/srv/irclogs.ubuntu.com/2013/11/20/#ubuntu-ci-eng.txt

rsalvetiwill trigger a new image so I can have a working emulator00:14
plarssomething happened with the download taking way too long for the latest image it seems... I restarted them, so results should start to show up in a bit03:25
Mirvgood morning05:12
Mirvrobru: I'll test unity8 AP too..05:18
Mirvrobru: I've gotten sometimes problematic results where another run(s) have been fluent05:18
Mirvcihelp daily-release-executor is reported offline05:31
Mirvso all builds halted as well05:31
rsalvetiplars: there was a regression in the recovery image, which is fixed in the archive but needs a newer image05:33
rsalvetinot sure if the regression would cause the device to fail in someway05:33
didrocksMirv: hey! around already? :)07:12
didrockscihelp: hey, all executors on q-jenkins are down, (so everything is blocked). Seems the reason why robru/ken couldnt publish Mir yesterday07:14
viladidrocks, Mirv: on it07:14
didrocksthanks07:15
didrocksvila: I guess the daily-release executor is the first one to investigate why it went down07:15
viladidrocks: yes, Mirv said that already, we all agree07:16
Mirvdidrocks: yeah, as written on langing plan07:21
Mirvdidrocks: robert also had some unity8 AP problem, but I couldn't reproduce it so it's probably his environment (like wrongly updated unity8 from daily-build, or click tests being run and messing up unity8 testing)07:22
didrocksMirv: ah, the notes were not from yesterday07:22
didrocksMirv: yeah, probably07:22
didrocksMirv: I'm not sure to understand the unity-mir rebuild?07:22
didrocksif it wasn't rebuilt already, you can't test, right?07:23
Mirvdidrocks: the mir 0.1.1 build-dep was not in, otherwise it's ok. same for u-s-c.07:23
didrocksMirv: ah, it's just the bump build-dep MP07:23
didrocksMirv: well, if we release without that one, I won't die TBH ;)07:23
didrocksas long as the things are rebuilt as expected07:23
Mirvdidrocks: ok ;) yeah the libmirserver10 dependency is there anyway, as it was built after the mir build07:24
didrocksMirv: ok, just don't stress on that, if you feel that we should skip the build, please do (if we are ready to publish)07:24
didrocksogra_: asac: did one of you requested a new image?07:25
didrocksogra_: asac: I see a new image and 2 android uploads in particular to the archive (the second to fix a regression it seems)07:25
Mirvdidrocks: ok. we're ready to publish, aside from that I don't see a note if unity-system-compositor was smoke tested.07:25
Mirvso that needs to be done still07:25
didrocksogra_: asac: but the image which was built was between the 2 android uploads, so the "recovery regression" is in the current proposed image?07:25
didrocksMirv: ok, maybe you should kill in jenkins if there are some mir/platform/unity8 stacks planned to be rebuilt?07:26
didrocksso that one q-jenkins is repaired, you are not trapped by a rebuild :)07:26
Mirvdidrocks: yes, I did that already07:26
Mirv(except for platform it seems, now did that too)07:27
didrocksthanks Mirv :)07:27
didrockswaow, it's snowing here07:30
Mirvdidrocks: cool, it's just rain currently here :)07:31
MirvI'd welcome snow already, it's been unusually warm (which means also dark)07:32
didrocksMirv: well, I think it'll transform soon in rain as well :)07:32
didrocksah ah, for some kind of "warm" ;)07:32
didrockshere, for us, it's freezing (like we have the temperature we do have generally in 3 weeks)07:33
viladaily-release-executor is back07:33
didrocksthanks vila07:41
Mirvok got to the desktop with updated mir + u-s-c too08:04
didrocksMirv: and still get graphics? ;)08:04
Mirvdidrocks: even that :) not much multimonitor though08:04
didrocksMirv: not a surprise, but not a bad surprise either :)08:04
didrockssweetness!08:05
Mirvdidrocks: the mir stack had already queued the prepare tasks before daily release executor broke earlier, so they got rebuilt. mir is a no-change rebuild and u-s-c got now the mir 0.1.1 b-d in. I'm rerunning unity8 AP still and after then asking for packaging acks to publish.08:31
didrocksMirv: sounds good, just do minor dogfooding08:31
didrocksdon't stress it too much in some words :)08:32
didrocksthanks!08:32
Mirvdidrocks: ok then, http://q-jenkins.ubuntu-ci:8080/view/cu2d/view/Head/view/Mir/job/cu2d-mir-head-3.0publish/lastSuccessfulBuild/artifact/packaging_changes_mir_0.1.1+14.04.20131120-0ubuntu1.diff + http://q-jenkins.ubuntu-ci:8080/view/cu2d/view/Head/view/Mir/job/cu2d-mir-head-3.0publish/lastSuccessfulBuild/artifact/packaging_changes_unity-system-compositor_0.0.2+14.04.20131120-0ubuntu1.diff + http://q-jenkins.ubuntu-ci:8080/view/cu2d/view/Head/view08:50
Mirvso the actual u-s-c session moved to ubuntu-desktop-mir which I already noticed08:50
didrocksMirv: sorry, my system is in autodestruction mode after removing the emulator, I can't use sudo, ls or any other tool like chromium, can you get someone to review it?08:50
didrocksI can't even open a webpage08:50
Mirvdidrocks: ok, have a nice debugging session..08:56
didrockswell, nice with no "ls", I hope that my home dir is still intact…08:57
asacdidrocks: from backlog:09:01
asac09:59 < -bip> 20-11-2013 00:08:09 -!- alesage!~alesage@c-98-212-4-39.hsd1.il.comcast.net has quit  [Quit: Leaving]09:01
asac09:59 < rsalveti> 00:15:21> will trigger a new image so I can have a working emulator09:01
asac:/09:01
didrocksasac: well, and no second image with "recovery fixed"09:02
didrocksright?09:02
didrocksor did you find anything in the backlog?09:02
asacdidrocks: let me look more careful :)09:02
didrocksasac: let's discuss in a few, my install is broken by the emulator I guess09:02
asac09:59 < plars> 03:25:24> something happened with the download taking way too long for the latest image it  seems... I restarted them, so results should start to show up in a bit09:03
asac09:59 < rsalveti> 05:33:30> plars: there was a regression in the recovery image, which is fixed in the  archive but needs a newer image09:03
asac09:59 < rsalveti> 05:33:42> not sure if the regression would cause the device to fail in someway09:03
asacdidrocks: next line spoken here was you pinging Mirv09:03
asacso guess not09:03
asacthough the need for an image was identified09:03
asacdidrocks: so feels we want another image? whatelse happened in the meantime?09:04
didrocksasac: yeah, I guess we want one (but again, I can't even ls right now, let me fix my box)09:05
didrocksand post a warning if it's really due to the emulator replacing libc6 with the armhf version on your system09:05
asacdidrocks: ok. guess just boot in a live session and fix it from there09:05
asac:/09:06
didrocksasac: I'm trying with busybox here09:06
asacor from initrd if you are brave09:06
asacyeah09:06
didrocksbut dpkg from busybox doesn't show which arch is installed :p09:06
asacdidrocks: guess check in /var/lib/dpkg on your own09:07
didrocks2013-11-20 09:48:34 status half-configured libc-bin:amd64 2.17-93ubuntu409:07
didrocksfrom removing the emulator09:07
* asac wonders what the emulator does that can cause this09:09
seb128didrocks, weird, usually app ask you to type the "yes, I'm sure I want to do that" before removing libc-bin09:09
didrocksseb128: I think it didn't remove, it has overriden with the armhf version09:10
didrocksseb128: I clearly didn't get "remove libc6"09:10
asachalf-configured probably means that things fell over during upgrade09:10
didrocksok, busybox didn't work from my needs, let me try to find an usb stick and fix from there09:10
didrocksyeah09:10
didrocksI guess there are some divert magic for the emulator09:10
asacdidrocks: where did you get the emulator from?09:11
didrocksand something when went on removing09:11
asacack09:11
didrocksasac: followed the wiki page09:11
didrocks(but then, tried to remove the android-emulator package)09:11
seb128I hope other people don't try the emulator09:12
didrockswell, it's not 100% confirmed, but it was the only thing I was doing at this time09:12
seb128didrocks, do you want me to follow up on the list to "be careful, Didier bricked his system with that"?09:12
asacdidrocks: you hvae a URL to the wiki?09:12
didrocksseb128: would be nice to have someone confirming (in a vm first)09:12
didrocksasac: well, I just have a weechat window here :p it was on the ML09:12
