[01:51] <vibhav> God morning
[01:51] <vibhav> Good*
[02:04] <smoser> cody-somerville, soren if you're interested, https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1136781 describes the issue i was seeing.
[03:57] <psusi> smoser, thanks for subscribing me to the bug report about loop partitions not being cleaned up after detach.. I duped it against the one I filed at the tail end of 2011 with patch attached ;)
[03:57] <smoser> nice
[03:57] <smoser> psusi, i jsut committed the 'update' spuport to growpart
[03:57] <smoser> but now realize it has an issue.
[03:58] <smoser> sfdisk is annoying.
[03:58] <smoser> it insists on calling the re-read blockdevice action
[03:58] <smoser> which fails
[03:59] <smoser> if there is a mounted partition on the block device
[04:01] <smoser> psusi, can you think of anything clever that i can do to avoid that ?
[04:01] <smoser> outside of patching sfdisk to actually not invoke the reread ?
[04:01] <psusi> smoser, and growpart runs sfdisk to make the change?  can't you just ignore that error and then run kpartx -u or invoke the ioctl directly?
[04:02] <smoser> well, i can't just ignore the error, becauase i can't differenciate the error from any other.
[04:02] <psusi> smoser, you could use parted instead and it will take care of both steps ;)
[04:03] <smoser> this is true.
[04:03] <smoser> funny, sfdisk actually has a '--no-reread' flag.
[04:03] <psusi> https://code.launchpad.net/~psusi/ubuntu/raring/parted/resize
[04:03] <smoser> which, one might think would say "do not issue re-read"
[04:03] <smoser> but it doesn't mean that.
[04:04] <smoser> it only means "dont issue it before re-writing partition table as a safety check"
[04:04] <psusi> build/install that ( build copy in my ppa ) and you can just parted /dev/sda resizepart 1 100%
[04:04] <smoser> that is pretty nice..
[04:05] <smoser> that is really nice, psusi
[04:05] <psusi> you think that's nice, you should play with my patches for gparted to do online expand and shrink ( shrink on btrfs only )
[04:06] <psusi> grab root partition, make bigger.. make smaller... apply, done..
[04:07] <psusi> I was using parted this week to resize my partitions on my server that were components of a raid5 array to shrink them down, use the new space to make partitions and build them into a raid0 array, all without rebooting with root on lvm on the original raid5
[04:11] <smoser> psusi, can i use parted command line to do what i want ?
[04:11] <smoser> (without your patches)
[04:11] <psusi> smoser, need my patches
[04:11] <smoser> ie, can i have it just write the partition table the way i want it to
[04:12] <smoser> all i need it to do is write the table
[04:12] <smoser> and *not* call 'reread'
[04:12] <psusi> you could make it write a new table without my patches, but you would have to explicitly rm the existing partition, then create a new one, then ignore the error syncing the table
[04:13] <psusi> which means grepping for that specific message in the output and choosing to ignore it
[04:13] <smoser> :-(
[04:13] <psusi> so, better to just use my patches ;)
[04:13] <psusi> much easier to just say parted /dev/sda resize 1 100% and be done with it
[04:14] <psusi> err, resizepart, not resize
[04:14] <psusi> ( the old resize command also tries to resize the fs, which won't work and has been removed )
[04:14] <smoser> right.
[04:15] <smoser> whats the situation with your patches?
[04:15] <psusi> so you just do the resizepart, it fixes the partition table, fixes the in kernel partition table, then you run resize2fs and you're all set
[04:15] <smoser> you've submitted upstream, but they're not accepted because its languished at the moment?
[04:15] <smoser> i'm just afraid of pulling code that isn't upstream.
[04:16] <psusi> the situation is that parted is in a bit of limbo atm... the old maintainer has left redhat and no longer seems to have time to work on it.. he appointed me and another as co-maintainers... my patches have been sitting in limbo thus far... at some point I'm going to have to give up on waiting and push them and make a new release if the original maintainer and my new fellow co-maintainer don't have time
[04:17] <smoser> hmm..
[04:18] <psusi> my patches were initially based on some rough work by peter uzel who I have been hoping will find the time to review my improvements, add a bit better commentary to his original patches that I started with, add a few more test cases, and then we could move forward
[04:19] <pitti> Good morning
[04:19] <smoser> you have an ubuntu bug for this?
[04:19] <psusi> the original maintainer asked him to be a third co-maintainer but he said he didn't have time for that responsibility
[04:19] <psusi> a bug for which part exactly? ;)
[04:20] <psusi> pitti, morning Martin
[04:22] <psusi> smoser, so far the only concern raised since posting the original patch set to the parted mailing list way back when has been resolve... that was that originally the new command was resize, which is the same command that used to also resize the filesystem contained within the partition, so I changed the new command to resizepart to avoid the conflict
[04:22] <smoser> psusi, just a "please pull this patch" bug.
[04:23] <smoser> right. i remember that.
