[00:36] <ev> slight ordering issue with this in the parallel logic.  It sets up the new debconffilter on the same fd as the previous, then disconnects the fd.  On it now.
[00:38] <cjwatson> discussion of USB issue in #ubuntu-devel
[00:38] <cjwatson> I have a plausible-looking hypothesis and a likely fix if it holds
[00:38] <cjwatson> not confirmed yet
[00:38] <cjwatson> hypothesis does explain why it seems random
[01:02] <ev> success!
[01:02] <ev> cjwatson: http://paste.ubuntu.com/487013/ - are the 'callback =' bits I added too ugly for your taste?
[01:04] <cjwatson> not really, but doesn't that still leave us calling process_line once incorrectly?
[01:04] <cjwatson> maybe the check should be at the top of the three helper methods, and just return?
[01:04] <cjwatson> oh, wait, no, I see what you mean
[01:05] <cjwatson> no, ignore me, that seems correct, go ahead
[01:05] <ev> cool, thanks for the ack
[01:08] <cjwatson> so just waiting for manjo to verify my USB patch
[01:11] <ev> okay
[01:14] <cjwatson> um, so bug 628317 looks like a regression in the recently-introduced hostname code
[01:14] <ubot2> Launchpad bug 628317 in ubiquity (Ubuntu) "Mavrick Kubuntu ubiquity installer crash if system name is blank (affects: 1) (heat: 6)" [Undecided,New] https://launchpad.net/bugs/628317
[01:15] <cjwatson> in lucid, ubiquity/components/ubi-usersetup.py had a check_hostname function that iterated through the return value of validation.check_hostname and constructed a newline-separated string with all the errors
[01:15] <ev> ugh, ffs
[01:15] <ev> on it now
[01:15] <cjwatson> it was kind of nasty but should we just reintroduce that?
[01:17] <ev> maybe - but let me try to fix this while we're waiting on manjo
[01:19] <CIA-71> ubiquity: evand * r4265 trunk/ (debian/changelog ubiquity/frontend/kde_ui.py):
[01:19] <CIA-71> ubiquity: Disconnect and destroy QSocketNotifiers when the dbfilter finishes.
[01:19] <CIA-71> ubiquity: Thanks Colin Watson (LP: #625586).
[01:27] <ev> fix in hand, testing
[01:27] <CIA-71> ubiquity: evand * r4266 trunk/debian/changelog: Fix changelog
[01:35] <CIA-71> ubiquity: cjwatson * r4267 ubiquity/ (debian/changelog scripts/plugininstall.py):
[01:35] <CIA-71> ubiquity: Set Debug::identcdrom=true for apt-cdrom if installing from a
[01:35] <CIA-71> ubiquity: non-ISO9660 filesystem, since otherwise apt-cdrom will include the free
[01:35] <CIA-71> ubiquity: blocks count of /cdrom in its database which is unreliable on read-write
[01:35] <CIA-71> ubiquity: filesystems, especially USB images with persistence enabled. Thanks to
[01:35] <CIA-71> ubiquity: Manoj Iyer for patient testing (LP: #627672).
[01:42] <CIA-71> ubiquity: evand * r4268 trunk/ (debian/changelog ubiquity/plugins/ubi-usersetup.py): add the check_hostname function back to ubi-usersetup (LP: #628317).
[01:42] <ev> cjwatson: ready for an upload?
[01:45] <cjwatson> yep
[01:46] <ev> done
[01:46]  * cjwatson pokes around for a reviewer
[01:47] <ev> and this is why release sprints are a good thing.  Reviewers are trapped in the same room.
[01:51] <Riddell> cjwatson: review needed?
[01:51] <cjwatson> aye, ubiquity 2.3.14 in the unapproved queue
[01:52] <cjwatson> no queuediff yet though
[01:53] <Riddell> what's queuediff? (I feel I should know this)
[01:53] <cjwatson> lp:ubuntu-archive-tools
[01:54] <cjwatson> scrapes the queue page for the link to the diff from the last version
[01:54]  * ev notices lp:~superm1/ubiquity/maverick-post-beta-fixes - superm1, you rock!