asackk09:13
didrocksok, see you (hopefully) in a few09:13
seb128asac, https://wiki.ubuntu.com/Touch/Emulator09:13
didrocksyou should try on amd64 to reproduce, as jibel noted, it's probably a i386<->amd64 issue09:14
didrocksso seb128 is safe, you can try in production! :p09:14
seb128lol09:14
asacit refers to an android-emulator package09:15
asacbut thats noot in the archive... which ppa is it?09:15
jibelasac, binaries are only available for i38609:15
jibelin the archive09:15
seb128asac, it's in multiverse for i38609:16
seb128https://launchpad.net/ubuntu/+source/android/20131120-0225-0ubuntu109:16
asacoh its part of android09:16
asacyeah i can imagine that that might cause accidential havoc09:16
seb128on i386 it wants to install libc6-amd6409:20
seb128not sure I want to risk trying on my system :p09:20
asacso the package doesnt look harmful by itself09:22
jibeltrying in an amd64 VM, it's installing 723MB of i386 packages09:23
asacseb128: it includes stuff like: ./usr/share/android/emulator/out/host/linux-x86/bin/emulator64-arm09:23
asacso it somewhat makes sense that it brings in amd64 stuff09:23
didrocksCommandline: apt-get install android-emulator09:23
didrocksInstall: android-emulator:i386 (20131120-0225-0ubuntu1), lib64gcc1:i386 (4.8.2-1ubuntu2, automatic), lib64stdc++6:i386 (4.8.2-1ubuntu2, automatic), libc6-amd64:i386 (2.17-93ubuntu4, automatic)09:24
didrocks(that was what I had)09:24
didrocksand it worked after that, it's really when I removed the package and autoremove the deps09:24
seb128asac, it's a bit weird that it's an i386 package and that it pulls in amd64 binaries09:24
jibeldidrocks, I reproduced your problem09:24
asachttp://paste.ubuntu.com/6447226/09:24
jibeland broke the VM after removing the emulator09:24
jibelasac, ^09:24
seb128jibel, can you paste the log of what apt did when you removed the emulator?09:25
asacright. that would be good to know09:25
jibelseb128, sure09:25
didrocksjibel: ah, so not rebooting immediatly, would be interesting to get your VM fix rather than me being offline :)09:25
jibeldidrocks, indeed, 1 minute09:25
didrocks(ok, just finished getting an usb stick ready with an old trusty image)09:26
asacso the emulator package has nothing in the postrm postinst scripts09:26
asacso the bustage must come genuinely from apt and libc6-amd64 removal09:26
didrocksright09:26
asacmy guess is that if i just instlal libc6-amf64 and remove it will also be broken09:27
jibelseb128, http://paste.ubuntu.com/6447229/09:27
asac-> e.g. if thats the case thats a doko bug, no?09:27
didrocksat least, I'm looking at busybox dpkg and it doesn't have a lot of options09:27
didrocksasac: you're probably right, indeed09:27
jibelseb128, asac to reproduce on an amd64 VM, dpkg --add-architecture i386 && apt-get update && apt-get install -y android-emulator && apt-get autoremove --purge android-emulator09:28
asaclibstdc++6:i386*09:28
asacodd.09:28
didrocksyeah, so it probably removed the libc6-amd64 files configured to be used against09:31
didrocksand then, can't run the postrm script to configure the new links?09:31
asacjibel: can you reproduce by just installing the libc-amd64 and purging that?09:32
asac(in case you have a VM image easily available)09:32
jibelasac, hold on, I'm trying to recover the system first to help didrocks save his machine09:32
didrocksI would appreciate, indeed :)09:32
asacsure09:32
seb128that package has a postrm09:33
seb128if [ "$1" = remove ]; then09:33
seb128    ARCH=${DPKG_MAINTSCRIPT_ARCH:-$(dpkg --print-architecture)}09:33
seb128    if [ "${ARCH}" = "amd64" ] && [ "libc6-amd64" = "libc6-i386" ]; then09:33
seb128if [ -h /lib/ld-linux.so.2 ] && [ ! -f /lib/ld-linux.so.2 ]; then09:33
seb128    rm /lib/ld-linux.so.209:33
asachowever, maybe first narrowing it down will help :)09:33
asacgreedilyu removing ld-linux09:33
didrockshum, I still have /lib/ld-linux.so.2 here09:34
asacseb128: that only happens on amd64, no?09:34
asaci think that might not be it09:34
didrocksbut the postrm failed itself09:34
didrocksso it's not executed09:34
didrocks(as per log)09:34
asacall those postrm fail09:34
didrocksright09:34
asacseems stuff is already busted after prerm09:34
didrocksyep09:34
asacand after just removing the files09:34
asacint he packages09:34
asacmaybe having the log from installing might show some diversions09:35
asacetc.09:35
didrocksI guess when removing libc6-amd64 files09:35
ogra_are you 100% sure the last android build made it into the image ?09:36
jibelasac, trying now09:36
* ogra_ sees that publishing and image build are very close 09:36
didrocksogra_: was it supposed to be reflected in your diff?09:37
ogra_no09:37
ogra_the android package isnt inside anywhere09:37
ogra_so the manifest generator cant pick it up09:37
ogra_Get:1 http://ftpmaster.internal/ubuntu/ trusty/multiverse android all 20131120-0225-0ubuntu1 [348 MB]09:37
ogra_Fetched 348 MB in 48s (7243 kB/s)09:37
ogra_ok, seems it made it09:38
ogra_(you can find it in the livefs build log)09:38
ogra_http://people.canonical.com/~ubuntu-archive/livefs-build-logs/trusty/ubuntu-touch/20131120.1/livecd-20131120.1-armhf.out09:38
didrocksogra_: I guess you need a new image then09:38
ogra_?09:39
didrocksogra_: see the followup android upload at 5 UTC09:39
ogra_20131120-0225-0ubuntu1 is the last upload09:39
didrocksah, you got with that one?09:39
ogra_see above09:39
didrocksogra_: can't click on links right now :p09:39
didrocksif you follow what was discussed, system hosed ^09:40
jibelasac, confirmed: apt-get install libc6-amd64 && apt-get remove libc6-amd6409:40
jibeland get a broken system for free09:40
didrocksI would have prefered to not have it at all than for free ;)09:40
didrocksthanks for digging jibel :)09:41
ogra_didrocks, yeah09:41
asacjibel: thats on i386?09:42
jibelasac, amd64 with i386 enabled09:43
jibelwhich is the same cofigiration than didrocks. I can try on i386 too09:43
asackk09:43
asacdont worry. guess file a bug so we can have doko etc. look at it09:43
jibelk09:44
Mirvdidrocks: I filed bug #1253008 about pitti's worry about u-s-c configuration file removal (or actually it's a move to another package)09:59
ubot5bug 1253008 in Unity System Compositor "Clean configuration file on upgrade" [Undecided,New] https://launchpad.net/bugs/125300809:59
didrocksMirv: thanks10:00
ogra_r26 looks ok on magureo10:35
ogra_(havent done any call or sms tests, but it boots fine)10:35
popeyblimey, 101MB to go to 2610:42
popeydid you add the kitchen sink to it?10:42
=== vrruiz_ is now known as rvr
didrocksogra_: mind updating the spreadsheet with latest infos? I'm a little bit lost as the android update is still to TODO11:08
didrocksso not sure if the upload was that one (it seems so though)11:08
* ogra_ can look, but i wasnt involved in any of the last three android uploads11:12
ogra_(i didnt even know about them until i saw them on the changes ML)11:12
didrocksogra_: I guess the planned image # needs to change as well11:16
didrocksogra_: hum, ok, I think, I'll bump everyting to 2711:26
didrockswell, system settings is in 25 from what I see11:27
ogra_didrocks, oh, you wanted me to bump the others ?11:27
ogra_sorry, i understood i should sort the android line11:27
didrocksogra_: yeah, as they are not exact, but no worry :)11:27
ogra_sorry11:28
didrocksjibel: lool: can one of you confirm https://bugs.launchpad.net/ubuntu/+source/ubuntu-download-manager/+bug/1240656? that will get it moved and then we can kick a saucy image build11:29
ubot5Ubuntu bug 1240656 in ubuntu-download-manager (Ubuntu) "disable debug logging by default" [Critical,In progress]11:29
psivaaogra_: didrocks: not sure if that's known, but with image 26, we see "failed to copy '/var/lib/jenkins/phablet-flash/imageupdates/pool/ubuntu-418162f2a9ba9a320a2b460938e3d333f84b6f35826200c2a6b6c342ba1e21ef.tar.xz' to '/cache/recovery//ubuntu-418162f2a9ba9a320a2b460938e3d333f84b6f35826200c2a6b6c342ba1e21ef.tar.xz': No space left on device" on mako11:40
psivaahave seen that in two devices11:40
didrocksinteresting, as the broken recovery was installed in image 25? that can be it, right?11:41
ogra_psivaa, the recovery from r25 was broken ... could be that you still havethat in place11:41
ogra_i just did an OTA upgrade on maguro, that one worked fine11:41
