[02:21] <ryoohki> when kick starting hardy server 8.04 amd64 "Loading kickseed-common failed for unknown reasons"
[03:12] <ryoohki> i think i found the problem... it's seems like it's trying to load additional installer components?!  that's not in my kickstart and it wasn't doing this earlier.  this is very frustrating - that ubuntu insists on "phoning home" during an install.  can this be turned off?  i have the server dvd mounted and shared via apache but it insists on not using it even though i specify it in the kickstart file.
[03:14] <ryoohki> has anyone been able to kickstart ubuntu with the lvm?  the kickstart "preseed" directive doesn't seem to work.
[03:38] <Zelut> ryoohki: lvm is still tricky with preseed and not supported with kickseed/kickstart.
[10:25] <cjwatson> ryoohki: use the current hardy-proposed image, and set apt-setup/proposed=true
[10:26] <cjwatson> ryoohki: talk of "phoning home" is rather ridiculous I'm afraid - the installer has always fetched components of itself at run-time
[10:27] <cjwatson> if it didn't, you'd have a rather useless installer ;-)
[11:06] <CIA-38> debian-installer: cjwatson * r933 ubuntu/ (build/config/sparc.cfg debian/changelog): Disable VERSIONED_SYSTEM_MAP again for sparc.
[11:08] <CIA-38> debian-installer: cjwatson * r934 ubuntu/ (3 files in 2 dirs): Fix some leftover calls to bootscreen-subst (renamed to bootvars-subst).
[11:12] <CIA-38> debian-installer: cjwatson * r935 ubuntu/debian/changelog: releasing version 20080522ubuntu3
[11:41] <CIA-38> debian-installer: cjwatson * r936 ubuntu/ (build/boot/x86/boot.txt.live debian/changelog): Remove build/boot/x86/boot.txt.live, no longer used.
[12:09] <xivulon> evand good news to cking
[12:10] <xivulon> see https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/204133/comments/35
[12:10] <xivulon> do you think we can push that into hardy.proposed?
[12:11] <xivulon> that sounds as a big improvement over the sysctl hacks and could address data loss situations
[12:13] <xivulon> it will require removing lupin-sysctl from lupin (nobody is going to cry for that) and add a root mounting argument for ntfs (can be done via wubi as a boot parameter or via initramfs-tools)
[12:14] <xivulon> + of course the ntfs3g patch
[12:44] <xivulon> evand see also the discussion on #ubuntu-kernel
[13:00] <xivulon> evand I have asked #ubuntu-testing to test the ntfs-3g patch as that should be the most invasive bit
[14:05] <evand> xivulon: We can try.  If you can do some initial testing, I'd appreciate it.  If that goes well, I'll make an SRU request.
[14:09] <xivulon> evand I cannot test it until tonight
[14:09] <xivulon> but i have already asked in ubuntu-testing
[14:10] <xivulon> as mentioned the only big chunk is ntfs-3g, which should be ok since the changes have effect only if you set the option
[14:10] <xivulon> for the rest in lupin it is a matter of removing lupin-sysctl and in wubi add the mount options to menu.lst
[14:10] <evand> indeed, though performance is also a cocnern.
[14:10] <evand> concern*
[14:11] <xivulon> as mentioned by coling, things should not be worse than current situation
[14:11] <xivulon> with the advantage that now you can choose whether to sync the nested fs or not for extra security
[14:11] <cjwatson> note that further kernel uploads to hardy-proposed will delay 8.04.1
[14:11] <xivulon> while retaining the sync on the host fs (pretty much a hard requirement)
[14:11] <xivulon> cjwatson, my understanding is that it only affects ntfs-3g
[14:11] <evand> ah, 8.04.2 then
[14:12] <xivulon> which is userspace
[14:12] <evand> Is there a better schedule? I looked on HardyReleaseSchedule and it was a bit sparse
[14:12] <evand> re cking> he said it *may* have little impact, but should be tested to be sure
[14:13] <xivulon> evand the sync on the loop, is something we can tune from wubi quite easily
[14:13] <xivulon> as a menu.lst kernel parameter
[14:13] <xivulon> the sync on the host (ntfs-3g) is a hard requirement, and the patch is a clear improvement over the syctl hack
[14:14] <xivulon> which in fact did not guarantee immediate flushes -> read data-loss
[14:14] <evand> right, but we should still know exactly what we're dealing with in terms of performance
[14:15] <xivulon> sure
[14:15] <xivulon> but there are 2 separate issues here
[14:15] <xivulon> lupin-sysctl vs ntfs-3g -o syncio
[14:15] <xivulon> loop -o sync vs loop -o nosync
[14:15] <evand> indeed
[14:16] <xivulon> the first one IMHO is a no brainer
[14:16] <evand> I'd like to see it tested before I make the same assumption.
