/srv/irclogs.ubuntu.com/2014/02/26/#ubuntu-ci-eng.txt

xnoxsergiusens: Saviq: bzoltan: didrocks: you've got mail, requesting a silo reconfiguration =) (and to steal branch from Saviq & bzoltan for sergiusens to land) I hope we can agree to do that =)00:06
ricmmSaviq: its a lot easier to crash after a while00:25
rsalvetiricmm: no deal?02:00
sil2100hm, my thunderbird went nanners08:47
sil2100Can anyone paste me the hangout link? ;)09:30
sil2100Ah, nevermind now!09:30
mardydidrocks: hi! Is it possible to have several branches for the same project in a silo? Will they get automagically merged?09:36
seb128mardy, yes, they get merged in the order you list them in the request09:44
mardyseb128: cool, thanks09:45
seb128yw09:45
dbarthhi, what's the weather like for the CI train this morning? ie do you think landings may resume today?09:46
didrocksdbarth: not now, we still have the dashboard failing09:47
dbarthok, nw; will land in a ppa for now09:48
psivaadidrocks: ogra: systemsettle failure on default set is because it was using an identical script but from a different location, which was not changed.10:13
didrockspsivaa: ahah10:13
didrocks:)10:13
psivaai'll propose a fix with symlinking to the original one10:13
ogra=== Image 209 Building ===10:13
didrocksyeah, seems better than duplication :)10:13
didrocksthanks psivaa10:14
didrocksand let's see if we can this in time for 20910:14
ografunny :)10:14
psivaayw :)10:14
ograthe system-image part takes significantly longer, you have plenty of time :)10:14
sil2100hmmm10:21
sil2100I might have some clue on what makes the terminal test fail, just need to figure out what is causing this10:22
Saviqhuh? stable == saucy still?10:37
didrocksSaviq: yeah, it's been a long debate, not sure we should focus our strength on that today :)10:38
Saviqdidrocks, nah, I just went for ubuntu-device-flash and was surprised it chose image 104 for me :)10:41
* ogra thinks we should drop these channels altogether ... 10:42
ogradefault and proposed is all we should have10:42
didrocksSaviq: I think we can bring that back on the table once the transition to 4.4 is done10:42
didrocksogra: +110:42
Saviqyeah, ETOOMANYCHANNELS10:43
ograstable was theoretically thought for developers to get a stable base for their apps ...10:44
ograbut we are moving way to fast for that imho10:44
didrocksindeed10:44
ograoooh10:45
ogradidrocks, the new logo is beautiful10:45
* ogra sees itfor the first time10:46
didrocksogra: heh, thanks ;)10:46
ogradancing friends :)10:46
didrocksthanks to my wife to have detour it (even with the svg, to downscale it and not having weird outer circle "moving"… wasn't that straightforward)10:47
didrocksheh :)10:47
seb128what logo?10:47
xnoxseb128: the new recovery, has a spinning ubuntu cof upon bootstrap flash.10:47
ograseb128, new recovery mode10:47
seb128oh ok10:48
ograthe dead android with rotating guts is gone10:48
didrocksthat was really annoying me \o/10:48
ogra++10:48
xnoxdidrocks: can we get rid of the choppy progress bar?10:50
ograthats harder than replacing the logo10:50
ogra(needs actual code changes)10:50
xnoxdidrocks: or like slow it down? i'm not resurrected robocop to register that it's working properly, instead of looking choppingly fast.10:50
didrocksxnox: slowing it down is easy, the issue is that we can get a progress bar with the number of steps, but some steps are taking seconds, other minutes :p10:51
ograwow10:58
ogramy device just rebooted in the middle of a unity8 test run10:58
ograhmm11:00
ograWARNING: in file "/usr/lib/python2.7/dist-packages/unity8/shell/tests/__init__.py", line 148 in setUp11:00
ograThis function is deprecated. Please use 'the Touch class to instantiate a device object' instead.11:00
ograi thought it is python3 now11:00
ograhmm, the camera fake thingie in unty8 doesnt fit on the screen on flo11:04
davmor2Morning all11:05
xnoxogra: nah, my branch to switch it to python3 is on hold.11:05
ograah, k11:06
xnoxogra: and it doesn't fix / change classes.11:06
xnoxogra: something about unity8 crashing, and that being priority to fix.11:06
ograyep, i know, i was in the meeting this morning :)11:06
didrockssometimes GAS will turn me crazy11:10
sil2100It's probably poison GAS11:12
sil2100What's up?11:12
didrocksoh, I found the fix11:13
didrocksbut using "name" and 'name' is different if you pass that to a function11:13
didrocksit's because of their internal caching creating bugs11:15
sil2100;/11:16
Laneygas the gnu assembler?11:18
didrocksgoogle apps scripts11:20
ograumm11:20
ogrash: 1: gcc: not found11:20
ogradpkg-architecture: warning: couldn't determine gcc system type, falling back to default (native compilation)11:20
ograwhy does the unity8 test use gcc11:21
didrocksogra: it's autopilot11:21
ograah11:21
didrocksand this is the case for a long time11:21
didrocksthat we have that error11:21
ogranever noticed it11:21
ogra(but i also rarely watched the terminal while the tests were running, it probably just scrolled offscreen)11:21
didrockssil2100: Mirv: robru-is-dying: cyphermox: ok, so now, when you add more lines to the spreadsheet, it should apply the formatting for you (phew)11:21
didrocksogra: yeah, it's at the beginning of a test or testsuite11:22
Mirvnice..11:22
ograright11:22
ograRan 46 tests in 1373.092s11:22
ograFAILED (failures=1)11:22
ograRestoring shell11:22
ograunity8 start/running, process 1358611:22
ogranot that bad11:22
didrocksyeah, same than in the dashboard11:22
didrocksbecause of the crash11:22
sil2100\o/11:23
ograERROR: unity8.indicators.tests.test_indicators.IndicatorTestCase.test_indicator_exists(Bluetooth)11:23
ograand thats expected ... since flo has no working BT yet11:23
didrockssil2100: Mirv: tell me if you find any bug, I had to rewrite some formulas for that11:23
ograhmm11:25
ograso http://ci.ubuntu.com/smokeng/trusty/touch/flo/208:20140226:20140224/6826/unity8/ looks quite different from ym test11:25
ogra*my11:25
didrocksogra: again those StateNotFoundError: State not found for class 'DefaultIndicatorWidget' and filters {'objectName': 'indicator-bluetooth-widget'}.11:26
didrocksanother one is the crash11:26
ogradidrocks, well, there is no BT indicator on flo11:26
ograi had the crash (which actually caused a reboot) the run after the crash was fine apart from the BT test11:27
didrocksogra: ok, so only the crashes11:27
didrocksthe 3 others11:27
didrocks(no process found with pid…)11:27
ograi only had one11:27
ograsee above11:27
didrocksyeah, it's random, so you can get lucky11:27
didrockstalking about the dashboard, seems there were 311:27
ograright11:27
* ogra re-runs11:28
ogradidrocks, oh, i totally forgot to mention in the meeting ... when we re-added click-update-manager (which had been dropped in 194) inn image 202, we didnt promote 202 which means users cant upgrade their apps at all atm11:39
didrocksogra: yeah, one of the issue is that we don't have full test results on 202 nor dogfooding11:40
ograi was wondering (sice we're stuck atm anyway) if we should put some effort into gettin 202 late promoted11:40
didrocksso, if the CI team can maybe rerun 202 in parallel?11:40
didrocksthat makes sense11:40
didrockspsivaa: will that take you a long time to get 202 running all tests? ^11:40
ogramight clash with 209 now though :/11:41
didrocksyeah11:41
didrockslet's wait for his feedback11:41
ogra=== Image 209 DONE ===11:41
psivaadidrocks: i get the setup going, it wont take long but for the tests to complete it will take around 4+ hrs.11:41
psivaai could run 202 on a different device though11:41
psivaabut we wont have the results on the dashboard11:41
ograthat might be good11:42
didrockspsivaa: if some tests are failing or need rerun…11:42
psivaadidrocks: that's fine. they will be isolated11:42
didrockspsivaa: you will still be able to just run the testsuite failing?11:42
didrocksok :)11:42
didrocksso yeah, if you have time, please :)11:42
didrocksthen, if the results are good, we can ask our lovely dogfooders11:42
psivaadidrocks: ack, will kick one11:43
didrocksthx!11:44
bregmahey didrocks, how do I get a project to start using ci-train... just add a MP and it will automatically climb aboard, or is there some manual adjusyment the Landing Team has to make?11:53
didrocksbregma: is that a new project or something that was under daily release before?11:53
bregmain this case, it's the unity8-desktop-session project, so, new11:53
bregmaI also need to get geis and libgrip releases out, but I'm pretty sure they need manual reconfiguration11:54
didrocksah ok, so a MP is enough to get it (just ensure we can bzr bd)11:54
didrockscan you please add it as well to https://docs.google.com/a/canonical.com/spreadsheet/ccc?key=0Au6idq7TkpUUdC05a2ZQSmgwU2NFYnJQOE9qMDRYa3c#gid=1?11:54
didrocksthat helps tracking11:54
bregmaI don't believe I have edit access, let me try11:54
didrocksyou should have11:55
didrocksbregma: btw, you probably want to take bamf (and some of the scopes?)11:56
didrocksas a Lander11:56
didrocksI see they still don't have any landers11:57
bregmathe scopes should all be mhr311:57
didrocksmhr3: taking them? ^11:57
didrocksthostr_: ^11:58
bregmaOK, I have edit, adding and updating11:58
didrocksI only see the mediascanner one11:58
didrocksbregma: thanks!11:58
didrocksbregma: it will by default collect from rev 0 for the changelog though, if you released one version, ensure it's tagged with the packaging version then11:58
bregmadidrocks, up to now I used standard debrelease techniques, it should be tagged12:00
didrocksexcellent, we should be fine then :)12:00
thostr_bregma: well, the old scopes is unity7 only that's why I though you'd take those12:00
bregmadidrocks, how about geis, it has pending changes that haven't landed in distro for some time and I have no new MPs at the moment, so it will need a trunk flush, correct?12:01
didrocksbregma: yeah, just propose an empty MP to flush those12:04
didrocksMirv: mind unwiring from daily release and update the items without CI Train "yes" on https://docs.google.com/a/canonical.com/spreadsheet/ccc?key=0Au6idq7TkpUUdC05a2ZQSmgwU2NFYnJQOE9qMDRYa3c#gid=1, please?12:04
thostr_didrocks: do we need to explicitly move all old scopes to ci, or can we rather do that once we have changes?12:07
thostr_didrocks: ... which I don't expect any time soon12:07
didrocksthostr_: we can do that once you have change, just be aware that we'll disable everything shortly though12:08
thostr_didrocks: sure12:08
didrocksthostr_: however, once you have change, it will take more time to migrate12:08
didrocksso be aware of that and don't ask the impossible later on :)12:08
thostr_didrocks: yes, understood12:09
* bregma always expects the impossible from didrocks12:12
didrockstssss ;)12:12
psivaaom26er: hey do you think bug 1276747 could also be a reason why turning the screen on failing in http://pastebin.ubuntu.com/6999455/12:16
ubot5bug 1276747 in autopilot (Ubuntu) "after starting a new app cannot get proxy object for unity8" [High,Confirmed] https://launchpad.net/bugs/127674712:16
psivaadidrocks: http://q-jenkins.ubuntu-ci:8080/job/psivaa-trusty-touch-mako-smoke-daily/3/console is flashing with 20212:17
didrockspsivaa: excellent, thanks!12:17
=== alan_g is now known as alan_g|afk
mhr3didrocks, what, what, what?12:19
mhr3what am i supposed to take?12:19
om26erpsivaa, no, this log seems to show it could not find proxy for unity8, could be unity8 crashed12:20
didrocksmhr3: backlog just before my hilight :)12:20
psivaaom26er: unity8 did not crash on those instances12:20
mhr3didrocks, ok, a sec, in meeting12:21
om26erpsivaa, let me try to reproduce that, not sure if 'com.canonical.Shell.BottomBarVisibilityCommunicator' exists these days12:22
