proposed is basically echo'ing the deb+deb-src into sources.list.d/
[15:55] <balloons> whoa, so many files changed :-)
[15:55] <elfy> I know :(
[15:55] <elfy> I'd have left them all tbh
[15:57] <balloons> merged and syncing
[15:59] <balloons> I see several merges from Ki7MT
[16:06] <balloons> elfy, do you have time / desire to review some of the others?
[16:06] <balloons> I see you've earmarked the xfce one for david so I'll leave it
[16:12] <balloons> nvm elfy I'm starting them.. they look fun
[16:17] <pitti> $ time adt-run -o /tmp/adt-libpng --setup "apt-get update" libpng  --- adt-virt-lxc --ephemeral trusty-cloud
[16:17] <pitti> real…………1m10.964s
[16:17] <pitti> $ ls /tmp/adt-libpng/
[16:17] <pitti> apt0t-build-stderr  apt0t-build-stdout  log
[16:17] <pitti> jibel: ^
[16:17] <pitti> jibel: so the empty binaries/ is gone as well now; I don't think there's anything else to clean up now
[16:18] <pitti> jibel: it's only installed on node02 for now, in case you want to play around
[16:20] <pitti> I'll install the official debs tomorrow when they are in Debian and synced into trusty
[16:20] <pitti> (althuogh they shouldn't look any different from my home-built one)
[16:28] <balloons> DanChapman, ping
[16:29] <elfy> balloons: the ki7mt ones - looked like server stuff from memory - no idea here whether they'd be right or wrong
[16:29] <balloons> elfy, yep no worries. I finished everything but those I believe
[16:29] <elfy> and I'm trying to ping my own qa team for in-house stuff so as to not bother others
[16:33] <DanChapman> balloons, pong
[16:36] <elfy> DanChapman: bit rude :p
[16:37] <DanChapman> elfy :-D
[16:40] <pitti> jibel: rolled out everywhere now; I'll start a loop which runs all our autopkgtests on the four nodes, so that I can have a look at the logs/success rate tomorrow
[16:41] <elfy> DanChapman: I see some green at jenkins for us now finally ...
[16:45] <DanChapman> elfy yeah finally running again. :-D
[16:50]  * DanChapman prods balloons 0--
[16:55] <pitti> jibel: ok, started (all except linux and libreoffice, and the known-bad maas* tests); if you see smoke coming out of those nodes, you know why :-)
[17:08] <balloons> hey DanChapman
[17:08] <jibel> pitti, Excellent, thanks!
[17:08] <balloons> so DanChapman there's a few ap tests out there, I saw you reviewing some and leaving comments about where to put them
[17:08] <balloons> shall we clean that stuff up?
[17:23] <DanChapman> balloons, sure :-) the desktop tests?
[17:24] <balloons> DanChapman, yes those things :-) I saw Jackson and some others left some MP's to review
[17:31] <DanChapman> balloons, cool :-) Yes so do you mean shall we clean up the MP's or something else?
[17:32] <balloons> DanChapman, I was hoping to merge or give feedback on all the mp's yes
[17:32] <balloons> I saw you had started on some so I thought I'd ping you
[17:34] <DanChapman> balloons, yes so adam/ disc0tech is working on rhythmbox, ive been in contact with him over that MP  via email which i will put a copy of the convo's into the MP for clarity. I believe he is going to improve on the current MP so I will note that
[17:34] <balloons> Yes, I assumed so as I remember having some chats with him
[17:38] <DanChapman> balloons, err lderan's branch I havn't ran myself so I can't comment on that one and senan's branch was still flaky last time we spoke, i'll run it again to see what it's like now
[17:52] <balloons> DanChapman, let me know if you need any help or confirmations / etc. I'd just like to provide feedback on all those mp's and merge anything that can be merged :-)
[18:05] <lderan> DanChapman, okay
[18:05] <DanChapman> balloons, Yuk emailing a convo to an MP comes out looking horrid :-S anyway feel free to confirm any of them, you can do lderan's if you want? ;p
[18:05] <DanChapman> lderan howdy o/
[18:06] <lderan> DanChapman, :P hello
[19:17] <balloons> DanChapman, sorry I didn't respond because I saw lderan appear :-)
[19:24] <elfy> I'm watching that balloons guy :p
[19:25] <teward> ... i have never had a program EVER take so long to build inside of an schroot before... o.o
[19:25] <balloons> lol elfy .. so much fun with tests today
[19:26] <elfy> :p
[19:26] <balloons> sometimes feels like for every fire you put out, two more appear
[19:27] <elfy> I know - I've done that lot for you, checked a bunch of our stuff for our next package call, changed stuff, did merge request, synced the tracker, 2 meetings
[19:27] <elfy> some other stuff
[19:28] <elfy> you get paid though :p
[19:28] <balloons> hehe.. you get more props
[19:28] <balloons> as you should
[19:28] <elfy> ha ha ha
[19:28] <teward> balloons: i take it the bug triager role is getting people's names on the QA team's wiki page, I'm seeing bug triage, SRU, etc. showing up more now.