[04:23] <psusi> in the upstream patches, I also added a "resize" command that yells at you for trying to use a removed command and exits with failure to catch scripts using the old command... when I backported it to Ubuntu, our old parted still has the original resize command so I left that part out
[04:24] <psusi> of course, our old resize command basically doesn't work worth a damn, which is why it was removed upstream
[04:24] <psusi> which is a shame, and something I hope to fix at some point....
[04:57] <smoser> psusi, is https://bugs.launchpad.net/ubuntu/+source/linux/+bug/899717 fixed upstream?
[05:00] <psusi> smoser, afaik, it never got applied upstream... I'll have to repost it
[05:06] <smoser> well, psusi i'm sure i've lost some of your respect.
[05:06] <smoser> http://bazaar.launchpad.net/~cloud-utils-dev/cloud-utils/trunk/revision/219
[05:07] <psusi> I can't believe it's been that long... time  flies when you have a newborn
[05:07] <smoser> that seems reasonable as a "we'll do this right with parted at some point"
[05:07] <psusi> now 1 year old today... sheesh...
[05:07] <smoser> well happy birthday.
[05:07] <psusi> well, yesterday I guess since it's after midnight
[05:07] <psusi> happy march 1st
[05:10] <psusi> smoser, yea, you can either detect the sfdisk error from trying to BLKRRPART and run partx -u, or instead you can just apply my parted patches, and run parted /dev/sda resizepart 1 100% instead
[05:11] <psusi> or /dev/xvda for xen
[05:12] <pitti> am I the only one who cannot see https://blueprints.launchpad.net/ubuntu/+spec/release-r-monthly-snapshots ?
[05:12] <pitti> it keeps coming up in the u-devel@ thread, but nobody except me seems to complain
[05:12] <psusi> pitti, gives me the lost something page
[05:13] <pitti> ok, thanks for confirming
[05:13] <pitti> I was just wondering as at least wendar and ScottK seem to be able to see it
[05:13] <ScottK> pitti: I can't see it now.  I could before.
[05:14] <wendar> pitti: gone for me too
[05:14] <wendar> pitti: though I still have it open in another tab
[05:14] <pitti> thanks for confirming
[05:15] <psusi> ok, wife says it's time for bed.. night all
[05:15] <pitti> night psusi
[05:15] <ScottK> So if this whole process had been a model of transparency, I'd probably credit it to an LP glitch.
[05:19] <micahg> maybe someone got mad since I added a comment to the whiteboard :-/
[05:22] <micahg> slangasek: ^^ know anything about the blueprint disappearing?
[05:23] <pitti> or maybe it got renamed?
[05:24] <pitti> hm no, still first result on https://launchpad.net/+search?field.text=monthly+snapshots
[05:24] <pitti> curious that I can see it in the search results, but not the page itself
[05:24] <pitti> ScottK: well, I certainly hope it wasn't deliberate to hide the page now
[05:24] <pitti> with Rick's discussion kick-off this wouldn't make any sense
[05:25] <ScottK> People sometimes make poor choices under stress.
[05:25] <ScottK> No idea really.
[05:40] <micahg> slangasek: nevermind
[05:40] <micahg> blueprint here: https://blueprints.launchpad.net/ubuntu/+spec/client-1303-monthly-snapshots
[05:41]  * micahg hugs wgrant for finding it
[05:44] <wgrant> (I found it by searching for 'rolling' on https://blueprints.launchpad.net/ubuntu, which unlike the global search page doesn't rely on Google to have updated its index)
[05:44] <micahg> right, I was thinking to do that, but it's late here :)
[05:48] <ScottK> So is the secret sauce for spec naming public?
[05:49]  * micahg would think track-UDS_Code-Name_Goes_Here
[05:49] <wgrant> That seems to be it.
[05:53] <ScottK> What does client mean as a track?
[05:53] <ScottK> Is there a list of tracks?
[05:53] <micahg> ScottK: http://summit.ubuntu.com/uds-1303/
[05:54] <ScottK> So "Ubuntu" is now reduced to "Client" at an Ubuntu Developer Summit.
[05:54] <wgrant> I would guess client == the old desktop track + mobile and tablet
[05:54] <wgrant> Hence the lack of a desktop track.
[05:54] <ScottK> And platform
[05:57] <ScottK> Ah, I see it's downgrades all around.  No sabdfl for the intro either.
[06:08] <pitti> TBH I'm still rather sceptical about hangouts
[06:09] <pitti> their limit to 10 people and its rather bad audio quality don't seem to make it a first choice for a virtual summit
[06:09] <pitti> the live on-air videos work really great, but don't allow participating
[06:10] <wgrant> The plugin also has a bad habit of hanging any browser I throw at it :)
[06:10] <pitti> but I guess we have to try with something and then see how it goes
[06:10] <pitti> I yet have to find something which scales as well and is as robust as mumble
[06:12] <micahg> why can't we use mumble?