psivaaom26er: thanks. http://pastebin.ubuntu.com/6999481/ contains some more information indicating unity8 is running12:23
om26erpsivaa, where can i find the unlock screen script ?12:24
psivaaom26er: and this does not occur all the time: http://bazaar.launchpad.net/~ubuntu-test-case-dev/ubuntu-test-cases/touch/view/head:/utils/target/unlock_screen.py12:24
om26erthat code looks familiar12:27
=== alan_g|afk is now known as alan_g
psivaaom26er: if it helps, this behaviour started only this week and during the weekend we had autopilot, unity and android updates.12:29
om26erpsivaa, i am trying on the very latest image, seems to work for me multiple times. did anyone reproduce the issue locally as well ?12:30
psivaaom26er: i dont think anyone tried this. and it happens about 1 in 5 times i'd say12:31
Mirvdidrocks: ok, so both updating daily release config for ci train 'yes' if not already plus adding possible entries in configs that are not in the sheet12:32
om26erpsivaa, i have a question is the script running as root in CI ?12:33
didrocksMirv: thanks!12:33
psivaaom26er: i think so, let me confirm12:33
om26erpsivaa, also i want a complete log, seems i have been able to reproduce the issue here12:34
psivaaom26er: it is being run as root12:36
didrockspsivaa: hum, default system settle for 209 didn't use your change?12:36
psivaaom26er: https://jenkins.qa.ubuntu.com/view/Trusty/view/Smoke%20Testing/job/trusty-touch-mako-smoke-daily/90/ has the necessary logs12:37
psivaaom26er: 'adb-shell /home/phablet/bin/unlock_screen.sh' is when it starts for all the app tests12:38
psivaadidrocks: the MP needs approval and merge. dint want to push it straight12:38
didrockspsivaa: ah ok12:38
ograhmm12:54
ograso i can enable BT on the android side of the flo12:54
ograand apparently it even comes up ... but the indicator doesnt seem to notice12:54
didrocksapw: hey, do you think it will be possible to have the same "core up" behavior than in 4.2? (so cores not shutting down)12:59
didrocksjust to run one full test suite with that12:59
didrocksand see if the flakyness that are revealed is due to that12:59
=== alan_g is now known as alan_g|lunch
* ogra files bug 1285146 for cyphermox 13:06
ubot5bug 1285146 in bluetooth-touch (Ubuntu) "Bluetooth on flo can be easily enabled but the indicator does not recognize it" [High,New] https://launchpad.net/bugs/128514613:06
mhr3didrocks, noone's touching the old scopes, so let's deal with it if/when they need to be touched13:07
didrocksmhr3: ok, just note my remark on the introduced delay then once you want to land it13:09
Mirvdidrocks: ok sync done. all packages having in CI Train already set were disabled correctly, but I was wary of setting any disabled-in-daily-release-config to be automatically CI Train 'yes'. instead, for you to review: http://pastebin.ubuntu.com/6999661/13:10
mhr3didrocks, if there's just a simple fix needed or something trivial can't it use no-train?13:10
didrocksmhr3: no, we are not going to support the old infra anymore13:10
mhr3didrocks, there's still manual uploads ;)13:10
didrocksmhr3: well, we don't do that for our projects :p13:11
mhr3didrocks, clearly i need to become ubuntu developer so i can screw with you ;)13:11
didrocksmhr3: better to keep one process only13:11
didrocksmhr3: don't worry, a lot of people are using their time to accomplish this :p13:12
mhr3didrocks, aaah, so i wouldn't be the only one? that sucks13:12
didrocksMirv: ok, I flipped the switch for those having a landers13:13
didrocksMirv: but it seems you basically missed the ## Stack oif.cfg ##13:13
didrocksMirv: and account-plugins13:14
didrocksclick-apparmor13:14
didrocksand another click13:14
Mirvdidrocks: hmm?13:14
didrocksMirv: basically, everything that has a lander is up to ci train13:14
sil2100bregma: so, unity7 is tested, right?13:15
Mirvdidrocks: oif is covered as being enabled in daily release config, so covered by pastebin's first line. the rest, I'm probably missing some pass you were wanting? I worked through the daily-release configs and compared those to the spreadsheet, but I didn't go through the spreadsheet by itself as such13:15
sil2100bregma: do any of those changes there could affect touch?13:16
didrocksMirv: yeah, so when there is a lander, we need to disable them from daily-release13:16
didrocksMirv: as well as upstream-merger, and then, set the "In CI Train" to yes13:16
Mirvdidrocks: ok, so if lander name is there, it means need to disable, ok13:16
didrocksyep :)13:16
Mirvdidrocks: that's the pass I'm missing13:16
sil2100Exactly ;)13:16
didrocksMirv: sorry, it seems they are just 6 of them :)13:17
didrocksfginther: please, when you add projects or change cu2d-config, please reflect the change in the spreadsheet as well: https://docs.google.com/a/canonical.com/spreadsheet/ccc?key=0Au6idq7TkpUUdC05a2ZQSmgwU2NFYnJQOE9qMDRYa3c#gid=113:17
didrocksfginther: and don't enable upstream-merger :)13:17
Mirvdidrocks: so actually just oif needs disabling in config, I think. click-* were not in daily release, and I now moved accounts-plugins also under the title I created "Other CI Train packages not formerly in daily release system". https://code.launchpad.net/~timo-jyrinki/cupstream2distro-config/sync_citrain_disable_oif/+merge/20836113:22
didrocksMirv: hum, you need to disable upstream merger as well for them, they were not in it?13:23
om26erpsivaa, i have come to a situation where the unlock script prints that unity8 has been restarted but it is actually not, probably due to a bug somewhere on unity side or our config13:23
didrocksMirv: hum, you only changed the daily-release tag, there is more than that, remember at Bluefinn?13:23
bregmasil2100, yes, tested and FFe acked, all changes are only in Unity7 (the removed config option was never used anywhere) so no Touch effects13:23
didrocksMirv: let me find you a commit13:23
Mirvdidrocks: in stacks/head, I find no trace of apparmor/click-apparmor/click in projects:13:24
didrocksMirv: you need to add:13:24
om26erat that time the screen is indeed turned on *but* starting unity with 'initctl start unity8' does not actually show it on my screen13:24
didrocks+     daily_release: False13:24
didrocks+     autolanding_template: False13:24
didrocks+     use_stack_ppa: False13:24
didrocksMirv: to every project that you are disabling ^13:24
Mirvdidrocks: right.. adding13:24
didrocks(so probably account-plugin, click-* as well)13:24
om26erplars, do you manage the unlock screen script ?13:24
sil2100didrocks: ^ can I publish unity7 then? :) I'll double check the FFe bits before that13:25
didrocksom26er: it happens quite often since the android 4.4 switch FYI13:25
didrockssil2100: yeah :)13:25
om26erplars, how important is it to restart unity8 after every test suite ?13:26
Mirvdidrocks: pushed to the branch for oif, still not finding accounts-plugins, click-* in the daily release config with grep13:26
sil2100didrocks: ok, Laney ACKed the FFe's, so it's indeed all green13:26
didrocksMirv: oh nothing? interesting13:26
sil2100bregma: btw.! Wow! I see you guys come from the future ;D "Unity7 bugfixes 2024.02.24" ;)13:27
didrocksMirv: I would still add the 3 lines for "dead" projects13:27
didrocksMirv: just in case :p13:27
bregmageez, one typo and they never let you forget13:28
ograsil2100, bregma, oh horror !!!!!13:28
* ogra really doesnt want to see us using unity7 in 2024 still13:28
sil2100bregma: ;)13:28
om26erdidrocks, plars psivaa we can probably add tests for the unlock script itself13:28
didrockspeople mixing numbers13:29
didrocksI never do that!13:29
ograexcept in lotteries13:29
didrocksom26er: yeah, but first, let's get it fixed :)13:29
ogra:)13:29
didrocksogra: ahah13:29
apwdidrocks, the change was not to whether they were up, it was how they are reported in the idle stats; right?13:30
didrocksMirv: and hum stacks/head/webcred.cfg:    account-plugins:13:30
Mirvdidrocks: aha, why was I grepping accountS-plugins13:30
Mirvjust a moment13:30
didrocksapw: not sure, no power management change as you know of?13:31
didrocksit's only on the reporting side?13:31
om26erSaviq, hey! unity8 does not start and if i start it from the terminal i see this: http://paste.ubuntu.com/6999748/13:31
om26erdoes that mean anything ?13:31
apwdidrocks, i don't have that information without waiting for this stupid thing to charge, but the issue with the test suites was about reporting _only_13:32
ograom26er, did you use upstart to start it ?13:32
Mirvdidrocks: ok now the merge request page is updated13:32
didrocksapw: yeah, but we have other "weird" new races and so on since the switch itself, so I was wondering if they changed something around CPU or scheduling (and if we can proove that theory)13:33
sil2100didrocks: packaging ACK needed: http://162.213.34.102/job/landing-012-2-publish/lastSuccessfulBuild/artifact/packaging_changes_unity_7.1.2+14.04.20140225-0ubuntu1.diff <- new dep in main, seems ok13:33
om26erogra, with initctl it gave: http://paste.ubuntu.com/6999760/13:33
apwdidrocks, i didn't see a heap of nasty looking change, but presumably we tested these kernels before switching13:33
ograom26er, as phablet user ? and is lightdm running (and thus unity-system-compositor)13:34
sil2100Great, I just broke my unity7, geh ;)13:34
om26erogra, yes as phablet user, i ssh'd in13:34
didrocksapw: ok, false hope then :)13:34
=== alan_g|lunch is now known as alan_g
om26erogra, lightdm have 2 instances and unity-system-compositor have 113:35
ograok, thats fine then13:35
ogra(something seems to hog /tmp/mir_socket or so)13:36
didrockssil2100: +113:37
didrocksMirv: hum, seems you missed my remark…13:38
didrocks14:27:56 didrocks | Mirv: I would still add the 3 lines for "dead" projects13:38
didrocks14:27:59 didrocks | Mirv: just in case :p13:38
om26erogra, rm -r /tmp/mir_socket ?13:38
ograom26er, no, thats privided by unity-system-compositor ... you would rip the carpet out underneath it13:38
ograom26er, rather check if there is another unity8 stuck or so13:39
ograit hooks into that socket13:39
Mirvdidrocks: sorry, there are 8 different cases. so add where for which dead projects? I somehow skipped that as being that "yes keep those mentions in the daily-release config proposal that I copied from spreadsheet"13:40
Mirvdidrocks: this syncing is complex :)13:40
om26erogra, seems our unlocker tries to remove mir_socket http://bazaar.launchpad.net/~ubuntu-test-case-dev/ubuntu-test-cases/touch/view/head:/utils/target/unlock_screen.py13:41
om26erline 2913:41
didrocksMirv: yeah ;) basically the idea is "disable daily release and upstream merger for the dead projects"13:41
didrocksMirv: so adding the same 3 lines in the configuration13:41
didrocksfor them13:41
ograom26er, hmm, that feels rather wrong, ask mterry once he comes online ... he designed the nested Mir mode we use today13:42
Mirvdidrocks: ah, adding the three "disabling lines"! great.13:43
=== psivaa is now known as psivaa-lunch
ogradidrocks, ^^^^ that could cause a lot of unity8 test issues actually13:43
didrocksyep13:43
didrocksogra: yeah, that's probably the cause of so many phablet-test-run which don't start13:43
didrockspsivaa-lunch: right? ^13:43
ograi'm not sure if /tmp/mir_socket gets dynamically re-created13:44
ograbut i doubt it13:44
Mirvdidrocks: pushed and LP is updated13:44
ograand unity8 kind of wants to connect to it13:44
jdstranddidrocks, Mirv: I don't know what this oif stuff is, but apparmor, apparmor-easyprof-ubuntu and click has me as a lander but getting them into the citrain fold is still in progress (I started it but got pulled away)13:44
ograjdstrand, you can always dput to a landing PPA13:45
ograwhich i guess is at least hafl a CI train ...13:45
ogra*half13:45
