[00:10] greearb: udebs should come out in ../, not in debian/ [00:21] I wasn't building correctly, it seems...trying to build with debian/rules now..but first to port my patches to the official .24 ubuntu kernel.... [01:14] oem-config: cjwatson * r453 oem-config/debian/changelog: fix bug number typo [01:35] ubiquity: cjwatson * r2663 ubiquity/ (configure configure.ac): bump to 1.8.6 [01:40] cjwatson: I'll be committing my fixes as soon as I've found the least amount of time things can wait before orca can load. I'm doing various test runs on different machines, either from CD, or via wubi, and getting different results. At this point, it looks like 20 seconds is the best bet. [01:41] For intrepid, I'd like to see if the sleep can't be changed to something that can detect when things are ready enough for orca to load. [01:43] sleeps are certainly usually a problem [01:43] but ok, as a workaround [01:44] if we can find something better, perhaps it can qualify for 8.04.1 [01:44] how are things going for release otherwise? Have you been picking things up off the milestone/hardy lists? [01:45] there are a bunch of build failures and things that it would be good to polish off (they'll just bite us later in security updates if we don't) [01:46] Yes, although I'm stumped with that openssl upgrade bug. Other than dropping the priority of the restart question, I've no idea as to how that can be solved... bug 91814 [01:46] Launchpad bug 91814 in openssl "libssl0.9.8 config asking me 'which services should be restarted to make them use the new lbraries?'" [Medium,Confirmed] https://launchpad.net/bugs/91814 [01:46] And I am guessing that dropping the priority below what is set for Ubuntu, i.e high is not a good idea... [01:47] have you talked with mvo about it? one option would be to preseed that in the dist-upgrader [01:47] I'll do that tonight. [01:48] (as a possibility) [01:48] However I can also reproduce it using apt-get dist-upgrade on the console. [01:48] right, but console users can expect a little more interactivity [01:48] Yep, fair call. [01:49] we don't actually know that people are doing a full OS upgrade, rather than just dist-upgrade within a release or even for security updates, unless they're using the update-manager [01:49] Right. [01:49] and for dist-upgrade people aren't (and shouldn't be) told to shut down all running programs [01:49] that's my reasoning, anyway [01:49] * TheMuso nods. [01:50] in 182446, Kees says "I ran into this as well. update manager did not notice the prompt (it was only visible in the "Terminal" output). In the past, we'd altered openssl's prompting levels to auto-restart when running within update-manager. Perhaps that patch was accidentally dropped?" [01:50] I'm not sure what that alleged patch might have been, but perhaps worth looking into [01:51] Yes indeed. [02:04] TheMuso: bug 198453 is on the milestoned list, in case you haven't seen it [02:04] Launchpad bug 198453 in pulseaudio "Default ALSA device must use PulseAudio, otherwise ALSA applications may fail" [High,Confirmed] https://launchpad.net/bugs/198453 [02:08] Yes I've seen it, I'll take a closer look. [07:44] Well, I got the udebs to be created, seemingly proper. I copied them into the debian-installer localdebs and ran 'make build_cdrom_isolinux' [07:45] I then coppied the dest/cdrom/initrd.gz and vmlinuz to the livecd/cd/casper/ directory [07:45] I installed the matching kernel in the livecd filesystem chroot and rebuilt the iso [07:46] the iso boots in qemu, but the installer goes to text mode and eventually fails to make progress because it cannot find the cdrom [07:47] from the busybox shell, it seems like things are ok..the proper kernel is loaded, there are lots of modules loaded, etc. [07:47] but, it certainly isn't like the official livecd experience. [09:37] <_ruben> hmm .. development chan, also the right place for asking d-i related questions? (how to setup raid10 in particular, mdadm in shell only allows for raid0 and linear, d-i doesnt cover raid10 either) [09:39] it's the right place in theory, but since as you say d-i doesn't have raid10 support, it's hard for us to help ... [09:40] <_ruben> seems to be a slightly different issue in fact .. it keeps reverting to my old raid layout .. guess i'll have to zero out some (super)blocks or smth [09:40] you sure mdadm only allows for raid0 and linear? d-i expects it to support -l raid1 and -l raid5 [09:41] <_ruben> cjwatson: that was my bad, used build instead of create, but now im getting device in use because it somehow picked up my old configuration