[14:02] i have 'finish-install/reboot_in_progress note' in my preseed file (and no lines containing 'exit') and i find my new machines in a shutdown state. i want them to reboot upon installation [14:03] I assume you have "d-i " at the start of that line, or it will have no effect [14:04] In any case, that only controls whether a dialog is shown just before exiting the installer [14:05] Rebooting is the default if you haven't preseeded otherwise. If the installer hasn't rebooted then it must not be able to ... [14:05] As in, it will have been calling the "reboot" command [14:10] yes, i have the d-i part. i figured reboot is the default. these are kvm guests so i guess they can't reboot for some reason [16:00] xnox: bug 1100694 is regarding UbuntuKylin and appears to have a patch [16:00] Launchpad bug 1100694 in ubiquity (Ubuntu) "display OS in existing partitions" [Undecided,Confirmed] https://launchpad.net/bugs/1100694 [16:00] interesting. let me look. [16:02] if you accept that, please remove the useless start/end comments, and advise them not to include them in future patches ... [16:04] cjwatson: yeah. I'm not sure what it displays and how... given that one OS can span multiple disks and partitions. [16:05] os-prober results, perhaps? Or maybe they actually mean partition type? [16:06] OMG what [16:06] they display strings from misc.grub_options() [16:06] So, uh [16:06] Not quite [16:07] They look up the device name for the partition, and try to match that against misc.grub_options() [16:07] device path I mean, e.g. /dev/sda1 [16:08] That AT BEST needs a big comment explaining what it's doing, but I can't see how it makes sense; the described purpose would be better served by something involving os-prober [16:08] But I'll let you figure it out properly and talk to them :-) [16:08] =)))))) [16:09] yeah, but it will print nothing for "D:/" or like separate "/home", "/var" partitions.... [16:09] "D:\" that is. [16:09] I don't think trying to say anything useful for non-system partitions is worth the effort. [16:10] But os-prober -> icon mappings (or something) would be friendly for the normal "This is your Windows boot partition, doofus". [16:11] infinity: I really don't want ooohhh i have this partition with windows and this one says nothing and i can reuse it, OMG i just formatted by D:\ with all of my Documents&Settings =(((( boo-hoo [16:11] Well, you already display labels, right? [16:12] (I'll admit, I haven't looked at this UI in a while) [16:12] If you didn't give your D:\ an NTFS label of "Data" or "Storage" or "DON'T DELETE ME, ARGH", I'm not sure we can help you. What's there to detect? [16:12] "We scanned all the files on your partition and noticed some pretty compelling porn, are you sure you want to delete that?" [16:15] xnox: In the modern world where Linux installers default to a single partition (not counting swap), and Windows is always on a single partition, anyone who has multiple partitions/disks did it on purpose, so they really should know. [16:15] xnox: But my parents would probably find "Windows is installed here, no deletey" pretty handy. [16:16] Saying "data" or "NTFS" or something is better than looking like it's empty. [16:16] oem preinstalls with recovery partitions, excluded. [16:16] true. [16:17] Yeah, you have recovery partitions, separate Windows boot partitions sometimes, EFI specials, all sorts of little oddities [16:17] * xnox over christmas "Dad, why did you wipe all data from your external hard disk?" "it asked me something & I click ok. There was no "D-E-L-E-T-E" in the prompt. It said something about "F-O-R-M-A-T" which is like defrag, right?!" [16:19] Heh. [16:20] Don't recovery partitions have a special type anyway? [16:20] I vaguely recall they show differently in the Windows partitioner. [16:21] cjwatson, I tested this morning's raring server iso for that 4k sector problem. Still crashes in parted_server. The right dosftstools deb is in the build, but, not the udeb...do I need the udeb, too? bug 1065281 [16:21] Launchpad bug 1065281 in partman-basicfilesystems (Ubuntu Quantal) "Installer crashed when trying to partition 4k/4k sector hard disks" [High,Triaged] https://launchpad.net/bugs/1065281 [16:21] Or maybe they just get the hidden flag set, usually. [16:21] they do have magic files on them for windows to treat them "hidden" (that's what I saw), but it was just a normal NTFS/FAT partitions. But that's from ~XP era. [16:21] kentb: Test this afternoon's, please :) [16:21] cjwatson, doh! ok. will do :) [16:21] kentb: I rebuilt recently for exactly that problem [16:21] And yes, you need the udeb [16:21] cjwatson, ok. cool [16:22] xnox: Are you sure they're not type 17 instead of type 07? [16:22] It went missing due to an LP snafu [16:22] I was going to wait for the respin before asking you to retest ... [16:22] infinity: that was like more than 8 years ago, I had to deal with that..... i don't have my parted logs handy. [16:22] Anyhow, this all sounds like stuff that os-prober partially