[00:00] HedgeMage: getting whitelisted can be done by opening a ticket at: https://launchpad.net/launchpad/+addquestion if we don't get it done quick enough, feel free to poke me and I'll nag the right people ;) [00:03] stgraber: I'm thinking of proposing a patch for 000-daemon-handling to also handle initctl of the upstart package, because I think it's the new acpid script that's using $ROOT/proc at the end of ltsp-build-client... is that OK? [00:03] (Hi all, btw :)) [00:05] Debian is also (slowly) switching to upstart, so that script would be again common, right? [00:06] alkisg: yeah, you'll probably need to verwrite initctl, start, stop, restart and status (assuming all these aren't simple symlink to initctl) [00:06] symlinks, so initctl should be enough [00:06] ok, they are symlinks to initctl ;) [00:06] right, sounds good [00:06] But it won't have the same name, so I'll have to call it with initctl.real $action "$@" [00:07] OK, will do that tomorrow, thanks [00:07] *tommorow [00:07] well, I'm wondering why we actually have to overwrite the binary ... [00:07] stgraber: thanks, I'll file it away and take a look as soon as I can. [00:08] stgraber@castiana:~$ initctl() { echo "It's not really initctl"; } [00:08] stgraber@castiana:~$ initctl blah [00:08] It's not really initctl [00:08] alkisg: ^ [00:08] stgraber: what if some C program tries to execute /sbin/initctl? [00:09] (or shell program, with an absolut path /sbin/initctl...) [00:09] alkisg: they shouldn't, upstart uses dbus for communication and library exists for C programs, initctl is only there for shell scripts. [00:09] yeah, calling with the full path is the only case where doing shell function overwrite won't work [00:09] Oooops so if someone uses dbus to start acpid, we can't block it? [00:09] but I don't know if we have any scripts doing that [00:10] alkisg: indeed [00:10] So we probably *need* umount -l to be safe... :-/ [00:10] ...or at the very least, to exclude /proc from the squashfs line [00:11] alkisg: I did a quick check, on my laptop I don't have any software that's calling start/stop/restart or start-stop-daemon with a full path [00:11] alkisg: well, then the issue is that we are actually starting process inside the chroot that should never have started [00:12] Well, if they're started from a postinst script, what else can we do? [00:12] alkisg: so ideally we should have: function overwrite for start/stop/restart/start-stop-daemon + kill all process started from within the chroot + umount -l [00:12] then we'd be safe [00:12] stgraber: I think dpkg-divert is a little safer, because it also helps *after* the ltsp-build-client script, i.e. when the user wants to update his chroot [00:12] export LTSP_HANDLE_DAEMONS=false, chroot, update etc [00:13] (but I'd prefer it if LTSP_HANDLE_DAEMONS was "autodetected", so that it worked automagically...) [00:14] alkisg: right, though I'd be very very careful with initctl as I'd suspect it gets copied to the initrd and so you may end up with a non booting thin client [00:14] Ooops :( [00:14] it's one thing to override a Debian helper software (start-stop-daemon), it's another to overwrite some part of init ;) [00:15] * alkisg thinks it would be safer if he left this for stgraber, and just did the umount -l / squashfs etc stuff :D [00:17] root@ltsp-root01:~# ls -l /proc/[0-9]*/exe | grep "/opt/ltsp/" | awk '{print $8}' | cut -d / -f3 [00:17] replacing /opt/ltsp by $CHROOT [00:17] that'll list you all remaining process running from the chroot, you can then kill them and /proc will unmount just fine [00:18] that's probably the easiest until upstart starts causing us more issues [00:18] ls -l /proc/[0-9]*/exe | grep $CHROOT | awk '{print $8}' | cut -d / -f3 | xargs kill -9 [00:19] * alkisg is thinking where that would fit best... [00:19] ls -l /proc/[0-9]*/exe | grep $CHROOT | awk '{print $8}' | cut -d / -f3 | xargs kill -9 2>/dev/zero || true [00:19] actually ;) [00:19] Right now, and excluding the fat client script, ltsp-update-image is what unmounts $ROOT/proc first [00:19] as if we don't have any process to kill we don't want ltsp-build-client to fail [00:20] ok, ltsp-update-image would make sense [00:20] as you also want all process to be killed before compressing an updated chroot [00:20] But then again, ltsp-update-image could be modified to not bother with $ROOT/proc at all, and just exclude that directory from the squashfs image [00:20] yeah but usually you don't want a second cron running on your system ;) [00:21] Uhm care to clarify that? [00:21] I mean for ltsp-update-image to pass an "exclude /proc" param to squashfs [00:21] if you update your chroot and get a cron update which uses upstart, it'll update cron inside the chroot and start it [00:21] so if you do a ps aux, you'll see two cron running [00:22] one inside the chroot, one outside [00:22] so even if you ignore /proc when doing the squashfs, you still end up with two cron running on your server [00:22] which can't be good ;) [00:22] umount -l and mksquashfs -e only workaround the fact that you have some process running that shouldn't be running [00:23] Well, that problem would also be faced by someone who doesn't use nbd/ltsp-update-image [00:23] so that fixes the issue of compressing the squashfs but not the fact that you