[01:55] <cjwatson> (alternatively, queuediff is a product of Martin Pitt's clinical aversion to ever typing the same thing twice)
[01:57] <Riddell> I see plugininstall.py is altered in the ubiquity upload but I can't match that to something in the changelog
[01:58] <cjwatson> "Set Debug::identcdrom=true ..."
[01:58] <Riddell> right you are, accepting
[01:58] <CIA-71> ubiquity: evand * r4269 trunk/debian/changelog: releasing version 2.3.14
[01:59] <cjwatson> thank you
[01:59] <ev> thanks Riddell
[01:59] <Riddell> me?  you're the ones who fixed it :)
[02:00] <cjwatson> ... we hope
[02:00] <cjwatson> :-)
[02:00] <Riddell> yes hopefully. I guess it won't be published until after 03:00 BST so no images until sometime after that
[02:01] <cjwatson> yeah, I'm going to set an alarm and hope it wakes me
[02:01] <cjwatson> and doesn't wake my wife
[02:01] <cjwatson> some things don't make you popular
[02:01] <Riddell> set for when?
[02:02] <cjwatson> 3:30 - but actually, I just remembered that cdimage has a magic thing for this
[02:02] <cjwatson> wait-for-package
[02:02] <Riddell> that sounds clever
[02:03] <ev> cute
[02:04] <ev> should we let -testing know to expect new CDs?
[02:05] <cjwatson> wait-for-package ubiquity_2.3.14 && { echo kubuntu; buildlive kubuntu && for-project kubuntu cron.daily-live; echo ubuntu-netbook; buildlive ubuntu-netbook && for-project ubuntu-netbook cron.daily-live; echo kubuntu-mobile; buildlive kubuntu-mobile && for-project kubuntu-mobile cron.daily-live; echo kubuntu dvd; buildlive kubuntu-dvd && for-project kubuntu cron.dvd; }
[02:05] <cjwatson> sound like a plausible list?
[02:05] <cjwatson> yes, as nothing will post them to the tracker unless I happen to wake up or somebody else does it
[02:06] <ev> looks reasonable
[02:06] <cjwatson> I'll tell -testing
[02:06]  * Riddell sets alarm for 04:00
[02:06] <cjwatson> ... or you can :)
[02:06] <cjwatson> since I see you just joined
[02:07] <cjwatson> Riddell: yeah, that's probably about right, might be ten minutes after that
[02:07] <cjwatson> or so
[02:07] <Riddell> oh I expect I'll press snooze on the alarm at least once :)
[02:08] <cjwatson> heh
[02:08] <cjwatson> right, I'll try to get some sleep then, SMS me if it all goes horribly wrong?
[02:08] <Riddell> guid nicht
[02:10] <ev> goodnight guys.  Thanks again for all of your help.
[02:11] <cjwatson> agus thú
[02:21] <ev> :)
[05:36] <rgreening> hey ev
[06:00] <jjk9> looking for ldapsam.so to use with samba 3.4.7 on 10.04 LTS server. anyone know where it is?
[06:02] <jjk9> maybe I am in wrong room for that q...
[08:04] <ev> rgreening_: hiya
[08:04] <ev> jjk9: indeed, you are in the wrong room.  Please ask in #ubuntu.
[08:04] <ev> Or use http://packages.ubuntu.com
[08:32] <jjk9> tks I already looked in http://packages.ubuntu.com but found no reference. will pursue further over at #ubuntu tks bi
[08:40] <ev> cjwatson: it's still broken
[08:40] <ev> I'm getting KeyErrors at line 1043
[08:40] <ev> investigating now
[09:03] <cjwatson> on which, Kubuntu or netbook?
[09:07] <ev> kubuntu for me
[09:08] <ev> it looks like if debconffilter_done spawns a new dbfilter that uses the same fd we were about to delete, then we don't delete it, but we also don't clear out the QSocketNotifier
[09:15] <Riddell> I've done three installs now without problems
[09:24] <ev> probably not the most elegant solution - http://paste.ubuntu.com/487134/
[09:25] <ev> just testing the hell out of that
[09:34] <cjwatson> ev: that looks ok, though you forgot a 'self.'