knows and could learn to be much smarter about. [16:23] treu. [16:23] Anyway, we don't have to do Kylin's work for them - get back to them, lay out the problems and invite them to do better [16:23] Duplicating the logic elsewhere would be madness. [16:23] ack. [17:46] kentb: Thanks. Nearly out of time for today - will probably have to return to it on Monday [17:48] kentb: I think I may need the same trick with a desktop installation in order to get a traceback [17:49] kentb: But you'd need to apply http://paste.ubuntu.com/1546175/ to /lib/partman/commit.d/50format_basicfilesystems by hand first [17:50] cjwatson, ok. np. [17:50] cjwatson, thanks for your help and have a good weekend [17:50] Ah, hmm, I thought mkdosfs would autodetect logical sector size but it doesn't look like it [17:51] That might explain it [17:57] http://paste.ubuntu.com/1546203/ would help, but I think I'll leave that for now and just do it with blockdev [17:58] cjwatson, ok. I'm pulling down the desktop iso now and will apply the basicfilesystems mods and see what I get [18:00] kentb: actually, before that [18:01] kentb: Would you mind trying a repeat of the server installation, but before you get to the partitioner, switch to tty2 and use nano to apply http://paste.ubuntu.com/1546211/ to /lib/partman/commit.d/50format_basicfilesystems? [18:01] then save and exit, return to tty1, and continue [18:01] cjwatson, ok. I'll try that right now. [18:02] great [18:03] * kentb whistles while the server reboots [18:03] cjwatson: I looked into bug 1093819 and did not find /var/lib/dpkg/arch in the image [18:04] Launchpad bug 1093819 in Wubi "Wubi installed 12.10 amd64 without configuring i386" [Undecided,New] https://launchpad.net/bugs/1093819 [18:05] how strange, I'll have to look into that (next week) [18:15] Oh, hrm. That could be because the dpkg postinst only adds i386 as an arch on fresh installs, but dpkg is never a fresh install. [18:15] (Since debootstrap unpacks it without running postinsts, then reinstalls it, IIRC) [18:15] We paper over that in both d-i and ubiquity by explicitly adding the arch, I believe. [18:17] Assuming my theory about debootstrap is right, I can't think of a sane way to fix that other than having wubi also do the requisite papering over. [18:17] ah, yeah, or have live-build do it [18:17] Or live-build, sure. [18:17] live-build probably makes more sense. [18:17] makes the livefs bigger due to the extra Packages files [18:17] Though, if... [18:17] so probably better not do it across the board [18:17] Was just going to say that. [18:17] maybe do it in livecd-rootfs just for wubi? [18:18] Could do it as the step right after the final apt-get update. [18:18] cjwatson, those modifications didn't make a difference :( [18:18] cjwatson, so, I'll continue with the desktop iso and get trace [18:18] kentb: can I have the new syslog? [18:18] cjwatson, sure. [18:19] infinity: right before ... [18:19] infinity: we don't want to do it after for wubi, because then people don't have i386 indexes until they update [18:19] wubi is a special case here due to the ghost-y nature of its installation [18:20] cjwatson: Oh, I meant in the general case, but sure, could just do it very late for wubi-only. [18:20] (And the apt-get update again) [18:20] s/the/then/ [18:22] cjwatson: Conversation above saved in the bug for posterity (and so one of us remembers later :P) [18:22] ta [18:23] cjwatson, attached to bug [18:27] hmm, I really wasn't expecting that ... [18:27] kentb: do you still have it booted? [18:28] Oh [18:28] I'm an idiot [18:28] partman-efi is the bit that's actually crashing [18:29] kentb: OK, so never mind about the desktop CD, let me just go round and upload some more stuff and you'll get something a bit more sensible in the next build [18:29] cjwatson, ok sounds good [18:29] kentb: Can you just run two commands for me to confirm ... [18:29] cjwatson, sure [18:29] cat /var/lib/partman/devices/=dev=sdb/device [18:29] blockdev --getss /dev/sdb [18:31] cjwatson, first command outputs /dev/sdb~ second command outputs 4096 [18:31] excellent [18:32] the ~ there is just an artifact of that file not having a newline at the end [18:42] kentb: Just as well you didn't go through the code path I thought you did. My previous paste was dangerously incorrect [18:44] cjwatson, ah! ok. got it. [18:54] kentb: OK, uploaded another set of changes - though the number of brown-paper-bag uploads I've made just now suggest that I need to call it EOW [18:55] in any case the next server build should be worth testing against [19:07] cjwatson: We can just get more bags and expense them. [19:18] cjwatson, ok. thank you! [23:30] wallpaper "app" is borked in raring, even launching it directry via xinit doesn't paint, yet gnome-settings-daemon (which we are running) can paint the background fine. [23:31] but gnome-settings-daemon is not available on flavours and may remove desktop painting code in the future. [23:31] poking lightdm to see if it's possible to fix wallpaper "app"