[06:12] <pitti> so https://blueprints.launchpad.net/ubuntu/+spec/client-1303-monthly-snapshots is the "monthly" thing, do we have a spec for "rolling release" where we can bring up things like library transitions, ARM PPA builders, and the like?
[06:13] <pitti> micahg: I guess we could; I just seem to remember having heard "hangout"
[06:13] <micahg> well, Canonical folk seem to be keen on hangouts, but 10 people at a time seems to be quite limiting
[06:13] <pitti> video is nice
[06:14] <micahg> sure, but personally, I'd rather have 100 people on audio and have people raising their "hands" in an IRC channel than video limited to 10 people
[06:14] <pitti> so audio-only in mumble would make the experience even worse (compared to IRL) than with hangout, but that won't help much if most of the listeners can't talk/show up at all
[06:14] <pitti> *nod*
[06:51]  * tumbleweed *really* dislikes that use of the word client
[07:14] <czajkowski> ScottK: naming for the blurptints is on the wiki. https://wiki.ubuntu.com/UDS/Create#Filling_Out_the_Blueprint_Fields
[07:15] <czajkowski> Laney: am still getting promted for partial upgrades again today.
[07:16] <Whoopie> arges: Hi, regarding bug 1074923, my debdiff also included a cleanup of the debian-changes which were maybe done unintended. Could you have a look again? Thanks!
[08:52] <geser> does somebody know where https://blueprints.launchpad.net/ubuntu/+spec/release-r-monthly-snapshots moved too? I get a 404
[08:53] <pitti> geser: https://blueprints.launchpad.net/ubuntu/+spec/client-1303-monthly-snapshots
[08:53] <geser> thanks
[09:00] <mitya57> geser: https://blueprints.launchpad.net/ubuntu/+spec/client-1303-monthly-snapshots
[09:06] <Laney> czajkowski: you said it worked yesterday?
[10:21] <zequence> I'm not considered a developer on the ubuntu-devel list. Is that a strict category? I'm a developer for Ubuntu Studio, but not a "Ubuntu Developer"
[10:21] <zequence> I'm referring to the list being moderated
[10:22] <zequence> cjwatson: You're the list admin, right?
[10:26] <maxb> zequence: My guess would be "someone who has upload rights to at least one package"
[10:26] <maxb> But that's just a guess
[10:33] <cjwatson> zequence: The category is "member of the ~ubuntu-dev team"
[10:33] <cjwatson> (Plus case-by-case whitelisting)
[10:34] <cjwatson> I've accepted your post in any case
[10:39] <zequence> cjwatson: Thanks
[10:50]  * mpt is disappointed to discover that en-GB doesn't translate "Display every two weeks" as "Display fortnightly"
[10:54] <pitti> mpt: I guess its LC_TIME also doesn't display 16:00 as "teatime"?
[11:00] <xnox> pitti: yeah and 11am as elevenses speaking of which.
[11:00]  * xnox goes to make tea.
[11:06] <xnox> mpt: which package?
[11:07] <xnox> mpt: seems like a bug cause it's "Every fortnight" for one string but not the other.
[11:09] <xnox> fixed, should be part of next lang-pack update.
[11:10] <Laney> see how fast we are to fix critical bugs?
[11:11] <xnox> I'm sure some people will find controversial for eastern european translating software into English GB locale. Something to do with taking away jobs from british or the like.
[11:11]  * xnox likes the fact that en_gb translation team is an open one.
[11:12] <Laney> that's what we're like over here
[11:34] <mpt> haha
[11:35] <mpt> xnox, and now we've got you suckered in to hacking software-properties-gtk, we'll get you to port it to a gnome-control-center panel
[11:35] <mpt> You know, just while you're there.
[11:35] <mlankhorst> Wait why does it need british translations?
[11:36] <mpt> Because America and Britain are divided by a common language
[11:38] <xnox> mpt: I totally did in launchpad with no source code branches locally =))))
[11:38] <highvoltage> Amerian English is the most different English though, most of the English in the rest of the world is pretty similar.
[11:42] <mpt> Be that as it may, Ubuntu is translated out of American English into other languages and variants, including en_CA, en_GB, en_AU, and even en_NZ.
[11:44] <Riddell> ogra_: any ideas on why this nexus won't talk to a computer any more?  plugging in usb all I get is this in syslog "unable to enumerate USB device on port 2" http://paste.kde.org/684758/
[11:44] <ogra_> Riddell, is it charged ? the n7 HW behaves very odd on low battery
[11:45] <Riddell> ogra_: maybe not, should I just leave it plugged in and hope it fixes itself?