[09:36] <ev> err yeah, I have it correct in the copy I'm testing with
[09:36] <ev> thanks for catching that
[10:01] <CIA-71> ubiquity: evand * r4270 trunk/ (debian/changelog ubiquity/frontend/kde_ui.py):
[10:01] <CIA-71> ubiquity: Clean stray QSocketNotifiers that were missed because their fd was
[10:01] <CIA-71> ubiquity: reused before they could normally be deleted.
[10:02] <ev> trying to see if I can run into this by more than just manual partitioning
[10:05] <cjwatson> rough ETA?
[10:06] <ev> another 10-15 minutes or so
[10:06] <ev> I've already confirmed that I can reliably reproduce the bug on manual partitioning and that the patch definitely fixes it for me
[10:07] <Riddell> ev: what actually is the symptom?
[10:07] <ev> Riddell: if I select manual partitioning, complete the install to the slideshow window, it will eventually throw up crash dialogs until it runs out of memory.
[10:09] <Riddell> mm, painful
[10:10] <Riddell> ev: how common do you think it is?  I haven't seen it yet
[10:11] <ev> hard to say, it's a bit of a race
[10:11] <cjwatson> might depend on system speed?
[10:13] <ev> perhaps
[10:15] <ev> it's a bit odd that it only seems to be happening when I go through advanced partitioning.  I wonder what aspect of that changes the order of events.
[10:21] <cjwatson> ev: 15 minutes up :) I'm hoping to get the binaries in for the next publisher run if possible?
[10:21] <ev> okay, releasing
[10:22] <cjwatson> thanks.  it's a 35-minute build
[10:24] <ev> uploaded
[10:24] <Riddell> is the publisher not on manual?
[10:25] <cjwatson> no
[10:25] <cjwatson> I mean, it can be
[10:25] <cjwatson> but that's yet more manual work
[10:25] <cjwatson> I will if I have to
[10:40] <Riddell> ev: ooh I think I got your crash on the DVD
[10:40] <ev> lots of crash dialogs?  Can you grab the /var/log/installer/debug log off there and pastebin it?
[10:41] <Riddell> sorry already rebooted
[10:58] <Riddell> ev: http://muse.19inch.net/~jr/tmp/debug
[10:58] <ev> yup, that's the bug
[10:59] <ev> what steps did you go through to produce this?
[11:11] <Riddell> ev: just did an install from the DVD
[11:11] <Riddell> twice, one with live session once ubiquity only
[11:11] <Riddell> manual partitioning
[11:11] <Riddell> no oem
[11:16] <ev> okay
[12:37] <jhesketh> Hi
[12:38] <jhesketh> I'm trying to build a copy of ubiquity but I'm stuck on this error:
[12:38] <jhesketh> make: *** source/console-setup/Keyboard: No such file or directory.  Stop.
[12:38] <jhesketh> does anybody know what I need to do (I suspect I need a copy of debian installer in the d-i folder or something?)
[12:38] <jhesketh> (I've branched ~ubuntu-installer/ubiquity/lucid-proposed/)
[12:49] <cjwatson> debian/rules update-local
[12:50] <cjwatson> or grab the corresponding source package version from archive.u.c
[12:51] <jhesketh> thanks cjwatson. I'll give that a try :-)
[13:16] <jhesketh> cool, that seems to work - now to make some changes
[13:17] <jhesketh> I'm building a liveCD whereby I don't need the install option, but I do however want to keep the first screen so that the language of the system may be changed before continuing into the desktop
[13:17] <jhesketh> any suggestions on the best way to do that (my thoughts was to just rip out the install option and reword the "try ubuntu" to "continue" or something)
[13:20] <cjwatson> you should mostly just be able to edit isolinux.cfg and possibly gfxboot.cfg
[13:20] <cjwatson> if you want to disable the splash-screen thing where you get just an icon at the start and if it times out then it boots with the maybe-ubiquity option, delete the hidden-timeout line from gfxboot.cfg
[13:22] <jhesketh> cjwatson: yeah I thought of that, but the interface for selecting a language is nicer in ubiquity and that would expose complications of kernel options to a very non-technical target group
[13:43] <jhesketh> cjwatson: sorry to bother you, but I don't suppose you know why bin/ubiquity-greeter-gtk sets up the interface rather than using gui/gtk/stepLanguage.ui ? (or am I reading it wrong)
[13:45] <cjwatson> that was an experiment which I think we don't actually use any more
[13:45] <cjwatson> we should probably remove it from the tree
[13:45] <jhesketh> ok
[13:46] <jhesketh> but the button label is replaced at some point right? Because I don't think it just says "Try Ubuntu" when booting (from memory, I'd have to double check)
[13:46] <cjwatson> er, I'm out of date on this stuff, you want ev really
[13:47] <jhesketh> no worries. Thanks for your help :-)
[13:47] <ev> it's set based on the release name in /cdrom/.disk/info
[13:47] <ev> "Try ${RELEASE}"
[13:48] <jhesketh> ev: cheers, I don't suppose you know where it sets that?