psivaai tried on two devices, not sure if both of them had attempted r25 as well, i'll try another device11:42
ogra_didrocks, you should probably not use IRC nicks (or make clear they are IRC nicks) in mails to ubuntu-phone :)11:42
didrocksogra_: good point11:42
ogra_"infinity is looking into it" sounds a bit cryptic for non IRC users11:43
ogra_:)11:43
ogra_(like "eternity will solve it")11:43
Mirvgo, builders, go11:44
Mirvwaiting for amd64+i386 on https://launchpad.net/~ubuntu-unity/+archive/daily-build/+sourcepub/3671940/+listing-archive-extra ...11:45
didrocksogra_: well, it can be true :)11:45
ogra_haha11:45
* didrocks doesn't want to hear about amd64 and i386 in the same sentence today11:46
Mirvoh, how rude of me11:46
ogra_yippiiee11:49
ogra_root@ubuntu-phablet:/# lxc-console -nandroid -t011:49
ogra_Connected to tty 011:49
ogra_Type <Ctrl+a q> to exit the console, <Ctrl+a Ctrl+a> to enter Ctrl+a itself11:49
ogra_root@android:/ #11:49
ogra_it works !11:49
Saviqcihelp, hey, unity8 otto is getting stuck due to some dependency issue http://s-jenkins.ubuntu-ci:8080/job/autopilot-testrunner-otto-trusty/817/console11:51
SaviqI *feel* like it might be caused by https://code.launchpad.net/~kgunn72/unity-mir/bump-mir-dep-0.1.1/+merge/19352111:54
Saviqif unity-mir built against mir > 0.1.1 somehow11:55
Saviqand is kept in the mbs repo11:55
Saviqbut then there's no mir > 0.1.1 in the archive yet, so I'm not sure how that could happen11:55
SaviqMirv, ideas ↑?11:55
Saviqhow did that even get through autolanding?11:55
didrocksSaviq: the transition is in the daily-build ppa normally12:00
Saviqdidrocks, right... but otto isn't using it12:00
Saviqgrr12:01
didrocksSaviq: otto for integration tests is12:01
MirvSaviq: aanyhow, once the amd64 PPA build of this https://launchpad.net/~ubuntu-unity/+archive/daily-build/+sourcepub/3671940/+listing-archive-extra finally starts, we're really close of having Mir 0.1.1 in archives12:01
didrocksmaybe not for upstream merger? not sure12:01
Mirvmy plan is 1. wait for amd64 build of that, 2. upgrade desktop, 3. reboot and login, 4. publish mir + platform-api + unity-mir + unity-system-compositor12:03
SaviqMirv, ok12:03
Saviqwe'll just have to wait then12:04
Saviqdidrocks, you mean in cu2d it does?12:04
didrocksSaviq: yeah, when releasing, we are adding the ppa in the otto integration tests12:04
Saviqdidrocks, in -ci -autolanding it doesn't, as I understood it using daily is unsafe 'cause you never know if what you've built against is really going to end up in distro12:04
didrocksSaviq: yeah, and we have issues when there are some transitions12:06
didrockslike today12:06
didrocksthat's why we need CI Airline :p12:06
Saviqdidrocks, ok, so nobody is doing anything wrong, just that our tools are not letting do us what we really need, gotcha12:06
Saviqwe'll just wait, then12:06
didrocksSaviq: I guess yeah12:06
=== alan_g is now known as alan_g|afk
psivaadidrocks: ogra_: so the issue of no space in /cache/recovery/ on mako is not due to image r25 remnants, i flashed a device which had r23 and it still occurs12:18
ogra_weird12:19
didrockspsivaa: ok, if you changed the recovery as well and you're sure you don't have image 25 recovery, one less ;)12:19
psivaaon maguro this is not happening though12:22
=== alan_g|afk is now known as alan_g
=== MacSlow is now known as MacSlow|lunch
psivaadidrocks: the only files in /cache/recovery/ are very small log files so they can not be the reasons.12:31
psivaai mean before flashing with r2612:32
ogra_check one level up in /cache12:41
ogra_i think /cache is the partition, recovery/ is only a dir12:42
psivaaogra_: 32k on /cache12:46
ogra_used you mean ?12:47
psivaayes12:47
psivaabefore flashing with r2612:47
ogra_thats not much ... and should really not cause out of space warnings12:48
psivaaright during the flashing on *maguro i see 1069 M in /cach12:49
psivaa*cache12:49
psivaaogra_: not sure what's the capacity though on each devices12:51
ogra_http://phablet.ubuntu.com/gitweb?p=CyanogenMod/android_build.git;a=log;h=refs/heads/phablet-trusty12:51
ogra_i see nothing that even touches recovery ... strange12:51
psivaabut as you said it could be anything that goes into /cache12:52
psivaanow it's 1386M12:52
ogra_aha12:53
ogra_http://phablet.ubuntu.com/gitweb?p=CyanogenMod/android_system_core.git;a=blobdiff;f=init/init.c;h=e5f265858d0e0890deff18424ab09b174a4ce3f0;hp=462130711e331820467e2b8c427c1e2f7be9cdf2;hb=a1527ce86c90842b693941723888f7739f3bf473;hpb=7a318cec11f017c5ae9c186dae2b01f2dc4d3ef412:55
ogra_psivaa, does /sbin/recovery exist ?12:55
Mirvsil2100: I hoped you wouldn't block the stacks, that's why I asked for publishing jobs only12:56
psivaaogra_: the installation on maguro is finished and now /sbin/recovery is not there.. but i dont know if it would have existed during the flash12:58
ogra_psivaa, well, it should be existent in the recovery mode only12:59
ogra_oh, meeting ...13:00
* ogra_ totally forgot ... gimme a sec to relocate13:00
didrockssil2100: coming?13:02
jdstrandhey-- I'm looking at landing no 303 and wondering if now is the time to upload13:02
jdstrand(it is marked TODO)13:03
sil2100didrocks: I'll be a bit late, start off without meee13:03
sil2100Mirv: you told me that too late I guess, I just redeployed unity8 though13:05
Mirvsil2100: you didn't wait for answer :) you also started a build, so it's now waiting for the scope build13:05
sil2100Mirv: ok, so I'm not building/doing anything until I get a greenish light from you now ;p13:08
sil2100Mirv: since I would need to rebuild ubuntu-settings-components as well...13:08
sil2100But I'll do that after you're done ;)13:09
Mirvsil2100: ok :) let's wait for the current build to finish, then I'll publish, then I'll let you a sign that you can publish / do what you want13:09
sil2100;)13:10
Mirvsil2100: when the build job is published, isn't it in theory ok if I cancel the check job since you need to rebuil more anyway?13:12
Mirvs/published/finished/13:12
sil2100Mirv: you can cancel everything besides build ;p13:13
sil2100Mirv: since it's not being tested anyway anywhere13:13
Mirvsil2100: yeah. what I think I'll do that I'll wait for build to finish, then cancel check, then do my publishing and then I can let it run check again via foo13:13
sil2100Mirv: ok, but I think that for my needs a check re-run won't be needed anyway13:15
sil2100Since for me it's enough that it builds, so I can just do a direct force publish after your publishing is done13:15
Mirvright13:17
ogra_rsalveti, could the above git commit have any impact on that "/cache running out of space" issue when flashing ?13:17
Mirvsil2100: and actually the check didn't run since unity8 build happens to fail13:18
sil2100Oh13:18
Mirvwe should look into that at some point too, and file a bug as usual, there just hasn't been time for that13:18
Mirvand we probably need to remind the other team mates too that the usual cu2d vanguarding starts again now that it's up13:20
didrocksMirv: quite, publish mir then!13:21
didrocksquick*13:21
didrocksas the stack stopped :p13:21
Mirvdidrocks: already ... done!13:22
didrockssooo fast, it's not an airline, it's a jet \o/13:22
Mirvsil2100: so I'll wait still to see with my own eyes that the packages are in LP, and ping you after that13:22
ogra_rsalveti, oooh, i think i see whats wrong with your patch ... you actually want the mkdir for /proc and /sys outside of the "if" in http://phablet.ubuntu.com/gitweb?p=CyanogenMod/android_system_core.git;a=blobdiff;f=init/init.c;h=e5f265858d0e0890deff18424ab09b174a4ce3f0;hp=462130711e331820467e2b8c427c1e2f7be9cdf2;hb=a1527ce86c90842b693941723888f7739f3bf473;hpb=7a318cec11f017c5ae9c186dae2b01f2dc4d3ef413:24
ogra_psivaa, could you boot one of the failing devices into recovery and check that ?13:28
psivaaogra_: the devices are in the lab and the failing devices are not reachable by adb.13:29
ogra_hmm13:29
psivaai dont know any other way to connect them.13:29
ogra_so we need someone in the lab to check them ... ok13:29
psivaaogra_: yea waiting till rfowler comes online13:30
=== MacSlow|lunch is now known as MacSlow
Mirvsil2100: so "ping", feel free13:36
rsalvetiogra_: that patch is fine13:45
rsalvetiogra_: the problem was the previous one that got us into a broken recovery13:45