[14:16] <xivulon> sure, see my posts on ubuntu-testing as well
[14:17] <xivulon> in any case we can try to keep things moving and not use the -o syncio option in the worst case
[14:17] <evand> ok
[14:22] <xivulon> was about to test #230716 but noticed that there are no virtualbox modules for 16+ kernel... ...and of course had removed 16...
[14:22] <xivulon> will resume that tonight too
[15:23] <xivulon> evand could you produce a build of ntfs-3g in ppa, that would simplify life for ubuntu-testing
[15:24] <evand> yes
[15:24] <xivulon> great thx
[17:13] <tormod> we have an issue with the live CD eagerly mounting swap from raid raw devices, can someone have a look at bug #136804?
[17:13]  * tormod asks in right channel this time
[17:13] <tormod> I don't know if it's limited to fakeraid
[17:15] <cjwatson> I expect it will be
[17:15] <cjwatson> can't really fix that until intrepid when we plan to do a proper job of dm-raid; at the moment it's basically a bit hosed
[17:16] <cjwatson> I'm definitely not interested in attempting to wedge dmraid into the desktop CD for 8.04.1
[17:16] <cjwatson> even targeted fixes are very difficult for 8.04.1 at this point; it's frozen
[17:16] <cjwatson> but if somebody wants to figure out the proper patch for casper/scripts/casper-bottom/13swap, be my guest
[17:17] <cjwatson> oh, you did :)
[17:17] <cjwatson> please post patches, not new copies of scripts
[17:18] <evand> http://bazaar.launchpad.net/~tormodvolden/casper/no-raid-swap/revision/511
[17:19] <cjwatson> looks fine to merge except that the version should be 1.133
[17:19] <evand> indeed
[17:20] <evand> shall I take care of that, or do you already have it?
[17:20] <cjwatson> please go ahead
[17:20] <evand> ok, will do
[17:20] <cjwatson> if tested, that might be ok for hardy, but at this point any further uploads are likely to delay 8.04.1 :(
[17:20] <cjwatson> so maybe for 8.04.2
[17:34] <tormod> thanks!
[17:38] <CIA-38> debian-installer: cjwatson * r937 ubuntu/ (build/config/sparc/netboot.cfg debian/changelog): Another System.map fix for sparc.
[17:55] <CIA-38> debian-installer: cjwatson * r938 ubuntu/ (build/pkg-lists/cdrom/common debian/changelog):
[17:55] <CIA-38> debian-installer: Temporarily add fs-core-modules to cdrom initrd, since isofs was moved
[17:55] <CIA-38> debian-installer: there. It should really be moved back.
[18:26] <mario_limonciell> before the system goes down for a reboot at the end of the install, should all modules already be properly unloaded (eg rmmod), or is it possible that some will  be left dangling?
[18:40] <mario_limonciell> er well i guess the reboot code should be the same as a regular reboot.  i'll repost in (the more populated) ubuntu-devel
[19:20] <Lrrr> cjwatson: ping
[19:23] <Lrrr`> err
[21:03] <hardwire> hey.. yo..
[21:03] <hardwire> Seems like installation methods requiring a fetch from repositories has been choking by downloading packages that are a bit newer than tha packages.gz for that pool.
[21:04] <hardwire> I've looked around on launchpad but couldn't find anything recent mentioning this.  Curious if there is any action on resolving network installation issues by making sure packages and udebs (and some debs) are correctly listed.
[21:05] <cjwatson> Lrrr: yes?
[21:05] <cjwatson> mario_limonciell: I don't think we take any special care to unload modules, and it might be quite inconvenient to do so
[21:05] <cjwatson> hardwire: it's fixed in hardy-proposed - bug 234486
[21:06] <cjwatson> also a pinch of bug 94398
[21:06] <hardwire> I just assumed it was a problem with the pools.
[21:06] <hardwire> ahha
[21:06] <hardwire> thank you, that first bug makes total sense
[21:06] <cjwatson> no, it was an installer bug - it wasn't dealing properly with packages showing up in multiple Packages files
[21:07] <cjwatson> one of those problems that by its very nature doesn't show up until after release
[21:07] <hardwire> so I should create a network boot image based on hardy proposed somehow?
