[00:43] evand have added a diff patch to bug #12596659 [00:44] bug #186114 [00:46] wubi rev 451 is up [00:53] ago: Ok that updated mounthost made no difference. [00:54] Let me reboot in recovery mode to see how things look... [00:57] ago: Still mounted ro. [00:58] hmm can you set -x in mounthost to see where it fails? [00:59] Yep, give me a bit so I can boot into a live disk to adjust it. [00:59] sure, might be that there are some earlier errors [01:01] * ago thinks about folders mount-moved in initrd [01:06] I'd assume that a loopdevice within a ro host is also ro [01:07] and in scripts/local we mount-move /host into a ro loopdevice (not sure if that is allowed) [01:08] TheMuso you might want to have set -x also in initramfs-tools/scripts/local to check that [01:09] ago: Ok, I'll get the script's contents dumped to a log file for both scripts, and I'll apstebin them both. [01:09] thanks [01:11] I am out of space and cannot create a vfat vm before backup/cleanup... [01:14] Launchpad bug 186114 in wubi "umountroot cannot handle fuse host mounts" [Medium,Fix committed] https://launchpad.net/bugs/186114 [01:16] * TheMuso re-generates the initramfs [01:16] ubotu, that must be a record.... [01:29] Wooho. It would help if I made the mounthost script I copied over executable. :S [02:04] ago: initramfs-tool scripts/local output: http://pastebin.com/m99dbdb6 [02:05] ago: mounthosts: http://pastebin.com/m1d0d3f2b [02:13] TheMuso: mounthost seems to have executed the following command successfully: [02:13] mount -n -o remount,defaults,rw /dev/disk/by-uuid/747E-E96F /host [02:13] ago: But /host is still ro. [02:13] Let me boot it again, just to be sure. [02:14] initramfs does not show any error during mount-move, I assume you stopped that to get the log [02:14] Yes [02:15] I know that for instance ntfs-3g cannot be remounted at the moment #186117 [02:15] I wouldn't expect the same issue to extend to vfat, but you never know... [02:16] can you try to mount/remount vfat from a live session? [02:16] Hrm ok maybe with the recent mounthosts update, host is now mounted rw, however things still crash out, which amkes me think its the initramfs code that is the issue. [02:16] makes [02:17] well if it is mounted rw, one less I guess... [02:17] Yes. [02:18] Yeah, even now I can't remount the root filesystem read-only, still says /host/ubuntu/disks/root.disk is wrte-protected. [02:19] So the initial mount from the initramfs sticks. [02:19] Even accross the pivot, which makes sense. [02:19] ah so /host gets remounted but / does not... [02:20] Yes, and cannot. [02:20] I went through that long time ago'... [02:20] Once we pivot, the remount of /host by mounthost is treated as a separate entity it seems. [02:20] IRC the solution is not to use mount -o loop [02:21] but use losetup instead and the mount /dev/loop0 [02:21] That doesn't appear to be used. [02:21] it is in initramfs-tools/scripts/local [02:21] mount ${roflag} -o loop -t ${FSTYPE} ${LOOPFLAGS} "/host/${LOOP#/}" ${rootmnt} [02:21] Right. [02:21] I still don't quite follow what you are saying. [02:22] you split mount -o loop in 2 operations [02:22] 1) you create a loop device with losetup [02:22] 2) you mount that normally without "-o loop" since now it is available as /dev/loop# [02:22] Um, so what difference would this make with initramfs tools, and vfat being mounted ro? [02:23] as mentioned I went through that ~1 y ago', but I think that for some reason when you use mount -o loop the hosted device is marked for life with the host settings [02:23] while if you split, the 2 are independent [02:24] sort of [02:27] Right. [02:27] But is there a reason why vfat gets mounted read-only in initramfs in the first place, an ntfs doesn't? [02:28] yes there was an exception for ntfs, since it could not be remounted [02:28] the relevant line is: [02:28] ([ -z "$LOOP" ] || [ "${FSTYPE#ntfs}" = "$FSTYPE" ]); then [02:28] [02:28] yeah I saw that [02:28] that makes sense. [02:29] 1 solution is to delete || [ "${FSTYPE#ntfs}" = "$FSTYPE" ]) [02:29] so that whenever there is a loop device it gets always mounted rw [02:29] Hrm, but wouldn't it be saner to use losetup as you previously mentioned? [02:29] other solution is to use the losetup + mount /dev/loop [02:29] And let mounthost remount it rw? [02:29] I think so [02:30] and when 186117 is addressed we can remove the specialcasing for ntfs [02:30] Right. [02:30] I'll research the losetup stuff now if you'd like, and I'll d a test. If it works, I'll upload a