[11:46] <ogra_> you need to charge it ewith the wall charger, an USB port of the PC only provides 500ma, thats about as much as the device actually usue when on
[11:46] <ogra_> the charger should provide something like 2A or above
[11:46] <Riddell> oh that might explain it, how unintuitive
[11:47] <ogra_> well, its fine if you have a desktop running that properly tells you the battery is low :)
[11:47] <ogra_> fi you dont have any measure because working on cmdline all the time or flashing over and over, it will fall over
[12:12] <davmor2> Riddell, ogra_: it will charge from the pc but an entire day gives you about 37% battery charge iirc,  from the wall will do it in about 1hour or so
[12:13] <ogra_> right, and if it runs while you charge it largely eats all that comes from the PC
[12:13] <ogra_> i would be surprised if more than 100mA get into the battery
[12:48] <cjwatson> bdmurray: http://reqorts.qa.ubuntu.com/reports/rls-mgr/rls-r-tracking-bug-tasks.html seems to have lost its mind^Wteam mapping
[13:06] <xnox> jodh: it's not like infinity and I compete on a number of badges, but it is one he doesn't have yet ;-)
[13:08]  * Laney blinks xnox 
[13:10]  * xnox Upstart Developers, Joined 17 minutes ago
[13:10] <Laney> very fancy
[13:10] <jodh> xnox: :-)
[13:15] <seb128> xnox, hum, can we trick you in doing extra desktop work if there is a chance to get a nice badge for it? :p
[13:16] <xnox> seb128: maybe...
[13:16] <seb128> ;-)
[13:16] <seb128> Laney, quick, find some work for xnox!
[13:17]  * xnox finds no artistic value in a haskell badge, just saying.
[13:18] <ogra_> seb128, a badge and a virtual beer at the virtual UDS
[13:18] <ogra_> vitually ...
[13:18] <jpds> ogra_: Nothing like that virtual German beer.
[13:18] <ogra_> ++
[13:19] <xnox> carlsberg probably the best virtual beer in the world
[13:19] <ogra_> lol
[13:20] <cjwatson> Better than actually drinking it for real
[13:20]  * xnox doesn't like beer anyway and not many places have cider outside of the Isles.
[13:20] <pitti> cjwatson: objection!
[13:21] <cjwatson> pitti: I'm sure we can find some common ground, just not with Carlsberg :)
[13:21] <xnox> pitti: you can raise that objection... virtually  =)))))
[13:21] <pitti> cjwatson: ah, that I can agree with :) I thought you meant beer in general
[13:21] <cjwatson> Oh, goodness me no
[13:22] <pitti> . o O { merci dieu c'est vendredi }
[13:22] <Nafallo> sooo. people will stop owe each other beers now? :-)
[13:22] <xnox> Nafallo: offlicense stores in UK accept orders over the internet with home delivery.
[13:23] <pitti> I tried pouring some into a hangout, but it didn't go well
[13:23] <pitti> even though I served it on a tablet!
[13:23] <ogra_> you didnt use the Gfunnel then
[13:24] <ogra_> only works with google provided devices :)
[13:24] <xnox> pitti: did it come with phablets as little salty snacks to go with beer?
[13:24] <jbicha> xnox: here have a badge: https://launchpad.net/~desktop-bugs
[13:24] <Nafallo> xnox: truth... but that would totally exclude Swedish people
[13:24] <pitti> xnox: that might have been the problem
[13:50] <geser> xnox: drinking virtual beer during a virtual UDS?
[13:50] <xnox> virtually yes
[13:51] <sladen> that's virtually a joke
[14:00] <tvoss> czajkowski, ping
[14:02] <bdcomp> Hi, I am trying to backport LibreOffice 4.0.1 to Precise and got this error "ERROR: error 512 occurred while making /build/buildd/libreoffice-4.0.1~rc1/tail_build/prj"
[14:02] <bdcomp> Any hints what is the issue?
[14:03] <bdcomp> The last output before the error is "[build LNK] Executable/idlc make[4]: *** No rule to make target `/build/buildd/libreoffice-4.0.1~rc1/src/0168229624cfac409e766913506961a8-ucpp-1.3.2.tar.gz', needed by `/build/buildd/libreoffice-4.0.1~rc1/workdir/unxlngi6.pro/UnpackedTarget/0168229624cfac409e766913506961a8-ucpp-1.3.2.tar.gz'.  Stop."
[17:10] <dholbach> cjwatson, pitti, soren, stgraber, kees, mdz: can one of you add me to ~uds-organizers? I'm going to be track lead together with Jono for the community track - cjohnston made the other necessary changes in summit already
[17:14] <stgraber> dholbach: doing that now
[17:14] <dholbach> thanks stgraber
[17:14] <stgraber> dholbach: done
[17:14] <dholbach> thanks a bunch
[17:41] <BenC> what's the URL for rootfs tar balls?
[17:43] <cjohnston> zyga: ping
[17:43] <BenC> Nm, found it
[17:43] <zyga> cjohnston: hey
[17:44] <cjohnston> zyga: your blueprints don't follow the correct naming convention, and more than likely won't get scheduled unless they are fixed.
[17:45] <zyga> cjohnston: where is the naming convention announced?
[17:45] <zyga> cjohnston: and is renaming them the only required thing?