[21:07] <cjwatson> there's one linked from those bugs
[21:07] <hardwire> I could kiss you on the libs
[21:07] <hardwire> lips
[21:08] <hardwire> and your libs
[21:08] <hardwire> wherever it may be
[21:08] <cjwatson> my wife might object. :)
[21:08] <cjwatson> note the boot option you need to use - apt-setup/proposed=true
[21:08] <hardwire> which only that network image groks?
[21:08] <cjwatson> no, all of them do, but if you don't use it then it'll be confused since it has a newer kernel than what's in -updtaes
[21:09] <cjwatson> -updates
[21:09] <cjwatson> actually, that may not be the case any more, so you might get away without it
[21:09] <hardwire> I have 30 crazy fists of qemu here waiting to try it out
[21:09] <hardwire> give me a sec and I'll let you know
[21:10] <cjwatson> apt-setup/proposed=true => grab both installer components and .debs from the not-yet-QAed queue
[21:10] <hardwire> using stock network.tar.gz
[21:10] <hardwire> netboot
[21:10] <cjwatson> that installer image should be able to move into -updates relatively shortly, at which point that won't be necessary any more
[21:10] <hardwire> sudo qemu -append "vga=normal apt-setup/proposed=true" -initrd ubuntu-installer/i386/initrd.gz -kernel ubuntu-installer/i386/linux -hda test.img -net nic -net user
[21:10] <hardwire> lets see how well this works
[21:19] <hardwire> hardy-proposed is in /usr/share/apt-setup/release-files/
[21:19] <hardwire> woot
[21:20] <cjwatson> lucky I was foresighted that day ;-)
[21:20] <hardwire> are you mr fixit for this bug?
[21:21] <cjwatson> yeah
[21:21] <hardwire> it's been haunting me for almost a month
[21:21] <hardwire> I was doing a lot of re-installs at work while QA'ing some install procedures.
[21:21] <cjwatson> unfortunately it'll probably continue to haunt people, as we generally don't change dists/hardy/
[21:22] <cjwatson> so we'll have to update documentation to point to dists/hardy-updates/ or something
[21:22] <hardwire> cjwatson: I was wondering on how fixable stuff like this is unless you constantly make sure people are using the most up to date install initrd's
[21:22] <hardwire> maybe there needs to be a businesscard for ubuntu soon.
[21:22] <hardwire> that way all these updates are pre-inserted
[21:22] <cjwatson> still have to make sure people include that ...
[21:23] <cjwatson> err ... update
[21:23] <hardwire> yar
[21:23] <cjwatson> really we just have to be perfect first time round ;-)
[21:23] <hardwire> I'm no man to say how it should be done
[21:23] <hardwire> a friend of mine followed my directions and somehow managed to brick his EEPC via network install :)
[21:23] <cjwatson> I'm not sure what the answer is; it may be that we should just decide to update the installer in dists/hardy/
[21:23] <hardwire> I'm not sure how he managed it, since it doesn't hit disk until way later
[21:23] <hardwire> but, this will help him regain faith in an OS he is not too familiar with
[21:23] <cjwatson> but then it becomes incompatible with the kernel referenced from dists/hardy/, so we'd really have to do the lot
[21:24] <hardwire> I told him geeks are nothing without technical issues.
[21:24] <hardwire> you can't be a geek around things that work.. you're just some random employee at that point.
[21:25] <cjwatson> job security! *ahem*
[21:25] <hardwire> now I just need to make him a bootable usb stick with a .iso on it..
[21:25] <CIA-38> debian-installer: cjwatson * r939 ubuntu/debian/changelog: clarify
[21:25] <hardwire> cjwatson: unless you can put pool/dist on the root of a usb drive then use file:/// somewhere in installation for the source.
[21:25] <CIA-38> debian-installer: cjwatson * r940 ubuntu/debian/changelog: releasing version 20080522ubuntu4
[21:26] <cjwatson> hardwire: evand is the installer hacker who's been playing with this most recently
[21:26] <cjwatson> the main problem is that grub tends to get a bit confused about where to install itself
[21:26] <hardwire> wonder if I can help out
[21:26] <cjwatson> needs to be taught about uuids
[21:27] <hardwire> the biggest challenge to jumping in on these projects is all the backstory on how you guys do releases
[21:27] <cjwatson> but, there's an isotostick.sh somewhere on the wiki, I believe
[21:27] <hardwire> I need to sit down and read up on all the angsty forums and figure it all out
[21:27] <hardwire> yeh
[21:27] <hardwire> that .. doesn't work
[21:27] <cjwatson> heh, ok
[21:27] <hardwire> at least the version I found.