rsalvetiogra_: http://phablet.ubuntu.com/gitweb?p=CyanogenMod/android_system_core.git;a=commitdiff;h=aefb9653a367a6fd528f2ad48a1cce8945aecb3c;hp=87b40ee4a93bd682506fa082f2c2d1a0f5a60f0413:45
ogra_rsalveti, well, that still  doesnt create /proc or /sys13:46
ogra_it tries to mount them to a nonexisting mountpoint13:46
rsalvetididrocks: I did trigger another one with the fix included in it13:46
didrocksrsalveti: hum, it seems to have never gotten anywhere (we only had one new image)13:46
ogra_rsalveti, though i still dont have any explanation for the out of space issue13:46
didrocksor ogra's script is broken :p13:46
ogra_didrocks, ?13:46
rsalvetiogra_: that is ok in our default image, because they are created by lxc13:47
ogra_there are 20 and 20.113:47
ogra_20 is the broken one13:47
rsalvetiyes13:47
rsalveti20.1 is fine13:47
didrocksogra_: but you did build 20.1, no?13:47
ogra_rsalveti, in recovery ?13:47
rsalvetiogra_: in recovery it'll be created13:47
ogra_oh, i see what you are saying13:47
ogra_yeah13:47
rsalvetiogra_: +    if (stat("/sbin/recovery", &s) == 0) {13:47
ogra_well, still, all makos run our of space when flaching13:47
rsalvetithat's the fix I did yesterday when I noticed the regression13:48
rsalvetihow?13:48
didrocks12:40:27   psivaa | ogra_: didrocks: not sure if that's known, but with image 26, we see "failed to copy13:48
didrocks                  | '/var/lib/jenkins/phablet-flash/imageupdates/pool/ubuntu-418162f2a9ba9a320a2b460938e3d333f84b6f35826200c2a6b6c342ba1e21ef.tar.xz' to13:48
didrocks                  | '/cache/recovery//ubuntu-418162f2a9ba9a320a2b460938e3d333f84b6f35826200c2a6b6c342ba1e21ef.tar.xz': No space left on device" on mako13:48
didrocksrsalveti: ^13:49
ogra_rsalveti, <psivaa> ogra_: didrocks: not sure if that's known, but with image 26, we see "failed to copy '/var/lib/jenkins/phablet-flash/imageupdates/pool/ubuntu-418162f2a9ba9a320a2b460938e3d333f84b6f35826200c2a6b6c342ba1e21ef.tar.xz' to '/cache/recovery//ubuntu-418162f2a9ba9a320a2b460938e3d333f84b6f35826200c2a6b6c342ba1e21ef.tar.xz': No space left on device" on mako13:49
rsalvetididrocks: why did you remove the emulator?13:50
rsalveti:P13:50
rsalvetiogra_: right, but is 26 = 20.1?13:50
ogra_rsalveti, he did an apt-get autremove ...13:51
ogra_rsalveti, it is13:51
didrocksrsalveti: yeah, I know you wanted to punish people doing that! :p13:51
didrocksI'll note this for next time "never ever remove what rsalveti installed on your system" :)13:51
rsalvetiogra_: right, not sure if that is indeed using the cache partition, need to give it a try here13:51
ogra_rsalveti, r26 works fine on my maguro with OTA upgrade13:51
rsalvetiogra_: I know the cdimage one works fine13:51
rsalvetijust flashed it13:51
rsalvetiogra_: are we flashing the lab devices with -b?13:53
ogra_could be, not sure13:53
ogra_doanac` should know13:53
ogra_or sergio13:53
psivaathe devices in the lab use - phablet-flash ubuntu-system -b --channel trusty-proposed13:54
plarspsivaa: I just got to my desk, looks like there's some trouble this morning?13:55
plarsrsalveti, ogra_: we are using -b for quite some time now13:56
psivaaplars: yes, a couple but the latest impacting one is that flashing on mako fail due to no space available error on /cache/recovery13:56
rsalvetilet me flash my mako to see13:56
rsalvetithe cdimage one worked fine13:57
MirvMir 0.1.1 now in release pocket13:57
Mirvand time for UDS13:57
plarspsivaa: we seem to have a lot of devices showing up with the 0123456789ABCDEF adb id again now14:04
psivaaplars: yes, vila mentioned about it. dont seem to be an id that's used for smoke14:05
plarspsivaa: it's a generic one that's used when it can't sort out the real one14:05
plarspsivaa: we'll probably need rfowler to reflash them14:06
psivaaplars: ahh ok, could be that we have a bunch of mako's that failed to flash due to the no space reason and they could be it14:06
psivaaplars: on another issue, we had adb protocol fault in kinnara in the morning14:07
plarspsivaa: just one? :)14:07
plarspsivaa: we do get it a lot less frequently I think, but it still happens from time to time14:07
psivaaeven running 'adb devices' from kinnara constantly threw this error14:08
psivaaplars: how do you normally recover from it14:08
plarspsivaa: oh, that's different14:08
plarspsivaa: in a case like that, I guess we'd just need to restart adb server... I'm not sure I've seen that happen before14:09
plarspsivaa: we added quite a few more devices to that system though14:09
ogra_plars, psivaa, this will be fixed within the next weeks14:10
psivaaplars: yea, it turned out to be one of the old adb processes was causing the issue14:10
rsalvetiplars: ogra_: mako worked fine here with system-image14:10
ogra_very weird14:10
psivaaplars: killing that process fixed it14:10
plarsrsalveti: you flashed with -b also?14:10
psivaaInstalled: 1.0+14.04.20131108-0ubuntu1 is the phablet tools in the lab14:11
rsalvetiplars: yes14:11
ogra_i dont think it has anything to do with phablet-tools, after all the out of space message seems to come from the device itself14:11
rsalvetiplars: http://paste.ubuntu.com/6448304/14:12
plarsbrb14:12
psivaaogra_: ok14:14
rsalvetiogra_: so I think the cache related issue could be a side effect from the broken image14:22
ogra_rsalveti, thats what i thought too, but -b pushes a new recovery14:23
ogra_so it should boot the fixed one14:23
ogra_(it actually boots the new recovery via fastboot)14:24
rsalvetiogra_: right, but that doesn't necessarily mean that it'll clean up the cache partition14:25
rsalvetiINFO:phablet-flash:Clearing /data and /cache14:26
ogra_h,mm, i thought we do that from ubuntu_commands or how thats called14:26
rsalvetibut it should in theory clean it as part of the flashing process14:26
plarsso was the bad recovery actually on image 25 then?14:26
ogra_right14:26
ogra_plars, yes14:26
psivaaplars: ogra_ i tried flashing from image 23 as well14:28
psivaaand the same thing happend14:28
ogra_psivaa, if you use -b it weill use the new recovery ... and if you were unlucky you had r2514:29
ogra_though looking at the above it smells more like the cleanup step failed14:30
plarsindeed14:30
ogra_but you should have seen that in the log14:30
plarsogra_: I don't see anywhere in phablet-tools where fastboot erase is called, could be that we need to really erase those partitions14:31
ogra_i dont think it calls erase14:31
rsalvetiplars: the clean is a shell call done in recovery14:32
ogra_yeah14:32
rsalvetinot via fastboot14:32
rsalvetiso that might be failing14:32
plarsrsalveti: yeah, it just does rm14:32
rsalvetiplars: you might want to force a clean with fastboot then and retry to see14:48
plarsrsalveti: yeah, I'm trying to reproduce locally. Flashing 25 right now.14:49
sergiusensfastboot -w created too many support issues14:55
ogra_yeah14:55
ogra_well, i can imagine that if there was no /dev and no /proc in r25's recovery that the rm might indeed have failed ... i just dont get why it would also fail in r2614:56
=== alex-abreu|afk is now known as alex-abreu
rsalvetiplars: were you able to reproduce the issue locally?15:22
plarsrsalveti: yes, but psivaa said he's seeing the issue going from 23->26 also15:23
rsalvetioh =\15:23
plarsrsalveti: I was also able to get it to flash just fine from 25->26 after it failed once... it was in recovery and not getting the bad adbid like we see in the lab15:23
plarsrsalveti: and I just added -d mako15:23
plarsrsalveti: it all just worked... so I'm trying to figure out what's going on15:23
=== alan_g is now known as alan_g|tea
=== alan_g|tea is now known as alan_g
plarspsivaa, rsalveti: so I just watched it get the same error when trying to go from 26->2515:55
plarswait, maybe not15:55
plarsthis one might have just been an adb failure15:55
plarsadb/mtp thing probably15:56
rsalvetiright15:56
rsalvetiI think you might be able to extract the logs from the broken flags if you're able to boot the device15:57
rsalveticheck /cache/recovery/last_log15:57
plarsrsalveti: ok, after coming back up, all I have is busybox and there's no fstab or /cache. If I mount mmcblk0p22 by hand though, I found what seems to be a cache partition there, and here's the last_log: http://paste.ubuntu.com/6448925/16:22
