[04:42] Good morning [05:15] hi === cpaelzer_ is now known as cpaelzer [08:00] morning all [08:03] hey willcooke! [08:10] it's quiet here today! [08:20] or is my IRC not working? [08:25] bonjour seb128! [08:25] hey willcooke [08:25] hey pitti, how are you? [08:25] seb128: quite fine, thanks! how about yourself? [08:25] I'm good thanks! [08:25] did a nightshift with reworking the autopkgtest web result browser [08:26] ah, nice [08:28] autopkgtest.u.c. takes ages to update right now and keeps running out of inodes [08:30] :-/ [08:41] SO ANGRY [08:42] Laney: ugh, what's up? [08:42] hey pitti! [08:42] I went to ride a new cycle route with some people from the council [08:42] but after I got there I discovered I had a puncture [08:43] so went to try and find one of the city hire bikes [08:43] annoying [08:43] EVERY location I went to had none available [08:43] urgh, annoying²! [08:43] so I have given up [08:43] even the one by the station [08:43] Laney: do you have kevlar tires? [08:43] sorry, visitors! [08:43] they are supposed to be "puncture proof", whatever that means [08:44] I've had some with a kevlar stripe inside, haven't managed to kill that yet [08:44] hey Laney! [08:44] what a way to start the day :-/ [08:45] at least it's sunny :P [08:46] oh also I have security bolts [08:46] so have to go home, get the tool and come back before I can fix it [08:47] and I never got to tell them the shortcomings of their new route [08:47] :(! [08:47] * Laney exhales [08:47] :-/ [08:47] how are you seb128 et pitti? [08:48] Laney: quite fine, thanks; have been hacking on a new autopkgtest results browser half of the night.. [08:48] http://autopkgtest.ubuntu.com/browse.cgi/packages/systemd [08:48] oh, how come? [08:49] Laney: you still remember the "running out of inodes" problem due to the gazillion static files? [08:49] I thought I had noticed some missing results [08:49] Laney: and current a.u.c. has a latency of several hours [08:49] afraid not [08:49] so this is dynamically querying swift? [08:49] Laney: so I changed the download from "save into files" to "save into sqlite" [08:50] Laney: and completely rewrote the results browser with flask and dynamic page generation/sql query [08:50] which has become absurdly small [08:50] Laney: no, that'd be way too slow [08:50] but sqlite is awesome [08:51] the whole ~ 700K test results are in a ~ 50 MB database, and you can query the table for any result (with a join) in a negligible amount of time [08:51] nice [08:51] so browse.cgi now has a latency of a few minutes only (real production data) [08:51] still missing "running" and "test list" [08:51] so it's not the default view yet [08:53] Laney: https://git.launchpad.net/~ubuntu-release/+git/autopkgtest-cloud/tree/webcontrol/browse.cgi?h=db-webui → that's all (plus a few HTML templates in https://git.launchpad.net/~ubuntu-release/+git/autopkgtest-cloud/tree/webcontrol/templates?h=db-webui) [08:53] quite impressive, should have done this right away.. [08:54] flask looks quite simple for this task [08:54] nice [08:54] running will still be the amqp thing though? just processed dynamically? [08:55] Laney: yeah, I want to do the brutal amqp flush on demand instead of every 10 s [08:55] Laney: listening to and collecting logtails still needs to stay a permanently running process, but that's not a biggie -- that can just continue to write a HTML include snippet [08:56] and the flask app just slurps that in [08:56] or maybe even the template, need to learn about that stuff more [09:03] pitti: good work! results latency was a bit annoying [09:03] more than once I used swift's api to find things before a.u.c knew about them [09:03] yeah, me too [09:04] and the box running out of inodes despite having two huge hard disks now is also a pain [09:04] Laney: the other day we discussed how to attach external storage to cloud instances (that was for this context) [09:04] the inode thing? [09:05] right [09:05] nod, that sucks [09:21] hey guys, anybody running unity8 on desktop? I got a fatal error trying to re-install a click === dpm_ is now known as dpm [09:51] Laney, seb128: Hey. I was trying the tedg's branches for systemd... But it doesn't seem to start here, do you have any clue why http://pastebin.ubuntu.com/23115333/ could happen? journalctl doesn't say much [09:52] hey Trevinho! [09:52] :) [09:53] Trevinho, what's the status of indicator-common? [09:53] sorry [09:53] indicator-pre [09:53] pre? [09:54] those are After=indicators-pre.target [09:55] seb128: indicators-pre.target loaded inactive dead [09:56] journalctl doesn't say much about