[09:45] <cjwatson> cwj: I'd suggest looking at the server logs to see which URL is 403ing
[10:48] <ogra> cjwatson, which installer part generates fstab, is that partman or something later (before i search my butt off)
[10:49] <ogra> (i dont want to write new code for the preinstalled images if something exists already)
[10:54] <ogra> ah, rootskel
[10:56] <ogra> hmm, no, that seems to be d-i only
[11:00] <cjwatson> partman-target
[11:00] <ogra> yeah, found it by grepping through ubiquity source now
[11:00] <cjwatson> you can't really run it in isolation from partman, if that's what you're thinking
[11:01] <ogra> nope, i just wanted to see the code to get an idea for jasper (the initramfs tool for preinstalled images)
[11:44] <CIA-4> pkgsel: cjwatson * r161 ubuntu/debian/ (changelog postinst):
[11:44] <CIA-4> pkgsel: Install tasksel before doing anything else. This means that we can take
[11:44] <CIA-4> pkgsel: it out of the base system, thus saving space on the live CD.
[11:48] <ogra> hmm, what part of the installer adds the default system groups ? (adm, dialout etc)
[11:48] <cjwatson> user-setup
[11:48] <ogra> ah, i was looking at user-setup-apply :)
[11:49] <CIA-4> pkgsel: cjwatson * r162 ubuntu/debian/changelog: releasing version 0.25ubuntu8
[12:00] <cjwatson> ev: this works out as about 2MB off the .deb size contributing to the live CD
[12:00] <cjwatson> so not quite as big as I'd hoped, but not to be sneezed at either
[12:01] <cjwatson> it takes 10MB off minimal; 7MB just moves out to standard, and another MB or so to desktop
[12:02] <ev> so, usb-creator is 1.9 MB :).  Mind if I steal that 2 MB away, or would you prefer it go somewhere else?
[12:02] <cjwatson> (methodology: two germinate runs with tasksel+aptitude moved from minimal/standard to d-i-requirements in the second run, and with a modified germinate that ignores the dependencies on those from ubuntu-minimal and ubuntu-standard; then diff -bru on the two outputs and scan visually)
[12:02] <cjwatson> seems a reasonable use for it
[12:03] <cjwatson> I have some other savings in mind anyway
[12:03] <ogra> hmm, does oem-config make any use of the passwd/user-default-groups value ? or is it expected that this is already set when oem-config runs
[12:03] <ev> awesome, thanks
[12:04] <ev> thanks for elaborating as well.  Always good to know these things
[12:05] <cjwatson> ogra: it uses the value, in that the first user will be created with those groups; it doesn't set the value though, it's typically just the value from the debconf template unless somebody preseeds it
[12:05] <ogra> ah, cool, so i dont need to bother as long as i run oem-config ?
[12:06] <cjwatson> right
[12:06] <ogra> perfect
[12:06] <ogra> my work gets easier every day :)
[12:59] <twb`> Minor gripe: hardy's netboot.tar.gz includes a pxelinux.cfg/default.9600, but lucid's doesn't.
[13:01] <twb`> Hm, interesting.  I have two four-port cards, and lucid has managed to assign eth[0247] to one card and eth[1356] to the other.  Hardy at least had the four ports of each card in a row.
[13:07] <cjwatson> pxelinux.9600> removed upstream, by the look of it.
[13:07] <cjwatson>   * Note that the pxelinux config file for serial terminals has been dropped,
[13:07] <cjwatson>     at least for now, since the split config files made it too difficult to
[13:07] <cjwatson>     set up.
[13:07] <twb`> upstream meaning debian?
[13:09] <cjwatson> yes, that's as far upstream as the installer goes.
[13:11] <ogra> cjwatson, am i right to assume that oem-config also sets up sudoers properly for me ?
[13:11] <cjwatson> user-setup does, yes
[13:11] <ogra> heh, my jasper-setup sctip boiled down to two functions :)
[13:11] <ogra> *script
[13:11] <ogra> great
[13:12]  * ogra cant really belive he only needs to setup fstab and /etc/network/interfaces ... but i dont seem to find anything beyond that
[13:14] <twb`> ogra: the latter should be magicked by NM for desktops
[13:14] <ogra> twb`, not for loopback
[13:14] <twb`> Ah.