jdstrandtrue, I'll keep that in mind13:45
didrocksjdstrand: yeah, but we can disabling them from daily-release and automerging :)13:46
jdstrandI'll get it done-- I just haven't had landings to request yet (things got delayed there)13:46
psivaa-lunchdidrocks: ogra: om26er: 'rm -f /tmp/mir_socket' is there for long time and 'rm: cannot remove '/tmp/mir_socket': Operation not permitted' has also been reported before13:47
jdstranddidrocks: I'm not sure I understand13:47
psivaa-lunchwithout much harm13:47
ograpsivaa-lunch, yeah, it shouldnt be removed ever with nested mode13:47
didrocksjdstrand: if you don't, basically don't worry, we just unplug from the old system :)13:47
ograbut i guess if you run it as phablet user you wont have the permissions13:47
ographablet@ubuntu-phablet:~$ rm /tmp/mir_socket13:48
ograrm: cannot remove '/tmp/mir_socket': Operation not permitted13:48
ograyeah13:48
jdstranddidrocks: ok, thanks13:48
ograom26er, so you iare safe as long as you dont run it as root13:48
psivaa-lunchogra: ack13:48
* om26er back in a few minutes, doctor appointment13:49
=== om26er is now known as om26er|doc
bzoltansomething is horrible wrong with my ISP ... I got 0.06MBs upstream13:54
plarsom26er|doc: I had nothing to do with that script, iirc you were the original author of it. :)  I really still think there needs to be some way to call into unity itself where it knows how to do all of that without having to use autopilot externally13:55
plarsom26er|doc: unlock has been occasionally failing to unlock the screen lately though13:55
plarsfor the last week or so it seems, before that it was pretty reliable13:56
didrocksplars: yeah, seems it's since the 4.4 switch13:56
bregmadidrocks, any suggestion what component unity8-desktop-session should go under in the CiTrain spreadsheet? (unity8.cfg makes some sense, but I don't know what these .cfg things are)13:57
rsalvetididrocks: I got quite a few ascii errors when running tests with flo as well13:57
rsalvetinot with mako though, not sure what causes it13:58
didrocksbregma: I would say "misc", but you can add to the end, it doesn't make sense anymore as we don't have stack configs13:58
didrocksrsalveti: ascii errors, like invalid locales?13:58
rsalvetiyeah, usual locale error with python13:58
didrocksinteresting, I guess no langpack difference and we do have utf8 there as well. Apart if we go to some error path and the output isn't correct binary/ascii convertion14:00
rsalvetibut nice that it's part of the dashboard now :-)14:00
rsalvetiright14:00
didrocksrsalveti: we really see various timeouts14:02
didrocksso, if the power management changed in the 4.4 kernel… maybe it just reveals some flakyness14:03
didrocks(that we have in our code)14:03
apwdidrocks, ok ... the old kernel on a very old image i happen to have here ... shows that we run with a single cpu normally also14:03
didrocksapw: oh, really?14:03
apwas i said it was just a statistics reporting change14:04
* didrocks doesn't understand "shows that we run with a single14:04
didrockscpu normally also"14:04
didrocksseems it's not reporting change14:04
apwdefine reporting change ?14:05
apwthe online cpulist from the kernel is '0' on my phone by defualt, ie a single cpu online called 014:05
rsalvetiyeah, I got the impression that we're no using 4 cpus14:05
rsalvetiand before we were using 2 the most14:05
rsalveti*we're now14:05
apwhard to tell without having some what to make the thing busy i guess14:06
didrocksrsalveti: at least, that would explain most of the flakyness we are seeing I guess14:07
apwdidrocks, i hate to think we write code which isn't smp aware14:07
apwits all in go right14:07
rsalvetididrocks: yeah14:08
rsalvetiwe can now run many more threads in parallel :-)14:08
apwrsalveti, what do you base it only using 2 on ?14:09
fgintherdidrocks, ack14:09
rsalvetiapw: when debugging pulse I was only getting cpu1 on/off14:09
rsalvetiwith 4.4.2 I was getting cpu 1,2 and 3 on/off events14:09
apwrsalveti, ok i've got an old installed here, and with a cpu loop i can get all cpus online no problem, so it is possible to use them all14:10
rsalvetibut we can check by flashing the older image14:10
ograwe should really try to run with 0 CPUs14:10
ograsaves so much battery14:10
rsalvetiright14:11
apwmy older image deffo can use all 4, and indeed is getting nice and warm doing exactly that14:11
plarsoh nice - flo seems to be getting past the webbrowser tests in 208 and 20914:11
plarshttp://ci.ubuntu.com/smokeng/trusty/touch/?show_all=114:12
rsalvetimaybe the governor is behaving differently now somehow14:13
ograrsalveti, is it still using the same governor ?14:15
ogra(we still force ondemand on the ubuntu side)14:15
=== retoaded changed the topic of #ubuntu-ci-eng to: Ubuntu CI Engineering Team | Vanguard: retoaded | CITrain support - US: robru, cyphermox, rsalveti - EU: sil2100, Mirv, didrocks | CITrain support no answer: use mup bot after 30 minutes, but choose right timezone | Landing instructions: http://goo.gl/8H1Du3 | Known Issues: -
=== om26er|doc is now known as om26er
om26erplars, the issue on hand right now is that sometimes unity8 does not start even initctl tells us it did14:20
elopioasac: can you please comment on what we are missing for https://bugs.launchpad.net/ubuntu-ci-services-itself/+bug/1275012 to happen?14:21
ubot5Ubuntu bug 1275012 in Ubuntu CI Services "Add a job to run all the image tests with qt5.2" [Undecided,New]14:21
plarsom26er: well that could keep us from unlocking it I suppose :)14:21
om26erplars, yep ;)14:22
didrockselopio: hey, have you seen my answer on the nostate error?14:23
elopiodidrocks: not yet. On the bug you mean?14:24
elopiodidrocks: oh, got it. I just marked as incomplete the one in the dialer, because I ran it 5 times on my phone and the dash is now green.14:25
elopioI've been trying to run the ones in weather app, but something is funny here. I'm reflashing to try again.14:25
didrockselopio: see my answers and mostly, look at the dashboard :)14:26
didrockselopio: it's our first input source14:26
didrocksand as you can see, things are really flaky (look at different results in different images)14:27
didrocksso we'll need your expertise I guess in how those tests are misbehaving14:27
asacelopio: who from CI team have you been working with on this until now?14:28
elopioasac: nobody.14:28
elopiodidrocks: yes, I'm trying to reproduce the errors here. I just started with the one that strangely fixed itself :)14:29
elopioabout the clock, I've made a branch that should improve the alarm tests a little: https://code.launchpad.net/~elopio/ubuntu-clock-app/page_object-alarm_tests/+merge/20798814:29
asacelopio: so i was told this was agreed work with the CI team. not really great to hear today that this wasn't done now.14:29
elopioI'll ask nik about landing it.14:29
didrockselopio: oh nice! balloons, sergiusens ^14:29
elopioasac: maybe somebody agreed on it, but didn't update the status of the bug.14:30
asacdoanac`: what do you think about https://bugs.launchpad.net/ubuntu-ci-services-itself/+bug/1275012?14:31
ubot5Ubuntu bug 1275012 in Ubuntu CI Services "Add a job to run all the image tests with qt5.2" [Undecided,New]14:31
asacev: ^^14:31
popeyMirv:                         text: "Author: Mike Sheldon <elleo@gnu.org>\n\nReleased14:32
popeyunder the GPL, version 3.0 or later.";14:32
popeyit's those \n's that are doing the boxes14:32
asacelopio: in the future, please dont committ to engineering team that you will make stuff happen that you can't happen alone without aligning engineering work done by other teams. this commitment from QA caused people to believe all is going fine and well, and the CI team is deep in the hot phase of an important customer project right now14:33
asacanyway, lets see14:33
Mirvpopey: file a bug against ubuntu-ui-toolkit first, maybe they can retarget if it's nothing they can affect14:34
Mirvpopey: I was able to see it now by installing TouchWriter14:34
popeyok14:34
didrocksMirv: did you push your changes? I missed your ping probably14:34
asacelopio: is this just duplicating the autopilot job or what does this entail?14:35
elopioasac: as far as I can remember, I never commited to this task. That's why I didn't assign it to myself.14:35
elopiobut well, I suppose I didn't make it clear, I'm sorry.14:35
asacelopio: right. be careful. thanks.14:35
asacelopio: so the "in the meanimte" comment ... is that still valid?14:36
elopioasac: duplicating the job, they also want a view like the dashboard to get a nice view of the errors.14:36
asacelopio: thats not going to happen i am sure14:36
asacbut guess its not critical14:36
asacas long as the data is there someone can write a perl script or something14:36
thostr_sil2100: can I get a silo for line 30?14:36
elopioand I think we need clear direction of when we can run this job, because I heard somebody saying that it might starve the other jobs that need hardware.14:37
Mirvdidrocks: I was thinking you'd approve https://code.launchpad.net/~timo-jyrinki/cupstream2distro-config/sync_citrain_disable_oif/+merge/20836114:37
elopioasac: yes, we can use the autopilot job whenever they are not trying to make a release.14:37
didrocksMirv: approved! remember you need to redeploy on q-jenkins14:38
asacelopio: sonuds like that could be mitigated through manual coordination for a few days14:38
Mirvdidrocks: yes, redeployed already. thanks!14:39
asacelopio: so http://q-jenkins:8080/job/autopilot-release-gatekeeper/42/ seems to already show what tests failed14:39
elopioasac: ok, we can do that. Who should we coordinate with in order to start this job? didrocks? Or we just run it whenever we want?14:40
asacelopio: guess thats good enough from dashboard point of view14:40
asacelopio: my understanding is that this autopilot job is something that the qa tools team (thomi/veebers) is using internally, so thjink its coordinating between qt task force and autopilot tools team14:40
asacand didrocks has no stake in this14:41
asacdidrocks: correct? are we using this job for anything in CI Train?14:41
didrocksMirv: thanks to you :)14:41
elopioI'm just wondering. If this job consumes all the phones we have, and they are trying to release a new image, both jobs will take twice the time.14:41
asacelopio: yeah, i dont think thats the case. its probably consuming at most one phone at the same time14:42
asacanyway, we have to wait for doanac`/ev to comment14:42
didrocksasac: I don't, indeed14:43
asacgood14:43
didrockselopio: feel free, the pool is static14:43
asacplars: there?14:43
elopiook, then. Thanks for the clarifications asac. Sorry for the troubles.14:43
plarsasac: yes14:43
asacplars: is there a need to coordinate running http://q-jenkins:8080/job/autopilot-release-gatekeeper/42/ and image testin?14:43
asacplars: or do they use separate phones for testing and dont step on each others toe?14:44
asac(i assume its all fine, just double checking)14:44
didrockselopio: it's another incentive to get the base image good again, to know what impacts really Qt 5.2 has :)14:44
plarsasac: I'm not familiar with that job, let me look14:44
asacelopio: dont worry. if you can help fixing those flaky tests StateNotFound things it would be an amazing contribution to the qt landing effort :)14:44
asacelopio: my understanding was that you guys understood the bad coding pattern used in those tests... if not check with thomi, i believe he knows exactly what needs to be done14:45
elopioI'm trying to fix ugly tests one project at a time. There are too many.14:45
elopioasac: it's pretty obvious. There are many many steps in some of the tests, with only one assertion.14:46
plarsasac: looks like it runs on a specific device, a mako14:46