indicators [09:57] ah, /usr/lib/x86_64-linux-gnu/unity-compiz-profile-select isn't there too [10:01] hey desktopers [10:04] hi ricotz [10:04] seb128: so...... fixing that unity file doesn't seem to change the things [10:05] seb128: if I manually do systemctl --user start unity7 it starts [10:05] but not otherwise [10:06] hey ricotz [10:06] Trevinho, you need pitti or Laney I guess [10:06] I don't remember how those systemd target are supposed to work [10:07] but you shouldn't have to manually start the unity7 unit [10:08] hi [10:08] eh, yeah... it's quite expected [10:08] hi Laney [10:08] is there a silo? [10:08] Trevinho, seb128, Laney, hi [10:09] is there something wrong with mdds which seems not actually be published according to apt-cache -- https://launchpad.net/ubuntu/+source/mdds/1.2.1-1 [10:09] Laney: yeah, 093 [10:10] Laney: although it has wrong paths in prestart scripts [10:10] Laney: so you should manually fix /usr/lib/systemd/user/unity7.service pre-start script to include the /unity subpat [10:10] ricotz: h [10:12] hi ricotz [10:12] you want #launchpad for that [10:12] ricotz, what binary are you looking at? [10:12] seb128, libmdds-dev [10:14] https://launchpad.net/ubuntu/yakkety/amd64/libmdds-dev [10:16] I don't seem to be able to parse this log or make sense of what it says [10:16] Laney, which means it is gone? although launchpad still lists it [10:16] $apt-cache policy libmdds-dev: https://paste.debian.net/plain/797664 [10:17] It didn't publish the binary or something [10:17] go ask #launchpad [10:18] Laney, will do [10:18] there was and maybe still is an old bug that made launchpad eat binaries on successive promotion/demotion [10:19] like it seems like that it was moving forth and back between main and universe [10:19] which might have hit that launchpad bug [10:19] in which case it needs a no change upload to get a new build [10:19] Trevinho: oh, is there an updated MP for unity7 now? https://requests.ci-train.ubuntu.com/#/ticket/1710 doesn't have unity, tedg removed it [10:19] could be, I don't know how that bug looks in logs [10:20] you can soemtimes copy binaries back over themselves too [10:21] Trevinho: I suppose there's an enablement symlink missing from ubuntu-session.target.requires (or ubuntu-sessin.target should grow a Requires=unity7.service) [10:22] Trevinho: in our staging git we have /usr/lib/systemd/user/ubuntu-session.target.wants/unity7.service -> ../unity7.service [10:46] pitti: mh, so probably it's in that silo.. I guess I should move that back there then [10:47] the thing is that I've already added libindicator... seb128 could you remove libindicator src from the landing 093 ppa? So that I can rebuild and get rid of the systemd stuff for now from there? [10:47] I asked tedg about the status of that PPA the other day, but didn't get a response yet, so I can't say what's up with this :/ [10:48] i. e. I guess this is blocked on $something [10:48] also wondering why it's targetted to xenial and vivid too [10:48] shouldn't it be only yakkety? [10:48] not sure, the usual triple-landing madness? [10:48] if the branches are diverged, then we surely only need it for y [10:48] I'd remove that, I guess we don't want to SRU it, right? [10:48] right [10:48] well, maybe at some point [10:49] but that shouldn't block landing in y [10:49] Trevinho: You're missing the symlink, and also Environment=COMPIZ_CONFIG_PROFILE=ubuntu [10:49] with those two it starts for me [10:50] Laney: symlink where? [10:50] as pitti just said [10:50] Trevinho: in our staging git we have /usr/lib/systemd/user/ubuntu-session.target.wants/unity7.service -> ../unity7.service [10:50] Laney: can you plase comment https://code.launchpad.net/~ted/unity/systemd-unit./+merge/300624 ? [10:50] but it should be requires rather than wants, I'd think [10:50] eh, but who should handle that? [10:50] yes, agreed [10:50] ubuntu-desktop package? [10:50] Trevinho: once we have unity ported we can add Requires=unity7 to ubuntu-desktop.target indeed [10:51] Trevinho: but this would be a lockstep transition, so shipping the enablement symlink in the unity7 package is easier for now [10:51] easier for now if unity adds the symlink, but it should more properly be in the target [10:51] pitti: so for now it's up to the unity package to setup such symlik? [10:51] we can clean this up later [10:51] snap [10:51] Trevinho: yes, as we try to decouple the steps of the transition as much as possible [10:51] mh, ok [10:51] at the price of needing two rounds (a second one for removing upstart jobs, and transitional things like this) [10:52] but still easier