plarsrsalveti: this was after installing 25, and then trying to install 2616:22
plarsoddly: 'Skipping missing file: version-25.tar.xz' was in the log16:23
rsalvetiplars: hm, right, but the log is from 25 I believe16:25
rsalvetiweird16:25
plarsrsalveti: yeah, so maybe we didn't get far enough for that log to happen16:26
ogra_if you are in busybox you are not in recovery16:26
ogra_so indeed there is no /cache16:26
rsalvetiyeah16:26
plarssince it happened on the adb push, that makes sense16:26
rsalvetiplars: let me try to flash 25 to see16:27
rsalvetiplars: got the issue here16:40
rsalvetiplars: so, what happens, when flashing 25 with -b it will first flash the recovery image with fastboot16:40
rsalvetiplars: then it'll try to load recovery, which will fail16:40
rsalvetiplars: meanwhile it waits for adb, thinking that the recovery will be up16:41
rsalvetias recovery fails to boot, the device boots the previous image instead16:41
rsalvetiand phablet-tools thinks that the recovery is then up, because adb is there16:41
plarsrsalveti: ah, I see. so we think we're booting 25 but really we're not16:41
plarsrsalveti: I see16:42
rsalvetihttp://paste.ubuntu.com/6449025/16:42
rsalvetisergiusens: we should be checking for /sbin/recovery when flashing16:42
rsalvetisergiusens: to check if we're indeed in recovery16:42
sergiusensrsalveti, sure, I'm already checking for something; checking for recovery specifically shouldn't be complicated16:43
sergiusensrsalveti, my wonder is, what did it boot?16:43
sergiusensplars, any indications of why it failed to boot?16:44
rsalvetisergiusens: it fails to load the recovery, then it tries to boot the previous image16:44
rsalvetibut it seems that the 25->26 migration worked fine16:44
sergiusensrsalveti, yeah, but why? :-)16:44
rsalvetiso the error when copying the file into /cache happens because it's not really in the recovery16:45
plarsrsalveti: really? for me, I get a broken system - just boots to busybox16:45
rsalvetisergiusens: that's the default behavior16:45
sergiusensrsalveti, how could've recovery been broken :-)16:45
rsalvetisergiusens: we had a regression16:45
sergiusensrsalveti, that's the how I'm asking :-)16:45
sergiusensack16:45
plarsrsalveti: you are flashing cdimage? or ubuntu-system image?16:45
sergiusensplars, rsalveti super easy to have a better check if we are in recovery, no worries16:45
rsalvetisergiusens: http://phablet.ubuntu.com/gitweb?p=CyanogenMod/android_system_core.git;a=commitdiff;h=aefb9653a367a6fd528f2ad48a1cce8945aecb3c;hp=87b40ee4a93bd682506fa082f2c2d1a0f5a60f0416:46
rsalvetiwhich was fixed at http://phablet.ubuntu.com/gitweb?p=CyanogenMod/android_system_core.git;a=commitdiff;h=a1527ce86c90842b693941723888f7739f3bf473;hp=7a318cec11f017c5ae9c186dae2b01f2dc4d3ef416:46
rsalvetiplars: ubuntu-system16:46
didrocksogra_: can we get a new image build?16:46
didrocksogra_: now that Mir entered :)16:46
plarsstrange... for me, I'm getting exactly what we have seen on multiple devices in the lab16:46
rsalvetididrocks: sure :-)16:46
didrocksrsalveti: thanks!16:47
ogra_didrocks, yeah16:47
sergiusenswhy does everyone get to push but me? :-P16:47
didrocksit's christmas16:47
ogra_rsalveti, do you already ?16:47
didrocksI saw some snow this morning ;)16:47
rsalvetididrocks: having newer images is always a good thing16:47
rsalvetididrocks: easier to test and revert broken changes as well16:47
ogra_we should just build one every hour !16:47
rsalvetiogra_: nops16:47
rsalvetiogra_: yeah16:47
ogra_then i'll do16:47
sergiusensrsalveti, ogra_ you will starve the testing infra though16:47
didrocksrsalveti: fully agreed, we just need to have a way to communicate those to you and us more easily16:47
rsalvetididrocks: why do we need the communication here?16:48
ogra_sergiusens, nah, we'll have our fast emulator and all will be fine16:48
rsalvetiwe don't16:48
rsalvetithe communication is the changelog :-)16:48
sergiusensogra_, big LOL16:48
ogra_:)16:48
didrocksrsalveti: well, we do plan landings on certain image/try to communicate that16:48
rsalvetididrocks: and that's wrong16:48
sergiusensogra_, the emulator would be mostly useful for apps, but not image testing ;-)16:48
rsalvetididrocks: you should only know when something lands, not planning when it'll land16:48
ogra_=== Image r27 building ===16:48
didrocksrsalveti: will be easier once we have the CI Airline I hope to know more on feature-based what landed16:49
sergiusensogra_, we can certainly only run on real hw if the emulator test pass16:49
ogra_sergiusens, well, i thought the plan was to use it in all testing infra16:49
rsalvetiwe don't need to know when it'll be landing16:49
rsalvetiwe just need to know when it *landed*16:49
didrockswell, we want to know when big thangs are landing at the same time16:49
didrocksto avoid clashes16:49
didrocksbut apart from that, yeah16:49
ogra_right, only have a test on real HW if all emu tests were good16:49
ogra_didrocks, and that ifno comes from the changelogs16:50
rsalvetiright, but even if we get clashes, we want more images to test all the steps16:50
didrocksogra_: note on a "feature-approach", and not planning clashes16:50
ogra_rsalveti, the point is that asac wants us to have stuff tested in full context before we upload16:51
ogra_nobody does that16:51
ogra_(or do you run the full AP suite for an android change you do)16:51
rsalvetiogra_: right, but that's not a blocker here16:51
rsalvetihaving more images is *always* a good thing16:51
ogra_thats true16:51
ogra_but sergiusens has a point16:51
ogra_we'd saturate the lab16:51
rsalvetiright, that's our only limitation16:52
rsalvetibut we can do as most images we can depending on how long it takes to validate an image16:52
rsalvetiwhat would that be?16:52
rsalvetiat every 4 hours?16:52
asacrsalveti: in theory we can produce images in parallel and validate them16:52
asacthe problem is not parallelizing the validation16:52
asacbut the building16:53
asacits not really clear what atomic snapshot you will take when you start the job i think :)16:53
ogra_asac, we only talk about the actual images16:53
ogra_not about any parallel test builds etc16:53
asac(though you could mitigate that by pulling the packages file out of the tool and passing itas an input)16:53
ogra_we currenntly only build two per day at most16:53
asacright16:54
rsalvetiand need to speed this up16:54
asacso for now we talk about the future anyway because emulator is not there etc.16:54
ogra_having a more fine grained view in the changelogs through building more of them will make identifying issues a lot easier16:54
rsalvetiright, but I think we can already improve by building at least a few images a day16:54
asachowever, if the emulator is fully landed, i anticipate that we can produce as manages as we want a day.16:54
ogra_from a builder POV we could build one every hour16:54
rsalvetias we always get stuff that lands in the archive which is not  part of the landing spreadsheet16:54
asacjust not test all on all phones16:54
rsalvetiso it's *always* good to produce and test newer images16:54
ogra_30min for cdimage + 15min for system-image post processing16:55
asacwe would basically validate all iamges in emlator and then pick those that we are interested in (for instance, because we consider them a release image) and test them  everywhere16:55
asacwe discussed hooking image productionm up with the publisher runs16:55
asacso we get one image somewhere stored for each injection of what really goes in the archive16:55
asachas to wait for cloud buildd which is worked on somewhere16:56
* cjwatson forces android-emulator back in despite p-m not quite understanding multiarch16:56
asacp-m?16:56
cjwatsonproposed-migration16:56
asacah16:56
cjwatsonasac: https://rt.admin.canonical.com/Ticket/Display.html?id=62272 and https://bugs.launchpad.net/bugs/124746116:57
ubot5Ubuntu bug 1247461 in launchpad-buildd "Move live filesystem building into Launchpad" [High,In progress]16:57
ogra_right, that will give us parallel builds16:58
josephtasac: once an image + new packages is promoted don't all previous images need to be updated and retested?16:58
asacnot really. all we do is snapshotting the archive after a publisher run in the form of an image16:59