[13:49] <ev> ubiquity/plugins/ubi-language.py:on_language_selection_changed()
[13:49] <jhesketh> hmm, I'm working on the lucid branch, is that meant to be  ubiquity/components/ubi-language.py ?
[13:50] <ev> yes
[13:50] <jhesketh> cheers, found it :-)
[13:50] <ev> good deal
[14:54] <ScottK> ev: If I have a system with an existing *buntu install and I pick "Guided - use entire partition", what is the intended result?  Two *buntu installs on separate partitions or the new install installs over the old one and wipes it out?
[15:07] <ScottK> Anyone know where I can find the design document for the new installer so I can check myself?
[15:24] <ScottK> Riddell: Any suggestions on where I can find the intended behavior for "Guided"?
[15:29] <ScottK> OK.  I'll just assume wiping out the existing install isn't intended.
[15:34] <ev> ScottK: think of it as "replace the copy of ubuntu on this exsiting partition with a new one"
[15:34] <ScottK> OK.
[15:34] <ScottK> Then autoresize is gone?
[15:35] <ev> resize is only available if a partition can be resized
[15:35] <ev> but hm, those two options should be shown together
[15:35] <ev> yeah, that may be broken
[15:37] <ScottK> Bug #628864 filed.
[15:37] <ubot2> Launchpad bug 628864 in ubiquity (Ubuntu) ""Guided" install does no guiding and wipes out existing install without warning (affects: 1) (heat: 6)" [Critical,New] https://launchpad.net/bugs/628864
[15:37] <ScottK> I don't see an autoresize option in the installer.
[15:38] <ScottK> Riddell: I see that as a release blocker.
[15:38] <ScottK> ev: ^^^
[15:39] <ev> as in beta release blocker?
[15:39] <ScottK> Yes.  I think wiping out existing installs and user data without warning qualifies.
[15:40] <ev> well, I'll leave it to you guys to determine what blocks your release, but the button does say, "Install Now"
[15:40] <ScottK> Right, but it also says "Guided"
[15:40] <ScottK> That leads me to think there will be some guidance.
[15:40] <Riddell> the option is the guidance
[15:41] <Riddell> that's always been the case
[15:41] <ScottK> I don't think that makes any sense.
[15:41] <ev> the KDE frontend should replace the text on those using a similar approach to the one taken by the GTK frontend for this
[15:41] <cjwatson> we did rename some of the options away from Guided a while back (the name "Guided" is a hangover from d-i)
[15:42] <cjwatson> ubiquity in general should not be describing options as Guided
[15:42] <ScottK> I'm at the Disk Setup page and the options I have are Guided, Erase and use entire disk, and Specify partitions.
[15:42] <cjwatson> (I have no opinion on whether or not this is a Kubuntu release blocker, aside from a residual "argh")
[15:42] <ScottK> If it's not new, I don't think it's a release blocker.
[15:43] <ScottK> ev: Should I also have an autoresize option here?
[15:43] <cjwatson> the "replace OS on this partition with a new one" behaviour is new since lucid
[15:43] <cjwatson> the presence of autoresize depends on your partition layout, but bug 628815 indicates that there may be a problem specifically with the KDE frontend here
[15:43] <ScottK> Right.  That's what's bit me.