Saviqom26er, rm /run/user/32011/mir_socket14:46
om26erSaviq, removing that every time is fine ?14:48
Saviqom26er, as long as unity8 isn't running14:48
Saviqom26er, or well, it might be slightly different now with system compositor, check initctl get-env --global UNITY_MIR_SOCKET14:48
Saviqom26er, as that one ↑ might be held by unity-system-compositor14:49
plarsasac: that device seems to also be part of the daily smoke runs, and is in the middle of a job right now. It won't interfere with that if someone starts a new one on it, but it will be stuck in the queue until the other job finishes14:49
om26erSaviq, it directs to /tmp/mir_socket14:49
Saviqom26er, so yeah, that's the one unity8 tries to create and fails, if it's there14:50
plarsasac: if you'd like, I can take this device out of the pool that runs daily smoke tests, but we're close to the end of the current run, so it would be better not to cancel it unless it's really urgent14:50
=== psivaa-lunch is now known as psivaa
om26erSaviq, whats the solution, should we clear it when we stop unity and expect for it to be recreated when unity8 is started again ?14:50
Saviqom26er, it happens already, but it's left hanging when unity8 crashes badly14:51
Saviqom26er, so... we might be nasty and remove it in unity8's pre-start14:52
Saviqom26er, but I'm not entirely sure that's a good thing to do14:53
Saviqom26er, mir folk really didn't want that14:53
asacplars: if we have a way to allocate a separate device for the autopilot/qt job14:54
asacthat would be good14:54
asacplars: but not sure what the capacity of phones in the lab is14:54
asacplars: its not urgent for sure.14:55
plarsasac: yeah, I'll take care of it as soon as this current job finishes on that device14:55
asacplars: just want to ensure that the AP/qt job does not interfere with the image testing14:55
plarsasac: it won't interfere for sure, but it could get delayed if, for instance, someone tried to start it right now14:55
asacright. at best there are no delays on image testing because of AP/qt job14:56
asacplars: does that increase risk that we run out of devices for image testing?14:56
plarsasac: no, we have 3 other mako devices in that pool14:56
asacplars: so we will have 1 hot/spare still for image testing? and in worst case we could easily bring back the AP device, right?14:56
plarsasac: exactly, we are doing pretty good on makos14:57
asacsound good14:57
plarsasac: and on flo we have 3 for smoke, we only have 2 mantas and rick is working on getting those hooked up. So hopefully that one should be rolling soon too14:57
psivaadidrocks: http://q-jenkins.ubuntu-ci:8080/job/psivaa-trusty-touch-mako-smoke-daily/3/#showFailuresLink has some failures with 202. dint look each of them though15:09
* didrocks needs to go for a run15:09
didrockspsivaa: do you have time to look at them? ^15:09
didrocksor plars ^15:09
didrocksmaybe :)15:09
plarspsivaa: what job is that?15:10
didrockshum15:10
psivaaplars: that was a job with image 20215:10
didrocksall failures seems to be unable to unlock the scren15:10
plarspsivaa: I guess you were trying new/old packages with it by hand?15:10
didrocksplars: no, only image 20215:10
didrocksbut most of the tests failed15:11
plarsah, ok15:11
didrocksbecause can't unlock the screen15:11
psivaadidrocks: yea, sorry so that means the actual app tests have been skipped. let me run them again15:13
ChrisTownsendretoaded: Hi.  I updated the branch for https://code.launchpad.net/~townsend/nux/fix-animation-crash/+merge/208218, but I've yet to see a result from CI and I also don't see any evidence a CI job is running.  IS this because it's in the globally approved state or is there something messed up with Nux CI?15:19
dbarth_hey again, do you guys mind if i continue stacking merge proposals into an already allocated silo? (silo-005)15:23
dbarth_i know it's not helping much fix the image, but this way we can continue our tests and copy the silo ppa binaries into the SDK PPA once tests pass15:23
dbarth_(this, to be ready on time for the app showdown kickoff)15:23
retoadedChrisTownsend, checking15:30
ChrisTownsendretoaded: Thanks15:30
doanac`asac: the autopilot-release-gatekeeper job should be able to test qt5.2 properly using the code as our daily-image testing. However, it wouldn't show up in the qa-dashboard. If that is really required, i think we need to bring in ev and priortize this effort. its something plars or myself could do.15:31
plarsdoanac`: I may have misunderstood, but I didn't think that's what he was asking for15:31
asacdoanac`: i looked at the jenkins job. that seems good enough15:32
plarsdoanac`: I think he was just trying to make sure the job wouldn't stomp on daily smoke testing if they both happened at the same time15:32
doanac`plars: oh - that should be fine. our new pooling should handle things correctly. I mis-read the bug, they want dashboard-like runs. not results15:32
=== alan_g is now known as alan_g|tea
sil2100thostr_: hi! Is that a touch landing?15:44
sil2100thostr_: since it says 'mostly desktop' ;)15:45
thostr_sil2100: mostly, but not all15:45
thostr_sil2100: it's converged, so applicable to everything in the end15:45
sil2100didrocks: ^ can I assign a silo for those? It's line 30, it's all indicators15:47
* sil2100 is normally a bit worried about those15:47
=== alan_g|tea is now known as alan_g
Laneysil2100: (tedg:) we were discussing fixing the OnlyShowIn lines in those yesterday, so i'm not sure they are ready15:50
sil2100Laney: ok, I'll switch it to 'Ready: No' - once there is a final decision, just switch that back to Yes and we'll see about assigning a silo ;)15:53
seb128sil2100, thostr_, Laney: changes to upstart jobs could impact on the phone I think15:54
LaneyI think they break !Unity users as is15:54
seb128Laney, tedg: do we have a summary of what issues we try to address with those changes?15:54
seb128I don't understand the problem/why we change15:54
Laneydidn't try to understand that part, just pushing back on the OnlyShowIn thing15:54
tedgWe're fixing a couple things, but mostly it's for consistency.  For instance indicator-application is basically broken in XFCE.15:55
tedgSome indicators didn't have respawn stanzas or limits.15:55
tedgetc., etc., with that set everyone behaves the same.15:56
psivaadidrocks: so with 202 the screen unlocking still fails with a number of tests. probably some conflicting packages to what's already in 202 causing this.15:56
tedgI need to adjust the conf file slightly for the XFCE folks.  I wasn't hurrying because of no-silos.  Can do now.15:57
psivaadidrocks: because the archive has moved on and during each test we download app, test AP packages from the archive and they might clash with the versions in 20215:57
retoadedChrisTownsend yes, the mp is already in an Approved state so will not trigger a  run.15:59
ChrisTownsendretoaded: Ok, thanks for confirming.16:03
retoadednp16:03
ChrisTownsendretoaded: Do you know if setting it back to Needs Review->Approved will trigger a CI job now that Nux is part of the CI train and automerging is disabled?16:04
ogracyphermox, https://launchpad.net/bugs/128514616:04
ubot5Ubuntu bug 1285146 in bluetooth-touch (Ubuntu) "Bluetooth on flo can be easily enabled but the indicator does not recognize it" [High,New]16:04
doanac`didrocks: FYI - bug 1284226 has been released. we are now doing setup/teardown for each test. this should hopefully help the ofono-simd issue16:05
cyphermoxogra: thx16:05
ubot5bug 1284226 in Ubuntu CI Services "touch daily image test: need to install and remove deps for each testsuite" [High,Fix released] https://launchpad.net/bugs/128422616:05
retoadedChrisTownsend, yes it should.16:05
ogracyphermox, if you need any other logs or info, just let me know16:05
ChrisTownsendretoaded: Ok, thanks.  I'll give it a try.16:06
retoadedack16:06
tedgLaney, sil2100, seb128, Updated with the change requested by the Xubuntu folks.16:11
Laneytedg: I think you also need to fix the OnlyShowIn conditions in the xdg files (remove it?) for non-upstart folks16:14
LaneyOr OnlyShowIn=GNOME;Unity;XFCE;\nAutostartCondition=GNOME3 unless-session gnome or something16:14
tedgLaney, I don't think so.  For instance if you installed Kubuntu, you wouldn't want it to launch the indicator services.16:15
tedgI don't think that GNOME wants indicator services, eh?16:15
Laneyfallback sure does16:15
tedgFallback?16:16
Laneypanel16:16
tedgMATE?16:16
Laneyno16:16
seb128tedg, "Ubuntu classic" if you prefer :p gnome-panel + indicators + compiz (or other wm if you prefer)16:16
tedgseb128, Laney, what does that set as its desktop session?16:17
dbarth_sil2100 or an EU guy, can i get a reconfig on silo-005?16:18
dbarth_we won't land but will use that for final tests and upload to public ppa16:18
dbarth_sil2100: that was silo-001 sorry16:19
Laneytedg: XDG_CURRENT_DESKTOP=GNOME; I think what I listed up there ought to work16:21
tedgLaney, Won't that screw Gnome Shell then?16:21
seb128tedg, GNOME atm I think, but they are talking about changing to Unity16:21
tedgHa, I'm in the future already :-)16:22
Laneywhy?16:22
LaneySee the AutostartCondition16:22
ograbug 128523416:22
ubot5bug 1285234 in lightdm (Ubuntu) "lightdm on touch leaves a unity-system-compositor process around" [Undecided,New] https://launchpad.net/bugs/128523416:22
tedgLaney, Doesn't that just check the desktop session to see if it's "gnome" ?16:23
Laneywhat more do you want to do?16:23
tedgI think you need to distinguish between not-gnome and gnome.16:23
tedgOr I guess, was-gnome16:24
Laney'gnome' means gnome-shell16:24
tedgYes, which is why the gnome-fallback shouldn't say that it's gnome.  It's not.16:24
LaneyI'm not sure what you're talking about, I'm afraid16:24
sil2100dbarth_: looking16:25
LaneyXDG_CURRENT_DESKTOP is what OnlyShowIn looks at and DESKTOP_SESSION is what the unless-session is considering16:25
LaneyIt's basically exactly what we're doing for gnome-screensaver already16:25
sil2100dbarth_: ok, so silo 001, right?16:26
tedgUhg, okay.  I think that's kinda a hack.  But if it makes fallback work I don't care.16:26
LaneyIt's all worse than dbus activation. :)16:27
tedgDBus activation is a hack, it's hard to rank them though.16:27
sil2100dbarth_: ok, reconfiguring 001 and setting to Testing to No16:28
didrocksdoanac`: excellent! so next run? :)16:28
didrockssil2100: ask them to rerun the tests on the phone as well (unity8 in particular)16:29
didrockssil2100: if so, ok16:29
didrockspsivaa: argh, so only click apps failed?16:29
didrocksasac: ogra: so, we can't have results on 202 either :/16:30
ogradidrocks, sad :(16:30
bregmasil2100, I can has silos for line 44 and line 45 pretty please?16:32
dbarth_sil2100: thank you16:32
asacdidrocks: 202? i see results16:35
asacalso its in the past16:35
ograasac, ignore the dashboard16:37
didrocksasac: the results were partials16:37
ograthis 202 test was running on a device that isnt attached to it16:37
psivaadidrocks: not only them but also the the apps whose tests download some deps from the archive16:37
ograasac, we were hoping we can get a gree 202 to fix the click upgrader issue16:37
didrockspsivaa: I doubt we had imcompatible changes here16:37
ogra*green16:37
didrocksasac: we wanted full results because latest image we promoted doesn't have the click upgrade UI16:37
asacdidrocks: is there a way at all to get those results still?16:38
didrocksasac: that's what I was asking psivaa for, but seems not16:38
ograasac, the original tests fell over heavily16:38
ograso we wanted it re-tested ... but that was not a dashboard capable setup16:38
ograand it seems it didnt work nyway16:39
ogra*anyway16:39
