[00:00] doko: even openjdk-6 fails [00:00] bdrung, file a report, even with patches [00:01] and fixing eclipse issues as well .... ;-P [00:03] slangasek, I'm working on spice MIR right now [00:03] well, not RIGHT now. it's dinner time [00:05] slangasek, MIR for spice was not feasible for oneiric cycle. qemu-kvm-spice was. Now, it's unfortunate that in the end we missed oneiric after all :) [00:07] doko: first step done: bug #879167 [00:08] Launchpad bug 879167 in openjdk-6 (Ubuntu) "Missing or broken multiarch support" [Undecided,New] https://launchpad.net/bugs/879167 [00:08] doko: next step for me: go to bed :p [00:10] doko: btw, should openjdk multiarch already work in oneiric? [00:11] bdrung, yes with 7, no with 6 [00:19] doko: doesn't work too: bug #879170 [00:19] Launchpad bug 879170 in openjdk-7 (Ubuntu) "Missing or broken multiarch support " [Undecided,New] https://launchpad.net/bugs/879170 === jdong- is now known as jdong === dendro-afk is now known as dendrobates [00:33] hallyn: ah. Do you know if qemu-spice could reasonably be built from the qemu-linaro branch? [00:35] i think so [00:35] its in universe right? [00:37] yes [00:37] i figured you were pinging me on lvm+udev :) [00:37] so maybe that would be better than having Yet Another Build? [00:37] nah, I was leaving lvm+udev for another day :) [00:37] well it is a different tree... [00:38] and we'd need a separate configure+make in any case [00:38] you saw the change proposed on the bug? I don't know what --noudevsync does [00:38] sure, but keeping it in a single source package means one fewer source package to be uploaded for each upstream release [00:38] true. [00:39] i don't know that the people who want spice woudl care much about it being qemu-based instead of qemu-kvm-based [00:39] so it might be doable [00:39] i personally dont' much care, apart from these poor folks being able to use spice at last :) [00:39] OTOH, if you're doing an MIR for spice anyway, maybe we should wait for the outcome of that before changing the packaging [00:40] but if we don't get spice in main, I'd like to pull this into qemu-linaro [00:40] ok, that sounds like a good plan [00:40] thanks :) [00:40] thank you! [00:41] i'll look into EXACTLY what --noudevsync brings and comment later [00:41] i fear it will just hide boot problems [00:41] but you need the kvm built to have the spice stuff enabled? [00:41] penguin42, kvm would be enabled. [00:42] the qemu and qemu-kvm trees are pretty darned close these days [00:42] minor differences wrt io threads and such [00:42] * penguin42 shudders [00:42] at what? [00:42] io threads :-) [00:42] heh [00:42] aio threads i should say [00:53] grr, yeah i guess bug 878162 showed me the dependency chain for spice MIR. That is NOT pretty :) [00:53] Launchpad bug 878162 in qemu-kvm (Ubuntu) "[MIR] qemu-kvm pulls packages from universe" [High,Fix released] https://launchpad.net/bugs/878162 === Guest56801 is now known as shirgall === shirgall is now known as Guest65598 === dendrobates is now known as dendro-afk === Pici is now known as Guest59733 === Guest59733 is now known as Pici === plars is now known as plars-afk === zyga is now known as zyga-afk === shirgall_ is now known as shirgall [04:15] Good morning [04:15] stgraber: pong [04:35] lool: can you please commit your udev changes to lp:~ubuntu-core-dev/ubuntu/precise/udev/ubuntu ? (there's no diff in LP yet) [05:45] good morning [06:08] pitti: hi [06:08] we were talking some time ago about updating g-i and pygobject in debian [06:09] and seb128 was mentioning, that you would take care of that [06:09] (for exp, ie) [06:10] hey mbiebl [06:11] mbiebl: I'm happy to do that, yes; but we really need a newer glib [06:11] 2.30.1 landed in exp some days ago [06:12] I assume that should be sufficient [06:12] oh, nice! [06:12] mbiebl: I'll update the packages today then [06:12] great, thanks [06:13] this way we can stage some of the 3.2 changes in exp until the current GNOME 3 transition is ongoing [06:20] slangasek: committed your udisks multi-arch fix to debian git, so we can sync again next time. thanks! [06:21] ah, I see your deiban bug, thanks [06:56] pitti: Done, thanks; odd that I don't see it in the package import queue [06:56] albeit, import-dsc didn't work because