[15:43] <ubot2> Launchpad bug 628815 in ubiquity (Ubuntu) "Cannot resize from kubuntu partitioner (affects: 1) (heat: 6)" [Undecided,New] https://launchpad.net/bugs/628815
[15:43] <ScottK> OK.
[15:43] <cjwatson> as in, the underlying logs there suggest to me that autoresize should be made available, but the screenshot doesn't show it
[15:44] <cjwatson> again, no opinion on whether that's blocker or release-note
[15:44] <ScottK> Sounds like the same thing I have then.
[15:44] <cjwatson> (largely because too tired to have opinions)
[15:44] <ScottK> If it was an alpha, I'd be totally fine with a release note.
[15:45] <ScottK> Riddell: Since you're still here it's your call, but this will cause users to lose data IMO.
[15:45] <ScottK> (despite all the warnings)
[15:46] <ev> so the "replace this partition with a new Ubuntu install" (some_partition) option should always appear alongside the resize option
[15:46] <ev> the KDE frontend appears to be showing the former and not the latter
[15:46] <Riddell> I don't see missing autoresize as a blocker, it's a pain but it's not a killer. and I don't see having an old label on the replace as a blocker either
[15:47] <ev> the option was created this release cycle for the GTK frontend's "[use entire partition] [use entire disk]" buttons when in resize mode.
[15:48] <ScottK> Riddell: I don't see missing autoresize as a potential blocker.
[15:53] <ScottK> Riddell: I can't make the manual partioner work either.  "Error!!! Partiion(s) 5 on /dev/sda have written, but we have been unable to inform the kernel of the change, probably because it/they are in use.  As a result, the old partition(s) will remain in use.  You should reboot now before making further changes.
[15:53] <Riddell> um, no idea what that's about
[15:54] <cjwatson> desktop automounter perhaps?  is anything mounted from /dev/sda?
[15:55] <ScottK> It claims to be installing anyway, so let's see what's up.
[15:55] <ev> I've got to step away for a call, back in an hour or so
[15:55] <cjwatson> if anything is mounted, then I refer to http://www.chiark.greenend.org.uk/ucgi/~cjwatson/blosxom/ubuntu/2008-04-12-desktop-automount-pain.html
[15:56] <ScottK> checking
[15:57] <cjwatson> could also be a race with a udev rule or similar
[15:57] <cjwatson> though I think nowadays libparted tries a couple of times to try to avoid that
[15:57] <ScottK> /dev/sda1 is mounted, but with the new partition type (made it ext3 so I could tell)
[15:58] <ScottK> So it may have been a race and was retried but succeeded or something.
[15:58] <cjwatson> bug but not a blocker then ...
[15:58] <ScottK> Yes
[15:59] <ScottK> The idea that "Guided" is going to replace the operating system is just still got me boggled.
[16:00] <cjwatson> I do find that rather odd
[16:01] <ScottK> The word guided implies to me that there will be help coming.
[16:01] <Riddell> that's probably why seele said it should be changed, but we have had several releases using that text
[16:02] <ScottK> Riddell: But it changed what it does this cycle.
[16:02] <ev> it does the exact same thing
[16:02] <ev> the option you've selected is a different option than the resize one
[16:02] <ScottK> Hmmm.  <cjwatson> the "replace OS on this partition with a new one" behaviour is new since lucid
[16:02] <ScottK> ^^^ what I'm going on
[16:03] <ev> I think Colin is confused, this is new since Maverick
[16:03] <ev> this is "some_partition" in partman-auto
[16:03] <ev> really got to run now
[16:03] <ev> back later
[16:03] <cjwatson> by "new since lucid", I mean "not in lucid, but in maverick"
[16:03] <cjwatson> exclusive since
[16:03] <ScottK> Riddell: So it's not "the same".
[16:04] <ScottK> cjwatson: Any logs you want for the partioning alert?
[16:04] <cjwatson> 'ubuntu-bug ubiquity' should be sufficient
[16:05] <ScottK> OK.
[16:05]  * ScottK goes to find a wire with internet in it.
