=== chihchun_afk is now known as chihchun === maclin_ is now known as maclin [06:36] jibel: so this morning I committed the .results format fix [06:37] pitti, I saw your fix, thansk for that [06:37] jibel: (sent you a google message yesterday, not sure whether you got it) [06:37] jibel: so this needs to be rolled out to tachash, and then we need to reconfigure all existing jobs, right? [06:37] pitti, the format is still something wrong when there is no *-package [06:37] pitti, I did it [06:37] jibel: infinity was pointing out that britney currently considers all jobs as RUNNING, presumably due to either missing or wrong .result files? [06:37] jibel: ah, great [06:38] pitti, right, it's because there were no result files, I regenerated them, and are waiting for britney to pull them [06:38] jibel: no -packages would be if the test bed failed? [06:39] pitti, yes, but we need to send something back otherwise the status in britney will stay running forever [06:39] jibel: oh, you mean it fails as cat doesn't have anything to cat? [06:40] yes [06:41] hmm [06:41] $ sort -u /nonexisting | tr -s '[\n\t]' [06:41] that at least exits with 0 as tr succeeds [06:41] it doesn't spit out anything of course [06:41] but the format of the result file will be wrong [06:41] jibel: or is testpkg-version missing as well? [06:41] jibel: i. e. it doesn't get along with zero dependencies, or is the package itself missing? [06:42] it will be release arch package result instead of release arch package version result [06:42] and it wron't be parsed correcly on the other side [06:42] pitti, test-version is missing as well [06:43] ah; so that would be if it doesn't even get to the point where it unpacks the package, like qemu failing to start or so [06:43] http://d-jenkins.ubuntu-ci:8080/job/utopic-adt-libreoffice/4/ARCH=amd64,label=adt/ for example [06:43] jibel: would something like "release arch package ERROR" be acceptable? to tell apart a test failure from autopkgtestr machinery failure? [06:44] http://d-jenkins.ubuntu-ci:8080/job/utopic-adt-libreoffice/4/ARCH=amd64,label=adt/ [06:44] jibel: btw, http://d-jenkins.ubuntu-ci:8080/view/Utopic/view/AutoPkgTest/job/utopic-adt-binutils/configure still doesn't have the updated config [06:44] pitti, yes that's what I did previously we could return an erro [06:44] r [06:44] I'll add the logic to the job [06:44] so if testpkg-version is not present, that's definitively an error [06:45] jibel: for the port redirection I suspect we run into a race condition and my logic to check whether a port is taken is flawed; I'll look into that today [06:45] pitti, if *-package are missing it's an error as well [06:45] jibel: it might be that all tests were skipped [06:46] pitti, but you'd generate the testbed-package in that case, isn't it? [06:46] jibel: oh, right [06:46] jibel: confirmed [06:47] so [ -e testpkg-version -a -e testbed-packages ] || MAYHEM [07:04] pitti, there is indeed an important risk of race in find_freeport because there is a lot of time between the moment the port is allocated and the moment qemu is really started and the port really open. [07:04] jibel: *nod*, that's what I need to fix [07:05] we don't even need to do this at all normally, only when using --shell-on-failure [08:19] jibel: I rolled out several fixes (including proper port locking) to the adt workers [08:20] jibel: when you do the mass-rollout of the new job configs all jobs will automatically re-run, right? [08:21] pitti, no, I requeue them manually [08:21] pitti, should I do it now ? [08:23] (so that we get proper .results files) [08:23] jibel: I guess we should wait until the "error" logic for the .results file is done? [08:25] jibel: just to confirm, the results for a successful run like http://d-jenkins.ubuntu-ci:8080/job/utopic-adt-apt/2/ARCH=i386,label=adt/console look ok now? [08:25] pitti, okay, pushing that fix now [08:26] pitti, it looks ok, I didn't notice the