[13:15] <ogra> actually i guess i also need to create /etc/hosts but that should be it
[13:16] <twb`> There's now a pam_thingyhost
[13:16] <twb`> Make that libnss-myhostname - nss module providing falback resolution for the current hostname
[13:16] <twb`> Meaning that you don't need to put yourself into /etc/hosts anymore
[13:17] <ogra> well, i'm talking about a preinstalled oem image
[13:17] <ogra> and localhost needs to be resolved
[13:17] <ogra> the tool i'm writing is run from initramfs to do the basic setup before oem-config runs
[13:18] <ogra> netcfg in ubiquity relies on the squashfs to have /etc/network/interfaces /etc/networks /etc/hostname /etc/resolv.conf /etc/hosts and copies them over, i need to check which of these i need to manually create on boot
[13:20] <ogra> hmm
[13:20] <ogra> i should probably just run netcfg in oem-config and let it do its job :)
[13:20] <cjwatson> netcfg can be awkward to run directly for one reason or another
[13:20] <twb`> Huh.  Last time I looked, casper/live-initramfs generated resolv.conf on the fly iff there isn't one in a module.
[13:21] <ogra> cjwatson, but oem-config can run the network part of ubiquity, no ?
[13:21] <ogra> shouldnt that run netcfg-common and create loopback and friends ?
[13:22] <cjwatson> look at the network bit of ubiquity, it actually does it itself with a comment about how it would be nice not to need to
[13:22] <ogra> ah, great
[13:24] <ogra> heh, that only leaves fstab then
[13:24] <ogra> which i sadly have to handle in my script since i change the UUID during boot and partition resizing
[13:39] <twb`> BTW, why does priority=low now prompt me to choose locales "#" and "#" in addition to the usual ones (such as aa_DJ.UTF-8)?
[13:59] <ev> cjwatson: what would be a good time for you to talk through this debconf stuff?  I'm free the rest of the day.
[14:00] <cjwatson> twb`: moderately-long-standing localechooser bug, sorry
[14:00] <cjwatson> it's been filed
[14:00] <cjwatson> ev: about 20 minutes?
[14:00] <ev> surely
[14:00] <twb`> cjwatson: OK, it's not bothering me.  So long as someone knows about it.
[14:01] <cjwatson> I think it'll be fixed by a merge, it's just a big ugly complex merge
[14:01] <cjwatson> bug 465120
[14:01] <ubot2> Launchpad bug 465120 in localechooser (Ubuntu) ""Choose language" dialog to select additional locales has two "#" locales (affects: 1) (heat: 6)" [Low,Confirmed] https://launchpad.net/bugs/465120
[14:02] <twb`> Does d-i use "PXEClient" as its vendor string in DHCP requests?
[14:02] <twb`> Because I told dnsmasq to only respond to that vendor, but it was hijacking d-i's DHCP requests
[14:02] <cjwatson> "d-i", I believe
[14:03] <cjwatson> ./dhcp.c:186:                fprintf(dc, "send vendor-class-identifier \"d-i\";\n" );
[14:03] <twb`> That's what I thought
[14:03] <twb`> Obviously I messed up, or dnsmasq is buggy.  Most likely I messed up.
[14:06] <twb`> Hm.  As well as being able to opt-in to some udebs (e.g. sshd), I should be able to opt out of udebs like xfsprogs.
[14:08] <twb`> "Based on your present physical location, your time zone is None."
[14:08] <twb`> I'm pretty sure I told it I was in Australia already.
[14:09] <twb`> I don't like the new chooser where I have to scroll down lots instead of just picking "australia" and "melbourne" separately.
[14:47] <ev> twb`: can you put set -x at the top of /usr/sbin/tzsetup, run through the timezone step, and attach syslog to a new bug report?
[14:48] <twb`> ev: unfortunately, not right now.  This is my precious new mission-critical router.
[14:48] <ev> twb`: no worries
[14:48] <twb`> Unless it's acceptable to manually re-run tzsetup after I get to the "hit enter to reboot" step?
[14:48] <twb`> I can do that fine, because at that point I don't give a damn what happens in the tmpfs
[15:09] <twb`> WTF?  I passed pkgsel/language-pack-patterns= pkgsel/install-language-support= tasks=standard, but it's installing language-pack-en-base
[15:10] <cjwatson> that'd be pkgsel/install-language-support=false - it's a boolean
[15:10] <twb`> I used what the pxelinux.cfg default was
[15:10] <cjwatson> (don't know if that's the problem)
[15:12] <twb`> Looks like it's =false in text.cfg, I just mis-copied it into my notebook.
[15:12] <twb`> I suppose I must have fat-fingered it when I was adding in console=ttyS0 and friends.
[15:14] <twb`> Grmph
[15:14] <twb`> It's installing ppp now
[15:14] <twb`> Once upon a time the minimal install was actually minimal
[15:16] <twb`> I should've bothered to write a preseed file.
[15:17] <twb`> Is extlinux gonna get into d-i now that pantera wrote kernel.conf.d hooks for it?
[15:18] <twb`> Heh, I like how the "grub legacy or 2?" prompt says "don't use 2 in production" but defaults to 2.