the upstream tag was "173" instead of "upstream-173" [06:56] LP has trouble generating udev diffs, because of the test/sys stuff [06:56] lool: this isn't an UDD branch [06:56] it's derived from the upstream trunk bzr import [06:56] oh right [06:56] i. e. a manual one [06:57] I guess we could replace the UDD branch with it, but I don't work on udev often enough that I should push for this myself [07:02] good morning [07:22] mvo: http://paste.ubuntu.com/714898/ <--- Does this sound familiar at all? (from #ubuntu-installer last night) [07:26] soren: no, sound like it failed to download the signature file for some reason? is this a common failure or was it a one-off thing? [07:26] mvo: That was two people reporting it. [07:26] mvo: And hour and a half apart. [07:27] mvo: So at least a two-off thing :) [07:32] soren: hm, hm, thanks! I wonder if maybe one of the archive.ubuntu.com servers was/is out of sync, I est now [07:35] soren: thanks, that is fixed onw [07:40] mvo: \o/ Thanks. [07:42] thank you! [09:53] slangasek: ping [09:58] cjwatson: ping [10:03] cjwatson, slangasek: i've filed https://bugs.launchpad.net/ubuntu/+source/apt/+bug/879324 for this supposed multiarch-misfeature of apt [10:04] Ubuntu bug 879324 in apt (Ubuntu) "apt-get dselect-upgrade prefers multiarch over native" [Undecided,New] === dholbach_ is now known as dholbach [10:06] cjwatson, slangasek: (the issue we talked about yeasterday) [10:07] xtian: ok, I was just helping you narrow it down, I doubt I'll be fixing it ;-) [10:07] so no need to ping me [10:07] but thanks for filing that [10:08] I can confirm joe is currently in sync between amd64 and i386 [10:08] cjwatson: ok thanks anyway :) [10:08] although it was changed relatively recently [10:09] cjwatson: yeah, depends on the mirror you're using. i tried with german mirrors today, but could not reproduce, since these mirrors caught up overnight [10:09] there was a period of about two hours on Monday when amd64 and i386 had different versions of joe in precise [10:09] cjwatson: that's why i picked a mirror with a reasonable lag for reproducing :) [10:11] cjwatson: interesting detail maybe, the misfeature shows with 'apt-get dselect-upgrade' only, not with 'apt-get upgrade' or 'apt-get install' [10:12] (it's all there in the report, detailed console logs of what i did included) [10:12] ...moving on... [10:15] yep, that definitely looks like a good way to narrow it down; I'll mark it Confirmed [10:15] or actually triaged === MacSlow is now known as MacSlow|lunch === doko_ is now known as doko === MacSlow|lunch is now known as MacSlow [12:21] !pilot in [12:21] argh [12:21] @pilot in === udevbot_ changed the topic of #ubuntu-devel to: Precise open for uploads | Ubuntu 11.10 Released! | Development of Ubuntu (not support, not app development) | build failures -> http://bit.ly/or6CHJ | #ubuntu for support and general discussion for hardy -> oneiric | #ubuntu-app-devel for application development on Ubuntu | http://wiki.ubuntu.com/UbuntuDevelopment | See #ubuntu-bugs for http://bit.ly/lv8soi | Patch Pilots: mdeslaur [12:23] heh [12:23] mdeslaur: TGIF? :) [12:23] nigelb: hehe :) === txwikinger2 is now known as txwikinger [12:55] * dholbach hugs mdeslaur [12:55] * mdeslaur hugs dholbach back [13:02] mvo: hi! [13:02] mvo: I installed a couple of maverick machines yesterday, and the "upgrade to new release" window that popped up points to an invalid URL: http://archive.ubuntu.com/ubuntu/dists/natty-proposed/main/dist-upgrader-all/0.150.2/ReleaseAnnouncement.html [13:03] mvo: are you aware of that? I also seem to recall someone saying the tarball couldn't be downloaded to upgrade maverick also [13:03] mdeslaur: Fixed earlier today. [13:03] soren: ah, cool. thanks! [13:03] mvo: never mind :) [13:03] mdeslaur: http://paste.ubuntu.com/715103/ [13:04] hmm...the web page still doesn't seem to be there though [13:04] unless the url got changed with whatever mvo changed [13:06] 0.150.2 has been garbage-collected; the current version is 0.150.5 [13:06] shouldn't it be using current anyway? [13:06] pitti: that ping was for yesterday's TB meeting :) [13:06] unless this is a version-specific tarball that's being downloaded, but then why is it an out-of-date version ... [13:06] stgraber: erk, I