have some process running from within the chroot and that should never happen [00:23] indeed and we won't be able to do much for them [00:24] what we can do is call that command line once at the end of ltsp-build-client and in ltsp-update-image [00:24] "ltsp-enter-chroot" "ltsp-exit-chroot" commands would be helpful :D [00:24] so for nbd users that will be working perfectly and for NFS users, that'll be fine during the chroot build process but they'll need to be careful when updating [00:25] OK, I'll try to put that in ltsp-update-image [00:26] I'll also try to bypass the daemon starting by defining an initctl in the fat client script [00:26] Thanks man :) [00:26] ok, if you use the shell hack, remember that you'll need to also define start, stop and restart [00:26] even if they are symlinks [00:27] OK, I'll try them and see if they work [02:20] evening all [02:29] evening sbalneav [03:03] Hmmm, new pulseaudio seems flaky [03:03] At least, I think it's pulse. [03:04] volume mixer was a bit broken here and video playback got a bit laggy afterwards (as if audio couldn't catch up with video and mplayer would drop frames to sync) [03:04] so I reverted to the previous version for now [03:05] When I play a game under Wine, pulse disappears entirely [03:05] it's still running, but only shows the dummy device. [03:15] thank goodness pulse is getting tested ... [03:25] Ahmuck: Anything else you'd like tested? [03:26] i deserved that [03:27] The sniping's getting a little annoying, quite frankly. [03:27] If you've got a complaint, I wish you'd make it plain. [03:27] So we can deal with it, and move on. [03:29] so there is a pulse issue in edubuntu? [03:29] There's a pulse issue, period. [03:29] nm [03:30] Pulse is a great peice of software, but like all peices of software, it has bugs. [03:30] sorry, i need a break for a while [03:31] You know, Ahmuck's passive-agressive baloney's really getting to be a nuisance. [03:32] If he has a complaint about something, I sure wish he'd just make the complaint, we could deal with the complaint, and move on. [03:32] Instead, he just makes vague hints and sarcastic remarks. [03:54] yeah ... it's getting boring [04:55] is anyone live? [04:55] * CarlitoI3rigante shout for a live person [04:59] CarlitoI3rigante, that's a tough question ;) [05:11] CarlitoI3rigante: Hello [05:21] I'm alive. Awake is debatable, but I'm definitely alive. [05:22] HedgeMage, hi there :) I thought you were going to sleep :P [05:23] Yeah, that didn't happen...well, sort of...I dozed off cuddling in front of a movie with LF, almost ruined dinner, rescued dinner, ate dinner, played with a treasure I found while unpacking, and now can't sleep. [05:23] HedgeMage, nice :) [05:24] heh [05:24] On the up side, there's something incredibly relaxing about typing on a good manual typewriter. [05:24] I love computers, but they can't come close to that sound and feel. [05:25] * HedgeMage gets all nostalgic and makes people think she's crazy. [05:29] HedgeMage, gotta go, bye and good night :) [05:31] Good night [05:31] oops, too late [05:32] sbalneav: which version of pulseaudio are you using? [05:33] sbalneav: the fixed one is 1:0.9.22~0.9.21+stable-queue-24-gfa64-0ubuntu1 [05:38] 1:0.9.22~0.9.21+stable-queue-24-gfa64-0ubuntu1 [05:38] crimsun: ^^^ [05:39] sbalneav: what sort of behaviour? [05:39] sbalneav: also, a verbose PA log might help debug (https://wiki.ubuntu.com/PulseAudio/Log) [05:40] Well, I've got an old windows game I like to play from time to time (homeworld), and after I play it, pulse kind of... stops responding. [05:40] crimsun: Ah, thanks, I'll maybe produce a log. [05:40] sbalneav: so this is using Wine /without/ Neil's Winepulse ppa, I presume [05:41] Yes. [05:41] sbalneav: and otherwise all completely current 10.04? [05:41] yup. [05:41] I updated an hour or so. [05:41] yes, a verbose log would be helpful [05:41] Okiedokie, I'll produce one tomorrow. [05:41] thanks [05:42] I was trying to find something else I could make it die with [05:42] try anything using alsa [05:42] Since "Dies on playing a windows game produced in 1999 under wine" isn't a use case that'll get taken too seriously :) [05:42] i.e., anything that would route through the pulse pcm+ctl alsa-plugin [05:42] ok, will do. [05:43] my most simple tests are speaker-test, ogg123, and mpg123 [05:44] ogg123's a good one. I got oodles of oggs. I'll try that. [05:44] Thanks! === highvolt2ge is now known as highvoltage [21:41] Hmmm dpkg-reconfigure edubuntu-artwork gives 3 choices: default, young, and plain. Would it be possible to also have an "ubuntu" choice there? :) [21:43] Or, could we modify "plain" to use the ubuntu icons instead of the gnome icons, etc? [21:43] Relevant files are in /usr/share/edubuntu-artwork/gconf [22:00] Maybe the best thing to do would be to lower "atomix, edubuntu-artwork, gnome-icon-theme-gartoon, khelpcenter4, kolourpaint4, xaos" from "Depends:" to "Recommends:" in edubuntu-desktop. [22:00] This way if someone doensn't want one of those, he can still keep the rest of the edubuntu-desktop. [22:12] alkisg: Makes sense [22:13] OK, filed there: https://bugs.launchpad.net/ubuntu/+source/edubuntu-meta/+bug/508923 [22:13] Ubuntu bug 508923 in edubuntu-meta "Lower some dependencies to recommentations" [Undecided,New]