ograbecause the click apps and autopilot are out of sync now16:39
asacpsivaa: plars: can we get a rerun of 202?16:39
psivaadidrocks: http://q-jenkins.ubuntu-ci:8080/job/psivaa-trusty-touch-mako-smoke-daily/4/testReport/junit/%28root%29/camera_app/setup/ is one example16:39
ograasac, it wont help16:39
=== robru-is-dying is now known as robru
plarsasac: I think psivaa was already having one in progress16:39
ograasac, we would have to roll back the archive ....16:40
didrockspsivaa: why can't it fetch python-gi?16:40
psivaaasac: the version that image 202 looks for is not in the archive.16:40
psivaadidrocks: ^16:40
didrockspsivaa: ah, you don't apt-get update before installing16:40
psivaai guess we could probably force fix like apt-get update and stuff.. but that will take time16:41
sil2100bregma: looking16:41
didrocksdoanac`: can you look at that? it's a valid case ^16:41
didrockseven if we still have the initial flaw, the autopilot are potentially not matching the version we have on the device16:41
ograright16:41
tedgLaney, Updated16:42
Laneytedg: cheers, LGTM16:42
LaneyI only checked the diff for bluetooth, assuming the rest are the same16:42
Laneyno, wait, the second one hasn't been updated16:43
sil2100bregma: 44 assigned16:43
psivaadidrocks: i could run apt-get update and rerun the tests to see if there is any improvement but i think then the results wont exactly be that of 20216:45
ograheh, on 209 the flo results are actually better than mako16:46
ogra(one crash less)16:46
bregmasil2100, that's the priority one, thanks16:46
sil2100bregma: 45 assigned as well o/16:47
bregmata16:47
=== gatox is now known as gatox_lunch
doanac`didrocks: catching backscroll. so we want to call: apt-get update; apt-get install <pkg> instead of just calling apt-get install <pkg>, correct?16:49
didrocksdoanac`: exactly16:50
ograbut only for re-running tests16:50
didrockspsivaa: well, will still be better than no results, I don't think as long as we don't dist-upgrade that we'll have different apps16:50
didrocksyeah16:50
doanac`didrocks: sure. that's a simple change16:50
plarsdidrocks: asac: I thought we said before that we never want to do apt-get update during a test run, because we could pull in newer versions of things than what we want from the archive16:53
plarsdoanac`: ^16:53
ograplars, right ... but this one test is a special case16:53
plarsso we are talking about a one-off here, not in general16:54
ograsince it cant find a version at all for this old image if you dont run apt-get update16:54
ograyeah, please dont apt-get update on normal tests :)16:54
doanac`ah - so we *only* do apt-get update for a "re-run"16:54
ograyes16:54
doanac`ogra: k, thanks for the clarification. sorry i misunderstood at first16:54
ograsince for that the version in the archive can have moved forward ... so the Packages file is out of sync with the actual archive and you will get 404s16:55
plarsdoanac`: psivaa: I think for this one, it would be easier to just go onto the device, apt-get update, upgrade/install those specific packages, and restart the test without reinstallation16:55
psivaaogra: doanac`: we could leave that out and may be add as a manual step not as part of the test16:55
psivaaplars: yea16:55
ograright. or a switch or whatever16:56
elopiodidrocks: now this makes more sense: http://ci.ubuntu.com/smokeng/trusty/touch/mako/209:20140226.1:20140224/6842/ubuntu_clock_app/819490/16:56
elopiowe are starting with no alarms, so we can't delete any of them.16:56
elopioultimately, that's because of bug https://bugs.launchpad.net/ubuntu-ui-toolkit/+bug/1275060 I'll try to fix it today.16:56
ubot5Ubuntu bug 1275060 in Ubuntu UI Toolkit "QQuickListView fails to click an element if part of the list is out of view" [Undecided,Confirmed]16:56
ograthe best would actually be if we could do a snapshot of the archive when we build an image and always have the test point at that16:57
ograbut thats quite some effort16:57
sil2100balloons: ping16:57
sil2100balloons: did you see my comment on the terminal-app flaky test bug?16:57
didrockselopio: ah ok, so clock-app understood :)16:58
didrockselopio: just all the others to understand now :p16:58
didrockselopio: good luck, do not hesitate to answer on the phone ML16:58
didrocksasac: doanac`: hum, however, setup and teardown are counted as tests on the dashboard, is that expected?17:00
elopio:) everything - 1 to go17:00
didrocksasac: doanac`: for instance, if we can't run any application test, we have 80% of tests passing: http://ci.ubuntu.com/smokeng/trusty/touch/mako/209:20140226.1:20140224/6842/ubuntu_filemanager_app/17:00
doanac`didrocks: yeah. i needed to report them now, as they could fail and we'd need to know17:00
didrocksdoanac`: ok, but the count is puzzling then ;)17:00
balloonssil2100, no, I can look17:01
didrockslike 80% "not that bad", oh no test were run :p17:01
sil2100didrocks: be right there17:01
doanac`didrocks: yeah. that's a good point.17:01
sil2100Had a prank call to the door just now17:01
sil2100balloons: your theory with input events might be right according to that ;)17:02
doanac`didrocks: not sure the best way to fix that. i'll open a bug and discuss with team17:02
didrocksdoanac`: thanks!17:03
=== alan_g is now known as alan_g|walk
* balloons reads17:03
balloonssil2100, where did you find the logs you looked at? And yes I think your questions are quite valid17:04
sil2100balloons: in the smoketesting test results, the ubuntu-terminal-app application logs are attached to the AP result17:05
balloonsahh, ok17:06
ograrsalveti, did you knwo that 4.4 changed the oomkiller ?17:11
ograwho knows, probably the breakage is related17:11
rsalvetishouldn't17:11
ogradidrocks seems ot have more info, seems to have been mentioned in some android podcast today17:12
rsalvetiinteresting17:12
rsalvetiwell, not for every kernel17:12
psivaadidrocks: so running the tests after apt-get update resolves the package mismatches but the screen unlocking still fails17:13
rsalvetion another call now, but we're still debugging the crash17:13
rsalvetiit happens with SF as well, so easier to debug17:13
ograrsalveti, yeah17:13
psivaadidrocks: ogra: i see that happened after unity 8 test runs on 202 but dont see any packages being installed just before untiy8 tests17:13
ograweird17:14
ograrsalveti, did you get anywhere with the non working calls on second call ?17:15
ograor is that on hold for unity ?17:15
rsalvetiogra: fully focued on unity8, but will get to that later today for sure17:16
ograok17:16
psivaaand i checked unity8_7.84+14.04.20140221.orig.tar.gz was the version when we originally ran unity8 and that is the same version that the test uses now as well17:16
ogra(just was asked in the meeting)17:16
rsalvetialways the feeling we're getting closer to know what is wrong17:16
doanac`psivaa, plars, didrocks: as per the misleading pass-rate calculations: https://bugs.launchpad.net/ubuntu-ci-services-itself/+bug/128526617:16
ubot5Ubuntu bug 1285266 in Ubuntu CI Services "dashboard pass-rates can be misleading" [Medium,New]17:16
sil2100rsalveti: we're keeping our fingers crossed o/17:17
rsalvetiyeah :-)17:17
didrocksdoanac`: thanks17:20
didrocksrsalveti: ok, so, what I mentionned during the meeting is that in the android 4.4 kitkat developer podcast (from google), they mentionned they have made adjustement to their oom killer17:28
didrocksrsalveti: (yeah, I'm listening to it while exercising)17:28
didrocksrsalveti: so, they told they are way more aggressive in killing cached applications17:29
rsalvetiinteresting17:29
didrocksnot sure how our apps are seen on the android side as there is no onstage/cached applications notion17:29
didrocksthat would map with other failures we see17:29
didrocks(no crash though)17:29
didrockslike some applications don't have any pid and the tests failed17:30
didrocksor .service don't respond17:30
didrocksasac: FYI as well ^17:30
=== alan_g|walk is now known as alan_g
rsalvetino pid and no crash?17:41
rsalvetibut I'd guess dmesg to say something17:41
rsalveti*expect17:41
* didrocks tries to refind one test17:42
didrocksrsalveti: hum, actually, maybe the dashboard was slow to update the artefacts, but I found one crash on that one17:44
didrockslet me look at other17:44
didrocks(I was looking at http://ci.ubuntu.com/smokeng/trusty/touch/mako/209:20140226.1:20140224/6842/gallery_app/818684/)17:44
rsalvetiyeah, I'd expect at least one crash17:44
didrocksno dmesg though in artefacts :/17:44
didrocksto see if oomkiller entered into action17:44
didrocksyeah, nothing without crash in 229 or 22817:45
didrocks209* 208*17:45
ogra2009 2008 ?17:46
ogra:P17:46
didrocksogra: you start trolling me again! :)17:49
ogralol17:49
didrocksogra: do you have a window dedicated to g+ btw? :)17:49
didrocksor even a monitor/device17:49
ograwell i read G+ before going downstairs after the meeting :)17:50
didrocksahah :)17:50
ograsince google accounts dont really work for hangouts anymore i cant really keep a dedicated window for it anymore ... i used to do that though17:51
didrocksrsalveti: btw, finally found an official (written) source: https://source.android.com/devices/low-ram.html:17:52
didrocks"Tuned memory use of low-RAM devices: tighter out-of-memory (OOM) adjustment levels, smaller graphics caches, etc.17:52
didrocks"17:52
didrocksogra: use 2 profiles on your browser17:52
didrocksand:17:52
didrocks"Kill processes (even ordinarily unkillable ones such as the current IME) that get too large in idle maintenance."17:52
ograhmm, never tried that17:52
ogradidrocks, you should point ricmm and tvoss at that17:53
didrocksI think they are pinged now :)17:53
ogramight actually have impact on our app lifecycle17:53
ograyeah :)17:53
didrocksricmm: tvoss: backlog the last 30 lines. Basically, oomkiller changed in 4.4, so that can explain all the flakyness we start to see ^17:53
rsalvetiyeah, so it's not on the kernel side17:53
didrocksit's another process we are not using?17:53
rsalvetithey just changed the threshold17:53
ograwell, probably also the defaults in kernel17:54
ograwe should at least take a look17:54
rsalvetinot so sure, as it's not needed17:54
rsalvetigrouper kernel for example is the same old17:54
rsalvetione17:54
ograi doubt it would affect testing17:54
rsalvetiemulator as well, nothing changed17:54
tvossrsalveti, yup, the default thresholds are adjusted17:54
ograif it affects anything it is long running apps17:54
tvossogra, well, that depends on us to mark long-running correctly17:54
didrocksDon’t allow large services to put themselves back into A Services (so they can’t cause the launcher to be killed).17:55
ogratvoss, well, if we dont change it but the kernel default does it doesnt depend on us :)17:55
didrocksnot sure if we try to do that (or if it's another way of marking, only on dalvik side)17:55
tvossdidrocks, nothing about dalvik here. can you check in logcat if you see anything suspicious?17:56
tvossogra, ^17:56
om26erretoaded, hey! can we disable apport 'report a problem' in otto ? it comes up during our tests and causes failures18:00
=== gatox_lunch is now known as gatox
=== alan_g is now known as alan_g|EOD
ogratvoss, http://paste.ubuntu.com/7001147/18:19
ogra(thats a flo (new N7)18:20
ogranothing that strikes me18:20
ogracyphermox, yay ... i got a BT indicator after the echo18:21
tvossogra, I see an egl context creation issue18:24
tvossogra, E/libEGL  ( 1762): eglMakeCurrent:775 error 3009 (EGL_BAD_MATCH)18:24
plarsbfiller: any ideas on why https://jenkins.qa.ubuntu.com/job/trusty-touch-flo-smoke-daily/4/consoleFull seemed to fail out early? This was on flo18:25
ogratvoss, hmm18:26
tvossogra, which process is 176218:26
ogratvoss, dunno, i rebooted already18:26