[17:45] <cjohnston> https://wiki.ubuntu.com/UDS/Create   the naming convention is the same as its always been..   and yes
[17:46] <cjohnston> jodh: ^
[17:47] <zyga> cjohnston: the naming convention differs each time as in $CURRENT_PREFIX-$THE_STUFF_I_CARE_ABOUT
[17:47] <zyga> cjohnston: thanks for the tip though
[17:47]  * zyga finds it totally silly that this is even important
[17:47] <zyga> cjohnston: so which should I select, my blueprint fits the first three options (client, servercloud and community)
, is how its always been
[17:48] <zyga> cjohnston: where track and cycle are totally irrelevant (launchpad knows which sprint I target it for) and I only care about name, the rest is arbitrary limitation from the scheduling system.
[17:48] <zyga> cjohnston: but let's get back to the point
[17:48] <cjohnston> zyga: that I couldn't tell you.. I would think probably client, but I am not sure..
[17:48] <zyga> cjohnston: ok, I'll pick client
[17:49] <cjohnston> your more than welcome to come up with another way for Summit to randomly guess which track it should be for if you don't like the naming convention requirement
[17:49] <zyga> cjohnston: thanks for the help and don't get my comments personally, I'm just pointing out the absurd abehind some of those choices
[17:50] <zyga> cjohnston: renamed
[17:50] <Laney> cjohnston: is http://fridge.ubuntu.com/2013/02/26/ubuntu-developer-summits-now-online-and-every-three-months/ or http://summit.ubuntu.com/ right wrt times?
[17:50] <cjohnston> I don't know of any other way to make summit guess which track it should be for. And since we use tracks, it has to be assigned to one.
[17:50] <Laney> I assume the latter
[17:50] <cjohnston> Laney: 1400-2000
[17:50] <Laney> rararrhgh
[17:50] <Laney> who can fix the fridge post?
[17:50] <cjohnston> I just pinged jono_
[17:51] <Laney> someone should correct me on -devel :-)
[17:51] <jono_> the fridge post is updated
[17:51] <zyga> cjohnston: I guess the trakc is something I could select on the scheduler, the identifier is really an implementation detail but that's just IMHO
[17:51] <jono_> it was updated a few days ago
[17:51] <Laney> only the start time
[17:52] <cjohnston> jono_: "2pm UTC – 10pm UTC"
[17:52] <jono_> yeah fixing now
[17:52] <jono_> thanks
[17:55] <jono_> Laney, fixed
[17:55] <Laney> merci
[18:14] <cjwatson> cjohnston: when you say "always" ... :)
[18:24] <cjohnston> cjwatson: atleast as long as I've been involved in Ubuntu... :-P
[18:26] <cjwatson> cjohnston: IIRC we started using tracks in spec names for the Natty UDS
[18:29] <cjohnston> http://summit.ubuntu.com/uds-l/meeting/14749/server-lucid-puppet-etckeeper-integration/  it appears as though 'l' followed it
[18:33] <mitya57> why do we have two blueprints for the same thing?
[18:33] <mitya57> https://blueprints.launchpad.net/ubuntu/+spec/foundations-1303-logind-migration
[18:33] <mitya57> https://blueprints.launchpad.net/ubuntu/+spec/client-1303-consolekit-logind-migration
[18:33] <mitya57> jodh: ^^
[18:34] <cjohnston> it would appear as such
[18:37] <jodh> mitya57: the 1st you mention can be deleted - summit failed to show either earlier, but now does :)
[18:37] <cjohnston> jodh: can you please mark the first one superseeded by the second
[18:38] <jodh> cjohnston: done. what happened to delete I wonder? :)
[18:39] <cjohnston> I don't remember there being a deleted
[18:39] <cjohnston> or delete
[18:39] <cjohnston> there may have been, I just never used it
[19:16] <xxiao> on ppc 12.10 when I do: qemu-nbd -c /dev/nbd0 raw.img and always says /build/buildd/qemu-kvm-1.2.0+noroms/nbd.c:nbd_init():L414: Failed setting flags
[19:17] <xxiao> redo it right way: qemu-nbd -c /dev/nbd0 raw.img  now it says /build/buildd/qemu-kvm-1.2.0+noroms/nbd.c:nbd_init():L382: Failed to set NBD socket
[19:17] <xxiao> looks like the first command failed meanwhile left nbd0 unusable for later on commands
[19:17] <xxiao> similar steps on x86 worked well
[19:18] <sarnold> xxiao: (what was the difference between the wrong-way and the right-way?)
[19:19] <xxiao> sarnold: sorry, i meant "right away"
[19:19] <sarnold> xxiao: ah! :D
[19:21] <sarnold> xxiao: does qemu-nbd by chance create sockets without using the SO_REUSEADDR socket option?