wasn't at home last night; sorry, I thought it was next week [13:07] cjwatson: another question that pops to mind is why the archive is hosting a .html file... doesn't look like the proper place to me [13:07] cjwatson: this was a fresh installation from cd media [13:07] without updates installed [13:07] hm, changelogs.ubuntu.com lists current [13:07] but that may be a matter of (subjective) opinion [13:07] the archive's a reasonable place in this case, given that it comes from a custom upload [13:08] mdeslaur: is it possible that http://changelogs.ubuntu.com/meta-release has been cached? [13:08] mdeslaur: force-refreshing that might help [13:08] well it feels wrong to me. I'm used to archive.ubuntu.com and ftp.debian.org only hosting .deb's, sources, Packages etc... [13:09] but again, subjective opinion :) [13:09] Chipzz: installer kernels, initrds, package index translations, ... [13:09] (to list a few other things that live in custom uploads) [13:09] cjwatson: I don't know where it would get cached...there's nothing on my network that would cache it, and this was two fresh installs, immediately after reboot [13:09] mdeslaur: "transparent" proxy at your ISP [13:09] possibly [13:09] cjwatson: right, but those are all still closely related, and binary data, instead of a document [13:09] hope not :) [13:10] mdeslaur: open that URL in a browser, look for the ReleaseNotes/ReleaseNotesHtml entries in the Dist: maverick block [13:10] er, Dist: natty I mean [13:10] cjwatson: for example, http://www.ubuntu.com/download/ubuntu/download isn't hosted on a.u.c either [13:11] cjwatson: looks fine here...it's the "current" url [13:11] Chipzz: *shrug* this doesn't cause any harm and it arises from the structure of the software involved, I don't see that it's worth rearchitecting [13:11] but I'll shut up and keep my opinion to myself :) [13:11] I wonder where it got the versioned url from [13:11] the current symlink points to 0.150.5 [13:11] TheMuso, pitti: not sure who handles audio stuff in general, please see bug 879434 [13:11] mdeslaur: have you done a test install today? [13:11] Launchpad bug 879434 in liboggz (Ubuntu Precise) "[MIR] mutagen introduces b-d's on faad2 and liboggz" [High,Confirmed] https://launchpad.net/bugs/879434 [13:11] cjwatson: I'll try a fresh install now to see [13:12] mdeslaur: because that file was last modified this morning [13:12] mdeslaur: so I suspect mvo edited it during the conversation soren pasted above [13:12] cjwatson: oh, d'uh...sorry, my installs were from yesterday [13:12] right, so I think it was fixed today [13:12] ok, cool. thanks cjwatson [13:15] sigh, I really must write an actual germinate test suite; this is tedious [13:17] anyone know why sudo forks and not execs ? [13:17] ie, 'sudo sleep 1m' [13:17] then i'll have 2 processes running, the sudo and the sleep. [13:18] why wouldnt sudo exec sleep? [13:18] To keep track of the pam session. [13:19] If you exec, there's nothing left to close the pam session. [13:19] smoser: ^ [13:19] that makes sense. [13:19] i was looking at rabbitmq-server, which on start up has like 8 differen't sh -c hanging around [13:19] (ok, 3) [13:20] and was trying to get rid of some of them, but i can't get rid of the 'su' [13:20] thats why i was asking [13:20] Yeah, we have the same problem in OpenStack. [13:20] smoser: are you working on the rabitmq-server remove bug? [13:20] The ideal solution (AFAICS) would be if upstart would support setuid'ing on its own. [13:21] Daviey, well, sort of. [13:21] For now, we have to resort to su which does the same. [13:21] smoser: surely that is a binary question? :) [13:21] well, i was aware of that bug. [13:22] smoser: handy, because you raised it :) [13:22] and that (bug 878597) and bug 878600 was what made me look at it. [13:22] Launchpad bug 878597 in rabbitmq-server (Ubuntu) "rabbitmq-server fails to uninstall" [Medium,New] https://launchpad.net/bugs/878597 [13:22] Launchpad bug 878600 in rabbitmq-server (Ubuntu) "service start rabbitmq-server' does not fully detach from parent" [Medium,Triaged] https://launchpad.net/bugs/878600 [13:23] soren, so this seems like the best i can do: [13:23] http://paste.ubuntu.com/715116/ [13:23] smoser: Yup. [13:23] Unless rabbitmq could learn how to