bfillerplars: which issue is this one?18:28
bfillerplars: hard to tell what's wrong in that massive log :)18:28
plarsbfiller: I'm trying to get complete results on flo, and there were a few tests that didn't run to completion18:28
plarsbfiller: ah, sorry about that - search for "testing messaging" and it should get you to the right section18:29
plarsbfiller: messaging app18:29
balloonspopey, davmor2, do you notice that apps tend to end white-screened after closing on the latest images?18:29
bfillerplars: ok18:29
davmor2balloons: ah so maybe not just a qt5.2.1 issue then18:30
balloonsdavmor2, yes, perhaps something else if you've noticed it18:30
popeyballoons: davmor2 yes18:31
balloonsdavmor2, popey any bug, if not I guess I'll file18:32
davmor2balloons: it seems to happen on qt.5.2.1 on  209  notably on ap tests18:32
balloonsI'm going to assume qmlscene18:32
popeyno, i thought it was an AP "feature"18:32
balloonshaha18:32
davmor2popey: I'll set thomi on you18:32
bfillerplars: do they fail if you run them manually on the device?18:33
ogratvoss, so next boot ... http://paste.ubuntu.com/7001213/ ... pid 1734 is unity8 ... probably something for either Saviq, ricmm or the Mir team to look at18:33
plarsbfiller: no idea, I don't have a flo, and my mako is tied up at the moment trying to reproduce the unlock screen errors18:33
davmor2popey: and you know he is from the mordor side of New Zealand18:33
ograoooh intresting ...18:34
ograpid 1759 is actually maliit-server18:34
balloonsok, so I'll check and see what it looks like18:34
ograwhy does that talk directly to the driver ?18:34
davmor2balloons, popey: is one of you writing a bug for that one?18:35
tvossogra, it talks to libEGL18:35
ograwhy ?18:35
balloonsdavmor2, I am.. However, it doesn't seem like the app is still running18:35
ograshouldnt it all go through the shell ?18:35
balloonslooking at the process list18:36
popeyyeah, the app is finished18:36
balloonsthat makes sense, but ...18:36
popeyit just leaves white behind18:36
balloonsright.. why the white?18:36
tvossogra, nope, apps use egl18:36
tvossogra, it's not a driver in the classical sense18:36
ograah, k18:37
bfillerplars: I can ask osomon tomorrow to try it on flo, he's the only one on our team who has one18:37
tvossricmm, Saviq I see an an eglMakeCurrent failing in Unity818:37
=== fginther changed the topic of #ubuntu-ci-eng to: Ubuntu CI Engineering Team | Vanguard: fginther | CITrain support - US: robru, cyphermox, rsalveti - EU: sil2100, Mirv, didrocks | CITrain support no answer: use mup bot after 30 minutes, but choose right timezone | Landing instructions: http://goo.gl/8H1Du3 | Known Issues: -
seb128cyphermox, hey18:37
davmor2balloons: it looks to me like the window that opens first and then you have the second or 2 wait for the app to open on it18:38
plarsbfiller: thanks!18:38
balloonsdavmor2, popey do you only see this with autopilot?18:39
davmor2balloons: pass I've only been running ap tests18:39
davmor2let me try an app noramlly18:39
popeyditto18:41
davmor2balloons: okay so taking a quick look at the browser, if I swipe the app out of the way then close via the dash I get no issues, however if I open the hud and hit quit I briefly get a black screen in place.  So I wonder if it is the way that ap is closing the app?18:43
balloonsdavmor2, I assume yes. The app launch for ap changed very recently18:43
balloonsI assume this is fallout, but not limited to ap18:43
=== psivaa is now known as psivaa-afk
balloonsdavmor2, popey https://bugs.launchpad.net/autopilot/+bug/128530518:45
ubot5Ubuntu bug 1285305 in qtdeclarative-opensource-src (Ubuntu) "Apps leave whitescreen behind upon exiting" [Undecided,New]18:45
davmor2confirmededed18:47
ogradededed ?18:48
popey→ food18:48
om26erfginther, hello :)18:58
om26erfginther, did you get a chance to look into 'report a problem' issue in otto ?18:58
fgintherom26er, can you give me some more context? I'm not sure what you are referring to19:00
bfillerom26er: hey, any idea what would cause this failure? do we need an eventually somewhere? http://paste.ubuntu.com/6993735/19:01
cyphermoxseb128: hey?19:01
om26erfginther, sorry about that, in some of the cases our tests fail because the apport 'report a problem' window comes on screen, so autopilot have a way to check the before-after of the system and it sees a new window being started and not closed19:02
om26erfginther, https://jenkins.qa.ubuntu.com/job/autopilot-testrunner-otto-trusty/3116/testReport/junit/ubuntu_system_settings.tests.test_datetime/TimeDateTestCase/test_searching_tz_not_found_with_mouse_/19:02
bfillerom26er: it's timing out just trying to switch tabs, seems pretty fundamental (gallery-app)19:03
fgintherom26er, I remember now. I have not had a chance to investigate yet. how often do you see this problem?19:03
om26erbfiller, i am trying to figure out, is there a way to reproduce the issue ?19:03
om26erfginther, almost everytime for ubuntu-system-settings tests19:04
bfillerom26er: I haven't been able to reproduce yet, but very reproducible on smoketests19:04
om26erbfiller, looks like tab switching went bad indeed19:05
om26erfginther, wouldn't 'sudo service apport stop' do it for us ?19:06
om26erbfiller, let me try to reproduce it on my phone19:06
fgintherom26er, that should work, I just need to make sure that doesn't impact actual crash collection. I'll give this a test today19:07
retoadedom26er, fginther: sorry, I forgot to respond to the issue but I was looking into it. only one of the nodes that are specifically marked as otto nodes had apport installed/running.19:07
fgintherretoaded, interesting... It's still possible that when the otto lxc container loads it re-enables apport. I'm not very familiar with how services are treated19:09
fgintherretoaded, thanks for checking19:09
retoadednp19:11
bfillerom26er: weird when running the test manually it first switches to the Photos tab, then to the Albums tab19:12
om26erbfiller, thats how its supposed to work with autopilot, the tabs are switched one after the other19:13
bfillerom26er: oh19:13
bfillerom26er: thought it would swtich directly to the requested tab19:13
om26erbfiller, i am running the test, its working fine for me. going to run the suite in a loop19:14
cgoldbergrobru, cyphermox, Hi.. I've got a landing for Autopilot, just added to spreadsheet (line 48) .. can i get a silo allocated today for that?19:17
cyphermoxcgoldberg: IIRC only if it fixed current known regressions19:17
cyphermox*fixes19:17
thomicyphermox: we have a FFE, and it contains bug fixes - is that enough, or...?19:18
cyphermoxit's slightly unclear to me, is this the fix for the state not found thing?19:18
thomicyphermox: no19:18
cyphermoxthomi: as I recall, didrocks mentioned we were mostly looking at fixing the things that are currently broken in the image19:19
cyphermoxI can allocate, but we won't publish it19:19
thomicyphermox: thanks. Hopefully by the time we're done with the testing the image will be fixed.19:20
cyphermoxhmm19:20
cgoldbergcyphermox, cool.. that would help19:20
cyphermoxchecking, perhaps it's even "don't allocate" from the message I had yesterday19:20
cyphermoxthe problem is, if you have that assigned and then you want to land the fix for the state thing, then we won't be able to19:20
cyphermoxrobru: what do you think?19:21
thomicyphermox: why not? Can't we just add additional MPs to the silo and re-spin?19:21
=== bfiller is now known as bfiller_afk
thomierr, re-allocate.. or whatever it is you guys do :)19:21
ogracyphermox, https://code.launchpad.net/~ogra/ubuntu/trusty/bluetooth-touch/add-flo/+merge/208461 for you19:23
asaccyphermox: thomi: think adding new MPs for components is fine. just remember that there is risk that as part of getting image fixed we might to invalidate silos as we have to shovel a cherry pick in or land something that makes you need to retest everything.19:25
thomiasac: sure, thanks.19:25
asacthomi: anyway, please get the state not found thing sorted. seems that elopio didnt know what to do to fix things19:25
cyphermoxthat's what I meant19:25
ogracyphermox, note that this MP also fixes a test failure on flo (BT indicator in unity8) so it should be fine to land it at any time19:25
thomiasac: elopio just told me that it's fixed already19:26
asacthomi: or has a different order of fixing bad tests than the ones that show problems19:26
asacoh19:26
thomiasac: elopio is the guy you want to talk to about that :)19:26
asacthomi: where is it :)19:26
cyphermoxogra: actually, no19:26
asacthomi: i talked today and the above is what i last heard19:26
ogracyphermox, why no ?19:26
asacelopio: there :)?19:26
ogracyphermox, works fine here19:26
asacelopio: gimme MPs that have the fixes19:26
cyphermoxogra: we discussed dropping any device-specific jobs, there is a better way to do this that should work on all devices just the same19:27
ogracyphermox, the chipset is the same as on mako and uses the same way of initialization19:27
elopiothomi: asac, I said that I'm not able to reproduce the StateNotFound errors that we got on image 206, and more recent images don't show StateNotFound errors.19:27
cyphermoxyes19:27
elopioso whatever was causing it, seems to be fixed now.19:27
cyphermoxogra: but you shouldn't have to poke hci_smd_set yourself either19:27
ogracyphermox, and we try to get to green images again19:27
* asac checks19:27
cyphermoxogra: gosh, I hadn't thought of this19:27
cyphermoxogra: give me a minute19:27
ograif you want to rework the whole way of starting thats fine19:27
elopioasac: the errors we are getting now make more sense, and I'm slowly working on them. First, the clock needs to create an alarm before deleting it.19:27
ogramy MP is good as an interim fix to get one failure down19:28
asacelopio: here is what didrocks sent around a bit earlier: "Work is going on. Leo wasn't able to reproduce it, some back and force on the mail comments shed some lights (with a whole list of tests that failed on #206, #207 and #208."19:28
cyphermoxogra: then don't bother sending an MP and just push it if that's what you'd prefer...19:29
asacelopio: ok you rock. sorry for making unqualified commments.19:29
ogracyphermox, well, its your baby, i'd like you to nod it off :)19:29
ograand its in CI, isnt it ?19:29
cyphermoxogra: I don't remember19:31
ograheh, k19:31
cyphermoxheh, whatever19:31
ograroot@ubuntu-phablet:/# grep hcismd_set /var/lib/lxc/android/rootfs/* 2>/dev/null19:31
ograroot@ubuntu-phablet:/#19:31
ogracyphermox, ^^^19:31
ograi fear there is nothing on the android side touching that19:32
ograguessing there is some UI tool that does the echoing19:32
cyphermoxogra: there should have been19:32
cyphermoxbut I guess for that particular device bludroid does it now19:32
ograwell, not in the init.*.rc files19:32
cyphermox>.<19:32
cyphermoxI can read19:32
cyphermoxapproved19:33
ograwell, i wonder if bludroid does it for all 4.4 devices now ...19:33
cyphermoxI don't really want to mess with this right now19:33
ograthen we would have that issue on all of the,19:33
ograok, i'll just uplaod then19:33
cyphermoxyeah19:33
cyphermoxapparently it's hard for android to decide how they want to do stuff19:34
ogra:)19:34
seb128cyphermox, hey!19:35
cyphermoxogra: the end result will have to be that we make the bluetooth.status job or whatever seems to be the "standard" on 4.4 do the /sys/module poking and just start the *.bt.sh job and then that one19:35
seb128cyphermox, can I haz slots?!19:35
seb128;-)19:35
cyphermoxogra: but I guess that will be for laterz19:35
seb128cyphermox, l46&4719:35
ogracyphermox, yeah19:36
cyphermoxseb128: shouldn't this stuff be blocked by beta?19:36
seb128cyphermox, there is no such thing as proposed block19:36
