[00:01] evand any other bug you want to go through tonight? [08:23] could I request a new d-i against -4 kernel? [09:18] tjaalton: sure, on its way [09:18] debian-installer: cjwatson * r869 ubuntu/ (10 files in 4 dirs): * Move to 2.6.24-4 kernels. [09:18] cjwatson: thanks! [09:20] debian-installer: cjwatson * r870 ubuntu/debian/changelog: releasing version 20070308ubuntu26 === cjwatson_ is now known as cjwatson === cjwatson_ is now known as cjwatson [14:54] hi evand [14:54] any q over remaining patches? [14:54] xivulon: I'll let you know, but none currently. [14:58] what's the status of bug #173659 and bug #151579? [14:58] Launchpad bug 173659 in wubi "Add /host/boot to fstab" [Medium,Fix committed] https://launchpad.net/bugs/173659 [14:58] Launchpad bug 151579 in wubi "umountfs must check whether a mountpoint contains a loopmounted root file" [High,Fix committed] https://launchpad.net/bugs/151579 [15:01] I'll take care of the former next. As for the latter, cjwatson, does http://paste.ubuntu-nl.org/52195/ look ok to you or would you prefer the solution you had originally suggested? [15:01] evand can you also look into bug #177868? Without the new hooks is very difficult for me to do releases (have to support 2 hooks) [15:01] Launchpad bug 177868 in wubi "When loopfiles are used mkfs has to target the file and not the containing device" [Low,Fix committed] https://launchpad.net/bugs/177868 [15:02] Are you sure that's the right bug? [15:02] That mostly involves merging https://code.launchpad.net/~ubuntu-installer/lupin/hardy [15:02] ehm nope [15:02] heh [15:02] I meant bug #144798 [15:02] Launchpad bug 144798 in casper "Merge lupin functionality + add external hooks" [Wishlist,Confirmed] https://launchpad.net/bugs/144798 [15:03] evand: it looks horrendously complicated and much harder than my suggestion, but it's your call [15:03] copied the wrong line [15:03] ok, I'll take some more time to look at that then. [15:03] cjwatson what was your proposal on that? [15:04] I can recode if you give some guidance [15:04] some derivative of: sed -n '/^rootfs/,$p' /proc/mounts [15:04] messing about with line numbers is almost always an overcomplex option in Unix [15:04] cjwatson one thing is not clear to me [15:05] do we want to skip all mountpoints above root, or all mountpoints above rootfs? [15:05] My understanding was the first case [15:08] oh, hmm, yeah, should be root. adjust the regexp accordingly [15:09] /^\/[^ ]* \// [15:09] /^\/[^ ]* \// [15:09] /^\/[^ ]* \/ / [15:09] my point is just to avoid having to read through the file multiple times, work out line numbers, use lots of commands - an elegance thing :) [15:10] Heh I am not a bash esthetist, getting things to work is good enough for me :P [15:10] sh [15:12] I think it's sort of important in the shutdown procedure to call as few external commands as possible, for efficiency [15:12] I agree [15:15] now I get the regex [15:15] I like it [15:33] yep much better http://paste.ubuntu-nl.org/52272/ [15:33] evand ^ [15:35] looks good [15:36] cjwatson, what is the sed command to print the top half of the file as opposed to the bottom half? [15:36] needed in umountroot [15:36] in fact not needed [15:36] 0,/regex/ [15:36] ah was trying $p,/regex/ [15:36] was close [15:36] well, sed -n '0,/regex/p' to be a little more verbose [15:37] alternatively sed '/regex/q' tells sed to quit as soon as it encounters /regex/ (before printing that line) [15:37] you can do similar things in awk too of course, sed just fits my brain better [15:38] cjwatson, evand in bug #151579 last comment is also a patch for umounroot, can you please review that too? [15:38] Launchpad bug 151579 in wubi "umountfs must check whether a mountpoint contains a loopmounted root file" [High,Fix committed] https://launchpad.net/bugs/151579 [15:39] will do [15:40] I have added colin suggestion as a comment, so the other one to check is now comment n-1 https://bugs.launchpad.net/ubuntu/+source/sysvinit/+bug/151579/comments/5 [15:40] Launchpad bug 151579 in wubi "umountfs must check whether a mountpoint contains a loopmounted root file" [High,Fix committed] [16:04] evand re lupin/#144798 https://code.launchpad.net/~ubuntu-installer/lupin/hardy, I am only intrested in the casper package, you can skip lupin-helpers alltogether [16:04] noted [16:05] lupin/casper by the way might become a separate package, called maybe custom-installation [16:05] with 2 packages, 1 for casper and 1 for the alternate initrd [16:06] that basically now provides external hooks to modify the installation using files on local hard disks [16:07] what's your view on that? [16:08] hrm, no complaints. But I'm more concerned with getting this done as we have it set up at the moment. [16:09] re casper, you can simply merge the old lupin/casper with the new one [16:10] suggestion only moves lupin/hardy/casper to a different launchpad project, + adds code for the alternate initrd to provide equivalent hooks [16:13] cjwatson, xivulon: May I ask that you both just