[16:05] <Riddell> you'll need to install konqueror for that to work
[16:14] <ScottK> Riddell: Nope.  Works with reqkonq.
[16:17] <ScottK> Bug #628894 (so I can find it later)
[16:17] <ubot2> Launchpad bug 628894 in ubiquity (Ubuntu) "Manual partition problem alert when there isn't an actual problem (affects: 1) (heat: 6)" [Undecided,New] https://launchpad.net/bugs/628894
[16:49] <Riddell> ScottK: hmm, it needed kfmclient when I tried it
[16:49] <ScottK> Odd.  Worked though.
[17:55] <ev> cjwatson: ah, sorry.  My bad.
[21:27] <dmarkey> cjwatson: would it be possible to release a (i386) server install CD that's PAE?
[21:30] <soren> dmarkey: The server install CD already installs a PAE kernel by default.
[21:30] <dmarkey> is the kernel booted off the CD PAE?
[21:30] <soren> dmarkey: Why is it important that the installer itself runs a PAE kernel?
[21:30] <soren> dmarkey: I don't believe it is, no.
[21:31] <dmarkey> to run paravirtualised on Xen/XenServer, Xen will only run i386 kernels that are PAE
[21:31] <cjwatson> I'd rather not do that with the server install CD, no - another CD there is pretty much out of the question
[21:31] <soren> I didn't think you could boot ISO's in paravirt Xen anyway.
[21:31] <cjwatson> that's many extra hours of human validation time for every milestone release
[21:32] <dmarkey> I see..
[21:32] <cjwatson> are we not getting to the point where you can just do amd64?
[21:32] <cjwatson> netboot and a local mirror should work reasonably well as well, and I'd be happy to help improve that
[21:33] <dmarkey> yea, netboot install is getting alot of use
[21:33] <dmarkey> but cdrom install was next on my checklist
[21:34] <soren> How does paravirt Xen and ISO booting work anyway?
[21:34] <cjwatson> there's a cdrom/xen d-i flavour, so it could be substituted by script
[21:34] <cjwatson> I think
[21:34] <dmarkey> bootloader basically extracts the kernel/initrd out of the ISO
[21:34] <soren> dmarkey: Ah, I see.
[21:36] <dmarkey> cjwatson: any other suggestions around i386 support?
[21:41] <soren> dmarkey: Well, it's not exactly elegant, but you could extract the pae kernel and pass it to Xen, but I understand that just solves the problem for you, not generally for everyone else.
[21:41] <soren> dmarkey: The ISO contains the pae kernel deb, so it just needs unpacking and a bit of Xen fiddling.
[21:42] <dmarkey> do we have a PAE initrd in there somewhere?
[21:43] <dmarkey> we neet xen-blkfront
[21:43] <soren> Ah, good point.
[21:43] <dmarkey> need*
[21:43] <cjwatson> there's a PAE initrd in the archive, yes
[21:44] <soren> What, really?
[21:44] <cjwatson> http://archive.ubuntu.com/ubuntu/dists/maverick/main/installer-i386/current/images/cdrom/xen/
[21:44] <soren> Oh.
[21:44] <soren> Imagine that.
[21:44] <cjwatson> blat that in, make sure the CD has generic-pae udebs in it
[21:45] <dmarkey> is this in lucid also?
[21:45] <cjwatson> it's not too hard to do, I just can't afford to try to release it as well
[21:45] <cjwatson> just maverick sorry
[21:45] <dmarkey> i see
[21:45] <cjwatson> lucid only has netboot/xen
[21:46] <cjwatson> it was the product of a merge from Debian
[21:46] <dmarkey> yup, so when did this cdrom/xen come into being? from EC2 or something?
[21:46] <dmarkey> oh ok
[21:47] <dmarkey> so, in 10.10 we should be good to go?
[21:49] <cjwatson> in that you can put the desired result together
[21:50] <dmarkey> one possible issue, will the installer check /dev/xvdd for a repo? it will be a normal block device
[21:51] <cjwatson> that I'm not sure, ask me when it isn't beta release day and I've had some sleep :-)
[21:52] <dmarkey> i'll give it a whirl
[21:52] <dmarkey> thanks!