seb128cyphermox, we stopped blocking upload, we just block britney/transitions to trusty19:37
cyphermoxyes, I know19:37
cyphermoxjust saying19:37
ograbt-touch uploaded19:38
ograone failure down on flo :)19:39
ograonly 3164323 left19:39
ogra:P19:39
rsalvetiogra: at least we just found out that the unity8 crash is not only happening with unity819:40
rsalvetiit seems to be qt related somehow19:40
rsalvetithe crash we got with unity8, gallery-app and weather-app (qmlscene) are all the same19:40
ograrsalveti, well, did you see my discussion with tvoss above19:40
ograthere seem to be EGL issues too19:40
cyphermoxogra: cool19:41
rsalvetiogra: which issues?19:41
ograrsalveti, http://paste.ubuntu.com/7001213/ EGL_BAD_MATCH ...19:41
rsalvetiwe always had one error19:41
ograrsalveti, pid 1734 is unity8 ... 1759 is maliit--server19:41
ograah, k19:41
rsalvetibut don't know which one :-)19:41
ograheh19:41
rsalvetiogra: check your mako19:41
rsalvetiogra: should happen for every app you load19:42
* ogra notices that he cant fing his mako 19:43
ograargh !19:43
ogradamned ... that slippery thing always slides down where i put it ... and then vanishes between some papers or so19:44
ogragrrr19:44
rsalvetihahah19:44
Saviqtvoss, there *was* a bug I believe19:55
tvossSaviq, hmmm ... worth investigating?19:55
ralsina_robru: can I get a reconfigure for silo 14? I added a new branch https://code.launchpad.net/~dobey/unity-scope-click/post-method/+merge/20821019:55
ralsina_ to the spreadsheet19:55
robruralsina_, sure19:56
Saviqtvoss, isn't it the one when your screen was off?19:56
robruralsina_, please just space-separate the URLs though19:56
ralsina_robru: ok19:56
tvossSaviq, hmmm, good point19:57
tvossSaviq, what about the egl swap interval thingy?19:57
ralsina_robru: fixed19:57
tvossSaviq, also: you might want to spin a build of u8 with address sanitizer enabled19:57
Saviqtvoss, https://bugs.launchpad.net/unity8/+bug/126146619:57
ubot5Ubuntu bug 1236525 in unity-mir "duplicate for #1261466 unity8 killed/crash then restart can result in mir unable "could not unblank display"" [Medium,Triaged]19:57
tvossSaviq, haven't come to it, yet19:57
robruralsina_, ok, please build19:58
Saviqtvoss, what about the swap interval thingy?19:58
tvossSaviq, there is an error reported here: http://paste.ubuntu.com/7001147/19:59
ralsina_robru: started, thanks19:59
robruralsina_, you're welcome19:59
tvossSaviq, line 427ff20:01
SaviqETOOMANYLINES20:01
Saviqtvoss, not sure what to do about that, is that happening at the same time the eglMakeCurrent, or?20:02
tvossSaviq, might well be ...  hang on, got a theory20:02
rsalvetiogra: the unicode error http://paste.ubuntu.com/7001623/20:08
rsalvetithis is when running tests with flo20:08
ograrsalveti, yeap, seen that20:09
ogra(and i have seen that before too ... like months ago ... and i know someone fixed it ... but cant remember when or who)20:09
ogra(or where even)20:09
rsalvetiright20:09
rsalvetiwas happening a few weeks ago20:09
ograyep20:10
=== robru is now known as robru-sick
cgoldbergcyphermox, line 16, windowmocker.. is that landing today?20:19
bfillerom26er: any luck reproducing?20:22
om26erbfiller, not in a reproducible fashion, I ran the whole suite and one one of the tests failed with a similar error message from dbus20:22
asaccyphermox: can you easily see how many silos are currently loaded?20:23
asackgunn: how many silos do you have in theory ready for landing?20:23
om26erbfiller, let me inspect the test code, I might find something there20:23
cyphermoxasac yes20:24
cyphermoxhttp://calypso.cyphermox.net/~mtrudel/silos/20:24
cyphermoxI made myself a graph20:24
cyphermoxit updates every hour20:24
asaccyphermox: ok, i think i am interested in loaded and green :)20:24
asachehe20:24
asacalways more and more20:24
cyphermoxbah20:24
bfillerom26er: ok thanks, yes I see the same failure in another gallery-app test from the nightly build: http://ci.ubuntu.com/smokeng/trusty/touch/mako/209:20140226.1:20140224/6842/gallery_app/818684/20:24
om26erbfiller, I saw the failure in test_add_photo as well20:25
cyphermoxcgoldberg: I guess it could land today20:25
kgunnasac: actually i have a question about that...20:29
kgunni have 1 unity-mir silo20:29
kgunnbut...20:29
kgunni really would like a silo in preparation for landing to begin...i'd like to have one for mir ?20:29
kgunnis it possible?20:30
asackgunn: so we need fast path silos20:30
asacwe have 17 allocated already.20:30
asacdont think we can really hand out more new ones :( ... otherwise we cant land the fixes for the regressions quickly once they came in20:30
asaccome in20:30
kgunnack20:31
asackgunn: we probably could hand out one more ...20:31
asaci would really like to see this unity8 heap crash fixed somehow. from what i understand its 40% chance of being mir and 30% chance of being android 4.4 and the rest something else20:32
kgunnit'd be great if i got a silo20:32
kgunni'm following phablet as well...seems like20:32
asaccyphermox: did anyone request a silo today?20:33
kgunnthere's lots of suspiscions around TLS and hybris20:33
asackgunn: so problem is that if a patch with mir comes out that fixes this, we would have to invalidate all your work if we want to land that cherry pickish20:34
ograthere is a long discussion about this going on atm on the internal channel (for whatever reason)20:34
kgunnack, happy to be invalidated if its mir's fault20:34
asackgunn: let me wait on what cypher says about my question above.20:36
asacneed to be somewhat fair20:36
kgunnyep...thanks for trying at least20:36
asacbfiller: how many silos do you have loaded?20:37
cyphermoxasac: yes, cgoldberg and seb12820:38
asaccyphermox: which landing topic?20:38
cyphermox46, 47, 4820:38
asacisnt cgoldberg autopilot?20:38
cyphermoxyou didn't specify20:38
asacme didnt specify?20:39
bfillerasac: you mean how many outstanding requests do we have in the train?20:39
bfillerasac: I think 4 or 520:39
cyphermoxyes, autopilot, unity-control-center and dbustest-runner20:39
asacbfiller: how many silos do you have loaded and validated20:40
asacor in progress20:40
asaccyphermox: ok, but those got a silo it seems? i was interested if we rejected requests for new silos20:40
bfillerasac: none loaded and validated, many waiting for silos20:41
asacbfiller: think you are close to gallery or dialer fix?20:44
asacthose are the ones you investigate right?20:45
bfillerasac: om26er helping with gallery, just sent an email about that. I don't know how to proceed on that one20:45
bfillerlooks like Chris gagnon just pushed a fix for some dialer-app flakiness20:45
bfillerasac: we can release that20:46
om26erbfiller, btw there is also a gallery-app.crash file as well so the app might have crashed20:46
om26erhttp://ci.ubuntu.com/smokeng/trusty/touch/mako/209:20140226.1:20140224/6842/gallery_app/20:46
asacbfiller: ok cool. you could piggyback your other landings in that landing20:46
asac:)20:46
asacas an idea20:47
asace.g. make one big silo request that also includes that fix... but please be extra careful about regressions in those other apps20:47
asacbfiller: what do you think?20:47
bfillerasac: yeah, was thinking that. but we have a bunch of other changes queued. might be better to just do the AP test fix first20:48
bfillerasac: just so we don't introduce any other weirdness20:48
asacdbarth_: i see you have two silos... are you still working on 005?20:49
asacbfiller: yeah, was just thinking how i can jusfity get your other stuff moving. but guess i would get harangued for such a move anyway20:50
asacbfiller: so for sure, if there is a dialer fix, land that asap20:50
dbarth_asac: 001 could land and free that slot; testing is finished20:50
bfillerasac: ok20:50
asacdbarth_: we cant really land20:51
dbarth_i know20:51
asacdbarth_: but want to ensure that others can continue preparing20:51
asacdbarth_: whats 001 about?20:51
dbarth_but if you want to recycle a ppa, can you take 005, but not 001 please20:51
dbarth_001 i will use to do pkg copies to the sdk20:51
asacdbarth_: will surely not recycle a ppa that is validated. dont worry20:52
asacdbarth_: just wonder if 005 is still being worked on20:52
dbarth_ok20:52
asacif you are active on that one i dont want to take that away20:52
asacdbarth_: can you explain to me more about 001?20:52
asacdbarth_: is that a landing in the centre of the stack?20:52
asacor rather a leave component?20:53
asacguess its not fully leave20:53
om26erbfiller, yes, the time stamps of the .crash file and our ci dashboard match exactly at the same time. the app crashed at 12:42:11 and just 1 second before autopilot raised that error20:53
bfillerom26er: interesting20:55
asacrobru-sick: 006 is in proposed?20:55
bfillerasac, robru: line 50 of the sheet has fix for dialer-app tests20:56
asaccyphermox: ^^20:57
bfillerat least some20:57
asacbfiller: good20:57
asacthanks!20:57
asaccyphermox: can you check if 006 is alreayd in release pcket?20:57
asacrobru-sick: cyphermox:  bregma: isnt 008 ready for landing?20:58
asacseems desktop only20:58
asacsame for 00920:58
asacrobru-sick: can we recyclke your cordova cli ppa?21:02
robru-sickasac, oh, yeah21:21
bregma0002 and 008 are definitely ready for landing, both exclusively desktop only21:21
robru-sickbregma, ok, desktop only we can do21:21
asacrobru-sick: cyphermox: ^ ... not sure if there is a directive to not land these, if not, lets do it and make room for kgunn and bfiller21:21
robru-sickasac, only touch is frozen as far as i know21:21
asacbregma: do you have more in the pipe to land right after?21:21
bfillerom26er: I downloaded the crash file and installed apport-retrace on the device21:21
bfillerom26er: getting this error when I try to run it: apport-retrace -s _usr_bin_gallery-app.32011.crash21:21
bfillerERROR: report file does not contain one of the required fields: CoreDump DistroRelease Package ExecutablePath21:21
=== ChrisTownsend1 is now known as ChrisTownsend
bfillerom26er: how do I get the backtrace from the crash file?21:21
robru-sickbfiller, what's the status of that MP in line 3? will you ever land that? if not, please delete the row21:22
om26erbfiller, we can probably try to report that bug to launchpad and it may retrace it for us21:22
sergiusensrobru-sick, it will land, but needs unity821:22
sergiusensit's in blockade mode now; right?21:22
bfillerrobru-sick: I think it's still valid21:22
robru-sickbfiller, oh ok, i thought it was abandoned. didn't realize you still wanted to land it eventually. thanks21:23
bfillerrobru-sick: yeah, just didn't want to land it at the time to not interfere with other MR's and possibly break something21:23
bfillersergiusens: you ever use apport-retrace with a crash file from smoketests?21:23
sergiusensrobru-sick, we need to reconfigure it with unity8 though21:23
sergiusensbfiller, nope21:24
robru-sicksergiusens, no worries21:24
bfillerom26er: would you mind doing that? we need the trace to figure out what's going on21:24
sergiusensbfiller, there's an email from ev titled "Manually retracing crashes on Touch images" on the phone list though21:25
om26erbfiller, trying to do that but seeing problem here as well21:25
asacrobru-sick: did you see bfillers silo request?21:25
asacoh i see you chattin gwith him21:25
robru-sickasac, on it21:25
sergiusensbfiller, it might be an incomplete crash21:25
asacrobru-sick: seems he submitted a regresison fix21:25
asace.g. flaky21:26
robru-sickbfiller, ok you got silo 13 for dialer-app21:26
bfillerrobru-sick: thanks21:27
robru-sickbfiller, you're welcome21:27
asacrobru-sick: the other question was whether 006 is already in release pocket and could be merged back21:28
asacseems not21:29