again [09:42] superblock zeroing sounds plausible, though also if you're running d-i make sure to exit the partitioner first! [09:43] <_ruben> my mdadm skills seems a bit rusty .. kept using wrong commands .. got me a raid10 now :) .. lets see if d-i picks up on it [09:44] d-i's own raid integration definitely doesn't support *building* raid10 arraying [09:44] er, arrays [09:45] but if you hit "Configure software RAID" in the partitioner and then immediately back out again, that should be enough to prod it into activating the devices [09:46] <_ruben> cjwatson: indeed, that's how im doing it now [09:47] <_ruben> seems to work sofar [09:50] <_ruben> another annoying thing btw (havent checked if there's a bug for it already) .. say you create a /dev/md0 for /boot and a /dev/md1 for lvm .. format /dev/md0 as ext2 and assign to /boot .. then go setup your lvm .. the reference to /boot will be gone afterwards [09:54] hmm, that's curious, I don't see why it should be doing that [09:54] let me see if I can construct a reproduction for that in kvm [09:55] <_ruben> similar behaviour on both physical and vmware installs, so i guess it should be easy to reproduce for you [09:56] not sure if I have time to fix it for hardy now even if I can reproduce it, but let's see [09:57] <_ruben> cjwatson: its not really a big issue, more a (small) annoyance ;) [10:19] ok, let's see, md0 and md1 [10:19] * cjwatson randomly opts for raid1 [10:20] <_ruben> raid10 install successful btw [10:22] cjwatson, reopened bug #188492 because of last user feedback (I did not test it myself yesterday, as was doing windows side tests) [10:22] Launchpad bug 188492 in ubiquity "console-setup/layoutcode settings are ignored" [Medium,In progress] https://launchpad.net/bugs/188492 [10:23] might also be that the wubi preseed is not adequate at this point [10:46] urgh [10:46] need a copy of the wubi preseed that's actually active (rather than a sample one) [10:49] yeah asked the chap to post all possible logs, as mentioned might be wubi side [10:49] the fact that he changed language, does not affect the keyboard layout for instance, since that depends only on the registry entry [10:49] which for germany is mapped to "de" [10:55] well, it could also be that preseeding is working just fine but ubiquity isn't propagating it to the installed system properly [10:56] any idea when the hardy installation guide is on help/doc.ubuntu.com? [10:57] I haven't prodded mdke about it yet [10:57] I can ask about that after RC [10:57] ok thanks [10:58] (remind me then?) [10:58] sure :) [11:00] _ruben: gosh, reproducible [11:12] <_ruben> cjwatson: as to be expected [11:13] well, you might be surprised about partitioner bugs [11:37] _ruben: hmm, it isn't specifically LVM, it's the process of restarting the partitioner that loses it [11:40] BLINK [11:40] partman-md/init.d/md-devices is HORRIBLE [11:44] so the problem is that when the partitioner restarts, partman-md sets up all the RAID devices from scratch [11:44] this is probably so that things work right first time round, but of course when you're using RAID+LVM the RAID device is already set up and this has the effect of trashing it [11:45] _ruben: would you file a bug on partman-md about this? Just copying and pasting this IRC discussion would be fine [11:50] <_ruben> cjwatson: figured as much [11:51] <_ruben> i'll file a bug in a bit [11:52] _ruben: link it to http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=398668 in Debian partman-md [11:52] Debian bug 398668 in partman-md "partman-md: should not recreated device dirs at each partman restart" [Normal,Open] [11:56] <_ruben> guess its ok to reuse that summary ? [11:58] sure, except recreated => recreate [11:59] <_ruben> i quoted you instead "partman-md sets up all the RAID devices from scratch" .. btw, it *might* be that partman-lvm has a similar bug .. would have to test that [12:00] <_ruben> crap .. forgot to link it to debian, lets find the button for that [12:02] <_ruben> cjwatson: i must be blind, cant see how to link against the debian bug :( [12:03] Also affects: Distribution/Package... [12:03] select Debian and partman-md, paste in URL above [12:03] <_ruben> ah [12:04] partman-lvm works a bit differently (it checks whether the LV is non-empty before initialising it), so I don't think it has the same problem [12:04] <_ruben> done .. bug #218141 [12:04] Launchpad bug 218141 in partman-md "partman-md sets up all the RAID devices from scratch" [Undecided,New] https://launchpad.net/bugs/218141 [12:04] <_ruben> cjwatson: ah, could be [12:05] obviously I might be misreading the code, but it looks OK [12:32] <_ruben> is there any way to automate an installtion based on lvm on top of raid .. neither sw raid and lvm seem supported in the various automation methods [12:36] err, there are specialised autopartitioning methods for both raid and lvm [12:36] I don't think it's possible to do the combination at present though, sorry :( [12:36] (which I regard as a major omission) [12:39] <_ruben> ic, guess i'll focus on automating the rest of the installation then (which isnt all that much for a server install, spending most my time in the partioner :P) [16:22] cjwatson, any idea on my lack-of-graphical installer problem with the respun live cd? [16:23] err, I didn't read closely before [16:23] you said you dropped a d-i initrd in and tried to use it with casper [16:23] I'm afraid that's totally wrong, sorry if I misled you - you can't use casper that way [16:23] yes, I think that's what I did [16:23] heh, oh [16:24] casper initrds are built on a regular system with the casper package installed, by running update-initramfs -u and fishing them out of /boot [16:24] have a look at the livecd-rootfs package [16:24] sorry if I wasted your time, looking back you did say that you were working on a live CD but then you mentioned udebs and I fixated on that [16:24] hrm, that is what I was doing at the very beginning..but I had troubles then too...maybe I should try again [16:25] those will be soluble troubles though :) [16:25] well, since I went through the trouble..what use is the udeb initrd & vmlinuz? [16:26] just in case I accidentally did something useful :P [16:26] it's for the alternate/server CDs [16:27] oh, ok [16:27] traditional installer rather than live CD-style [16:27] that makes sense, that seemed to be what it was trying for [16:27] is there any benefit of those over a live cd though? I can't think of any offhand... [16:28] sorry, afk for a few min...getting kid ready for school [16:29] several pros and cons [16:29] much lighter on memory, more flexible partitioning, much more flexible package installation [16:30] but more complex and you can't try it out before installing [16:30] so, we still ship both, and I expect will continue to do so [16:31] and slower [17:06] ok, I built the kernel with the debian/rules, which does the mkinitrd automatically when you apt-get install it (in the squashfs chroot file-system in my case) [17:07] It gets to the part where the slider bounces back and forth, and then it abruptly puts me into a (initramfs) prompt [17:08] Maybe I need to hand-build the initrd special? [17:15] greearb_: have a look in /casper.log to see what broke [17:16] greearb_: did you have the casper package installed in the chroot when you built the initramfs? [17:19] cjwatson: FYI: I'm running into a bug wherein partman fails on the latest daily live CDs on account of something locking /dev/sda. [17:19] not fixed by seb's gvfs upload by any chance? [17:19] thanks for the information though; let me know if you need help [17:19] was that post-20080416? [17:20] in my case I manually partitioned, created a new partition table and a single partition, but it fails as it cannot stat /dev/sda1. [17:21] running hdparm -z /dev/sda says that it's busy, so I killed all the user processes except bash, but still no dice. [17:21] it was today [17:22] cannot mount the //filesystem.squashfs: no such device [17:22] ah [17:22] hmm, ok, that doesn't sounds like a race [17:22] but, dev/loop0 exists (it seems), and so does the filesystem.squashfs [17:22] greearb_: it ought to be /casper/filesystem.squashfs, normally [17:22] er, filesystem.squashfs a directory [17:22] ! [17:23] and empty one at that :P [17:23] that sounds perhaps suboptimal [17:23] no casper/ dir at all [17:23] right, your ISO's a bit toasted then [17:23] have a look at how Ubuntu desktop CDs are laid out, is probably the easiest way [17:24] or use a hacked version of livecd-rootfs to build it (that's probably what I'd do; a little more capital outlay in setup, but much easier long-term) [17:24] funny though, the exact same iso image will boot fine if I just copy over the initrd and vmlinuz from an official kernel build [17:24] well, I don't know what it's booting if filesystem.squashfs is empty [17:24] probably your normal system [17:25] (at best) [17:25] can't be...I'm running it on a fedora 8 machine inside of qemu [17:25] as in, the CD is acting as a complicated chain-loader for the nearest operating system it can find [17:25] if there's no squashfs, then it doesn't have anything else it