[12:05] should work fine for me [12:05] it's nothing extremely urgent.. so if we can't make it, tough, we can always reschedule it [12:06] good night Kamion :) [12:06] sure, night :) [08:02] bah [08:03] it looks like that libselinux in lvm2-udeb is not lvm2 fault [08:03] there is something wrong with the build environment === pitti [n=pitti@ubuntu/member/pitti] has joined #ubuntu-boot === sivang wonders if this channel shares any of the properties of #debian-boot :) [08:32] Well, it has -boot in the name. [08:32] yeah but here there are only a bunch of bastards [08:32] :) [08:33] hehe [08:33] well, it's probably specially created for speeding up the boot process? [08:33] err, [08:33] thqat is for discussing that stuff [08:46] Kamion: ok the problem is way more complicated than it did look in the beginning [08:46] lvm2-udeb manages to link to libselinux via libdevmapper [08:46] even if --disable-selinux is forced [08:46] so libselinux is pulled in indirectly [08:47] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=315505 <- [08:47] but they didn't realize that device mapper will still break the world [09:57] Kamion: ok.. i got lvm2 back to normal.. just waiting for the packages/dep-wait & co to propagate [10:06] sivang: #debian-boot's name is historical, from boot-floppies; it should really be #debian-installer now, but we can't be bothered to move [10:07] Kamion: ofcourse :) [10:07] we created this channel primarily to coordinate the move to 2.6.15 and the new udev and co. for dapper; no doubt it will be used for other related things [10:11] (on unrelated ones) [10:13] Side note: The only thing worse than not being able to make something work for hours on end is MAKING IT WORK, BUT NOT BEING SURE HOW. [10:13] I think Ineed a brain-break. === Mithrandir break infty's brane [10:17] s/break/&s/ [10:37] infinity: indeed. [11:10] yeah [11:10] good new steps forward [11:11] Kamion: pal is still broken on ppc :).. but at least it seems that manual LVM works.. so that makes my life still very simple. I think i am hitting a very stupid problem but i will work on it after the dentist [11:49] so, I suck. This won't work, since cdrom-checker complains about not finding the cdrom device. [11:51] oh yeah, you'll need to make cdrom-detect run first [11:53] * Make ext2=m on all arch's. [11:53] oh great, more d-i changes ... [11:54] hmm, should I add a hack to cdrom-chooser for that, or should I wrap it in a small shell script which just does udpkg --configure cdrom-detect ; udpkg --configure cdrom-checker? [11:56] cdrom-chooser depends on cdrom-detect ... [11:56] so perhaps if your menu hook is udpkg --configure cdrom-checker rather than running /var/lib/dpkg/info/cdrom-checker.postinst directly, it'll work? [11:56] s/chooser/checker/ above [11:57] I'll try. [11:58] (can't remember if udpkg --configure follows dependencies) [11:58] it does, iirc [11:59] hm, doesn't seem to [12:00] maybe udpkg --configure cdrom-checker cdrom-detect [12:00] er, other way round [12:01] if DODEPENDS is #defined, it should [12:04] which is not default [12:04] ./configure --without-depends --without-debug \ [12:04] --without-remove --with-admindir=/var/lib/dpkg [12:04] only for --install [12:04] not --configure [12:04] (in the current code) [12:05] yeah, and it's compiled out anyway, so.. [12:07] "export: 5: --configre: bad variable name" [12:07] configure, even [12:08] quoting? [12:08] yeah, in rootskel [12:08] I'm sure joeyh is going to love me for first breaking rootskel, then uploading it twice in a day. [12:09] oh, boot faster, xoog [12:20] heh, debconf complains about --configure. I think I'll just write a tiny wrapper in cdrom-checker instead. [12:36] ok [01:17] gnnr, I just get a black screen with a single grey line on it. [01:18] hw-detect seems to be hanging. [01:23] Kamion: any great ideas? Running it by hand works, but well, is kinda wrong. [01:42] I'd boot with BOOT_DEBUG=3, use the "before init" debugshell to replace /sbin/debian-installer in /etc/inittab with /bin/sh, and then run through it step by step [01:43] maybe debconf