[15:27] <twb`> ev: looks like geoip.ubuntu.com/lookup gives <TimeZone>Australia/Victoria</TimeZone> for 203.7.155.64, but it should be Australia/Melbourne.
[15:30] <cjwatson> we're on the point of getting down to a single boot loader - not wild about introducing extlinux to the mix
[15:31] <cjwatson> yeah, the prompt text needs to be updated
[15:31] <twb`> Grmpg
[15:31] <twb`> grub sucks
[15:32] <twb`> It's too complicated and flaky to be doing something as low-level as bootstrapping the kernel
[15:32] <cody-somerville> cjwatson, Can grub do isohybrid images?
[15:33] <twb`> Anyway, my router is now booting 10.04, so I'm A-OK.
[15:33] <twb`> cody-somerville: hybrid as in multi-track?
[15:34] <cody-somerville> twb`, http://syslinux.zytor.com/wiki/index.php/Doc/isolinux#HYBRID_CD-ROM.2FHARD_DISK_MODE
[15:34] <twb`> Oh, http://syslinux.zytor.com/wiki/index.php/Doc/isolinux (the BIOS pretends the CD is a ZIP/HDD)
[15:35] <twb`> Oh!  It means a single file is both an ISO 9660 .iso and a bootable USB mass storage .img.
[15:35] <twb`> That's neat
[15:36] <cody-somerville> Very neat.
[15:36] <cjwatson> cody-somerville: yes, in fact if you install grub-rescue-pc you'll find it's already done
[15:36] <cody-somerville> schweet
[15:37] <cody-somerville> cjwatson, so I don't have to post-process the iso like I do with isolinux?
[15:37] <cjwatson> we need the complexity of grub in many places (sure, most people only need 10%, but they're different 10%s all over the place ...), and it's better to use the same code everywhere to reduce maintenance costs
[15:37] <cjwatson> speaking as the muggins who gets to do most of the maintenance, at least in Ubuntu, I think I'm not unqualified to comment :)
[15:37] <cjwatson> cody-somerville: grub-mkrescue does it out of the box
[15:39] <cjwatson> put another way: I'm fed up of all the different boot loaders working in four out of five situations.  I want to concentrate on one thing and make it work everywhere.
[15:40] <twb`> Yeah, I understand.
[15:40] <twb`> It's just annoying that when one drive in my md RAID1 dies, the systems fails to even BOOT because grub is trying to load (hd1,0)
[15:41] <ev> twb`: ah, nice catch!  I'll file a bug for that now.
[15:41] <twb`> Or when I can't install a new kernel because update-grub is trying to scan a CD drive that doesn't exist for Windows dual-boot partitions
[15:41] <cjwatson> yes, obviously those kinds of things are stupid and I want to get them fixed
[15:42] <cjwatson> but the difference I've found is that I generally actually can fix them with grub2.  with everything else I generally had to give up.
[15:42] <twb`> The other awesome one was when I asked it to install the dos disk label, raid1, ext3 and lvm modules into the MBR, and it exploded because they didn't all fit
[15:43] <cjwatson> yes, well, not much to be done about that except use a bigger embedding area (which in fact we do, if you partition from scratch with >=lucid) or use GPT and a BIOS Boot Partition
[15:43] <twb`> It's a PITA to use GPT with a conventional BIOS
[15:44] <ev> right, bug 589708 created
[15:44] <ubot2> Launchpad bug 589708 in tzsetup (Ubuntu) "Wrong timezone location: Victoria (affects: 1) (heat: 6)" [High,Confirmed] https://launchpad.net/bugs/589708
[15:45] <twb`> ev: ty
[15:45]  * twb` wonders what the lp-shell equivalent of "bts subscribe" is.
[15:45] <twb`> Urk, lp-shell just opens w3m
[15:46] <cjwatson> though that said the collection of modules you listed seems to work fine right noow
[15:46] <cjwatson> now
[15:46] <twb`> Goddamn web-based apps :-/
[15:46] <cjwatson> $ ./grub-mkimage -O i386-pc -d . -o cjwtest.img biosdisk ext2 part_msdos lvm raid search_fs_uuid
[15:46] <cjwatson> $ ls -l cjwtest.img
[15:46] <cjwatson> -rw-r--r-- 1 cjwatson cjwatson 29463 2010-06-04 15:45 cjwtest.img
[15:46] <twb`> cjwatson: yes, I got that particular failure back around 2005 :-)
[15:46] <cjwatson> I wasn't under the impression that GPT was that hard to use with a normal BIOS.  You just need a protective MBR with an active partition in it, which is part of the GPT spec anyway.