asacor maybe the status in that sheet is off21:29
asacrobru-sick: actualy, is that in for a day already? (e.g. 006)21:29
robru-sickasac, yes, that got in yesterday.21:30
asacrobru-sick: so we can recycle 006?21:30
robru-sickasac, yes, but bzoltan will be upset about it.21:30
asace.g. you have to hit merge and clean? please triple check. the commit message in MP looks different21:30
asacbzoltan: show up21:30
asacrobru-sick: but there is no way back anyway21:37
asacfeels odd21:37
asaclet me find someone from sdk team21:37
robru-sickasac, agreed, it's weird to have a published silo go unmerged for so long21:37
robru-sickasac, i guess i should just merge it, and if sdk team finds some problem with the code, we can rush through whatever fixes they feel appropriate.21:37
asacgive me a sec21:37
robru-sickok21:37
asacrobru-sick: ok go ahead and hit the button. next time dont do that without sdk folks unless its a firedrill thingy21:38
asackaleo gave an ack21:39
asacrobru-sick: ok cool21:40
asacrobru-sick: so 006 is free then afterwards21:40
asacalso we have two silos freed by bregma21:40
asaclets give kgunn and bfiller each one21:40
robru-sickasac, ok sorry, thought it was a firedrill situation ;-)21:41
asacbfiller: do you wnt to change your landing request to just include all that you want to land21:41
asacbfiller: if you merge them you can have everything in one silo21:41
asacbfiller: (except the dialer-app which has its own silo)21:41
bfillerasac: sure, so combine all requests into one? (except dialer-app)?21:42
asacbfiller: i would think in this way you can at least continue staging changes for those apps there21:42
asacand hav ea place to basically maintain a temp branch for all21:42
asacbfiller: risk is tha tif one app is infected and you notice dduring testing you have to ask to drop that21:42
asacand rebuild everything21:42
asacbfiller: oh wait21:43
asacbfiller: at best dont put an app in that is currently named as flaky21:43
asaci would like to see if we can land green apps tomorrow21:43
asaceven if the image isnt good yet21:43
om26erbfiller, you need apport-retrace -sR21:43
robru-sickbfiller, kgunn just ping me when your requests are ready to be assigned.21:44
kgunncool thanks guys21:44
asacrobru-sick: lets keep one of the freed silos for bregma/desktop landings ... e.g. jus thandout one for bfiller and one for kgunn after those three are freed21:44
kgunnrobru-sick: sorry you're sick...21:44
robru-sickkgunn, thanks21:45
asacrobru-sick: 002 008 and 006 (after the merging)21:45
bfillerasac, robru-sick : ack, let me review what I have in the pipeline and rearrange into one21:45
bfillerom26er: thanks, trying21:45
asackgunn: bfiller: so 006 will be avail really soon. 002 and 008 will only happen after stuff is through proposed21:45
asacyou guys decide who goes first :)21:45
* ogra waits for asac to hand out boxing gloves21:46
asacthink bfiller could potentially land green apps tomorrow potentially21:46
asacso maybe maybe he might be the one that could benefit the most from being ready today21:46
asac(but no promisses)21:46
kgunnbfiller: you can rock n roll first if you wnat21:46
seb128asac, there is a flag to ignore the "going through proposed" step if you want, I'm using it for desktop stuff atm, beta freeze keeps stuff even if they are ready, no reason to not merge back21:47
asacseb128: i think our touch stuff goes in by bot21:47
asacnot?21:47
ograit does21:48
seb128"in by bot"?21:48
ograjust not the stuff that overlaps with desktop/flavours21:48
seb128right21:48
asacseb128:  we have an auto approve bot21:48
asacso i think lets not use this option21:48
seb128well, you can clean the silo even if stuff are still in proposed, that's on box to click on21:48
seb128no we don't21:48
asacyeah good to know21:48
asacoh right its bregma :)21:49
seb128we have to click a "clean and merge back" button21:49
asacso yeah lets use that for those landings21:49
seb128right21:49
seb128no reason the unity8 session needs to lock a silo21:49
seb128just merge back21:49
bregmait's having trouble merging back, seems to want to push to the wrong place21:49
asacrobru-sick: ^^ thast valid for 002 and 008 i guess.21:49
robru-sickasac, ok, on it21:50
asacbregma: hmm21:50
seb128bregma, where?21:50
seb128hum, weird21:50
bregma2014-02-26 21:47:45,715 INFO Trying to push to https://code.launchpad.net/~bregma/unity8-desktop-session/trunk21:50
* bregma is resigned to forever have problems21:51
bregmathe merge is not lost, so I'll go ahead and clean the silo, I;m done with it21:51
seb128bregma, trunk is owned by you, whoever set that project up didn't do the checking it seems, the jenkins bot needs commit access to the trunk21:52
bregmathe merge can get clean up later21:52
robru-sickbregma, well, the problem there is that you personally own lp:unity8-desktop-session, so citrain doesn't have permission to push there21:52
bregmadoesn;t make sense why that would be, but I'll figure it out21:52
seb128bregma, https://code.launchpad.net/~bregma/unity8-desktop-session/trunk21:52
seb128that's owned by you21:52
seb128lp:unity8-desktop-session21:53
robru-sickbregma, well when you created the project, you would have set up the trunk branch as a branch that you owned, instead of a team-owned branch21:53
bregmaI understand that's how it looks, but I'm fairly sure I didn;t do that, I must have messed it up some other way21:53
bfillerom26er: getting corrupted stack trace, nothing useful from it21:53
bregmadoesn;t matter, I can recover21:53
om26erdarn!21:53
om26erbfiller, maybe try the crash file from the other failure ?21:54
robru-sickbregma, please take this branch specifically: https://code.launchpad.net/~ps-jenkins/unity8-desktop-session/latestsnapshot-recup and push it to something like lp:~unity8-team/unity8-desktop-session/trunk and then reconfigure the lp project trunk to point at that21:54
bfillerom26er: I tried both21:54
om26er:/21:54
om26erbfiller, thats basically a dead end21:55
bfillerom26er: wondering if this might be the same crash that dialer experiences due to unity-mir bug21:55
om26erbfiller, no, thats probably a different crash, with different symptoms21:55
bfillerom26er: this the bug I'm talking about https://bugs.launchpad.net/ubuntu/+source/mir/+bug/124040021:56
ubot5Ubuntu bug 1240400 in unity-mir "dialer-app crashed with SIGSEGV in __GI___pthread_mutex_lock()" [Critical,In progress]21:56
bfillerom26er: basically we need to be able to reproduce the crash on a device to see what's going on I think21:57
bfillerom26er: going to try and reflash with 209 image and try again with that latest crash file21:57
om26erbfiller, dialer-app handles that crash through an exception, it happens on the end of the test normally but yess we need to see on the phone what really happens21:58
bfillerom26er: can you get it gallery to crash when you run in a loop or run the entire suite on the device?21:58
om26erbfiller, it failed once, so not exactly reproducible but it did happen, i'll run it again21:59
bfillerom26er: run "ulimit -c unlimited" first from a shell before you start the tests21:59
bfillerthat way a core file will be gerneated if crash21:59
bfillersergiusens: what's the new command to reflash with the 4.42 based image?>22:00
bfillercan't keep track of it all22:00
* bfiller getting old22:00
ChickenCutlassbfiller: just use devel-propsed22:00
bfillerChickenCutlass: phablet-flash?22:00
ChickenCutlassbfiller: ubuntu-device-flash -channel=devel-propsed22:01
bfillerChickenCutlass: as I said, thanks22:01
asacbregma: i guess you should merge manually and then move the trunk to a better team22:02
asacnot exactly sure what rules to obey for the merging though, but maybe seb128 knows22:02
seb128seems like bregma figured it out, the silo is being cleaned (didrocks included enough checkboxes/override in the system to let us unblock)22:03
robru-sickasac, no need to merge manually, merged branch exists at https://code.launchpad.net/~ps-jenkins/unity8-desktop-session/latestsnapshot-recup already22:03
bfillerrobru-sick: silo 13 tested, ready to be released22:10
om26erbfiller, I have been able to reproduce the crash with albums tests here. what happens is that its able to switch to the albums tab but that tab is shown empty, i.e. no buttons on the toolbar and neither is there any 'dummy' album shown22:11
om26erand after a few seconds ofcourse the app has to go22:11
bfillerom26er: cool, did it generate a core file?22:12
om26erbfiller, i see a core on /home/phablet22:12
bfillerom26er: try running "gdb /usr/bin/gallery-app core"22:13
bfillerom26er: then "bt" once in gdb22:13
om26erbfiller, says corrupt stack22:14
bfiller(:22:14
asacrobru-sick: oh nice22:14
bfillerom26er: can you tell me how to reproduce the crash?22:14
asacnice feature22:14
om26erbfiller, run this multiple times. autopilot run gallery_app.tests.test_album_view22:15
bfillerom26er: let me try22:16
om26erbfiller, i reported a crash report from the file https://bugs.launchpad.net/ubuntu/+source/gallery-app/+bug/128538722:16
=== seb128_ is now known as seb128
ubot5Error: ubuntu bug 1285387 not found22:16
om26erbut its pretty much going to be useless as well22:16
robru-sickbfiller, thanks22:20
sergiusensbfiller, might want 't a a bt' for a bigger picture view22:48
bfillersergiusens: I've tried that22:48
sergiusensare looking at the crash on start?22:48
sergiusensbfiller, cause rsalveti was thinking that they all have the same root cause22:48
bfillersergiusens: no, just a very hard to reproduce autopilot failure on gallery tests22:48
sergiusensah, ok22:49
bfillersergiusens: don't think it's on startup though22:49
bfillersergiusens: http://ci.ubuntu.com/smokeng/trusty/touch/mako/206:20140224:20140224/6796/gallery_app/22:49
rsalvetisergiusens: bfiller: it's the same one22:49
rsalveti  #0 0x3810d8f8 in ?? ()22:49
rsalvetithat's the same offset we get with the qmlcrash and unity822:49
bfillerrsalveti: really?22:50
rsalvetiyes22:50
rsalvetiv8 code22:50
rsalvetiqt v822:50
bfillerrsalveti: you're looking at the .crash file?22:51
rsalvetibfiller: stack trace22:51
rsalvetibfiller: happens when you start gallery-app22:51
rsalvetisame for qmlscene and unity822:51
bfillerrsalveti: how do you reproduce?22:51
bfillerI can't make it happen22:52
rsalvetibfiller: open/close it until you get it22:52
rsalvetiyeah, it's painful22:52
rsalvetiand not going to help you either22:52
rsalvetiwe're debugging this crash like 2 days already22:52
rsalvetidoesn't happen with qt5.2 though22:52
bfillerI guess that's good22:52
bfillermove on then22:52
bfillerrsalveti: is the dialer-app crash the same too?23:00
rsalvetibfiller: check the stacktrace23:01
bfillerrsalveti: k23:01
rsalvetiif it ends with d8f8, then yes23:01
ChickenCutlassrsalveti: if this crash does not happen with 5.223:01
ChickenCutlasswhy are we wasting time23:02
* ChickenCutlass is confused23:02
bfillerChickenCutlass: +123:07
bfillermove one23:07
bfilleron23:07
bfillerrobru-sick: per asac earlier, I've combined a few asks into line 36. so that is ready for a silo23:33
robru-sickbfiller, alright23:42
kgunnfginther: surely you're eod....23:43
kgunnbut whoever is taking up vanguard23:43
kgunni've been hawking this for a while23:44
kgunnhttps://code.launchpad.net/~afrantzis/mir/update-valgrind-armhf-suppressions/+merge/20842123:44
kgunnand jenkins seems to be ignoring23:44
kgunni suppose i can manual merge...but23:44
kgunnstepping away, bbiab23:44
robru-sickbfiller, ok you got silo 2, please build.23:44
bfillerrobru-sick: ack23:48
bfillerrobru-sick: argh, merge failure23:51
bfillerrobru-sick: will try and resolve later, dinner now23:52
=== bfiller is now known as bfiller_afk

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