2008/11/21/#ubuntu-installer.txt

[09:11] <evand> superm1: your patches will be in there
[09:29] <CIA-2> debian-installer: cjwatson * r995 ubuntu/debian/changelog: releasing version 20081029ubuntu2
[11:32] <davmor2> cjwatson: I just tried an alternate install and it failed :(
[11:34] <davmor2> logs are at http://www.davmor2.co.uk/install
[11:36] <cjwatson> davmor2: yes, reported last night, already being fixed, thanks
[11:36] <cjwatson> davmor2: bug 300426
[11:36] <ubottu> Launchpad bug 300426 in udev "alt-install /dev fails to mount" [Critical,Fix released] https://launchpad.net/bugs/300426
[11:37] <davmor2> cjwatson: Cool :) I'll try again sunday :)
[11:40] <cjwatson> should have fixed images in an hour or two
[11:40] <davmor2> Oh cool you're going to reroll then :)
[11:40] <cjwatson> yes
[11:40] <cjwatson> under pressure to get alpha-1 done
[11:40] <davmor2> if you give me a ping I'll dl it and test :)
[11:42] <cjwatson> will do
[11:43] <davmor2> cjwatson: What's happening about Live-fs?  Is that being dropped for now or will that be fixed too?
[11:45] <cjwatson> not for a1
[11:45] <davmor2> Okay cool thought that might be the case but thought I'd better  check :)
[11:50] <cjwatson> ubiquity isn't ready yet anyway
[12:12] <CIA-2> debian-installer: cjwatson * r996 ubuntu/ (5 files in 4 dirs):
[12:12] <CIA-2> debian-installer: Require mouse-modules again on lpia and powerpc. rootskel-gtk depends on
[12:12] <CIA-2> debian-installer: it anyway, so we might as well just fail to build until linux-lpia and
[12:12] <CIA-2> debian-installer: linux-ports are updated.
[12:15] <cjwatson> davmor2: up now
[12:15] <cjwatson> amd64 is fractionally oversized, but I'm not going to worry about it since the new mono will shave off a good chunk once it lands
[12:22] <davmor2> np's thanks
[12:23] <cjwatson> I really wish my mirror would get on with it so that I could use jigdo rather than rsync for this
[12:23] <cjwatson> I mean, like I care about nvidia drivers
[12:52] <evand> It occurs to me that we never tested wubi on USB disks.
[12:55] <evand> ah, if I'm reading the NSIS code correctly, it looks like it would still work without issue.
[12:55] <evand> davmor2: xivulon Do you have a USB disk and some free time :) ?
[12:56] <davmor2> I do but I want to test the cd's first I can look at usb later though
[12:57] <evand> ok, thanks
[13:33] <kirkland> cjwatson: one more issue i need to get your take on, regarding encrypting $HOME ...
[13:33] <kirkland> cjwatson: i need a place to store $USER/.ecryptfs that I can ensure won't be encrypted
[13:33] <kirkland> cjwatson: i have two ideas ....
[13:34] <kirkland> cjwatson: I want to avoid having /home/$USER and /home/.$USER, as it sounds like you didn't like that idea much
[13:35] <kirkland> cjwatson: that's okay, i've worked around it
[13:35] <kirkland> cjwatson: but, i do think I'll need to put $HOME/.ecryptfs somewhere
[13:35] <kirkland> cjwatson: and I'm thinking /var/lib/ecryptfs/$USER
[13:35] <kirkland> cjwatson: that directory being perm'd 700, owned by $USER:$USER
[13:36] <kirkland> cjwatson: kees and jdstrand didn't like the idea of making /var/lib/ecryptfs itself be 777
[13:36] <soren> kirkland: Yikes! No, I can understand that :)
[13:36] <kirkland> cjwatson: what do you think about me patching either useradd or adduser to create and chown /var/lib/ecryptfs/$USER on user creation?
[13:37] <kirkland> cjwatson: just in the case where they want to encrypt all of $HOME
[13:37]  * soren has always dreamed that adduser some day would grow a plugin system for that sort of thing..
[13:37] <kirkland> cjwatson: and if they want to just encrypt $PRIVATE (or some other location less than $HOME), .ecryptfs stays in $HOME
[13:37] <kirkland> soren: "what" sort of thing?
[13:38] <soren> kirkland: Stuff that should happen on user creation.