isn't connected up right [01:44] it blows up when I run debconf -o d-i /bin/cdrom-detect-menu (which just runs udpkg --configure cdrom-checker cdrom-detect) [01:44] even from tty2 === jbailey [n=jbailey@modemcable139.249-203-24.mc.videotron.ca] has joined #ubuntu-boot [01:47] you'd want UDPKG_QUIET=y in cdrom-detect-menu at least [01:47] (perhaps the witter is confusing debconf) [01:48] yeah, I'll try that [01:49] I dunno, in general I'd be trying to imitate what main-menu does [01:50] hmm, you might also like to look at the setup/* scripts in kickseed; they're run using MENU [01:50] export UDPKG_QUIET=y [01:50] udpkg --force-configure --configure "$1" 2>&1 | logger -t kickseed [01:50] that could become log-output these days, but otherwise that works in a MENU in kickseed [01:50] # without this, debconf clients will talk debconf protocol to syslog [01:50] . /usr/share/debconf/confmodule [01:51] hah, right [01:51] Mithrandir: source the confmodule at the top of cdrom-detect-menu; that's probably your problem [01:51] doh. [01:51] I knew I'd solved this problem before ;-) [02:04] Kamion: I'll buy you a beer on Sunday. === pitti [n=pitti@195.227.105.180] has joined #ubuntu-boot [02:05] Sunday? [02:12] you're around Cambridge, aren't you? [02:12] Karianne and I are in London from Saturday until Wednesday, with a small detour to Cambridge on Sunday. === pitti_ [n=pitti@195.227.105.180] has joined #ubuntu-boot [02:14] oh, right, cool [02:14] I'll check with Kirsten what we're doing on Sunday; I'd forgotten about your visit :) [02:14] I mailed debian-uk about it last afternoon [02:14] iirc [02:15] no idea if it managed to traverse sauce, though [02:16] 6 Tollef Fog Heen 1 [Debian-uk] Beer in Cambridge Sunday 2005-11-04 [02:16] ^-- trn on chiark [02:16] goodie [02:36] Kamion: hmm, main in cdrom-checker seems to include crack such as system("touch /tmp/test") system("logger init"), etc. I'll just get rid of those if you don't mind. [02:38] sure :) [02:38] Anyone here bored and willing to take the grub merge from me? [02:38] I suspect it was a very early udeb [02:38] jbailey: I'll grab it [02:38] -- Martin Sjogren Sat, 22 Feb 2003 00:40:06 +0100 [02:38] not bored as *such*, but I've merged grub before [02:39] Kamion: Thanks. My bit of distro time left is best spent finishing the glibc bits. [02:58] so, it works, except reboots don't. [03:09] jbailey : Will that include the creation of libc6-dev-i386, or -ENOTIME for that? [03:11] infinity: I started looking at that last night. The next upload won't have it. I want that to be primarily things that are really needed for the locales-in-langpacks stuff. [03:11] infinity: We're aiming to have that all in for flight-2. [03:12] infinity: Do you have a more urgent need for tha libc6-dev-i386 stuff? [03:29] Nope. I couldn't care less if it happens soon, or even by release, it's just a wart I'd like to see go away "sometime" [04:11] yo [04:11] ok [04:12] right [04:12] i am having a bunch of problems with p-a-l and i need some help/advices on how to fix them [04:12] if you can get the code handy, it might help [04:12] ok; note I merged partman-auto last night [04:12] yes i saw that [04:12] but there shouldn't have been much relevant in there [04:13] i need to retest on i386 because ppc is farting on me [04:13] and i can't see why [04:13] but that's not the main issue for now [04:13] i will figure it later on [04:13] the first problem i wanted to talk with you was the way to restart parted [04:14] yeah, I noticed you disabled my change there [04:14] you did change the hugly code i did copy from somewhere [04:14] to something like: [04:14] stop_parted [04:14] start_partsomething [04:14] but i have issues with that [04:14] but i can't understand why [04:14] perhaps you can help me to shed some light on it [04:15] I'm just hunting down the svn changeset [04:15] if you have a test machine that you can trash [04:15] sure [04:15] Kamion: how can I get a status message from cdrom-checker through udpkg to cdrom-checker-menu? Any ideas? [04:15] no test machine handy unfortunately [04:15] you