[15:47] <cjwatson> the problem is more that people usually have some other OS installed already and it set things up with MBR
[15:47] <twb`> Back then I also had to patch the C code to not think LVM was a physical device
[15:47] <cjwatson> we couldn't reasonably have used grub2 in 2005 (we did look), but it's come on massively since then
[15:48] <twb`> The CD scanning thing, though, that happened a month ago
[15:48] <cjwatson> bug?
[15:48] <cjwatson> bug#, I mean
[15:48] <twb`> Um, I was in a hurry so I just dpkg-diverted the "look for other OSes" script
[15:48] <cjwatson> so probably not even grub2 - it'll be a bug in os-prober
[15:49] <twb`> That's true
[15:49] <cjwatson> p.s. GRUB_DISABLE_OS_PROBER=true in /etc/default/grub is easier
[15:49] <cjwatson> (and will be documented in 'info grub' with the next upload)
[15:49] <twb`> Or just uninstall os-prober
[15:49] <cjwatson> whatever
[16:06] <cwj> cjwatson: i don't control the apt repo i am trying to preseed into my install, so i can't reasonably get at the logs. can i turn up the debug in d-i somehow?
[16:06] <cjwatson> cwj: DEBCONF_DEBUG=developer is sometimes useful but I don't know if it will show this
[16:07] <cjwatson> you could try that though
[16:08] <cwj> ok
[16:08] <cwj> i did review the apt-cacher-ng logs, nothing stood out there
[17:17] <ogra> cjwatson, do you think it would do any harm if i added a dep on oem-config to jasper ?
[17:17] <ogra> that will definately make sure its removed with oem-config
[17:18] <cwj> cjwatson: now that i look at the installer syslog again, it appears that the 403 error is happening before the preseed configuration is retrieved. https://gist.github.com/8cdc39663420d986f7bd
[17:19] <cjwatson> ogra: sounds ok
[17:19] <ogra> great
[17:19] <ogra> i think then i have the core done :)
[17:19] <ogra> just need to test it
[17:19] <cjwatson> cwj: yeah, but it was happening during apt-setup too
[17:20] <cjwatson> I didn't even look at that part of your log earlier - I searched for 'apt-setup (leading apostrophe intentional) and went from there
[17:38] <cwj> well i am going to have to address this via preseed/late_command for now, running out of time to debug
[17:51] <cwj> which question do i need to preseed to avoid being prompted about how much of the disk to use? i think its LVM that is asking. here's what i have now: http://gist.github.com/425636
[17:55] <cwj> size of volume group to be used
[17:56] <cjwatson> partman-auto-lvm/guided_size
[17:56] <cwj> thanks
[18:25] <CIA-4> cdrom-detect: cjwatson * r462 ubuntu/debian/changelog: releasing version 1.32ubuntu5
[18:32] <CIA-4> cdrom-detect: cjwatson * r462 ubuntu/debian/changelog: releasing version 1.32ubuntu5
[18:32] <cjwatson> whoops, sorry for the dup there
[18:34] <CIA-4> cdrom-detect: cjwatson * r459 ubuntu.lucid/debian/ (cdrom-detect.postinst changelog):
[18:34] <CIA-4> cdrom-detect: usb-storage sometimes seems to take a while to settle to the point of
[18:34] <CIA-4> cdrom-detect: actually exposing block devices, despite the 'udevadm settle' in
[18:34] <CIA-4> cdrom-detect: hw-detect. I don't see an easy way to solve this, so, with
[18:34] <CIA-4> cdrom-detect: reservations, retry the detection loop three times with a 'sleep 3' in
[18:34] <CIA-4> cdrom-detect: between in the event that no devices are found, in the hope that that
[18:34] <CIA-4> cdrom-detect: improves matters (LP: #586036).
[18:36] <CIA-4> cdrom-detect: cjwatson * r460 ubuntu.lucid/debian/changelog: releasing version 1.32ubuntu3.10.04
[18:44] <CIA-4> cdrom-detect: cjwatson * r452 ubuntu.karmic/debian/ (cdrom-detect.postinst changelog):
[18:44] <CIA-4> cdrom-detect: usb-storage sometimes seems to take a while to settle to the point of
[18:44] <CIA-4> cdrom-detect: actually exposing block devices, despite the 'udevadm settle' in
[18:44] <CIA-4> cdrom-detect: hw-detect. I don't see an easy way to solve this, so, with
[18:44] <CIA-4> cdrom-detect: reservations, retry the detection loop three times with a 'sleep 3' in
[18:44] <CIA-4> cdrom-detect: between in the event that no devices are found, in the hope that that
[18:44] <CIA-4> cdrom-detect: improves matters (LP: #586036).
[18:44] <CIA-4> cdrom-detect: cjwatson * r453 ubuntu.karmic/debian/changelog: releasing version 1.31ubuntu3.9.10