[21:27] <cjwatson> https://wiki.ubuntu.com/InstallerDevelopment may help
[21:27] <hardwire> it failes to use fdisk correctly
[21:27] <cjwatson> (not for USB, for the other bit)
[21:27] <hardwire> much like I fails to spelling
[21:28] <cjwatson> oh, yeah, Evan did mention something about that I think - rather needs to be converted to parted
[21:28] <cjwatson> anything particular you're missing about how we do releases? I have a fairly complete picture, and nobody else seems to be vying for attention in here ...
[21:30] <hardwire> thats sad
[21:30] <hardwire> I'd say ToyKeeper needs to join this channel and chum it up with you about the installer.  He's interested in a canoniacal position if possible.  Already done quite a bit of the interview work.
[21:31] <hardwire> canonical.. spell check ftw
[21:31] <mario_limonciell> cjwatson, ah that's what i was afraid of.  unfortunately ricoh_mmc modifies the PCI address space when it's loaded.  it cleans up after itself when its unloaded, assuming you unload it properly
[21:31] <mario_limonciell> the very late stages of factory validation don't seem to like it when PCI hardware starts to not show up the same after you do an install
[21:32] <cjwatson> surprised that reboot() wouldn't go through and unload everything
[21:33] <cjwatson> you could have early_command write out /usr/lib/finish-install.d/98rmmod or something ...
[21:33] <cjwatson> (usual sort of grotty hack)
[21:34] <hardwire> cjwatson: has their been chatter on a netinst/business card alternate installer for ubuntu?
[21:34] <mario_limonciell> cjwatson, yeah at this point we're planning on dropping something in the late_command stuff - in any which case it will end up being a gross hack no matter what we have to do
[21:34] <cjwatson> hardwire: not really, I've sort of regarded the netboot/mini.iso as filling that gap, though I realise it's not quite the same
[21:34] <hardwire> mini.iso does wonders
[21:35] <cjwatson> hardwire: it probably wouldn't be very hard to do, I'm just slightly wary of further increasing the number of objects we ship (and therefore have to test)
[21:35] <hardwire> I'm wondering how well a livecd with squashfs root and unionfs in ram would work
[21:35] <hardwire> is it squashfs now?
[21:35] <cjwatson> squashfs, yes
[21:35] <hardwire> I haven't looked
[21:35] <cjwatson> there is actually a "base" live CD around somewhere
[21:35] <hardwire> I wonder what the installer size is sans packages.
[21:35] <hardwire> live installer\
[21:36] <cjwatson> http://cdimage.ubuntu.com/livecd-base/current/
[21:36] <cjwatson> not packaged up into an .iso though
[21:36] <hardwire> interesting
[21:36] <cjwatson> and not tested for ages :-)
[21:36] <hardwire> directoryindex ftw
[21:36] <cjwatson> doesn't have the installer in it, though that could be done (it sort of stops being all that "base" at that point)
[21:37] <cjwatson> I just made the index a bit more standard, though it doesn't have a HEADER.html
[21:38] <hardwire> I've been thinking of using ubuntu for an ebox like project that I'm working on (along with several hundreds of other people doing their own thing)
[21:38] <hardwire> I'm gonna have to play with the preseed stuff soon
[21:38] <cjwatson> preseeding is awesome
[21:38] <cjwatson> once you grok it :)
[21:40] <hardwire> willdooo
[21:40] <cjwatson> now, can I face working out what's wrong with the boot loader configuration on the intrepid daily, or is a bacon sandwich more appealing
[21:40] <hardwire> a qemu instance of the install appears to be working, still downloading the files but I'll live.  It didn't die at kickstart-common
[21:40] <cjwatson> good-oh
[21:41] <hardwire> I like fried egg and bacon sammich w/ mayo
[21:41] <hardwire> on toast
[21:41] <cjwatson> for some reason, all the menu labels have gone AWOL, and on startup it pops up a dialog saying "vesamenu.c32"
[21:41] <cjwatson> I think it may have developed AI by accident and gone a bit mad
[21:42] <hardwire> dangit
[21:42] <hardwire> it just threw an error at kickseed-common
[21:42] <hardwire> I'll try the newer initrds
[21:43] <cjwatson> oh, you were trying with the dists/hardy/ initrd and apt-setup/proposed=true? yeah, that definitely won't work
[21:43] <cjwatson> sorry if I wasn't clear - the bug fix is actually embedded in the initrd so you really do need the newer one - apt-setup/proposed=true is just to stop it being confused by a kernel version mismatch
[21:44] <hardwire> cjwatson: I asked if it would :( *sniff*
[21:44] <hardwire> gotcha
[21:44] <Lrrr> cjwatson: Got a question.  Is LVM preseed even half supported in the Hardy installer?  It's hard to gather information on that and it doesn't work well in my tests.