ogra_why would they ?16:59
asacso whatever is in there is already finally committed16:59
asacuntil overwritten for which a new image would exist16:59
josephtwe create an image for archive+A and archive+B in parallel, archive+A passes all tests so now archive = archive+A, if archive+B passes all tests we don't know that archive+A+B will pass all tests and need to test that scenario, right?17:01
rsalvetiasac: right, but let's try to improve our current situation17:02
asacjosepht: yeah you are right, but you are thinking about a different level17:02
rsalvetihow long does it take to test an image currently?17:02
rsalvetiplars: ^17:02
asacjosepht: you are thinking about the premerge level... there indeed we need to invalidate A after we land B17:02
plarsrsalveti: somewhere around 3 hours I believe17:02
asachwoever, we can try to be smart and only invalidate if the changes are in the same dependency tree for instance17:02
plarsrsalveti: assuming everything goes right :)17:02
didrocksplars: but then, you rekick jobs, right?17:02
plarsdidrocks: yes17:02
asacjosepht: we talk about images produced after the fact from whatever landed in the real archive17:03
didrocksso how long until we get a clean state, I guess17:03
didrocksand you rekicked enough? :)17:03
plarsdidrocks: if I'm up and see it needs to be done, and there's not another image already coming17:03
josephtasac: ah I see17:03
rsalvetiright, can we build a new image at every 6 hours then?17:03
* asac hope that makes sense17:03
rsalvetior 8?17:03
plarsdidrocks: for instance, if the devices could work right now, I wouldn't bother retesting because I see there's a new image on the way, and it would just delay17:03
asacrsalveti: well. so i dont want random images17:03
asaci want precisely cut images17:03
rsalvetiasac: why not?17:03
rsalvetiasac: that's *wrong*17:04
asacits not :)17:04
asacnot sure how that can be wrong :)17:04
rsalvetiasac: we want one image per change17:04
rsalvetior the most images we can17:04
asacright.17:04
rsalvetiotherwise it's a pita to find regressions17:04
rsalvetione image a day is *wrong* and only causes pain17:04
asacone image per chagne... but now that we can't get that (for various reasons) short term, it doesnt mean we should give up17:04
cjwatsonan image isn't any more precisely cut just because somebody asked for it manually17:04
asacand cut images in the middle of changes :)17:04
cjwatsonthere shouldn't be a "middle of changes"17:05
rsalvetiasac: give me one example17:05
asacrsalveti: i am supportive of more images per day. just not at cronjob kicked17:05
rsalvetiexactly17:05
asaccjwatson: depends on which level you define changes :)17:05
rsalvetionce it's in the archive, there's no "middle of changes"17:05
cjwatsonget your packaging metadata right so that proposed-migration can migrate it properly17:05
rsalvetiexactly17:05
cjwatsonand then roll17:05
rsalvetiwe should put it in cron17:05
cjwatsonmanual image building requests are a workaround for lazy metadata maintenance17:06
rsalvetiimage is just an archive snapshot17:06
asac"get your packaging metadata right" ... please give up on that hope. all you can hope for is "get youir packaging meta data right most of the times please"17:06
asacbut even if everyone does it right most of the times, if the engineering team is big enough you will always have problems :)17:06
cjwatsonI'm not going to give up on that hope, because working on that basis made, empirically, a *massive* difference to the day-to-day quality of Ubuntu.17:06
ogra_cjwatson, *my* manually built images are *allways* more precisely cut, i sand and olish them :P17:06
ogra_*polish17:06
cjwatsonIn fact I think it's mostly reality rather than hope.17:07
rsalvetias long we're not saturating our test lab, we should be fine to produce the most images we can17:07
asacexactly... its mostly17:07
asac:)17:07
cjwatsonYes, there's the odd exception, but they're, well, exceptional17:07
cjwatsonWe shouldn't structure processes around exceptions17:07
rsalveti+117:07
asacand will alwayus be mostly... but 600* mostly might mean rarely :)17:07
ogra_the point of automated builds is to simply have an as small changeset as possible ...17:07
ogra_doing only two manual builds completely goes against that17:08
ogra_and causes lots and lots of guesswork on our side until we found the actual issue17:08
rsalvetiyeah17:08
asacits human. noone is perfect. hence designing the process must work in continuous exceptions due to human failures induced17:08
rsalvetihorrible17:08
rsalvetisee the udev one we had a few weeks ago17:08
ogra_which wastes more developer time than having more images per day17:08
rsalvetithe package was uploaded at wednesday and we found the issue at saturday17:08
ogra_asac, its wasting human resources17:08
asacogra_: the problem can be solved by using a cronjob or by kicking builds again at a 3 times a day schedule17:09
asacwith shbuman smartness17:09
ogra_asac, how often did we get onto the wrong path in the recent times if an image was broken17:09
ogra_wasting hours to find it was the other change that we didnt look at yet17:09
rsalvetiright, even if we build it at every 12 hours, it's already better than what we have now17:09
ogra_rsalveti, we do17:09
asaci agree that we should revisit the image schedule :)17:09
rsalvetiogra_: not currently17:09
ogra_we usually have two manual built ones per day17:10
rsalvetiwell, it's manual17:10
asacnot sure if crontab is the right approach. lets talk about it tomorrow on the standup17:10
ogra_(if there isnt UDS)17:10
asaci would love to see at least 3 images :)17:10
ogra_so the timing is already 12h17:10
rsalvetiogra_: right, I want it to be automatically triggered17:10
rsalvetiogra_: as the archive is always moving17:10
ogra_right17:10
ogra_4 per day should be fine17:10
ogra_automated and all17:10
cjwatsonI'm completely boggled at the insistence that things must *not* be automated, frankly17:10
rsalvetican we try to put it back to the cron and see how it goes?17:10
cjwatsonIt goes against all the engineering practice I know17:11
rsalvetihaha, exactly17:11
ogra_cjwatson, all but asac and didrocks in here say we want automated17:11
didrocksogra_: I didn't say we don't want automation, I said that we want to cut after each big change as a first approach17:11
cjwatsonI understand "automation is hard and we can't quite manage it yet" (have said that myself on various occasions) - I just don't understand "the ideal is manual"17:11
ogra_right, but why ?17:11
asacfor the record: noone says "we dont want automation" :)17:12
ogra_didrocks, you land all your stuff in one chunk anyway17:12
asaclets discuss it tomorrow17:12
asaci will join you guys there17:12
ogra_where ?17:12
cjwatsonOK, so you're arguing against regularity but not against automation, I see17:12
asacstandup17:12
cjwatsonI disagree but at least that isn't so boggling :-)17:12
asachehe17:12
asacits also not exactly what i am saying :)17:12
ogra_asac, ah, the landing team one17:12
asacbut closer17:13
rsalvetiasac: why tomorrow?17:14
rsalvetiasac: let's just put it back to cron today and see how it goes17:14
didrocksogra_: if you land all the stuff in one chunk, well, you have to debug it, you decided the size of the chunk :)17:14
rsalvetiwhy would that cause us any pain now17:14
asaci want to first ensure people understand what we are saying :)17:14
didrocksogra_: we need to be able to produce images after each "chunk" landing, but a chunk is defined by the developer itself17:14
asacI dont see us having any pain from having manual image building17:14
asacit gives us many good17:14
rsalvetiyou can still manually trigger new images17:15
rsalvetibut we should *always* have it in cron17:15
ogra_didrocks, thats not true ... your chunks are controlled in the spreadsheet17:15
rsalvetibecause the archive is *always* moving17:15
ogra_not defined by a developer17:15
didrocksogra_: I just want to avoid A + B entered, "A says I don't care, it's B", it's not me, and B says "I don't care, it's because of A"17:15
asacso ... i know for sure that we will spin more than enough images once the landing stuff is fully back operational again17:15
rsalvetiall I'm asking is to add it back in cron at every 8/12 hours17:15