[19:29] <balloons> are you seeing output too?
[19:29] <balloons> yes, more people are adding there names, it's good to see
[19:30] <teward> i'm going to add my name there, but I have to put a "NOTE: THis wiki page is outdated, and may be updated at any time" section on my wiki page first.
[20:06] <balloons> Letozaf_, buonasera
[20:06] <Letozaf_> balloons, buonasera :)
[20:32] <Letozaf_> balloons, I'm carrying on fixing other locale test failures in calendar app, failures that don't have to do with the bug found yesterday
[20:33] <balloons> Letozaf_, ahh
[20:34] <Noskcaj> balloons, Thanks for the merge review. I have a few more packaging things to do first, but i'll get to fixing that eventually
[20:34] <balloons> Noskcaj, your most welcome. Trying to get through all the autopilot reviews this afternoon
[20:34] <balloons> * you are
[20:34] <balloons> Noskcaj, I believe I can merge this one though? https://code.launchpad.net/~noskcaj/ubuntu-autopilot-tests/xfce4-screenshooter/+merge/199723
[20:35] <Noskcaj> balloons, yeah, that's all good
[20:35] <balloons> oO it's using processmanager.. ugh
[20:35] <balloons> no introspection eh? ohh well
[20:35] <Noskcaj> most xfce stuff doesn't introspect
[20:35] <balloons> yea, sadness
[20:35] <Noskcaj> still, gthumb is huge and needs tests. That will keep me busy
[20:35] <balloons> can you add anything else to the test?
[20:36] <Noskcaj> not that i know how
[20:36] <Noskcaj> even my gthumb thing is as much as i can do
[20:37] <balloons> Noskcaj, with process manager it gets a bit more fun, but I'm wondering if you could check a hotkey or something
[21:20] <balloons> Noskcaj, it seems lderan is also working on gthumb; https://code.launchpad.net/~lderan/ubuntu-autopilot-tests/ubuntu-autopilot-tests/+merge/202335
[21:21] <Noskcaj> i think i'll leave xubuntu autopilot to him, i still can't make stuff work
[21:29] <balloons> Noskcaj,  I did leave a comment on your screenshooter.. can we hit a hotkey and check the app response?
[21:30] <Noskcaj> balloons, I don't think so
[21:30] <balloons> Noskcaj, I'm not familar with the app, but does it change at all after pressing the hotkey to take a shot for instance?
[21:31] <balloons> something simple and basic.. if possible.. beyond checking that it runs without crashing
[21:31] <balloons> which is quite useful ;-)
[21:31] <Noskcaj> maybe using PrntScrn to open it?
[21:34] <balloons> Noskcaj, yes, good example
[21:35] <Noskcaj> gulp, now i have to do that, don't i
[21:36] <balloons> I believe Noskcaj :-)
[21:37]  * Noskcaj add that to the todo list
[21:40] <balloons> ty Noskcaj
[21:40] <Noskcaj> now to try and unbreak the multiplayer games of trusty
[21:50] <lderan> Noskcaj, you can use process manager to find if a window is open with a specific name if that helps :)
[21:51] <Noskcaj> lderan, trust me, it's faster you do it than someone teaches me
[21:52] <lderan> okay :) tho if you do want it teaching then let me know