/srv/irclogs.ubuntu.com/2013/05/17/#ubuntu-quality.txt

=== meetingology` is now known as meetingology
=== Ursinha is now known as Ursinha-afk
=== Ursinhal is now known as Ursinha
=== _salem is now known as salem_
smartboyhwballoons: Have you slept yet?02:40
balloonssmartboyhw, sleeping now.. good night02:42
smartboyhwballoons: Good night! :P02:44
=== salem_ is now known as _salem
=== _salem is now known as salem_
=== salem_ is now known as _salem
=== _salem is now known as salem_
=== salem_ is now known as _salem
jibelgood morning06:18
jibelpitti, you were right about loop devices, I created 1000 yesterday without problems, I don't know why I had this limit in mind, maybe linux 1.2.x times ... :)06:29
pittijibel: with /dev/loop-control?06:35
pittijibel: as I tried with mount -o loop, and that indeed doesn't seem to request new loop devices if the existing ones run out06:35
jibelpitti, with mknod06:37
pittiah, that works, too :)06:38
pittimy hope is that a less ancient util-linux will actually ask loop-control for more by itself06:39
=== smartboyhw_ is now known as smartboyhw
jibeldidrocks, I moving prepare_fs from the pre-mount hook to a pre-start hook, then will unmount manually in a post-stop hook, see if it fixes the umount issue07:09
didrocksjibel: oh, nice idea!07:10
didrocksI'll be able to confirm if needed07:10
* didrocks continues on the archive thingy07:10
jibeldidrocks, pitti I fixed the loop devices leak in otto by moving all the mounts to a pre-start script and unmounting in a post-stop08:21
didrocksjibel: oh08:22
didrocksjibel: I have a post-stop as well08:22
jibelfurthermount I use umount.aufs instead of umount to fix a problem with permanent links between the delta and the rootfs after the umountr08:22
didrocksjibel: please push so that I reconcile :p08:22
jibeldidrocks, np, it is just a function08:22
didrocksjibel: ok, I handled the copy of those scripts in a better way as well08:22
didrocksI'm archive, finish testing the restore now ;)08:23
didrocksarchiving*08:23
jibeldidrocks, oh, and now lxc-start works without pre-mounting the iso08:26
didrocksjibel: excellent!08:29
jibeldidrocks, r5908:29
didrocksjibel: ok, reconciling with this :)08:30
jibeldidrocks, I dropped all the checks that verifies a squashfs exists on the iso, and they should be moved to the pre-start scripts08:31
didrocksjibel: hum, maybe my merge will be hard then :p08:32
* didrocks tries08:32
pittijibel: rock!08:33
pittijibel: that gets rid of the atexit lazy unmount, too?08:33
jibelpitti, yes08:33
pittireally nice08:34
pittijibel: so in general, we want to push as many smarts as possible into the pre/post-mount script, right?08:34
pittias there they have the potential for getting partial upstream adoption08:35
didrocksjibel: instead of you redoing iso_mount="/run/otto/iso/$(echo $IMAGE | tr '/' '_')" which double the logic, mind if I put that as a parameter?08:35
jibelpitti, I think so, we should stay as close as possible of native lxc commands, what do you think?08:35
jibeldidrocks, go08:35
pittijibel: yes, my feeling as well08:35
pittididrocks: ah, I used that tr (in python, formerly) to allow running several instances in parallel and still knowing which image is which08:36
didrockspitti: sure, that's fine, it's just the logic is in two places now :) otto and the post-stop script08:37
didrocksI'm passing it as a parameter to post-stop08:38
pittididrocks: perhaps we should move the mount point into the /var/lib/lxc/<container> directory?08:39
pittiand call it /iso08:39
pittithat's even more clear, and it doesn't need parameterization08:39
pittisimilarly to /squashfs08:39
pittiwhy didn't I think of this yesterday..08:39
didrockspitti: but we'll be able to mount the same iso in two containers?08:39
pittisure08:39
pittiyou can loop-mount an image as many times as you like08:39
didrockspitti: ok, let me first get back and reconcicle with jibel last changes08:40
didrockslet's see afterward to mount within the container08:40
pittithe "mountpoint" check was just to avoid mounting an iso over the same mount point twice (it's possible, but might lead to refcount issues008:40
pittibut /var/lib/lxc/container/iso seems more elegant to me08:40
didrocksjibel: argh, you failed me08:40
didrocksjibel: as you mount now in pre-mount…08:40
didrocksI can't get access to the iso infos08:40
didrocksso all my consistency check is not possible :/08:41