[13:38] <soren> kirkland: ...that might not always be wanted.
[13:38] <soren> kirkland: People who don't want to use ecryptfs at all won't want this to happen.
[13:38] <kirkland> soren: sure
[13:39] <soren> Another use case is about adding users to samba then they're added to passwd as well.
[13:39] <persia> kirkland, Why special-case it?  If you're going to use /var/lib/$USER, it makes sense to use that for any class of ecryptfs user.  I don't have an opinion on whether it's a good idea, but I'd prefer consistency.
[13:39] <soren> (that's where I originally got the idea)
[13:39] <persia> Er.. /var/lib/ecryptfs/$USER
[13:40] <kirkland> persia: current 'ecryptfs-setup-private' is something that any user can do, non-privileged
[13:40] <soren> Personally, I don't like having $HOME essentially split between /home and /var.
[13:40] <kirkland> persia: mkdir /var/lib/ecryptfs/$USER would be a privileged operation
[13:40] <soren> For the same reason, I put my users' maildirs in their home directory.
[13:40] <kirkland> soren: i don't like it either
[13:40] <cjwatson> adduser already has a plugin system
[13:41] <kirkland> soren: but if a user chooses to encrypt all of $HOME
[13:41] <cjwatson> adduser.local or whatever it's called
[13:41] <soren> cjwatson: Oh. Is that a recent addition?
[13:41] <cjwatson> no
[13:41] <kirkland> soren: i need somewhere to put their configuration information that can be read prior to $HOME being mounted
[13:41] <cjwatson> I've been using it since at least 2001 or so
[13:41] <cjwatson> however, it's intended for use by the sysadmin not by the distro
[13:42] <cjwatson> kirkland: I don't have a problem with patching adduser provided that we don't break in the event that the sysadmin used useradd
[13:42] <cjwatson> (as in, we degrade gracefully)
[13:42] <soren> cjwatson: Ah, that sort of thing. Truth be told, I didn't know about that, but I was thinkgin more along hte lines of a directory of scripts to be called by adduser. That way, packages could add scripts there that would be called. It's a bit hard to do that with the .local thing.
[13:42] <soren> Man, my typing is horrible today.
[13:43] <cjwatson> kirkland: however it sounds like you and soren have some talking-out of this to do
[13:43] <cjwatson> kirkland: I'm a bit too distracted by other things to think about it properly ...
[13:43] <kirkland> cjwatson: okay, no worries
[13:43] <cjwatson> kirkland: also, you could put it in /home/$USER and then just overmount it; the trick would be dealing with syncing it back on umount
[13:43] <soren> kirkland: I only joined this discussion now... Why are we putting stuff in /var that we both think belongs in /home? :)
[13:44] <cjwatson> kirkland: (or some similar kind of invention)
[13:44] <kirkland> soren: your .ecryptfs directory contains configuration data
[13:44] <kirkland> cjwatson: hmm, i've been trying various ways of doing that
[13:44] <kirkland> cjwatson: and they're breaking when a user, say, changes their password
[13:44]  * persia likes the overmount suggestion
[13:44] <kirkland> cjwatson: which has to re-write .ecryptfs/wrapped-passphrase
[13:45] <kirkland> cjwatson: which goes through the ecryptfs kernel encryption
[13:45] <cjwatson> understood
[13:45] <kirkland> cjwatson: and the next time the user logs in, it's not readable (yet)
[13:45] <kirkland> chicken -> meet egg
[13:45] <persia> kirkland, Maybe use a union mount to remount the .ecryptfs directory on top of the unencrypted $HOME ?
[13:45] <cjwatson> I don't particularly mind /home/.user. It's not very aesthetically pleasing but it would work
[13:45] <cjwatson> but /var is fine by me too
[13:46] <soren> I don't like the aesthetics of it either.
[13:46] <cjwatson> whatever, argue it among yourselves :-)
[13:46] <soren> Why make it a hidden directory? That just seems odd.
[13:47] <kirkland> persia: what's the syntax of a union mount?
[13:47] <kirkland> persia: i can go try that
[13:47] <soren> /home/$USER.ecrypts wouldn't look very pretty either, though.
[13:48]  * persia goes to refresh knowledge of the documentation.
[13:48] <persia> Also, is this the right channel?  Should this be -devel or -server or something?