can see it almost immediatly becuase you run pal [04:15] and the visuals are all wrong [04:15] Mithrandir: exit status, ideally ... [04:15] all wrong? [04:15] while using my restart code looks ok [04:16] if you are familiar with LVM a bit [04:16] and you have seen how partitions show up at partman? main menu [04:16] you will see that with your code the partitions do not show up the same way as when you do manual LVM [04:16] that's first "optical" symptom [04:17] and if you ask to commit the changes, it will fail [04:17] returning back to the main menu, everything looks ok again (like if parted is restarted the same way i do) [04:17] and a second commit will work [04:17] well, that's bizarre, because the function calls I replaced that code with contain identical code [04:18] ah [04:18] it's restarting in precisely the same way - the code's just in a common place now [04:18] hmmmm [04:18] if you can see a difference there, I'm all ears - I've got the diff in front of me and can't find anything [04:18] do you remember from what pkg is coming? [04:18] partman-base/definitions.sh [04:18] the common code i mean [04:19] hmm, unless 'exit 255' behaves differently in a function [04:19] eh [04:19] that could be [04:19] in a function you should return [04:19] seems unlikely; I'll test [04:19] i can test it [04:19] you can exit from a function [04:19] don [04:19] don't worry [04:19] i am only seeking ideas === jbailey [n=jbailey@modemcable139.249-203-24.mc.videotron.ca] has joined #ubuntu-boot [04:20] ~ $ foo () { echo; exit 255; echo; } [04:20] ~ $ foo [04:20] $ echo $? [04:20] 255 [04:20] seems to work as desired [04:20] also in dash or the busybox built-in? [04:20] fabbione: to be honest my first resort would be to stick 'set -x' at the top of perform_recipe_by_lvm and see what it's doing ... [04:20] fabbione: that's in busybox sh [04:20] yes i can stick set -x but where the hell does the output go? [04:20] /var/log/syslog [04:21] ok [04:21] perfect [04:21] it's not amazingly conveniently formatted but it should work [04:21] i am ok with that [04:21] given that it works :) [04:21] you can probably 'exec 2>/tmp/randomfile' if you like [04:21] i am happy if the out goes somewhere [04:21] i can figure the formatting [04:22] ok [04:22] the next problem is: [04:22] # check if the recipe contains lvmok tags otherwise fail [04:22] if [ -z "$(echo "$scheme" | grep "lvmok")" ] ; then [04:22] XXXX: how do we bailout properly? [04:22] exit 0 [04:22] fi [04:22] let say that we are asked to perform a recipe on arch foo [04:22] where arch foo has recipes that contains no lvmok tags [04:22] i need to bailout [04:23] what's the proper way to do it? [04:23] at that point in time i didn't write anything to the disk [04:23] so it's easy [04:23] no actually [04:23] i already trashed the disk [04:23] but i need to bailout [04:24] you're supposed to do that check before trashing the disk! [04:24] ok i can do that [04:24] that's not an issue [04:24] but i need to know how to bailout properly [04:24] i don't think exit 0 is the way it's meant to be [04:25] make automatically_partition/some_device_lvm/choices only print the LVM choice if it can actually do it [04:25] see resize_use_free, which does that sort of thing [04:26] the problem is that we don't know the recipe yet at that time [04:26] do we? [04:26] choices is right before do_option ... [04:26] so we already have the recipe.. === jbailey [n=jbailey@modemcable139.249-203-24.mc.videotron.ca] has joined #ubuntu-boot [04:26] if so ok... [04:26] they're both run from ask_user, and there's no recipe selection in between [04:26] no, you probably don't [04:26] hmm, maybe [04:27] let me hunt around [04:27] ok [04:27] your script chooses the recipe itself [04:27] so there's no issue [04:27] choose_recipe "$free_size" || exit $? [04:28] that's from do_option [04:29] to hide the selecion, it needs to be done from choices [04:29] ah, I see what you mean, you don't want to ask questions from choices [04:29] well in that case just do the lvmok check before trashing the disk [04:30] yes and i am