setuid. [13:24] right. [13:24] Or upstart could. [13:24] Well, either that, or you write a custom wrapper thing to do it, but that's an entirely different kettle of fish. [13:24] https://bugs.launchpad.net/upstart/+bug/586942 [13:24] Launchpad bug 586942 in upstart "init: support dropping privileges" [Wishlist,Triaged] [13:25] soren, or modify sudo so it could exec for me (or su to do the same). thats why i originally asked about sudo. [13:25] cjwatson: Oh, so that'll probably have the same sort of issue. [13:25] but i assumed there was good reason. [13:25] soren: no, I don't think so [13:26] cjwatson: No? How would you take care of closing the pam session? [13:26] Scott isn't saying "setuid needs to have a PAM session", he's saying "use a different keyword (i.e. setuid not user) because the user keyword is going to imply a PAM session" [13:27] basically just a quibble with the text of the bug description [13:27] cjwatson: Ah, right, yes. I meant the "user" directive would exhibit similar behaviour (i.e. the "extra" process hanging around just for this purpose). [13:27] cjwatson: I didn't realise there were plans for both. [13:28] Probably, unless pid 1 learned about PAM (which might be a bad idea) [13:28] cjwatson: YEah, I think it's unsafe to assume that pam modules won't mess things up when closing the session. [13:29] I mean I suppose you could marshal the PAM state back to pid 1 and then it could start a new process just to close the session, but (a) I don't know if that would even work and (b) it sounds like a lot of bother to save a process [13:30] Yeah. It sounds like no fun at all. [13:31] ok. related question. [13:32] other than not depending on sudo, is there any reason i would not want to use sudo to change permissions rather than su ? [13:32] su is always present. [13:32] sudo might not be. [13:33] and sudo's config is more prone to being screwed. [13:33] i dislike the fact that su executes things through sh [13:33] as root to non-root, sudo is not likely screwed i would think. [13:34] Nope. [13:34] Try it :) [13:34] yeah, you're right. [13:34] that does suck. [13:34] it should just get out of your way. [13:35] am i right that there is no sane way to get su to execute things without re-shell-quoting like at http://paste.ubuntu.com/715120/ [13:37] I'm not entirely sure why it doesn't just use "$@". [13:37] because su sucks [13:37] su -c "command" [13:37] not [13:37] su -c "command" "arg1" "arg2" [13:37] smoser: su is awful for this. Personally I use a tiny perl wrapper for this job. [13:38] su -- "command" "arg1" "arg2" ? [13:38] Wouldn't that work? [13:38] (I used to use start-stop-daemon; that's still OK for things that aren't liable to be installed via debootstrap. [13:38] ) [13:38] su sucks, soren. try it. [13:38] Well, really: su -- "command" "$@" [13:38] # start-stop-daemon isn't available when running from debootstrap. [13:38] (man-db.postinst) [13:38] perl -e '@pwd = getpwnam("man"); $( = $) = $pwd[3]; $< = $> = $pwd[2]; [13:38] exec "/usr/bin/mandb", @ARGV' -- "$@" || true [13:39] But modifying rabbitmq to have a drop-privileges option surely wouldn't be that hard. [13:40] # su cjwatson -c echo foo bar [13:40] Sessions still open, not unmounting [13:40] # su cjwatson -- echo foo bar [13:40] /bin/echo: /bin/echo: cannot execute binary file [13:40] Sessions still open, not unmounting [13:40] su really has an awful command-line syntax. [13:40] I've switched most of my use of start-stop-daemon to daemon with upstart (although that is not in main) [13:40] # su cjwatson -c 'echo foo bar' [13:40] foo bar [13:40] Sessions still open, not unmounting [13:41] Madness. [13:41] it wants it all in one arg, which is a quoting nightmare. [13:41] So that happens to the stuff after "--" ? [13:41] cjwatson, so for the perl magic novice.. the stuff up there is going to change perms to the man user ? [13:42] Interfaces designed sometime in the last decade or two tend to be sensibly adverbial and not require differing levels of quoting. [13:42] smoser: yes [13:43] smoser: I should probably change mandb to have a drop-privileges option instead, though, given that it already has all the code for that. [13:43] * smoser imagines cjwatson sitting there thinking "duh, smoser, that was *so* obvious! $( = $).." [13:43] no, I have to look those up too :-) [13:43] there are more friendly spellings of it but they require 'use English' [13:48] please mark merged: https://code.launchpad.net/~paulbrianstewart/ubuntu/oneiric/visualvm/813165-formattingTypo/+merge/68469 === dholbach_ is now known as dholbach [14:33] mdeslaur: yeah, I'm really sorry for the trouble with this last night, a new SRU broke it, it used the explicit version to workaround a old issue that launchpad does not copy the release upgrade from -rpopsoed to -updates. but now this is done manually be the archive-admins so the current symlink in -updates works [14:33] soren, cjwatson. just fyi, i realize that getopt is completely capable of correctly quoting variables for you. [14:33] so [14:33] sh -c 'x=$(getopt -s sh -o "" -- -- "$@"); x=${x# --}; exec su -c "set -- $x; \"\$@\""' arg0 echo howdy world [14:33] mvo: no problem, I just wanted to make sure you were aware of it [14:34] mdeslaur: indeed, thanks for this, if something like this happen, you can always mail the issue to me too if I'm not on irc :) [14:35] mvo: cooll, thanks === gord is now known as Guest8492 [14:41] smoser: My eyes began watering when I read that. Seriously. [14:41] well, it works, and its more reliable than the sed in the pastebin i showed earlier. [14:42] but other than being in /usr/bin, getopt is probably pretty likely to be around. [14:43] "Priority: required" agrees. [14:43] As does "Essential: yes" [14:44] Hello, is there a way to control the logo position in unity-greeter ? [14:44] FreakinInstallMe: uh-huh is the only one stronger than that [14:44] AnAnt: you probably want #ubuntu ;) [14:45] smoser: Isn't 'getopts' built in to dash/bash ? [14:45] it is not. [14:46] oh. wait. getopts is. [14:46] getopt is not [14:46] smoser: can you make it work w/ getopts ? [14:46] getopt is IMO actually usable for parsing parameters. but getopts is not. [14:46] but yeah, maybe you could. [14:46] smoser: heh [14:47] smoser: a page I just read yesterday actually claims the compelete opposite [14:47] smoser: getopts claims to be more advanced than getopt ;) [14:48] The getopts command deprecates the older getopt(1) utility due to its [14:48] handling of arguments containing whitespace. [14:48] smoser: getopt might be buggy wrt spaces in arguments, so you shouldn't use it [14:48] bah. [14:48] might be. [14:48] but is not [14:48] is on some OS'es [14:48] and its not like 'sh' is changing its parsing all that often. [14:48] so don't use it if you want your script to be portable [14:49] smoser: come on you can't make it work with a builtin? ;) [14:49] SpamapS, they're completely different beasts. [14:49] the reason that the builtin can do what it does is because it doesn't attempt to understand shell quoting. [14:49] it uses a global basically. [14:55] agateau: are you here ? === Guest8492 is now known as gord === gord is now known as Guest52713 === Guest52713 is now known as gord === dholbach_ is now known as dholbach [16:08] @pilot out === udevbot_ changed the topic of #ubuntu-devel to: Precise open for uploads | Ubuntu 11.10 Released! | Development of Ubuntu (not support, not app development) | build failures -> http://bit.ly/or6CHJ | #ubuntu for support and general discussion for hardy -> oneiric | #ubuntu-app-devel for application development on Ubuntu | http://wiki.ubuntu.com/UbuntuDevelopment | See #ubuntu-bugs for http://bit.ly/lv8soi | Patch Pilots: === deryck is now known as deryck[lunch] [16:22] slangasek, can you help me with my flashplayer? it doesn't seem to work with pulse [16:22] DBO: apt-get install libasound2-plugins:i386? [16:22] already done [16:23] still doesn't work === Quintasan_ is now known as Quintasan [16:23] tried installed adobe-flashplugin instead of flashplugin-installer also [16:23] that just made it not work all together [16:24] DBO: this is amd64? [16:24] slangasek, yep [16:25] restarted nspluginwrapper / the browser after installing libasound2-plugins:i386? [16:25] slangasek, restarted the entire machine [16:25] slangasek, its a fresh installed too [16:25] I remember it working once, but after installing my development environment [16:25] it stopped working [16:26] sorry, what do you mean by "development environment"? [16:27] * ogra_ glares at the title of bug 879448 [16:27] Launchpad bug 879448 in ubiquity (Ubuntu) "during upgrade from 4.11 to 10.11 the command grub-install/dev/sda failed error" [Undecided,New] https://launchpad.net/bugs/879448 [16:28] i hope that doesnt mean a typoed upgrade from warty to whtever 10.11 might be :) [16:28] *boggle* [16:28] DBO: Wait, you're using a flash that requires nspluginwrapper still? [16:29] Doesn't the partner repo contain a native amd64 package? [16:30] it's not yet integrated into flashplugin-installer [16:30] DBO: but, if you installed adobe-flashplugin and it still didn't work, I don't know what the problem is [16:30] DBO: how did you determine it's not talking to pulseaudio? [16:31] adobe-flashplugin simply didn't give me flash [16:31] like chrome didn't think it had flash [16:31] heh [16:31] you might need to make sure you uninstall flashplugin-downloader as well? [16:31] okay hold on [16:31] let me try that [16:32] as for not using pulse, lets assume I know what i am talking about :) [16:32] (I do) [16:32] slangasek: Right, I know the downloader package still does 32-bit and nspluginwrapper but adobe-flashplugin in partner is the native 64-bit 11.0 release [16:33] Oh, and I took too long to type that [16:33] * Amaranth goes back to watching paint dry^W^Whis panda board compile [16:34] slangasek, okay that worked [16:34] thanks :) [16:34] you're a champ [16:34] ok, good :) [16:34] I still don't know why the 32-bit one isn't working for you [17:21] is readahead standard in ubuntu now? [17:21] ureadahead is [17:21] since several releases [17:22] normal readahead was dropped in favor [17:22] ogra_: thanks. is there some documentation anywhere for how it works? or is it a 'Use The Source, Luke' thing? :) [17:23] both i think. there should be a blueprint on launchpad and possibly a spec on the wiki for its implementation [17:23] beyond that there surely is also source code to inspect :) [17:24] http://ubuntuforums.org/showthread.php?t=1434502 is also helpful [17:24] ogra_: thank you [17:24] broder: also :) === AaronMT is now known as AaronMT|lunch === deryck[lunch] is now known as deryck === AaronMT|lunch is now known as AaronMT [18:00] cjwatson, so where do you use that perl snippet? [18:01] it seems odd to add a perl dependency on a pakage just for that, although its very helpful. [18:06] smoser: as I say, in man-db.postinst. perl-base is Essential: yes so no dependency is required. [18:08] smoser: I do think having an option in the daemon to drop privileges itself is a strictly better approach. [18:08] yes. [19:13] i have ubuntu 11.10 and have problem with booting fakeraid (dualboot with windows 7), everything installed like a charm, it even booted after installation to both windows and ubuntu but after updating and rebooting one more time i get kernel panic and message that root= is wrong, didn't get any help on #ubuntu.. [19:23] soakda: This is not a support channel. You'll just need to be patient in #ubuntu, or check out our other support resources. [19:23] soakda: htps://help.ubuntu.com or http://ubuntuforums.org or http://askubuntu.com/ [19:25] Pici: ok, thanks [19:29] cjwatson: How would you rate the chances of bug 833994 being resolved this cycle? [19:29] Launchpad bug 833994 in debian-installer (Ubuntu) "debian-installer does not support https when using with preseed files" [Medium,Triaged] https://launchpad.net/bugs/833994 [19:45] superm1: can python-mysqldb be sync'd from sid? It seems it includes your fixes, dh_python2 and change to quilt. [20:03] is it possible to re enable module loading after 'echo "1" > /proc/sys/kernel/modules_disabled' ? === genupulas is now known as rajagenupula [20:21] penguin_03: no, that's the whole point of this option [20:22] stgraber, assuming you have physical access the the machine there *has* to be a way to disable it though [20:23] penguin_03: rebooting it is the easiest way [20:23] thanks [20:23] <__keybuk> just had one off those annoying upgrade problems where after upgrade mdadm is picking up /dev/sdb rather than /dev/sdb1 :( [20:25] "one of those"? I thought those were all in the distant past === yofel_ is now known as yofel === Quintasan_ is now known as Quintasan [20:53] <__keybuk> slangasek: apparently not [20:53] <__keybuk> this was on installing natty [20:53] <__keybuk> so it may be fixed in oneiric? [20:54] I don't remember any reports of anything like this happening in several years [20:54] it's not something anyone fixed in oneiric [20:54] TTBOMK [20:56] <__keybuk> if you have a partition that fills the entire disk, how does mdadm know to ignore the whole disk? [21:06] Daviey: *shrug* I don't care about it, but if you escalate it in the usual way with a justification then I guess I'll have to bloat up the initrd to do it :-P [21:07] Daviey: I think I checked and Debian wasn't going to do it [21:07] (but that's only a vague memory) [21:08] __keybuk: I don't know, but I thought it was solved :) [21:08] so if not, well, doh [21:08] __keybuk: I thought that was fixed by using a more modern md format; also the installer deliberately leaves a gap at the end of the disk to avoid ambiguity [21:09] <__keybuk> cjwatson: the installer didn't make these partitions, I did\ [21:09] in 1.1/1.2 the md superblock is at/near the start of the device rather than the end [21:09] <__keybuk> and I didn't leave a gap [21:09] well then :) [21:09] <__keybuk> cjwatson: hmm, the mdadm version is 0.9 [21:09] yeah, the ambiguity is inherent then unless you leave a gap [21:09] any syncpackage (for fakesync) experts around? [21:10] I can't remember how to deal with the situation where the partition already exists and can't conveniently be shrunk a bit [21:12] <__keybuk> I wonder how you update the mdadm version [21:13] barry: I know things about syncing, if that counts :) [21:14] slangasek: that's okay, i think i figured it out. fwiw, after i uploaded python-support 1.0.14ubuntu1 we discussed just dropping the ubuntu delta. i'llmanually prepare a 1.0.14ubuntu2 that drops it and make it clear in the changelog that next debian upload can do a sync for real [21:14] barry: right, that's the thing to do there [21:14] slangasek: cool, thx [21:14] I don't know of a way to change the metadata version in place [21:15] you could probably force the disk to be avoided by tweaking mdadm's udev rule locally [21:15] <__keybuk> yeah [21:16] that's probably the simplest local fix (but not generally suitable since some people deliberately use entire disks as components, of course) [21:16] <__keybuk> except that gets lost each upgrade [21:16] <__keybuk> I suspect that's what I did [21:16] dpkg-divert? [21:16] <__keybuk> indeed [21:16] <__keybuk> I think I actually dropped a replacement in /etc/udev/rules.d [21:16] or /etc/udev/rules.d/ I guess, you probably remember the details better than I do :) [21:16] <__keybuk> but it didn't get copied into the initramfs after upgrade [21:16] ah [21:16] initramfs-tools ought to have a helper function to DTRT there [21:16] (probably doesn't, but) [21:23] <__keybuk> ok, a little ENV{DEVTYPE}=="partition" seems to have solved my problem === kentb is now known as kentb-out [22:09] barry: yes (i wrote the syncpackage fakesync stuff) [22:09] bdrung: thanks, i figured it out though (can't use syncpackage for this particular use case) [22:12] barry: fakesync building assumes that the content is the same of the mismatching tarballs [22:24] bdrung: so, here's a new situation. i'm trying to sync python-peak.rules. it's blacklisted, so it suggests --no-lp --force, but --no-lp is "not recommended" (lp sync'ing was rejected). should i just use --no-lp --force? [22:25] barry: if blacklisted, use only --force. if this does not help, ask the archive admins to remove this package from the blacklist [22:26] bdrung: okay, thanks [22:27] barry: but --no-lp --force is the right way, if it needs a fakesync [22:27] bdrung: nod [22:27] barry: the comment is clear to me: http://paste.debian.net/138672/ [22:28] barry: and it seems to need a fakesync [22:29] bdrung: right, that part makes sense. it's a little confusing that it's recommending to use --force and --no-lp to override the blacklist, but the manpage warns against --no-lp [22:29] or should i say "seems to recommend" [22:29] barry: --no-lp is the only way for fakesync (patches to make that clear are welcome) [22:30] barry: --no-lp for fakesync are not discouraged by the admins [22:32] WARNING [22:32] The use of syncpackage --no-lp, which generates a changes file to be [22:32]