[19:21] <xxiao> sarnold: don't really know yet
[19:22] <sarnold> xxiao: in that case you may need to wait a little while for the socket to fall out the CLOSE_WAIT state (iirc)
[19:22] <xxiao> sarnold: i waited 10 mins and it's the same :(
[19:22] <xxiao> ioctl32(qemu-nbd:6441): Unknown cmd fd(13) cmd(2000ab0a){t:'\xffffffab';sz:0} arg(0000002d) on /dev/nbd2
[19:23] <xxiao> that's from dmesg when i do 'qemu-nbd -c /dev/nbd0 raw.img'
[19:23] <sarnold> xxiao: eek
[19:23] <xxiao> some ioctl thing?
[19:24] <xxiao> both have "ii  qemu-utils                         1.2.0+noroms-0ubuntu2.12.10.2                amd64        qemu utilities"
[19:24] <xxiao> ii  qemu-utils                       1.2.0+noroms-0ubuntu2.12.10.2                powerpc      qemu utilities
[19:25] <sarnold> xxiao: the error in dmesg seems to tell me something bigger has gone wrong, outside of my experience by a lot... (when I thought it was just tcp, well, that's one thing.. :)
[19:57] <czajkowski> tvoss: pong
[20:06] <xxiao> anyone that has ppc machine can do me a favor. on 12.10 run qemu-nbd -c /dev/nbd0 raw.img?
[20:49] <xxiao> per the suggestion from qemu guys, qemu-nbd 32bit running on 64bit kernel could have strange issues
[20:57] <xxiao> http://paste.ubuntu.com/5577430/  strace qemu-nbd
[20:58] <slangasek> xxiao: um, you can't strace sudo and expect it to do anything useful
[21:02] <xxiao> http://paste.ubuntu.com/5577449/
[21:02] <xxiao> slangasek: thanks ;)
[21:02] <xxiao> this time runs with root permission
[21:03] <xxiao> http://paste.ubuntu.com/5577454/  x86 working version
[21:03]  * xxiao hopes to see a diff button on paste.ubuntu.com that can diff two pastes
[21:03] <sarnold> xxiao: heh, the error may be happening on the other side of that clone() call
[21:03] <sarnold> xxiao: strace -f   :)
[21:07] <xxiao> http://paste.ubuntu.com/5577466/  strace -f
[21:09] <sarnold> aha now it got your error, line 676
[21:09] <xxiao> sarnold: right
[21:09] <xxiao> ioctl 13 does not know the magic number 0x2000ab0a
[21:09] <xxiao> seems like hand-coded magic #
[21:10] <xxiao> that kernel has no idea?
[21:11] <sarnold> I wonder, 0x20 is space, 0x00 is ascii NUL, 0xab is (nothing, really), 0x0a is a newline ...
[21:11] <xxiao> kernel/include/linux/nbd.h
[21:11] <xxiao> did not see that magic # of course
[21:12] <xxiao> #define NBD_SET_BLKSIZE_IO( 0xab, 1 )
[21:12] <sarnold> hmm :)
[21:12] <xxiao> #define NBD_SET_TIMEOUT _IO( 0xab, 9 )
[21:12] <xxiao> so they just  bypass the define and directly supply the number to kernel
[21:13] <xxiao> and kernel did not recognize
[21:13] <xxiao> actually the kernel definition stops at 0xab, 9, so next one is ab0a then
[21:14] <sarnold> maybe it is in newer kernel sources?
[21:14] <xxiao> sarnold: checking that...
[21:14] <xxiao> it is
[21:14] <xxiao> let me check when it's in
[21:14] <sarnold> #define NBD_SET_FLAGS   _IO( 0xab, 10)
[21:15] <xxiao> right, they moved it to uapi/linux/nbd.h
[21:15] <xxiao> now i have to find out the ioctl call and back-port it...
[21:18] <xxiao> yes it's in nbd.c, the driver
[21:30] <zyga> barry: thanks for working on command-not-found
[21:38] <lool> slangasek: what kind of tests do we run before updating /current in the livefs builds?
[21:38] <barry> zyga: np
[21:41] <kees> hallyn: around?
[21:41] <stgraber> lool: none, /current simply means the build succeeded
[21:41] <slangasek> lool: none; in the case of livefs builds, if the build succeeds it's "current"
[21:41] <lool> ok
[21:43] <toordog> I created my gpg key 2h ago and ran this command : gpg --keyserver keyserver.ubuntu.com --send-keys  I still cannot import my gpg key on launchpad
[21:43] <toordog> Launchpad could not import your OpenPGP key
[21:44] <zyga> toordog: you need to pass the ID of the key to export
[21:44] <zyga> toordog: see man gpg
[21:44] <zyga> toordog: search for '--send-keys'
[21:45] <zyga> toordog: you may want to use seahorse, a graphical wrapper around gpg instead
[21:45] <toordog> i use console only
[21:45] <zyga> toordog: then send the key you created and launchpad will be able to import it soon thereafter
[21:48] <sarnold> toordog: what key did you send? can you see it on any other other keyservers yet? https://sks-keyservers.net/status/
[21:48] <toordog> ok, i think it's going to work, i didn't put the ID.