[13:50] <kirkland> persia: we can move ...  this conversation was originally intended to have an -installer flair and involve primarily cjwatson
[13:50] <kirkland> persia: -devel is fine
[13:51] <persia> kirkland, Right.  The beginning seemed well targeted.  "whatever, argue it among yourselves :-)" makes me think we should go elsewhere :)
[13:51] <cjwatson> well, I just wanted to say that I don't feel strongly about it from an installer POV
[13:51] <kirkland> cjwatson: thx.
[13:51] <kirkland> persia: agreed.
[13:52] <cjwatson> I think there's a useful discussion to be had on the placement, and don't want to prejudge that
[13:52] <cjwatson> the other possibility is a small set-id thing that creates /var/lib/ecryptfs/$USER
[13:52] <cjwatson> i.e. say that any user has the right to create a directory by their name there
[13:53] <cjwatson> think about how that interacts with quotas (hopefully not much)
[13:53] <cjwatson> but that doesn't seem a particularly unreasonable thing to do
[14:09] <davmor2> cjwatson: fix seems to be working so far, it got past where it was anyway :)
[14:34] <cjwatson> it's working for me
[14:47] <davmor2> cjwatson: Has it gone all the way through?
[14:48] <cjwatson> yes
[14:48] <cjwatson> freecell's up now :)
[14:49] <davmor2> meh I might have hd issues then wipe and try again :)
[15:17] <cr3> cjwatson: ping, have you had an opportunity to upload the netboot image for the kernel in -proposed? just checking with you in case I'm looking at the wrong place in the archive
[15:20] <cjwatson> no, not yet sorry
[15:24] <cr3> cjwatson: no problem, I was worried for a moment that I might not be looking at the right place
[15:28] <davmor2> cjwatson: Something is not right here at all.  I dd if=/dev/zero of=/dev/sda the drive and now all I get is a flashing cursor top left :(
[15:30] <cr3> davmor2: that's weird, shouldn't you get some error from the bios saying that the device is not bootable?
[15:32] <davmor2> cr3: I reinstalled Jaunty sorry not very clear
[15:32] <cjwatson> cr3: I'll try to remember to explicitly tell you once it's in
[15:34] <CIA-2> debian-installer: cjwatson * r997 ubuntu/ (3 files in 3 dirs): Make serial-modules optional on ia64; it's currently built-in.
[15:35] <cjwatson> davmor2: hmm, I really don't know at this point - all I can say is it's working for me :(
[15:35] <cjwatson> davmor2: that symptom sounds like the kernel isn't making it up
[15:35] <davmor2> I'm not entirely convinced grub is I'm just checking it now
[15:41] <davmor2> cjwatson: I'm trying it on my other test system incase it's an issue with hw I'll ping you in a bit and let you know
[15:42] <davmor2> running in vm too just to tripple check
[15:47] <cr3> cjwatson: hopefully, you won't have to remember for long, I'm spending my spare time automating detection of changes on archive.u.c
[16:03] <cr3> cjwatson: by the way, thanks for the installer fix. I noticed some test results trickling in automatically for 20081121.1
[16:04] <cjwatson> oh, no problem
[16:04] <cjwatson> I'm pushing a bzr branch up for intrepid-proposed, just taking a while
[16:11] <davmor2> cjwatson: definitely a weird ass thing going on here.  On the other install I had already got kubuntu intrepid.  After trying to install jaunty I now get Grub loading please wait Error 17
[16:12] <davmor2> So that is 2 sets of hw not working
[16:14] <davmor2> cjwatson: vm hanging too
[16:14] <davmor2> and that is from the iso
[16:16] <davmor2> cjwatson: Cd burn and iso both match md5sums with the image md5sum
[16:17] <cjwatson> phone.
[16:18] <cjwatson> 17 : Cannot mount selected partition
[16:18] <cjwatson>      This error is returned if the partition requested exists, but the
[16:18] <cjwatson>      filesystem type cannot be recognized by GRUB.
[16:19] <cjwatson> davmor2: what partitioning method did you use? were you starting from a blank disk?
[16:19] <davmor2> cjwatson: I am guessing that the grub is the old grub from the kubuntu install I'll check
[16:20] <davmor2> whole disk
[16:20] <cjwatson> why would the grub be the old grub from the Kubuntu install?
[16:21] <cjwatson> the normal behaviour is for grub to be reinstalled, unless you overrode that manually
[16:21] <davmor2> no I didn't override at all
[16:22] <davmor2> cjwatson: Does any of grub get installed to mbr?