all around, look at how long this ticket has baked already [10:52] Trevinho: commented [10:52] IMHO you could branch tedg's branch and fix it yourself [10:52] be quicker [10:53] yeah, I was thinking the same [10:53] Laney: however, the environment thing should be set by the prescript, not by the .service, since it's dynamic now [10:54] does it do that now? [10:55] yeah [10:55] Laney: but you need to fix the path [10:56] I did [10:56] symlink is there... but nothing starts here too -__ [10:56] not sure how you pass back environment from the ExecStartPre [10:56] told you, you need to fix the environment [10:56] maybe it has to be a different service then [10:56] or target I mean [10:57] or whathever... in systemd language [10:57] on which unity7 depends on [10:57] so that the env is set [10:57] Laney: I don't get anything in the journalctl though... and if I manually systenclt start it the env is set... So.... [10:57] maybe make that script call the update-activation-environment thing [10:57] I thought it was working [10:58] in theory it should be done by systemctl set-environment, isn't it? [10:59] most jobs use dbus-update-activation-environment [10:59] I mean, without environment compiz should start anyway... but it doesn't [10:59] --systemd [11:03] Laney: you also still see [11:03] unity7.service loaded inactive dead Unity Shell v7 [11:03] inactive because? It doesn't parse the .wants dir? [11:04] anyway... the environment thing isn't needed... maybe i didn't reload the daemon [11:05] and.... I don't get integrated menu working... [11:05] gtk module not loaded [11:06] well, for dbus-activated stuff... So iguess the dbus-update-activation thing has to be used [11:06] for gtk module [11:06] oh, it does it [11:06] :o [11:07] haha [11:07] you made it work? [11:07] yeah, i had to daemon-reload it seems [11:08] however... it feels slower than upstart [11:08] Laney: I've this problem with the terminal though [11:09] when loaded from the dash [11:09] not from Alt+f2 [11:11] Laney: https://usercontent.irccloud-cdn.com/file/pILG7vb4/ [11:11] and dash uses nothing but gapplication launch.. [11:11] so it should take care of dbus environments [11:13] pitti: you have any clue what could be? [11:13] it's right here [11:13] look in the journal and you should be able to see the order it was started in === hikiko is now known as hikiko|ln [11:14] I always got such of these problems in this VM... but.... they were always actual bugs (like this always loses the races) [11:15] Laney: oh, yeah... it's after unity7 [11:16] that is weird [11:16] https://www.irccloud.com/pastebin/Le2O4BPq/ [11:16] it has Before=unity7.service [11:16] yeah... [11:16] you need a pitti [11:17] did you put it in ubuntu-session.target.requires/ or somewhere else? [11:17] that one no... I didn't change nothing a part that unity7 thing in the wants dir [11:18] oh [11:18] I've it at /usr/lib/systemd/user/ubuntu-session.target.wants/unity-gtk-module.service though [11:18] who added that? [11:18] unity7 [11:18] I mean [11:18] unity-gtk-module puts that symlink there, it's right [11:19] meh I restarted like 10 times and it's always been correct [11:19] Laney: it does it since it has [11:19] [Install] [11:19] WantedBy=ubuntu-session.target [11:19] ? [11:19] that would do it if you called systemctl enable, but the package itself probably makes it [11:19] (and the same should be done in unity7 target or you want it to be at debian links level, as a side request?) [11:19] ah, ok [11:20] but RequiredBy for unity [11:21] it worked now... meh. maybe it was because just after the daemon-reload or something like that? [11:23] and... back again in 2nd restart [11:23] this after a [11:23] sudo service lightdm restart [11:23] Not full restart [11:25] tried it like 20 times now [11:26] Laney: so confirmed here... it only happens when forcing a session restart [11:26] with sudo service lightdm restart [11:26] it always happens in such case to me [11:26] not if I logout/login back [11:26] not for me, sorry [11:28] Laney: if it helps, this is what I get in journalctl (log starts just before I restart lighdm): http://pastebin.ubuntu.com/23115687/ [11:29] what does systemctl --user status unity-gtk-module say? [11:31] Laney: it seems correct http://pastebin.ubuntu.com/23115693/ [11:32] Laney: however, timestamps looks the ones of the last good time I ran my session (without restarting lighdm) [11:32] that time is right? [11:32] ... [11:32] Laney: in fact, in the log I sent you before there was no mention of that unit [11:32] probably because it was already active, no? [11:32] Laney: yes, but restart happened at ago 31 13:26:56 [11:33] yeah, so it should be stopped when the desktop stops [11:33] RemainAfterExit=yes ?! [11:33] in the gtk-module [11:34] that is correct [11:37] Laney: I think this is the problem: [11:37] http://pastebin.ubuntu.com/23115712/ [11:38] We need to put it down when the session closes... [11:38] it might fail, bu there's that || true which would still make it go down, isn't it? [11:38] PartOf=graphical-session.target should do that [11:38] that's what I thought [11:39] aha [11:40] I reproduced it! [11:40] it happens if you are logged in on a vt [11:40] Laney: and the same is for systemctl --user status gnome-keyring-ssh.service [11:40] I did it from ssh... maybe that's why [11:40] I've a ssh session open, no vt opens [11:41] probably the same [11:41] guess so [11:41] not exactly sure why [11:41] el martinus of the pitt [11:41] systemctl --user status graphical-session.target [11:41] Active: active since mer 2016-08-31 13:26:37 CEST; 15min ago [11:41] so, it's wrong indeed :-) [11:44] yeah and if I actually log out properly instead of killing lightdm [11:44] it's good [11:44] yeah, that's the thing [11:46] It's all in here... http://pastebin.ubuntu.com/23115727/ (pitti, when possible, if you can have a look...) [11:46] any problem when you don't kill lightdm? [11:46] no [11:46] because if that's the only time... [11:46] if I logoit is fine [11:46] probably more of a minor bug [11:46] I don't know if the same if X is killed [11:48] ok, if gnome-session is killed (segfaults for example), it's fine [11:49] not with X though, and this should be the reason [11:49] sudo killall -9 Xorg [11:49] marco@ubuntu-vmware:~:0$ systemctl --user status graphical-session.target [11:49] ● graphical-session.target - Current graphical user session [11:49] Loaded: loaded (/usr/lib/systemd/user/graphical-session.target; static; vendor preset: enabled) [11:49] Active: active since mer 2016-08-31 13:48:21 CEST; 54s ago [11:49] Docs: man:systemd.special(7) [11:49] ago 31 13:48:21 ubuntu-vmware systemd[6159]: Reached target Current graphical user session. [11:49] ouch, paste didn't work [11:50] * Trevinho goes to lunch... Then fixes. [11:50] oho [11:51] now I get why dropbox told me to change my password [11:51] https://www.theguardian.com/technology/2016/aug/31/dropbox-hack-passwords-68m-data-breach [11:51] ah... fuck them [11:52] bring back ubuntuone! [11:52] The mail was stating [11:52] «This is purely a preventative measure, and we’re sorry for the inconvenience.» [11:52] preventative... [11:53] Laney, you didn't have your password changed since 2012?! ;-) [11:54] teehee [11:54] that was before I started using a password manager too [11:55] What, just got an amazon mail about unlimited cloud storage? [11:55] ah, no it was just marketing [11:57] there [11:57] now it's more like oiw5oitu hehe [12:07] hello peeps [12:07] * desrt goes streaking [12:08] HI [12:08] hot [12:10] trying to keep up my duolingo... [12:10] i have a 6-day streak, which i think is a personal best :p [12:15] hey desrt [12:15] sup sup [12:16] looking forward to visit you soon =) [12:16] ;-) [12:16] the location is still not confirmed right? [12:16] correct. [12:16] wonder what they are doing... [12:17] same thing as always :) [12:18] eating the budget for more frequent team events by delaying the flight booking for everyone? :-/ === hikiko|ln is now known as hikiko [12:29] Trevinho: re from lunch, sorry for the delay [12:30] Trevinho: so what's the remaining problem now? (we already talked about the requires and it seems the env variable is settled too?) === JanC is now known as Guest73215 === JanC_ is now known as JanC [13:03] has his brain broken: there is mdds 1.2.1-1 in yakktey for a day now and it provides libmdds-dev 1.2.1-1 as a binary in main. Yet https://launchpad.net/~bjoern-michaelsen/+archive/ubuntu/libreoffice-staging/+build/10686537 says depwait: "Missing build dependencies: libmdds-dev (>= 1.2)" .... [13:03] huh? how is libmdds-dev 1.2.1-1 not "libmdds-dev (>=1.2)"? [13:04] * Sweet5hark1 feels stupid as there should be an obvious answer ... [13:04] pitti: If you're logged in elsewhere, then the targets aren't stopped properly if you brutally kill lightdm [13:05] Laney: yes; they should be restarted on login [13:05] try it - you get $GTK_MODULES without unity-gtk-module in it [13:07] althought the service remains active so I'm not exactly sure what happens there [13:07] ah, I see it in my VM [13:07] on clean boot even, that smells like a race condition [13:07] with unity7 in systemd? [13:08] no, just with plain