[21:51] <hallyn> kees: what's up?
[22:03] <xxiao> strange with a new kernel that built with that ioctl i'mstill getting the same error
[22:04] <xxiao> #define NBD_SET_FLAGS   _IO( 0xab, 10)
[22:04] <xxiao> #define NBD_SET_FLAGS   _IO( 0xab, 10)
[22:04] <xxiao>  #define NBD_SET_FLAGS   _IO( 0xab, 10)
[22:05] <xxiao> in ubuntu's qemu-utils, qemu-kvm-1.2.0+noroms/nbd.c
[22:05] <xxiao> if (ioctl(fd, NBD_SET_FLAGS, flags) < 0
[22:05] <xxiao> failed, other similar code are shown like
[22:06] <xxiao> ioctl(13, NBD_SET_SOCK, 0xa) = 0
[22:06] <xxiao> why the failing line shown as ioctl(13, 0x2000ab0a, 0x2d) = -1 EINVAL (Invalid argument)
[22:07] <xxiao> not sure when the 0x2000ab0a happened, that does not happen to other args
[22:07] <xxiao> wierd
[22:08] <xxiao> I would expect it's something like ioctl(13, NBD_SET_FLAGS,0xa) = -1 EINVAL (Invalid argument)
[22:08] <xxiao> search ubuntu source code found no such hex either, all are using NBD_SET_FLAGS instead, how could this happen?
[22:10] <xxiao> case NBD_SET_FLAGS: lo->flags = arg;
[22:10] <xxiao> the kernel just sets a flag, that's it, easy to back port
[22:10] <xxiao> case NBD_SET_FLAGS:
[22:11] <xxiao> lo->flags = arg; return 0;
[22:37] <jcastro> slangasek: hey so I'm watching the video you guys had ... how do I explain to normal people what britney is
[22:37] <slangasek> oh man
[22:38] <slangasek> jcastro: I'm sure there are some suitable results if you search for 'britney' on youtube
[22:38] <stgraber> do "normal people" need to know what britney is? :)
[22:38] <jcastro> stgraber: no, they don't need to know, but it helps people understand when you say "oh don't worry britney will handle that for us..." or whatever
[22:39] <slangasek> jcastro: "the software that makes sure packages are built on all architectures and don't cause installability problems before we push them out for the whole world to see"?
[22:39] <barry> jcastro: there's something on youtube about britney???
[22:40] <jcastro> slangasek: good enough.
[22:40] <jcastro> barry: yeah, but they're not very good. :p
[22:40] <jcastro> slangasek: so also the thing that makes sure when I dist-upgrade that half my desktop doesn't get removed?
[22:40] <jcastro> like the old sid days
[22:40] <slangasek> it doesn't guarantee that at all ;)
[22:41] <stgraber> barry: actually: http://www.youtube.com/watch?v=3ctC4VQVh-Q appears to be relevant for the Debian implementation
[22:41] <jcastro> that's another woman's name isn't it.
[22:42] <slangasek> jcastro: nope, there's just nothing that guarantees you won't still hit such a problem... there are plenty of client-side bugs, or confusing local packages, that could still cause that
[23:00] <kees> hallyn: hi! I've been playing with user namespaces, and I was trying to figure out how to actually gain capabilities in them.
[23:00] <kees> hallyn: from what i can tell, it requires a uid mapping be in place, but that requires having had CAP_SETUID at some point.
[23:01] <hallyn> kees: yes, hence setuid-root binary to set up mappings.  did you see eric's shadow patchset?
[23:02] <hallyn> see lp:/~serge-hallyn/ubuntu/raring/shadow/shadow-userns/
[23:02] <hallyn> that is, lp:~serge-hallyn/ubuntu/raring/shadow/shadow-userns/
[23:02] <hallyn> i need to push that set, soon.
[23:02] <kees> ah, hm
[23:03] <hallyn> kees: if you could review that set too that'd be helpful for dure :)
[23:03] <hallyn> sure
[23:03] <kees> hallyn: yeah, reading now...
[23:04] <hallyn> (also see Pkg-shadow-devel@lists.alioth.debian.org m-l)
[23:05] <kees> hallyn: is there any overview anywhere for what the "expected" behavior should look like?
[23:08] <kees> hallyn: hrm, it looks like "newuidmap" is called by... the parent?
[23:10] <hallyn> kees: overviews in several places - man page updates (from eric), s3hh.wordpress.com, lwn.net user_namespace summary (this week)
[23:11] <hallyn> kees: without privilege you should be able to (i *think*) just map uid 0 in the container to your unpriv userid on the host
[23:11] <hallyn> (range of 1)
[23:11] <kees> hallyn: oh?