ok with doing that check... [04:30] all that logic is in your do_option so you can easily move it up [04:30] exit 1 should be fine to bail out [04:30] ok [04:30] perfect :) [04:30] exit 1 is all i was searching for ;) [04:30] you should probably db_input/db_go an error template of some kind first [04:30] that would make sense [04:31] look at perform_recipe, which does that [04:31] i guess we have some kind of standard error template [04:31] ok [04:31] i will [04:31] or in the choose_recipe function for that matter [04:31] ok [04:31] i can still move code up [04:31] hm, no, you'll probably need a new one since the text will need translated [04:31] i have no issues for that [04:32] the last problem i would like to address are unclean disks [04:32] my advice from bitter experience is to get new template text upstream as soon as possible, otherwise merging is HELL [04:32] ok. i want to finish pal no later than this week [04:32] modulo bugs [04:32] because i have other tons of stuff to do as well [04:33] sure [04:33] so let say you install breezy/dapper or whatever on lvm [04:33] you decide to reinstall dapper on it using pal [04:33] there is a big problem with LVM that finds the old install [04:33] pal in that case is dumb [04:33] it doesn't fail [04:33] it doesn't clean [04:34] the result can be a mess [04:34] so i was wondering what kind of approach i should use to that [04:34] like: [04:34] - send big fat warning to dd the disk before attempting [04:34] - try to cleanup automatically [04:34] - do it and hpe evertyhing will work right afterwards [04:34] my advice would be to just not offer it [04:35] ok [04:35] well, I guess you do want to be able to wipe the disk === fabbione didn't think about it [04:35] well the point is exactly that one [04:35] there is no clean way to wipe a disk that had LVM other dd or a bunch of lvm commands that might fail [04:36] really? what goes wrong? [04:36] you do a NEW_LABEL ... [04:36] the LVM metadata are all around the disk [04:36] lvm doesn't care about the LABEL data [04:36] i did try that.. trust me [04:36] hitted my head on it for 4 hours [04:36] is there a command to wipe that? I thought there was [04:36] I remember we discussed this before and found one [04:37] you need to do a set of commands [04:37] lvremove -> vgremove -> pvremove [04:37] then you can add the new LABEL [04:37] that was for the RAID [04:37] ah [04:37] not for lvm [04:37] RAID has metadata in one place only for each disk [04:37] lvm is all around the disk [04:37] sort of inodes [04:38] surely it won't look through the whole disk if it doesn't find its superblock-equivalent though [04:38] of course not [04:38] so dd over that :) [04:38] dd means dd the entire disk [04:38] you don't know where the inodes are [04:38] not all of them at least [04:38] but yeah.... dd would be the best :) [04:39] can we use pvremove --force? [04:39] sorry, -ff [04:39] /* we must have -ff to overwrite a non orphan */ [04:39] that will kill the metadata for the partition [04:40] or indeed -ffy - should be a big enough hammer [04:40] i can try that.. yes [04:40] i mean i can test it easily [04:40] right, so if we do that for each partition, then NEW_LABEL, then autopartition, that should probably work [04:40] given i need to cleanup my disk each time i try pal [04:40] that should be sufficient i assume [04:40] yeah, I think s [04:40] so [04:41] ok [04:41] i guess i will start with these stuff [04:41] do we already have a big "you are about to WIPE your DISK!" warning? [04:41] hmmmmm [04:41] i don't think so [04:41] probably partman-auto does [04:41] i don't [04:42] ah, you do, but only on sun disklabels because they need to be committed [04:42] "it sucks to be you if you choose LVM and it didn't work" [04:42] I think you should probably just remove the two label_type = sun conditionals [04:42] that's code i did copy from autopartitioning i think [04:42] or from somewhere inside partman-auto [04:42] why should i? [04:42] right, but you're now becoming a different case because you're doing non-undoable stuff to the disk [04:42] autopartition is only doing stuff that partman can unwind [04:43] right [04:43] up until you actually accept the changes [04:43] true [04:43] ideally p-a-l would be the same, but that sounds non-trivial [04:43] you can't really... [04:43] well, actually, where does stuff go wrong? at the autopartition stage, or later? [04:43] it depends... [04:43] as it is now it works [04:43] if it only goes wrong later, then you could do the pvremove -ffy in a commit.d script [04:44] and then you can preserve the ability to undo changes, which is really nice to have [04:44] yeah but you won't be able to unwind the main partitions [04:44] why not? [04:44] do you actually write them to disk in autopartition? [04:44] because you need to COMMIT to access the partition that will be lvm [04:44] oh, ok [04:44] if you see also manual LVM, it does exactly the same [04:44] it asks you to commit before accessing the LVM menu [04:44] and it makes sense [04:44] right, just pvremove -ffy up front then, and display the warnings unconditionally [04:45] yeps [04:45] if it fails.. well sucks to be you [04:45] buty [04:45] better partman/LVM integration would involve having a representation of LVM partitions in /var/lib/partman/devices/, so that you wouldn't need to commit beforehand [04:45] yeah that's true [04:45] but there are also other problems [04:45] like roundings [04:46] all sizes will be highly incorrect [04:46] but well.. minor details [04:46] mm [04:46] (sorry i need to wake up my wife) [04:47] re [04:48] ok i think for now i am all set [04:48] at least... [04:48] i think i am :) [04:48] thanks a lot Kamion [04:48] it's really nice to work with you [04:50] hope it all works :) [04:50] oh don't worry [04:50] i will come back if it doesnt :P [04:50] heh [04:50] right now i want to get the code to be in a decent state and work [04:51] the next cleanup step will be merging and removing all the duplicate with partman-auto [04:51] because there is really a tons [04:51] that doesn't need to be there [04:51] but for now it's easier for me so that i can fork and merge as i need without disrupting all of partman-auto [04:53] have you got upstream commit access yet? you probably should [04:53] ask joeyh [04:56] Kamion : Will d-i be using the same udev firmware-loading mechanism as the base system? (ie: should I be installing udeb firmware to the same filesystem location?) [05:00] infinity: yeah, more or less the same mechanism anyway [05:00] Kamion: no i don't and i would like not to. I will push the patches as they come [05:00] i am keeping the interdiffs [05:07] pvremove -ffy seems to do the trick :) [05:28] cool! === sivang [i=sivan@ubuntu/member/sivang] has left #ubuntu-boot [] === zyga [n=zyga@2-mi2-1.acn.waw.pl] has joined #ubuntu-boot === sivang [i=sivan@ubuntu/member/sivang] has joined #ubuntu-boot === Mithrandir scratches his head and wonder if he should just use a flag file instead of the return value. Not so pretty, but certainly less work, since something keeps eating the exit status. === seb128 [n=seb128@ubuntu/member/seb128] has joined #ubuntu-boot [11:26] it's so weird coming back into #debian-boot. Nothing has changed, kinda, except I'm not part of the crowd yet/any more.. I guess it'll work itself out over a few days or weeks. [11:56] Mithrandir: mostly the effect of trying to get used to the now-rather-enormous codebase again? [11:57] I do love design discussions like that one just now with joeyh though ... makes me feel like d-i is a real project ;-) [11:57] (real as in big commercial blah) [11:57] Kamion: nah, more the "I'm not sure whose feet I tread on when doing random crap" [11:58] Mithrandir: http://people.debian.org/~fjp/d-i/authors.html might help [11:59] has tausq or tbm done anything to d-i lately? [11:59] or Marvin? [12:00] (not accusing, more "are they still around?" [12:00] ) [12:00] tausq hasn't, tbm has done various bits and pieces [12:00] don't think sjogren has been around much at all [12:01] he was around _a lot_ in the early phases [12:01] yeah [12:01] tsauter's mostly vanished too [12:02] shame that. :-/