didrocks(like if you want to keep the delta, and so on…)08:41
pittiah, good point08:41
didrocksjibel just removed my 2 last hours of work :p08:42
jibeldidrocks, you just need to extract .disk/info from the iso isn't it ?08:46
jibeldidrocks, alternatively, you can mount in otto and check in the pre-start if it's already mounted08:47
didrocks.disk/info, README.diskdefines and the dist directory08:47
jibelthis way you can keep the logic in otto and lxc-start will still work08:47
didrocksjibel: let's do this08:47
jibeldidrocks, I found the problem with autopilot and zeitgeist08:52
jibeldidrocks, .local mode is set to 60008:52
jibeldidrocks, I think it's because it exists on the ISO, so we need to override it08:55
didrocksjibel: ah… makes sense08:56
didrocksjibel: hum, I wonder if my bzr rebase didn't failed on some on your file08:56
didrockslike I don't see the pre-start.sh08:56
didrockshum, there is none…08:57
didrocks(on trunk)08:57
* didrocks digs where mounting the iso is08:57
jibeldidrocks, r6008:58
didrocksjibel: ok, I was at r59 :p08:58
jibelYou'd see it if I added it08:58
didrocksI need to rerebase08:58
jibeldidrocks, did you add support for proxies and custom uris in sources.list ?09:01
didrocksjibel: not yet09:05
didrocksjibel: I'm fighting to get something working again, it's weird, I'm mounting in otto, but then, it seems pre-start sees it as unmounted and can't mount it again…09:06
didrocksah, was a transient issue it seems (like it happens sometimes if I don't reboot)09:06
didrocksno, after a manual run, of lxc-start, it's failing09:07
* didrocks should have pushed first :p09:07
jibeldidrocks, https://bugs.launchpad.net/ubuntu/+source/lxc/+bug/118113609:41
ubot5Launchpad bug 1181136 in lxc (Ubuntu) "Empty log file when a container is started with the API" [Undecided,New]09:41
=== xgoo is now known as Guest58857
jibeldidrocks, http://paste.ubuntu.com/5673638/10:21
didrockspitti: ok, FYI, apparmor was the guilty one from our discussion :)10:32
didrockspitti: pushed, if you use otto, you should be able to save/restore old run10:32
didrocks(also, check for image consistency is done)10:32
pittididrocks: oh, which discussion?10:39
didrockspitti: was that discussion above about mounting not working ^10:40
pittididrocks: nice!10:40
pittioh, that10:41
=== smartboyhw_ is now known as smartboyhw
jibeldidrocks, finally on the host: sudo ln -s /etc/apparmor.d/usr.bin.lxc-start /etc/apparmor.d/disable/; sudo /etc/init.d/apparmor reload11:43
didrocksjibel: ok, we need to write that in a procedure :)11:58
elfyphillw: posted and stuck it now12:59
smartboyhwelfy: Did I miss anything?13:03
phillwelfy: thanks!13:06
chrisccoulsonjibel, you seen these firefox failures in saucy?13:32
chrisccoulsonhttps://jenkins.qa.ubuntu.com/job/saucy-adt-firefox/26/ARCH=i386,label=adt/13:32
chrisccoulsonnot sure what is going on with that one13:32
chrisccoulsonalso, https://jenkins.qa.ubuntu.com/job/saucy-adt-firefox/23/ARCH=i386,label=adt/ looks like it's using the source package from proposed, but the old binaries in the release pocket13:33
chrisccoulsonoh, actually13:33
jibelchrisccoulson, yes I have seen that failure for lot of different packages yesterday, it doesn't really make sense to me.13:33
chrisccoulsonactually, the second failure makes sense now. it actually wasn't published in proposed when that ran13:34
chrisccoulsonwhich you already know about :)13:34
chrisccoulsonjibel, did you see my last few comments in #ubuntu-desktop? it seems seb uploaded a broken fontconfig which caused a lot of firefox tests to fail13:35
chrisccoulsonwhich is good ;)13:35
elfysmartboyhw: not that I know of13:35
jibelchrisccoulson, hm, this is possible only for PPAs, for -proposed we really start the test when the binary hits the archive. That's really weird.13:35
chrisccoulsonyeah, that is a bit weird13:36
jibelchrisccoulson, yeah, it's great when tests find bugs :)13:36
chrisccoulsonbut the second error is definitely due to running the test scripts from the new source with the old binaries13:36
chrisccoulsonas it's showing a file missing that was only introduced in to the latest upload13:36
=== francisco is now known as Guest99200
balloonsxnox, are you still about?14:11
=== _salem is now known as salem_
=== francisco is now known as Guest36497
=== francisco is now known as Guest85075
=== salem_ is now known as _salem

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