didrocksogra_: well, not really in our current model, but yeah, we would need to define priorities rather than image #, agreed17:15
asachence, i believe the issue you try to solve is the low velocity due to the 1ss move paired with a UDS going on now :)17:15
didrocksogra_: and I think we do align with what rsalveti told seeing it that way17:16
asacand yes, we should build at least 2 images a day even if not much is going on17:16
ogra_didrocks, so we can switch cron back on and set it to say 6h ?17:16
rsalvetiI just want to make sure we always have a fallback in case we don't have humans to trigger a new image17:16
ogra_(to have 4 images / day)17:16
asacwell. as i said befeore the capacity on testing is really bound to emulator :)17:17
didrocksogra_: well, for me, it would be a +1 but rather 8 hours TBH, to let the CI team to rekick the tests jobs and giving some slack for that). 3 images a day will be a start17:17
ogra_asac, we should have as many as the infra allows imho17:17
asacso ... dont hurry17:17
rsalvetiogra_: yeah17:17
didrocksogra_: but I don't think I'm the one able to take the decision :)17:17
rsalvetiogra_: I'd first try with 8h17:17
ogra_didrocks, why do you care if it is 4 or 8 ?17:17
rsalvetiogra_: and we can improve that as we go17:18
didrocksogra_: because then, the testing infra can't rekick the jobs17:18
ogra_if your team only lands something every second image thats fine17:18
asaci care because i want to ensur3e the CI folks that support those images whenthey hit validation17:18
didrocksif a new image is produced17:18
asacknow what to do and how to prioritize17:18
didrocksogra_: they can't "restest an older image"17:18
ogra_asac, but the CI team knows when cron runs17:18
asacanyway17:18
didrocksogra_: it's not only that, most of their time is just "relaunch the jobs"17:18
didrocksbecause the whole thing is flacky17:18
asacnot sure why is it suddenly such a big issue todya?17:18
rsalvetithat's why let's try with either 8h or 12h17:18
ogra_and as the one who does the manual builds i must say that the requirement for manual builds is largely always at the same time17:18
didrocksasac: and so you need to reboot the device17:19
didrocksrsalveti: +117:19
ogra_so why not just have images inbetween too ?17:19
ogra_rsalveti, i dont get it17:19
didrocksogra_: because you don't let the CI team to get test results17:19
rsalvetiogra_: we can, I just want to make sure we're not saturating our test lab17:19
ogra_why does having 2 images help more than 417:19
didrocksif you kick a new image17:19
didrocksbecause we give them more time to rekick the jobs17:19
didrockswhen you finished building a new image17:20
didrocksthey can't retest an older one17:20
ogra_didrocks, but if they dont make this image they will make the one in 4h17:20
ogra_i dont get it17:20
rsalvetiyeah17:20
asacif noone can tell me why this is so much more important today than yesterday, lets really talk tomorrow in the standup.17:20
asac:)17:20
ogra_there is nothing that puts time pressure on anyone17:20
didrocksogra_: but then, there is the same issues, they need to rekick17:20
rsalvetithat's why as long we're not saturating the lab, we're fine17:20
didrocksogra_: and you just rebuild the next next image17:20
didrocksand the story repeats :)17:20
ogra_just make sure you land your stuff inbetween two builds if you want to make sure all of it lands at the same time17:21
rsalvetiasac: why can't we just add it there and see what happens17:21
asacrsalveti: right now until we have the emuilator fully rolled out its about saturating support folks17:21
ogra_and the schedule is known17:21
ogra_didrocks, why do they need to rekick17:21
ogra_or even care for the images17:21
didrocksogra_: repeating "because it's flacky"17:21
asacrsalveti: understand taht i have to figure how to deal with that with CI process17:21
rsalvetiasac: sure, but all I'm asking is more images than we have today, and that's for sure something we can improve even without the emulator17:21
ogra_they just need to check the results for the image their change landed in17:21
didrocksogra_: see a vanilla test run17:21
ogra_no matter when they landed their stuff17:21
didrocksbefore plars or psivaa starts looking at it17:21
didrocksyou have a lot of tests not running17:22
asacrsalveti: not without undersanding and clarifyuing with my team what they should focus on testing etc.17:22
didrocksor with just 1 test17:22
asacrsalveti: on technology side its easy17:22
didrocks(and beeing representated as "green" btw :/)17:22
didrocksogra_: so, they restart the jenkins jobs as of now17:22
asacif you guys feel so strongly about more images17:22
rsalvetiright, but you don't get that we're also wasting resources by not triggering more images17:22
asaclets really check how we can get that17:22
rsalvetibecause it's really a pain to find and revert regressions17:22
asaci am not sayuing we shouldnt have more images17:23
rsalvetiif we get tons of changes at every image we produce17:23
asacjust lets discuss how to do that brest17:23
didrocks(btw, while we discuss, we still can't test an image I guess and powerpc is broken in distro ;))17:23
ogra_asac, it wastes immense amount of manpower to just make the CI team happy17:23
ogra_while they would just have to obey to a cronned build schedule17:23
ogra_finding an issue in an image that has 50 packages changed is lots and lots harder than finding it in one that only has 1017:24
rsalveti+10017:25
rsalvetiit's just stupid to fight automation17:25
ogra_right17:25
rsalvetias the archive is always moving17:25
ogra_asac, i worked with this system for 4 months now ... and thats the result of my experience17:25
ogra_we waste manpower and that doesnt need to happen17:26
asaci am the first who wants more images. if the landing team would be operational they would at least spin 2 images a day17:27
asacit should be closer to 317:27
ogra_and its a trivial change ... just a crontab entry17:27
asacand as we add more "smart automation" we woulc increase the pace17:27
ogra_but why do we have to make it depend on the landing team17:27
ogra_the landing team can land stuff even with automated builds17:27
rsalvetiyeah, that's the wrong, as the landing team doesn't have the power to freeze the archive :-)17:27
rsalveti*that's wrong17:28
ogra_its not like it is hard, the schedule will be known so you know when your stuff gets into the next image17:28
rsalvetididn't make at one image? wait for the next17:28
ogra_right17:28
rsalvetijust don't make everyone to wait on your call17:29
jibelcihelp: publisher on d-jenkins fails to create new jobs on public instance17:29
rsalvetiabout triggering a new image17:29
jibelexample http://d-jenkins.ubuntu-ci:8080/job/trusty-adt-python-misaka/1/?17:29
retoadedjibel, looking17:29
jibelretoaded, it is only the creation of a new jobs, publication of new builds work fine.17:32
retoadedjibel, it wouldn't happen to be a matrix job would it?17:34
tedgNot sure who to blame, but having DNS on the CI VPN is pretty cool.  Thanks!17:35
jibelretoaded, it is17:37
retoadedjibel, http://10.98.3.6:8080/plugin/build-publisher/instance/0/publisherThread/currentState/output  makes me think that it could be related to the just updated matrix-reloaded plugin.17:37
retoadedjibel, i will probably have to revert it back.17:37
jibelretoaded, I'm not sure, I see no reference to the plugin in the trace. But if it's a recent change revert it back and we'll see17:40
retoadedjibel, it near the bottom: Caused by: java.lang.NullPointerException at hudson.matrix.MatrixProject.onLoad(MatrixProject.java:474)17:43
retoadedjibel, and as soon as the currently running jobs complete I'll revert17:43
jibelretoaded, it is a reference to a Matrix object from jenkins core, not the plugin.17:49
ogra_=== Image r27 DONE ===17:51
* popey updates17:51
ogra_(totally forgot that over the abive discussion)17:51
ogra_*above17:51
jibelretoaded, it is not a new problem apparently17:53
jibelretoaded, trusty-adt-simplejson failed to publish too17:53
jibelretoaded, oh, and it vanished from d-jenkins17:56
jibelWTF17:56
Ursinhaogra_, can I update my phone then? :P17:56
jibelretoaded, so on the filesystem there is a /var/lib/jenkins/jobs/trusty-adt-simplejson but not on the UI17:57