[21:44] <hardwire> Lrrr: can't you run scripts from preseeds?
[21:44] <cjwatson> hardwire: ah, you asked whether it was just the dists/hardy-proposed/ initrd that understood apt-setup/proposed=true, and I did answer accurately :)
[21:45] <hardwire> cjwatson: I was confused..
[21:45] <cjwatson> Lrrr: it ought to be, but I confess it isn't well-tested. What's going wrong?
[21:45] <cjwatson> (nor is it especially well-documented)
[21:45] <cjwatson> hardwire: you can, but the partitioner is special and you do tend to have to drive it in the way it expects rather than by scripting around it
[21:45] <Lrrr> cjwatson: It doesn't cleanup the LVM volumes on the target disk
[21:46] <Lrrr> partman-lvm/device_remove_lvm doesn't seem to do anything
[21:47] <hardwire> I sorta had questions about this too... because I'm kinda evil and want to script it so that it sets up LVM striping after wiping disks.. :)
[21:47] <cjwatson> Lrrr: I'll probably need to see /var/log/syslog and /var/log/partman, possibly in a bug report on partman-lvm describing the problem in detail
[21:47] <cjwatson> this IRC channel might be too narrow to contain it :)
[21:48] <Lrrr> hardwire: I wouldn't want to make ugly scripts.  I'll use what's implemented or implemented myself.  I'm sick of ugly hacks :/
[21:48] <hardwire> but I love those.
[21:48] <cjwatson> I generally have to do this by following code though - I admit I don't use the installer's LVM support very often myself
[21:49] <hardwire> cjwatson: I usually use the LVM autopart but break to a shell and lvresize root (or home) sans 1 gig
[21:49] <hardwire> that way I can at least use LVM's snapshot features
[21:49] <Lrrr> cjwatson: well, I've put it aside right now but if LVM is retained as a solution to our problem I'll need to come back to it eventually.
[21:50] <Lrrr> still, what you said kinda confirms what I discovered trying to wire our installer to use it.  It doesn't look as supported as I hoped it to be.
[21:50] <hardwire> I'd be interested in a "headspace" feature to the lvm autopart.. hmm..
[21:50] <cjwatson> Lrrr: I'm very much dependent on user reports for LVM, so anything you can supply would be appreciated
[21:50]  * cjwatson -> bacon
[21:51] <hardwire> toast!
[21:51] <Lrrr> cjwatson: Thank you for taking the time to answer.  Have a nice sandwich.
[21:55] <hardwire> Emacs IRC.. that's insanity.
[21:55] <xivulon> cjwatson any luck with the daily?
[21:55] <hardwire> shhh he's eating bacon.
[22:43] <cjwatson> xivulon: no, I need to poke the archive still
[22:57] <xivulon> evand timing copying an ISO with/without syncio on a real partition makes almost no difference
[22:58] <evand> ok
[22:58] <evand> PPA is up, by the way
[23:00] <xivulon> http://paste.ubuntu.com/21243/
[23:00] <xivulon> in fact it takes even less with syncio (don't ask why)
[23:08] <xivulon> evand I will ask again ubuntu-testing
[23:09] <xivulon> can we try to release the ntfs-3g for proposed? I added a comment with my tests on 204133
[23:21] <evand> I'll talk to slangasek about it.  I believe we're getting a little late in 8.04.1 to be making changes, it may have to wait for .2
[23:27] <xivulon> well I have to add wubi anyway
[23:27] <xivulon> but am waiting for the dailys
[23:27] <xivulon> and we need the casper / autopartition-loops
[23:48] <xivulon> evand can we release 238701 and 230716?
[23:48] <xivulon> we might also want to push wubi rev 503
[23:49] <evand> they're already in proposed