eye this over one last time before I send it to the buildds? http://pastebin.ubuntu.com/3623/ [16:14] looks good to me [16:16] thanks a lot [16:19] evand: hmm, <(...) is a bashism [16:19] that's a /bin/sh script so it'll break [16:19] maybe use a temporary file if that's possible at that stage? [16:19] hrm [16:22] so: [16:22] sed -n '/^\/[^ ]* \/ /,$p' /proc/mounts > tempmounts [16:22] exec 9<&0 < tempmounts [16:22] well, put it in a known-safe temporary directory (*not* /tmp; /var/run maybe?) [16:23] ok, indeed I wasn't sure where to put it and was just going to ask [16:23] skip_mountpoints=$(sed -n '0,/^\/[^ ]* \/ /p' /proc/mounts) [16:23] maybe [16:36] http://paste.ubuntu-nl.org/52281/ [16:36] evand, cjwatson ^ [16:37] the above those not use a file and I believe is also safer code [16:37] if for instance a mountpoint appears below root using the same device of a mountpoint above root will also be skept [16:37] which is the correct behaviour IMHO [16:41] err ... no, it's still important to umount such a device even if it happens to be mounted multiple times [16:41] This is required when you have mountbinds for instance, [16:41] quite the opposite [16:41] don't agree [16:41] you need to unmount the bind mount [16:41] you don't need to unmount the thing it's a bind-mount of [16:41] true [16:42] therefore it is not correct to fail to unmount something just because the same device appears before root [16:42] remembered wrongly that it would umount the device [16:42] but it does umount -f -r -d $REG_MTPTS [16:43] yes, that's fine [16:43] yep [16:43] you can't umount a device, anyway [16:43] the umount system call takes a mountpoint [16:44] umount [-dflnrv] dir | device [16:44] that is not a system call [16:44] ah [16:44] that is a command [16:44] man 2 umount [16:45] but doesn't umountfs use the command? [16:46] anyway moot point [16:46] if you pass a device node to umount(8), it will simply look up the device node in mtab and figure out the mountpoint [16:47] therefore it's not really unmounting by device, it's just a convenience [16:47] my point is that you cannot tell the kernel "unmount everything mounted on this device" [16:48] though it's true that if you pass a device to umount(8) it will try to unmount it from all the places where it's mounted [16:50] http://paste.ubuntu-nl.org/52284/ [16:50] don't uses a file, but requires cut and grep [16:53] that's incorrect. If /usr was mounted for some reason before the real / was mounted, and then a new /usr mounted over the top, the new /usr *must* still be unmounted during umountfs. [16:54] true [16:55] that's what I call a corner case [16:55] but you seem to be putting serious effort into breaking this corner case when it's not necessary [16:55] making this corner case work is *simpler* [16:56] Evan's code above will take care of it just fine [16:56] one less then [17:00] evand you mentioned you were also working on #173659 any comment there? I'll be leaving in about 30m [17:01] nothing to report yet, I'm tied up with this sysvinit stuff at the moment. [17:02] sure [20:25] argh, xivulon's /host/boot to fstab patches fail for ubuntu inside ubuntu as you'll always have a boot directory inside /host. === mtp_ is now known as matttp [22:35] hi [22:35] can someone help me with raid1 and lvm? [22:36] I've setup via a howto on youtube [22:36] testing to see if failover works [22:37] disconnect one disk from the raid1 set and kernel loads but just gets stuck on either usb module load on cdrom [22:37] on either disk [22:39] anyone? [22:39] yes? [22:39] cool [22:39] hi - hope you can help [22:39] probably not [22:40] maybe u can suggest where to go? [22:40] I've installed ubuntu server with raid1 and lvm [22:40] raid1 part does not quite work [22:41] disconnect either disk of raid1 array and the system does not fully boot [22:46] I think you need the kernel people rather than here [22:46] #ubuntu-kernel [22:47] cool - thank you [22:47] oem-config: cjwatson * r397 oem-config/ (5 files in 4 dirs): * Activate appropriate input methods when changing language (LP: #181857). [22:48] ^- fear the nasty code [22:48] oem-config: cjwatson * r398 oem-config/lib/im_switch.py: remove stray debugging code [22:50] evand: I think we might want something very similar to oem-config r397 in ubiquity at some point, though I'll want to test it more first [23:08] evand any bug you want to go through tonight? [23:10] if you use wubi-installer to install linux to a folder on a windows os, doesnt that mean linux will run off of windows [23:10] nope [23:11] you can delete c:\windows and still boot wubi [23:11] oooh... ahhh [23:11] you cannot remove the c: partition though [23:12] i see [23:12] Thugacation: it's just running from the Windows filesystem, that's all [23:12] the Windows operating system is not running or otherwise involved [23:12] yeah i had a brainfart for a second [23:12] like windows is in c:\windows and wubi would be c:\wubi or something [23:12] basically a linux sees a file as a whole hard disk