can reasonably boot ...? [17:26] so if that's not it then there is something weirder going on [17:26] cdrom/casper/ has the real thing it seems [17:27] 615MB squashfs file [17:31] err, right, I meant /casper within the ISO9660 filesystem [17:31] sorry for being unclear [17:31] I tried: mount /cdrom/casper/filesystem.squashfs /filesystem.squashfs [17:31] it gives: Mounting /dev/loop0 on /filesystem.squashfs failed: Invalid argument [17:32] casper should do that (better) itself [17:32] before it, it says 3 times: /dev/loop0 is write-protected, mounting read-only [17:32] you didn't forget to build loop into your kernel, did you? [17:32] or squashfs? [17:32] loop is definately there...maybe missing squashfs [17:33] (they can be modules) [17:33] does that require an outside patch, or is it there by default in 2.6.24? [17:33] external [17:33] it's in linux-ubuntu-modules-2.6.24 in Ubuntu [17:33] ahhh, bummer..that would explain it [17:34] I'll figure out how to patch that in === evand_ is now known as evand [18:31] yeah, it's definitely not fixed by that gvfs upload :/. I chrooted into /root and upgraded and it's still saying that /dev/sda is busy once I get to the desktop. Still looking. [18:37] actually, I can reproduce this reliably on the Kubuntu live CD as well. [18:53] ah, it's not swapoff'ing the swap partitions before trying to re-read the partition table. [18:55] err rather, swapoff'ing the partitions first allows it to proceed without issue [20:16] Is there a maxmium limit on the length of the late_command in a d-i preseed ? [20:24] cjwatson, evand, I have also reopened #195905 (http://ubuntuforums.org/showthread.php?t=757040) :( [20:28] ughhh, hopefully they can get those logs to us in a timely fashion. [20:31] any news on 188492? [20:32] still trying to trace the cause of another bug at the moment [20:32] I'll be home in a couple of h and give it a shot [20:34] much appreciated [22:03] cjwatson: whenever you have a moment, can you please take a look at bug 218394? [22:03] Launchpad bug 218394 in partman-base "partman fails to swapoff all swap partitions on the target device" [Undecided,New] https://launchpad.net/bugs/218394 [22:04] evand small one, progressbar still reaches 128% in ubiquity [22:05] * xivulon testing layoutcode [22:09] console setup is not correct while ubiquity runs... [22:09] not sure if it has to be set then [22:19] postinstall layout is also wrong [22:20] will reinstall in verbose mode to get better log [22:20] cjwatson pls let me know if you need any special info [22:22] evand: oh, meh, I see - the idea's definitely correct, let me recaffeinate and digest the patch [22:38] xivulon: I desperately need the preseeded information that's actually being set [22:38] xivulon: because I did test my fixes (admittedly before upload) and they were working for me [22:38] one sec [22:38] so you must be doing something subtly different, and I need to know what [22:39] http://paste.ubuntu.com/7232/ [22:40] and what's happening with that preseed file? [22:40] is it leaving the layout at us or something? [22:40] I think it is us yes [22:40] (bizarre in itself, even if it screws up the preseeding it should take the default for en_GB.UTF-8 surely, which is gb) [22:41] what does shift-3 give you? [22:41] # [22:44] ok, so not gb at any rate [22:44] what kernel parameters are you using? [22:46] kernel /ubuntu/install/boot/vmlinuz boot=casper debian-installer/custom-installation=/ubuntu/install/custom-installation iso-scan/filename=/ubuntu/install/hardy-desktop-i386.iso quiet splash ro automatic-ubiquity locale=en_GB.UTF-8 noprompt -- [22:46] i [22:47] this is my vm rig, this is slightly different from the real iron test (no locale in there) [22:52] ok, let's see [22:52] I'll leave off all the custom-installation stuff; I assume it isn't relevant here [22:52] so you said it was definitely broken while ubiquity was running too? that will make it easier to attack [22:54] oh heck, custom-installation is where the preseeding is done, ok [22:55] but it just shoves it into the initramfs [22:56] hmm, casper preseeding is really a bit broken as far as the special cases are concerned [22:57] I can use kernel boot args if that helps [22:57] I was about to ask, yes [22:57] I think it's likely to help, but I'll just do a bit of testing [22:58] I'll try to install then with locale=en_GB.UTF-8 layoutcode=gb [22:58] (that will need to be console-setup/layoutcode=gb of course) [22:59] yep [22:59] unfortunately a proper fix is either (a) big pile of ugly special-case hacks or (b) restructure casper's argument parsing [22:59] neither of which fill me with love and sparkles at T-8 days :-/ [23:01] I could use some of my hooks to hack it in via wubi/custom-installation too... [23:01] please don't :) [23:01] kernel args> much easier and simpler [23:02] if that is not enough of course [23:03] oh, god, my system is absolutely crawling while this DVD downloads, I'm going to go and do something else for a little bit [23:03] one sec [23:05] at break=bottom layout is wrong [23:05] also after run-init [23:06] kernel /ubuntu/install/boot/vmlinuz boot=casper debian-installer/custom-installation=/ubuntu/install/custom-installation iso-scan/filename=/ubuntu/install/hardy-desktop-i386.iso quiet splash ro automatic-ubiquity debian-installer/locale=en_GB.UTF-8 console-setup/layoutcode=gb noprompt -- [23:06] i [23:17] it's right in debconf but not applied. hmm [23:18] this is with a preseed file though [23:18] using a preseed file for locale and keymap is definitely busted and not easily reparable for hardy [23:19] so let's ditch that plan for now [23:19] in config.dat layoucode has value us [23:19] did you remove the bits from the preseed file? [23:20] if not I suggest you do so - this code is convoluted enough that I wouldn't want to guarantee that it behaves in a way you might expect [23:20] saying locale= rather than debian-installer/locale= should be fine for shortening purposes though [23:22] no will do so now [23:23] also locale? [23:25] hmm? [23:26] with kernel parameters, keymap is wrong in the console (and to be honest I'm not surprised, very tricky to fix that one because it's hard to switch the keymap while usplash is running), but correct in X [23:27] did you mean "also layoutcode?"? No, leave that as console-setup/layoutcode= [23:27] I normally preseed locale, layoutcode, variantcode, will comment all 3 [23:27] move all to the command line, and remove/comment them from the preseed file [23:28] the keyboard test box does the right thing in ubiquity when I set it up this way too [23:34] battery run out... [23:34] one sec [23:37] hmm ubiquity is still with wrong layout [23:42] will try to boot into Live Desktop skipping preseeding [23:43] ah wait I tested in console only in the last run... [23:44] right, the console is going to be wrong [23:44] in fact in the last 2 runs... missed your comment [23:45] the problem is that in order to set the keyboard layout you have to do so before usplash starts [23:45] which is ridiculously early in the boot process, well before casper starts [23:46] we do this for regular systems with a working /etc/default/console-setup, but not for the live CD [23:46] now, it would be possible to just special-case /proc/cmdline in the console-setup initramfs-top script [23:46] that might be the best answer [23:46] still wouldn't (and couldn't possibly) deal with preseed files for it, but actually I think that's ok [23:47] just like d-i, there are certain things you can't do in preseed files and that have to go on the kernel command line [23:48] hmm launching gedit during ubiquity and typing gives me wrong chars [23:49] what does XkbLayout in /etc/X11/xorg.conf say? [23:49] is that already in initramfs-top? [23:49] never mind initramfs-top for now [23:49] I was thinking out loud [23:49] us [23:50] meh [23:50] cat /proc/cmdline (just so I am sure)? [23:50] debconf-get console-setup/layoutcode 2>/dev/null [23:51] kernel /ubuntu/install/boot/vmlinuz boot=casper debian-installer/custom-installation=/ubuntu/install/custom-installation iso-scan/filename=/ubuntu/install/hardy-desktop-i386.iso quiet splash ro automatic-ubiquity debian-installer/locale=en_GB.UTF-8 console-setup/layoutcode=gb noprompt -- [23:51] also, is this before or after you reach the keyboard step? what exact point during ubiquity is this? [23:51] please confirm that you have removed debian-installer/locale and console-setup/* from the preseed file [23:52] config.dat is locked because ubiquity is running [23:53] yep no locale in preseed (unless it sourcing it from somewhere else) [23:53] I let ubiquity automatic run [23:53] it was in the copying files stage [23:53] ok [23:56] in config.dat locale is set correctly, but layoutcode have value us [23:56] not reliable if debconf is running of course, could be only in memory [23:56] I wonder if this is the fault of the timezone page [23:57] I noticed it incorrectly defaulted to New York even after preseeding locale=en_GB.UTF-8 (this is an old problem, been around in one form or another since dapper) [23:58] timezone appears correct in config.dat [23:59] let me try in the live cd environment