yakkety [13:08] I don't think these are related really, gtk-module already needs to run in graphical-session-pre [13:09] it's racy when unity is under upstart of course [13:10] I filed bug 1618886, will look at this ASAP [13:10] bug 1618886 in unity-gtk-module (Ubuntu) "unity-gtk-module.service is racy" [High,Triaged] https://launchpad.net/bugs/1618886 [13:11] well, "of course", it shouldn't be, this is supposed to work with the status quo too [13:11] what no [13:11] how can it? [13:11] I see it ran correctly and it's in show-environment, but not in a terminal [13:11] because unity started after systemd started u-g-m [13:12] not sure, need to think about it (but can't right now) [13:12] well you can remove the override for unity-gtk-module [13:12] and just have the upstart and systemd versions run [13:13] that's not the bug we were talking about [13:14] Laney, seb128, pitti: please tell me how I am stupid, why does it depwait on libmdds (>=1.2) when there is libmdds-dev 1.2.1-1? [13:15] Sweet5hark1: not sure, looks fine; it's even in main [13:15] hum. [13:16] might take a while for launchpad to notice [13:16] Laney: so what is the bug you are talking about? with our PPA (i. e. unity started by systemd) my $GTK_MODULES is right [13:16] Sweet5hark1, see backlog [13:16] that package was only published recently [13:16] even after sudo systemctl restart lightdm [13:16] Sweet5hark1, the binary went missing due to a launchpad bug after an archive admin promoted/demoted it [13:17] ric_otz asking about it earlier on #launchpad and cjwatson rescued the binary by doing some sort of copy [13:19] kudos to cjwatson then. also I feel better for not being insane (however hiding, because Im a lazy bum that didnt read the backlog). thanks for the answer, seb128! [13:20] Sweet5hark1, yw! [13:21] pitti: you need to have another session open, e.g. a vt login or presumably SSH [13:23] Sweet5hark1, hey [13:26] Sweet5hark1, https://paste.debian.net/plain/798056 [13:28] Laney: oh, because unity-gtk-module.service stays running, but something else pokes a new $GTK_MODULES into the session (via Xsession.d or so) [13:29] pitti: it's not just that... [13:29] pitti: systemctl --user status graphical-session.target just gives me things as active [13:29] pitti: see [13:29] https://www.irccloud.com/pastebin/BIcRaSDb/ [13:31] Trevinho: yes, this is expected; it gets restarted on login [13:31] ricotz: thx, already did that in the morning. weird that the dupe was already there before and the cherry-pick added a third. I fixed that and actually set the enable pkgkit var instead. [13:31] i. e. if nothing actually closes the session and just X goes away, there's nothing that would stop g-s.t [13:32] isn't there any way to monitor that? [13:32] Sweet5hark1, I was about to say better check ENABLE_PACKAGEKIT as well [13:32] I mean the graphical session should be off at that point [13:33] Trevinho: yes, I think there is; I'll add that to the above bug [13:34] Trevinho: summary sent to #1618886 [13:34] Trevinho: so I suggest for simulating a "clean" logut use "kill -9 -1" [13:35] which will work around this until this gets fixed [13:35] or just use the logout button :) [13:35] pitti: you can add something like if systemctl --user --quiet is-active graphical-session.target; then systemctl --user stop graphical-session.target; fi [13:36] pitti: I can just stop the graphical session manually.. But I noticed the issue, so I reported. It's not critical, but I gues it could cause some troubles on X crashes [13:36] not even sure if you need the if [13:36] Laney: I'd think that at least gnome-session.service shoudl die, so something in the BindsTo= etc. doesn't seem right [13:36] it should die with it [13:36] I will look at it, just not right now, sorry [13:36] and this really isn't a blocker [13:36] Trevinho: right, appreciated; thanks for pointing this out, this needs to become more robust [13:37] as you say, ideally things should stop when X goes away; at least we should restart everything properly [13:37] ok [13:37] Laney: we already restart the session target, but this won't help for services which are already running [13:37] i. e. stop/start should make those bound servcies actually stop [13:37] like u-g-m.s [13:38] pitti: explicitly stopping it will [13:38] *nod* [13:39] * Laney commented [13:40] desrt: your wish has come true [13:43] :-( [13:43] why sad? [13:43] nowhere exotic? [13:44] yeah, and octobre in den haag is windy and grey and wet and not much fun [13:44] haha [13:44] but escher! [13:44] If it's any consolation, it's like that in June as well. [13:44]