[23:12] <hallyn> but indeed, ISTR you may end up without privileges - need to straighten that out with eric one day if that's the case
[23:12]  * hallyn goes to reread emails
[23:12] <kees>         /* Require the appropriate privilege CAP_SETUID or CAP_SETGID
[23:12] <kees>          * over the user namespace in order to set the id mapping.
[23:12] <kees>          */
[23:12] <kees>         if (cap_valid(cap_setid) && !ns_capable(ns, cap_setid))
[23:12] <kees>                 goto out;
[23:12] <kees> that's what I see at the top of map_write() handling the uidmap file writes
[23:13] <hallyn> yup, ok, so nm
[23:13] <hallyn> guess i was wrong :)
[23:13] <hallyn> offhand it seems like it would be a safe and useful thing to allow
[23:14] <hallyn> but we're not exactly looking to live on the edge here :)
[23:14] <hallyn> and if shadow will auto-generate subuid ranges for all users, then it becomes moot pretty much
[23:15] <kees> I don't see how to call newuidmap, though. I guess from the parent.
[23:16] <hallyn> kees: you're talkinga bout the program?
[23:17] <hallyn> sorry, i've not really *used* eric's tools much...  but yes, from the parent, or anyone in the parent ns
[23:17] <kees> yeah, something needs to call newuidmap. right now, a process will perform the clone(CLONE_NEWUSER); wait().
[23:17] <hallyn> kees: you can look at lxc for an example :0  it does it by hand...
[23:17] <kees> I love that user ns isn't available if you build NFS. ;)
[23:19] <slangasek> eew what?
[23:19] <hallyn> kees: which source are you using
[23:19] <kees> 3.8
[23:19] <kees> config USER_NS
[23:19] <kees>         depends on UIDGID_CONVERTED
[23:19] <kees> config UIDGID_CONVERTED
[23:20] <kees>         default y
[23:20] <kees>         depends on NFS_FS = n
[23:20] <hallyn> kees: yeah 3.8 is not complete
[23:20] <kees> yup
[23:20] <hallyn> there's a kernel at https://launchpad.net/~ubuntu-lxc/+archive/kernel you can use,
[23:20] <hallyn> or grab eric's sources from kernel.org
[23:20] <kees> I'm cool; I'm just playing with this under kvm.
[23:20] <hallyn> it should all go into 3.9 though
[23:20] <hallyn> well you're missing other things,
[23:20] <kees> took me a while to figure out why clone() kept failing, though :)
[23:20] <hallyn> but ok :)
[23:21] <kees> hallyn: I take it the default is to not create any subxid mappings?
[23:21] <hallyn> kees: hm, i'm not sure.
[23:22] <hallyn> it *really* is meant to be safe to allocate them by default
[23:22] <kees> ... :)
[23:22] <hallyn> but yeah, i suspect
[23:23] <kees> is there any ETA for the userns patch series landing in raring?
[23:23] <hallyn> kees: it'll go in with 3.9.  in other words, not into raring
[23:24] <kees> okay
[23:24] <hallyn> that's why i finally put up the ubuntu-lxc/kernel ppa :(
[23:24] <hallyn> (before that i just pointed to my personal ppa)
[23:26] <kees> this build behavior means that 3.8 actually regresses for anyone that _was_ using CLONE_NEWUSER ...
[23:29] <stgraber> was anyone actually using CLONE_NEWUSER?
[23:29] <kees> no idea! :) I doubt it.
[23:30] <hallyn> kees: we went to great lengths to warn people, over a period of years
[23:30] <kees> hallyn: hehe
[23:31] <hallyn> some ppl may ahve used it for separating peruser quotas at one point, but i think 2008 or 2009 was when i warned ppl first
[23:33] <hallyn> kees: let me know how it goes!  ttyl
[23:33] <kees> hallyn: thank, I'll Cc you on this (weak) vulnerability report...
[23:34] <hallyn> d'oh.  thanks :)
[23:34] <stgraber> hallyn: what did you expect, you're talking with kees ;)
[23:37] <kees> hehe
[23:38] <kees> stgraber: so, where does lxc do the uid mapping? I don't see anything in 0.9.0~alpha3-0ubuntu2
[23:40] <hallyn> kees: see src/lxc/conf.c:add_id_mapping
[23:45] <kees> hallyn: ha ha, grep failed ... %cid_map ;)
[23:45] <hallyn> yeah :)  i almost feel bad about that
[23:45] <hallyn> :)
[23:48] <kees> ... %s", idtype == ID_TYPE_UID ? "uid_map" : "gid_map") ...  ?  :)
[23:48] <cjwatson> jcastro: This is why I always call it "proposed-migration" (which is actually the directory under which our installation runs), not "britney"
[23:49] <cjwatson> jcastro: It does make it much less *likely* that half your desktop will get removed on upgrade - and especially it insulates us against *most* multiarch library-sync problems, which are incredibly confusing for normal people to resolve
[23:49] <cjwatson> well, s/especially/also/
[23:57] <kees> hallyn: actually, you need to check the return code of fclose() too.