fix. [02:31] http://paste.ubuntu-nl.org/59320/ [02:32] replace the mount -o loop with the above (I assume that losetup is available in initrd, do not remember now) [02:33] I can check [02:35] Yes, it is in the initramfs. [02:36] Ok, I'll make a change, and test. [02:53] TheMuso, really need to go to bed 3am here... [02:53] ago: Thats fine, I'll let you know how it goes. [02:53] thanks a lot for everything [02:53] 'night [02:53] You're welcome. [02:53] Night === cjwatson_ is now known as cjwatson [08:19] The noninteractive oem-config bug seems to not stop there. When noninteractive is used it cannot find any of the packages on the local cd mirror. I'll take a further look in the morning. [08:24] evand, oooh yuck. that's worse than expected :) [08:34] that's ... special [10:36] TheMuso any luck yesterday? [10:37] xivulon: It still crashes out. I need to get a complete boot log for you. [10:37] thanks [13:37] is there a way to preseed an installation so that deb-src repositories aren't generated in the sources.list? [13:37] also, is there a way to prevent hardy-updates from being generated too? [13:48] no, sorry (to both) [14:32] cjwatson: that's fine, I'll workaround in late-command. [14:32] when installing a new package, during the Setting up phase, I sometimes get: Not replacing deleted config file /etc/... [14:33] I'm not sure if this is the right place to ask but this is breaking packages being installed :( [14:34] or just running dpkg-reconfigure package_name returns the same error message [14:35] err, warning message :) === ebel_ is now known as ebel [15:08] aha! it was a ucf problem which can be fixed with ucf -p /etc/... === soren is now known as soren_not_joking === soren_not_joking is now known as soren_joking === soren_joking is now known as soren === ebel_ is now known as ebel === bilbo is now known as nerxgas [18:26] This noninteractive bug is horribly confusing. gtk_ui and noninteractive are using the same code during the actual install routine, yet the former is able to install packages from the CD and the latter complains about them not being in pool. [18:27] Curiously, if I interrupt the process with a sleep and run apt by hand (setting the appropriate options) it works just fine. [18:27] ugh. [18:47] that would make it rather difficult to sort out - === cjwatson_ is now known as cjwatson [19:05] after the installation has completed, in late-command, is there a way to add repositories to sources.list and apt-get a few more packages? [19:06] the reason for wanting to apt-get new packages so late is that I want to be very strict on the repositories used during the installation [19:06] use apt-setup/local0/repository etc. [19:06] and, the problem with running apt-get in late-command is that /proc doesn't exist on the target system so apt-get fails horribly :) [19:06] cjwatson: the problem is that repositories specified that way are used during the installation of the system proper [19:07] no they aren't [19:07] they're used for language packs, but that's all [19:07] relevantly, they're used for pkgsel/include [19:07] which is what you really want [19:07] cjwatson: indeed relevantly, thanks a lot for that info! [19:08] you can bind-mount /proc if you need to (necessary when not using in-target), but it shouldn't be needed with that different approach [19:08] and then just preseed the things you need to preseed in the normal way, using a preseed file [19:15] cjwatson: Why bind-mount /proc rather than just mount it? [19:53] hi [19:53] need help installing boost [19:53] is anyone here? [20:49] soren: either works [20:55] cjwatson: Ok. === ceeka1 is now known as seekay [22:17] I'm getting an error code 100 during pkgsel because the repository used during the installation doesn't have deb-src nor hardy-updates. Is there a way I could hook into preseed before pkgsel is called to strip sources.list or should I emulate deb-src and hardy-updates on the server side? [22:23] and it seems that pkgsel is forcing an upgrade, or maybe I could be wrong: The following NEW packages will be installed:... [22:54] there's a /usr/lib/pre-pkgsel.d [22:54] new packages doesn't sound like an upgrade [22:59] ls: /usr/lib/pre-pkgsel.d/*: No such file or directory; WARNING **: Configuring 'pkgsel' failed with error code 1 [22:59] mkdir [22:59] (if you want it); it won't be failing for that reason anyway [22:59] oh, there's also this line: The following packages will be upgraded: [22:59] right, just a warning. at what point should I be creating /usr/lib/pre-pkgsel.d/? [23:00] upgraded> maybe needed in order to install language packs; it certainly doesn't run a real upgrade [23:00] I know that for an absolute fact [23:00] some of the files under the upgraded packages include: apt apt-utils dhcp3-client dhcp3-common gcc-4.2-base initramfs-tools libgcc1 libsasl2-2 libsasl2-modules libssl0.9.8 libstdc++6 ntpdate python python-minimal python2.5 python2.5-minimal tasksel tasksel-data [23:01] if you want to hook in before pkgsel to do something, create the directory from preseed/early_command and drop an executable script in it [23:01] could I see the full log, please? it's hard to guess from snippets [23:02] cjwatson: sorry, I removed the log and reverted to doing all my mangling in late-command. I will try again to polish my preseeding once all the tests have completed. [23:02] I don't see anything in there that could obviously be forced, but if it were a significant enough mismatch between image and archive then it's possible [23:04] ok, so I might try to strip hardy-updates and deb-src in a pre-pkgsel script. I think that would be clean [23:04] pkgsel does: install appropriate language-pack-$LANGUAGE for basic localisation; run tasksel to install bulk of packages; install rest of language packs (e.g. language-pack-gnome-$LANGUAGE) and language-support-$LANGUAGE [23:05] and other miscellaneous (non-apt) bits and pieces around that [23:05] is this a CD or net install? [23:05] cjwatson: net install [23:05] ok, in that case I apologise for misleading you earlier [23:06] repositories specified with apt-setup/local* are indeed used for installing the system as a whole in the case of net installs [23:06] but not for CD installs [23:06] I get a bit CD-centric sometimes [23:06] cjwatson: heh, don't worry about it, I need to understand exactly what's going on and learning from mistakes helps tremendously [23:06] you could cheat in a potentially interesting way [23:07] in pre-pkgsel.d, remove deb-src and -updates as you wish [23:07] but also copy /target/etc/apt/sources.list to /target/etc/apt/sources.list.apt-setup, and add the extra local repositories you want [23:07] pkgsel will move the latter file into place before processing pkgsel/include [23:07] cjwatson: so perhaps I'll also add my other repositories in pre-pkgsel [23:08] this is complete undocumented cheating, and piggybacks on how CD stuff happens to be implemented, but ought to work [23:08] it's the easiest way I can think of to add local repositories in a net install but not have them used for the bulk package installation stage [23:09] 1. cp /target/etc/apt/sources.list /target/etc/apt/sources.list.apt-setup; 2. mangle /target/etc/apt/sources.list.apt-setup; 3. let 'er rip [23:09] right [23:09] cjwatson: this certainly beats my current workaround: have my test suite apt-get packages :( [23:09] you will probably want to remove deb-src and -updates from /target/etc/apt/sources.list before copying [23:09] that is totally evil and I'm ashamed to even mention it [23:10] cjwatson: so I might as well invert steps #2 and #1 [23:10] unless /target/etc/apt/sources.list is actually used by the installer to install the system [23:11] it is [23:11] ok, I really need to jet, taking spiv out for a beer :) [23:12] it uses /target/etc/apt/sources.list up to just after tasksel finishes [23:12] oh, thanks for the info, I'll be careful to get those steps straight [23:12] then copies sources.list.apt-setup into place, then installs pkgsel/include stuff, then installs rest of language packs [23:13] crystal clear :) [23:13] night, enjoy beer :) [23:13] meeting brad bollenbach as well, nice evening ahead :) [23:51] hmm I do not seem to be able to remount a loopfile inside vfat.... [23:52] if I start with a ro mounted host, I can remount the host but not the hosted loopfile [23:52] if I start with a rw host+hosted. I can remount ro the hosted fs, but not the host [23:55] xivulon: Whats interesting, si when I made those initramfs changes locally and tried again, it crashed, but the loop mounted root fs was mounted rw. I'd need to check to be sure, but thats what I can remember. [23:56] strange I am tru [23:56] trying to replicate on a vfat partition [23:56] I mount vfat partition ro (that contains an ext3 file) [23:56] then do losetup on the ext3 file in there [23:56] then mount the loop device ro [23:57] then remount the vfat partition rw [23:57] then remount the loopdevice rw [23:57] and... ...it does not work [23:57] but iirc it used to work [23:57] doublechecking