ogra_Ursinha, well, my maguro just updated fine here17:57
jibelretoaded, and publication fails with the same error 50017:57
ogra_bah, since when is facebook on the home lens17:58
Ursinhaogra_, are you using -proposed? read-only?17:58
ogra_yes17:58
Ursinhaogra_, at least since yesterday when I flashed mine with the current devel17:58
ogra_well, i do OTA updates17:58
ogra_Ursinha, heh, that was ricardos fault ... r25 was broken17:59
Ursinhaogra_, it rarely is rsalveti's fault, hehe17:59
ogra_haha17:59
ogra_never ever :)17:59
Ursinhararely ;)17:59
rsalvetilol18:03
Ursinhaogra_, so I heard you test all the images :P I have a nexus 4 and whenever I put some music on and change to another app the music stops playing after a few seconds, is that a known issue?18:41
popeyit shouldn't18:42
ogra_Ursinha, i test only on maguro ...18:42
UrsinhaI just updated with the latest -proposed image18:42
popeywhat image are you running Ursinha ?18:42
ogra_popey, is the makoman18:42
popeyUrsinha: are you using the music app or something else?18:42
Ursinhapopey, Home lens, click on music and then play18:43
=== fginther changed the topic of #ubuntu-ci-eng to: Ubuntu CI Engineering Team | Vanguard: fginther | Landing instructions: http://paste.ubuntu.com/6292280/ | Known issues: Some services are down (1SS move), network slowness
Ursinhaabout this phone says r2718:43
popeymy phone is dead.. one moment18:43
popeyworks fine here18:45
Ursinhapopey, music app has the same behavior, but the interesting thing is that the music stops playing but when I return to the music app it resumes playing as if it never stopped18:45
Ursinhathere's no sound but it seems the app was playing it18:45
popeyi am listening to music which is playing in the music app but I'm on the home screen18:45
popeyit even plays when locked18:45
popeyand continies to play after a song finishes18:46
Ursinhapopey, what should I do then?18:46
popeyis it launching the music app ?18:46
Ursinhapopey, yes, it launches and I'm able to listen to music, the only problem is when I change screens it stops after a few seconds18:47
popeyhttp://popey.com/~alan/phablet/device-2013-11-20-184641.png18:47
popeylike that18:47
Ursinhait's playing as it should (I think), displaying the cover and band/song name18:47
popeywas this a clean install or an update to an existing install?18:47
Ursinhapopey, it was a clean install yesterday, with the current devel image18:48
popeyodd18:48
UrsinhaI switched to -proposed today and the problem is the same18:48
popeyI dont understand why it's working for me and not you18:48
Ursinhahehe18:48
Ursinhapopey, btw it keeps playing when screen is locked18:49
Ursinhapopey, I'll try a clean install and let you know18:50
cjwatsonhttp://people.canonical.com/~cjwatson/tmp/emulator-yay.png18:51
cjwatsondidn't *quite* manage to get it up on the hangout just now ...18:51
ogra_:)18:51
ogra_thats quite a wide screen18:52
cjwatsondual monitor18:52
* ogra_ has triple ... but screenshots always only show one screen 18:52
cjwatsonimport(1) doesn't really understand that and just gives me the whole thing.  I should probably use the screenshot thing in the desktop but old habits die very hard18:52
ogra_might be nvidias fault18:52
ogra_ah18:53
ogra_yeah, i only use alt+print18:53
fginthercyphermox, kenvandine, robru, we need to restart q-jenkins as soon as possible19:09
robrufginther, fine by me...19:10
kenvandinefginther, go for it19:10
cyphermoxyeah, fine19:10
fgintherthere is a build in progress (for 11 hours) is it safe to kill ti?19:10
fgintherit19:10
robrufginther, yeah, whatever. we can restart it if necessary. i'm not waiting on anything myself19:11
robrufginther, also 11 hours usually indicates a problem anyway19:11
fgintherrobru, I kinda figured that :-)19:11
fgintherrobru, kenvandine cyphermox thanks for the input19:11
=== Ursinha is now known as Ursinha-afk
=== Ursinha-afk is now known as Ursinha
rsalvetiUrsinha: background music working fine with image 27, let me help you debugging the issue20:19
Ursinharsalveti, okay, thanks20:19
fginthercyphermox, is it difficult to re-deploy all of the cu2d jobs?20:47
cyphermoxwell, i'd love to avoid it if possible, why?20:48
fginthercyphermox, the jenkins root moved from /var/lib/jenkins to /iSCSI/jenkins. and some of the scripts had the old path hard coded20:49
fginthercyphermox, I have MPs to update the scripts and the job. I just want a plan for how best to do this20:49
fgintherwould prefer to actually do the update tomorrow when more help is available20:50
cyphermoxare you saying you would prefer or are you asking my opinion?20:58
fginthercyphermox, I just want some idea of what is involved? does it require lots of downtime, do we need to make special arrangments to make it happen, etc.?21:00
cyphermoxnot really, I think we just need to update the paths in lp:cupstream2distro and lp:cupstream2distro-config (if any there), redeploy all the jobs, etc, make sure all the necessary machines have their copies of those branches updated21:03
cyphermoxit would take an hour of work maybe?21:03
fginthercyphermox, good. I do have MPs here:21:03
cyphermoxbut I could see it being done much more cleanly if we use a symlink as a temporary step to avoid breaking things21:03
fgintherhttps://code.launchpad.net/~fginther/cupstream2distro/jenkins-home/+merge/19584821:03
fgintherhttps://code.launchpad.net/~fginther/cupstream2distro-config/update-stack_status/+merge/19595621:03
fginthercyphermox, things are already broken :-(21:05
cyphermoxhow so?21:06
fgintherthe check jobs are looking for job data under /var/lib/jenkins. it's not there21:06
cyphermoxright21:06
cyphermoxthat's partly why i mentioned a symlink21:06
cyphermoxif not /var/lib/jenkins -> /iSCSI/jenkins then one for just cu2d21:07
cyphermoxthat allows a smooth transition, but not meant to stay in place21:07
fginthercyphermox, hmm, ok21:07
cyphermoxso, both branches look fine21:07
fgintherthanks, I'll email out a plan21:08
cyphermoxfwiw I see other places in cupstream2distro-config that need a path change21:10
fginthercyphermox, ack! can you add them to the bug: https://bugs.launchpad.net/cupstream2distro/+bug/125275021:47
ubot5Ubuntu bug 1252750 in Canonical Upstream To Distro "latest_autopilot_results hardcodes a jenkins root directory" [Undecided,New]21:47
cyphermoxdo you have an equivalent bug for cupstream2distro-config?21:48
fginthercyphermox, no21:48
robrufginther, no CI after 5 hours? can you look please? https://code.launchpad.net/~robru/friends-app/autopilot-py3/+merge/195979 thanks21:50
fgintherrobru, all the phones are blocked due to an image flash problem21:52
fgintherrobru, it is being worked21:52
robrufginther, oh ok thanks. just really curious about what happens with that branch ;-)21:53
robruno worries21:53
Saviqfginther, afraid makos started to get stuck in flashing, too: http://s-jenkins.ubuntu-ci:8080/computer/mako-04cb53b598546534/? http://s-jenkins.ubuntu-ci:8080/computer/mako-04cbcc545f5328a5/?23:01
fgintherSaviq, thanks for the ping. Someone is working on it right now, but they have to manually reflash each device this time23:04
fgintherSaviq, all three of them were dead a couple hours ago23:06
asaccurious: what happened to them?23:06
Saviqfginther, ah23:06
fgintherasac, AIUI there was a problem with the image today, plars do you know more?23:06
asacok thanks. nevermind then23:08
fgintherwere also still being hit by https://bugs.launchpad.net/phablet-tools/+bug/124916223:09
ubot5Ubuntu bug 1249162 in android-tools (Ubuntu) "Devices lose adb connection after phablet-flash loop" [Undecided,New]23:09
* asac wonders what a phablet-flash loop is :)23:10
asacok i think i see what it tries to describe23:11
fgintherthe phablet-flash loop is a way to reproduce the problem we see on our devices23:13
asaccheck my comment please23:13
=== fginther changed the topic of #ubuntu-ci-eng to: Ubuntu CI Engineering Team | Vanguard: cihelp | Landing instructions: http://paste.ubuntu.com/6292280/ | Known issues: Some services are down (1SS move), network slowness
sergiusensasac, fginther Saviq recovery was broken, that was the problem23:47
sergiusensI'm adding a safeguard to phablet-flash to avoid flashing if recovery is broken23:48

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!