[16:24] <cjwatson> yes
[16:24] <cjwatson> grub installs its first stage to the mbr
[16:29] <davmor2> cjwatson: I've pulled a copy of mbr using dd should it read gibberish?
[16:32] <cjwatson> gibberish to you ;-)
[16:33] <cjwatson> it should not be human-readable, if that's what you mean
[16:33] <davmor2> cjwatson: bin file yes
[16:36] <cjwatson> davmor2: 'od -Ax -tx1 -N512 /dev/sda' please?
[16:36] <cjwatson> (somewhat easier to handle than dd)
[19:33] <davmor2> cjwatson: you around still sorry I had to go out I hadn't clocked the time
[19:38] <davmor2> cjwatson: When you get chance http://www.davmor2.co.uk/mymbr is the file created from od -Ax -tx1 -N512 /dev/sda
[19:53] <CarlFK> is there something like /etc/foo that has ibex/jaunty in it?
[19:54] <CarlFK>  /etc/lsb-base/
[19:55] <cjwatson> CarlFK: lsb_release -cs
[19:55] <CarlFK> thanks
[20:02] <CarlFK> in the installer: ~ # echo|sed 'aA\nB'
[20:02] <CarlFK> AnB
[20:05] <cjwatson> yes. that isn't portable.
[20:05] <cjwatson> -> POSIX
[20:05] <CarlFK> rats.  so how do I get a \n?
[20:05] <cjwatson> echo | sed 'aA\
[20:05] <cjwatson> B'
[20:06] <cjwatson> or something along those lines
[20:06] <cjwatson> http://www.opengroup.org/onlinepubs/009695399/utilities/sed.html
[20:06] <CarlFK> thaks - ill figure it out
[20:59] <CIA-2> grub-installer: cjwatson * r760 ubuntu/ (debian/changelog grub-installer): 'dmraid -c -s' changed its output format; cope with both old and new.
[21:03] <CIA-2> grub-installer: cjwatson * r761 ubuntu/debian/changelog: releasing version 1.35ubuntu2
[21:06] <CIA-2> base-installer: cjwatson * r338 ubuntu/ (debian/changelog library.sh): 'dmraid -c -s' changed its output format; cope with both old and new.
[21:07] <CIA-2> base-installer: cjwatson * r339 ubuntu/debian/changelog: releasing version 1.86ubuntu8
[21:09] <CIA-2> grub-installer: cjwatson * r762 ubuntu/debian/ (changelog isinstallable): Nail another instance of dmraid output-checking.
[21:10] <CIA-2> grub-installer: cjwatson * r763 ubuntu/debian/changelog: releasing version 1.35ubuntu3
[21:11] <CIA-2> os-prober: cjwatson * r224 ubuntu/ (debian/changelog os-prober): 'dmraid -c -s' changed its output format; cope with both old and new.
[21:12] <CIA-2> os-prober: cjwatson * r225 ubuntu/debian/changelog: releasing version 1.28ubuntu2
[21:13] <CIA-2> partman-base: cjwatson * r116 ubuntu/ (debian/changelog lib/base.sh): 'dmraid -c -s' changed its output format; cope with both old and new.
[21:14] <CIA-2> partman-base: cjwatson * r117 ubuntu/debian/changelog: releasing version 128ubuntu2
[22:15] <CIA-2> hw-detect: cjwatson * r96 ubuntu/ (debian/changelog discover-mac-io.sh):
[22:15] <CIA-2> hw-detect: Reinstate snd-powermac for i2s-a devices since apparently the lock-up
[22:15] <CIA-2> hw-detect: has been fixed (LP: #87652).
[23:06] <CIA-2> debian-installer: cjwatson * r984 debian-installer/ (3 files in 2 dirs): Move mainline architectures to 2.6.27-8 kernels.
[23:06] <cjwatson> (that was intrepid-proposed branch, silly cia)
[23:09] <CIA-2> debian-installer: cjwatson * r985 intrepid-proposed/debian/changelog: releasing version 20080522ubuntu24
[23:53] <CIA-2> grub-installer: cjwatson * r764 ubuntu/ (debian/changelog otheros.sh): Fix syntax error in otheros.sh.
[23:54] <CIA-2> grub-installer: cjwatson * r765 ubuntu/debian/changelog: releasing version 1.35ubuntu4