double package name yesterday, sorry about that. [08:26] jibel: no worries, I didn't think about it either; I just wanted to check that it propagated to the job config [08:27] jibel: looking at results.history it seems the results files still don't propagate? apt is still RUNNING there [08:28] pitti, last run of britney was more than 90min ago [08:28] ah [08:33] pitti, http://paste.ubuntu.com/7328311/ actually the format was right when there is no *-packages but the extension was wrong, the rest is unchanged and an error will be reported [08:34] err, without $PACKAGE in the first part [08:39] jibel: line 4: [08:39] [ $RET -gt 0 ] && result="FAIL" [08:39] jibel: this should be [ $RET -eq 0 ] [08:39] ... || [ $RET -eq 2 ] || result=FAIL [08:40] pitti, if adr-run return 0 or 2 it is a success, isn't it? [08:41] jibel: correct (2 is "some tests skipped") [08:41] pitti, the case where the resturn status is 2 is handled earlier, just not in the pastebin [08:41] jibel: ah, so an error doesn't go into .results, but a separate .error [08:41] jibel: ah, ok [08:42] jibel: line 5 still has the additional $PACKAGE, that needs to go [08:42] pitti, I dropped it [08:42] jibel: good; LGTM then [08:42] jibel: the [ -f "$resultfile" ] && bit is a Should Not Happen™ though :) [08:42] pitti, I think I'll put all of this in a shell script rather that updating the XML [08:43] jibel: ah good, then we don't need to update the job config to fix the script [08:44] pitti, espacing shell scripts and encapsulating them into xml file makes me sick :) [08:45] that's how it looks after encoding http://paste.ubuntu.com/7328364/ horrible! [08:45] Morning all [08:46] jibel: yeah, I know; +1 for shell script, also for easier updates [08:46] and if jenkins goes AWOL we don't lose it [08:52] pitti, I pulled the new config on tachash [08:52] and the queue is empty, I think we can do a mass retest [08:53] * pitti puts finger in ears === maclin__ is now known as maclin [08:55] pitti, 61 pacakges in the queue [08:55] they should appear on jenkins soon [08:56] jibel: ok, let's see what I broke this time :) [08:58] at least the syntax of the shell script seems correct. Jobs start :) [09:01] pitti, is test output buffered? eg. http://d-jenkins.ubuntu-ci:8080/job/utopic-adt-apt/ARCH=i386,label=adt/3/console don't display the output of the run-tests [09:04] jibel: there is no live output yet for adt-virt-qemu, that's rather tricky todo [09:04] jibel: that's on my list, though [09:04] pitti, ok [09:05] jibel: that one just succeeded [09:05] pfff, no fun [09:05] oh the queue length [09:06] we'll immediately see if there is still a race in port allocation :) [09:08] jibel: it properly uses flock() now, that really ought to do :) [09:09] jibel: ok, do you see anything else burning ATM? if not, I'll go ahead with eliminating dh_listpackages (I installed a local copy on the workers which don't have debhelper installed, but that's a temp workaround) [09:09] pitti, no, sounds good. I'll monitor the flow to see if anything else needs fixing [09:10] jibel: I guess all of that didn't affect the bug that results fall off excuses and it sometimes promotes failures, right? [09:11] jibel: I think the other day you said it was a "stupid sorting bug" or so [09:11] pitti, right, I'll push the fix today [09:11] jibel: oh, you have a fix for that? awesome [09:13] pitti, I hope, I simplified the history with just 1 line per tuple package@version/dependency@version and the result [09:14] that'll remove the problem when there are several causes that triggered a build and different results for the same package/dep [09:15] I planned to do it earlier but didn't anticipate I'd have to do so much manual testing this cycle again :( [09:26] pitti, http://d-jenkins.ubuntu-ci:8080/view/Utopic/view/AutoPkgTest/job/utopic-adt-fftw3/ARCH=amd64,label=adt/2/console [09:27] pitti, it generated an error file, so at least it validates thios point [09:27] -o [09:27] jibel: but isn't "echo utopic amd64" missing $PACKAGE? [09:27] jibel: indeed, I'll put that tar error on my list [09:28] I made a typo [09:49] jibel: do you have an idea about http://d-jenkins.ubuntu-ci:8080/job/utopic-adt-gzip/5/ARCH=i386,label=adt/console ? [09:49] pitti, yes, I moved the script from jenkins to a shell script. I'm on it [09:50] jibel: ack, thanks; I need to go offline in 10 mins to change trains [09:50] jibel: anything I need to do urgently? [09:50] I eliminated dh_listpackages and rolled it out, so the workers are once again hack free (i. e. clean git head checkout) [09:50] pitti, no, I'll go through the failures [09:50] and let you know if anything is not right [09:51] jibel: ack, thanks; so for now I know about the tar error (no idea yet) [09:52] uh, that sounds really wrong qemu-img: Could not open '/home/auto-package-testing/cache/disks/adt-trusty-amd64-cloud.img [09:52] but probably better to fix the usage of the shared dir to avoid the tar hack altogether [09:52] trusty?? [09:52] on aldebaran. looking [09:53] jibel: oh, perhaps we are running the new job template for trusty SRUs? [09:53] but these jobs should all exist [09:55] on aldebaran: $ distro-info -d [09:55] ubuntu-distro-info: Distribution data outdated [09:55] ah, so that hits the "trusty" fallback [09:55] it's been updated everywhere else [09:55] curious why it doesn't happen on the others [09:55] ah [09:56] ok, AFK for ~ 20 mins or so === _salem is now known as salem_ === salem_ is now known as _salem [10:44] jibel: back online [11:32] jibel: sorry, internet in the Thalys just sucks :/ will need to do offline work [12:04] jibel: ah, I can reproduce the "tar: Unexpected EOF in archive" bug locally \o/ [13:14] pitti, for linux {standard input}: Fatal error: can't write drivers/scsi/isci/.tmp_host.o: No space left on device [13:14] pitti, in trusty, there was a list of package to run on disk instead of memory [13:14] packages [13:15] pitti, for mysql, fork failed sleep 1 second and redo: Cannot allocate memory at lib/My/SafeProcess/Base.pm line 53. [13:17] pitti, I'll increase to 2G [14:16] pitti, I reviewed test failures. http://paste.ubuntu.com/7330086/ is the list of tests that potentially fails because of the new test environment. I am not sure about u-drivers-common though [14:17] pitti, I'll fix 1 and 2 [14:18] pitti, BTW indicator-datetime doesn't like comments in /etc/timezone. It displays all the content of that file in the indicator === zoopster_ is now known as zoopster [19:10] balloons, where would you send someone to report an upgrade an LTS-to-LTS upgrade bug? [19:10] balloons, elopio hi [19:28] Letozaf_, hello [19:28] balloons, hi [19:28] alesage, against the source package and/or the upgrade tool used [19:29] balloons, unknown, he's having boot-trouble [19:29] balloons, sorry to assail you, just don't know where exactly to take this q :) [19:29] balloons, and he's asking "how to help", etc. [19:30] balloons, I am having the same problem I had time ago in reminders app on seeing the objectNames I put in the qml not sure if I have to tell you or elopio [19:30] alesage, ohh.. is it live? [19:31] no worries, yes we can troubleshoot. how did he upgrade? and can he get console access? [19:31] I would want to use ubuntu-bug to send a log to lp with th ebug [19:31] Letozaf_, what was the original resolution to that? [19:31] do you remember how we fixed it? [19:32] balloons, I think it was that the reminders binary didn't point at the correct qml file [19:32] balloons, so -q was added to the binary [19:33] Letozaf_, ohh right! yes, it was the launching issue [19:33] balloons, could be I am doing something wrong as it was quite some time ago I used reminders [19:33] balloons, getting some info from him thanks === roadmr is now known as roadmr_afk === roadmr_afk is now known as roadmr === roadmr is now known as roadmr_afk === roadmr_afk is now known as roadmr