[01:09] cjwatson_: it has hung up at 94% and nothing in the syslog. Last entry was about grub and that is just before debconf reconfigures popularity-contest - I haven't killed reconfigure yet but it has been running for a while - stuck [01:21] kirkland: hrm, I remember seeing this with a netboot install. Will need to dig deeper. [01:22] TheMuso: thx [01:22] TheMuso: i've asked for a full syslog [01:26] kirkland: Ok great. [02:08] cjwatson_: On your official livecd squashfs filesystem, do you include the /var/cache/debconf fodler and data? [02:11] cjwatson: I didn't see anything in the log during the period where it hung up on the reconfigure but I did see messages later about it failing to copy over the /var/cache/debconf/ files to /target - going to have to check an official ubuntu livecd filesystem.squashfs and see if it si there. wonder if the 2 are related [03:23] hey guys, would someone please explain what a live filesystem image is in the livecd cdimage process [03:31] acoc: The live filesystem contains a pre-installed image of the ubuntu desktop. [03:33] TheMuso: so if you were building an xubuntu installer, would the live filesystem contain ubuntu or xubuntu [03:35] acoc: xubuntu [03:37] TheMuso: so is this correct: the livecd cdimage process creates a supplemental list of packages that get installed on top of the live filesystem when installed on the hard drive? [03:40] Actually, it's usually the other way about : typically the live filesystem contains a superset of that to be installed. [03:42] ok, so the filesystem is what is copied to the hard drive during an install, and cdimage creates the image that is loaded by the cdrom? [03:46] acoc, I'm afraid I don't understand the question. The ISO contains the live filesystem and some other stuff (documentation, boot hints, a few optional packages). The live fileystem contains the environment loaded in the live session which is also a superset of what is installed by ubiquity. [03:48] after the cdimage process (for daily-livecd) is completed, is the ISO created for the livecd [03:56] acoc: The live filesystem si created on another machine, separate from cdimage. [03:57] I remember I was told that before, what machine is this? [03:57] acoc: Varies for the architectures, and can and has been changed. [03:58] are there tools to create a "standard" environment so it can be created on my machine (is this livecd.sh) [03:59] acoc: livecs-rootfs [04:00] i think [04:02] ok, so this creates the base system, and cdimage would borrow from this filesystem image for the livecd and also for the hard drive install? [04:05] livecd-rootfs creates the live filesystem and the manifest files, which control how much of that ends up in the install. [04:06] acoc, You may find mounting a live CD to be instructive, perhaps followed by a loop-mount of casper/filesystem.sqashfs from the mounted live CD. [04:07] ok, I'll check that out next, but I'm still a little confused about what the cdimage scripts do after taking the seeds and germinating out the package lists [04:11] acoc, For that, you want to look at livecd-rootfs, ubuntu-cdimage, and debian-cd. These three packages contain just about everything that gets done. You may also want to look at casper and ubiquity, if you are curious about the boot and install process. [04:14] who controls livecd-rootfs, is that a debian project or ubuntu? [04:16] It's native to Ubuntu, but "control" is an odd word there : it's open source so each project that uses it may well branch. [04:19] Unlike cdimage however, its a package that gets uploaded like any other package. [04:20] is that considered up to date? [04:20] What do you mean? [04:21] should livecd-rootfs be checked out of a bazaar or is the package in apt acceptable [04:22] They should be identical [04:22] They are identical in intrepid at the moment. [04:23] it's the same script for both intrepid and hardy correct? [04:23] It's the same script for any of them [04:23] The versions may differ between releases though. [04:25] is there any good places for documentation of livecd-rootfs short of examining the script [04:27] Why are you wanting to use livecd-rootfs directly, rather than updating an existing LiveCD? [04:27] Not really, but the script isn't that large. [04:29] ultimately I'm trying to create a derivative with e17 as the windows manager, so I'd guess there would be some fairly major changes [04:30] I'm trying to make it a bit more official by developing a seed and had figured whatever would create the image would use germinate to make changes to the ubuntu image based on the seed [04:31] from what I understand that's what alternate install does (but I might be wrong about that as well) [04:32] acoc: Essentially, livecd.sh installs a task or a meta package, so you want a eubuntu-desktop package or so [04:32] acoc, I'd recommend starting with a set of seeds and metapackages, then building an alternate CD, and only thereafter looking at the liveCD : you'll find the fewest number of things you need to go back and change that way. [04:33] StevenK: the project I'm starting with OzOs currently has a meta package which contains everything e17 wise [04:34] persia: ok, currently I'm just trying to get xubuntu working [05:36] Is there a bzr repo for the partman-efi Ubuntu variation? [05:44] persia: the source package info doesn't give you an URL, and there is no launchpad project? [05:44] There is a launchpad project, it just only contains the Vcs-imports [05:49] Well, I'll just make the change locally now, and perhaps ask about branch history in several hours, when the last uploader is more likely to be about. [05:50] persia: I don't think the import had been done when I made the uploads [05:51] cjwatson, heh. I thought you'd be asleep now :) Shall I construct a branch from 17, apply ubuntu1 and ubuntu2 as two commits, and then layer my changes? [05:52] yes please [05:52] and push it to lp:~ubuntu-core-dev/partman-efi/ubuntu [05:52] I was about to do the same, but have no objection to you doing it ;-) [05:52] I can't do the last, but I can prep it. [05:52] persia: What are your changes BTW? [05:52] today is one of my visiting-the-office-in-London days, so I'm early [05:53] TheMuso, adding "lpia" to the list of architectures [05:53] persia: Oh. [05:53] persia: Is that really necessary? I thought lpia wouldn't have anything to do with efi... Or does it? [05:54] TheMuso, Intel claims EFI is supported by Atom processors. [05:54] And actually, given what lpia is, as opposed to what Intel claims it is, using lpia+EFI should result in something like a 1.5% speed increase on most Intel Macs, due to compiler optimisations. [05:54] persia: Right. [05:55] (mind you, this is a bad idea for Ubuntu Desktop right now, due to arch-specific compilation changes for some packages) [06:02] kirkland: Hrm ok, it appears that while the dmraid udebs are on the server disk, dmraid and libdmraid proper aren't. Checking logs and seeds to work out why now. [06:05] really? the seeds are fine [06:09] Ok seems that server-ship has the non-GUI related apps from d-i requirements, and d-i-requirements is only pulled in on desktop disks. [06:10] persia, do you have a link referring to what lpia is versus what intel is claiming it is? [06:10] So adding dmraid to server-ship would solve the server disks not having dmraid problem, but not knoing enough about netboot, I am not sure if that solves no dmraid for netboot. I tried a netboot install the other day and it failed in a similar fashion to what kirkland pointed out. [06:12] definitely a bug if d-i-requirements isn't being included on server CDs [06:12] Well from reading the latest server log, it seems that way. [06:12] fixing [06:12] the bug is in ubuntu.intrepid/STRUCTURE [06:13] Right, but I wasn't sure whether that was a bug, or whether that was intentional. [06:13] netboot will be something different [06:13] it's a bug, d-i-requirements is meant to be common to every installation method that uses d-i [06:13] (directly, anyway) [06:13] superm1, No. I've been following the literature from the Intel site in a limited fashion for the past 10 months or so. My understanding was that it was supposed to be a low-power chip with restrictions to 1G ram, direct memory mapping to some HW resources, and other stuff, and that the A1xx and Atom processors were lpia. I have seen these chips attached to 2G ram, and with vmx enabled. Aside from the CPUID string, I can't tell them apa [06:13] rt from i686. [06:14] cjwatson: Ok, I'll wait for everything to filter down, and will try another netboot install in the next day or so to see if anything has changed. [06:14] the main difference they talked about when first introducing it was the in-order execution thing [06:14] persia, hm that's quite interesting [06:14] TheMuso: this change will definitely not affect netboot in any way; I'm interested in what the symptoms were there [06:15] cjwatson: Ok, I'll try and grab a log dump from that. I don't have that install around any more however, but wil see what I can find. [06:24] cjwatson, Diff from partman-efi 17 to partman-efi 18 appears to be all translations. Is it worth merging these as long as I'm touching the package? [06:28] fine by me [07:31] I think lp:~persia/partman-efi/ubuntu is a correct representation of the package history, plus the merge, plus my changes, with appropriate layering, etc. Could someone take a look, to see if I've missed something? [07:48] I think I made the right set of changes to user-setup to enable passwd/allow-password-empty, and correctly reconstructed ubiquity sources (either ubiquity or partman-efi needs adjustment or ubiquity FTBFS). Testing this with a preseed file results in the final screen of ubiquity having the Install button disabled. Does anyone have any pointers to where I might hunt this? [09:37] persia: ubiquity/frontend/gtk_ui.py:info_loop() [09:38] (and probably similar stuff in other frontends) [09:41] cjwatson, Thanks. I also had some templates issues which I'm sorting, but I'll check that too. [09:43] persia: I think it's XS-Original-Vcs-Svn rather than XSBC- [09:45] persia: otherwise partman-efi looks fine [09:45] persia: want me to push it to ~ubuntu-core-dev? [09:51] Actually, just using XS-Original... sounds much better. If you wouldn't mind fixing that before pushing, I'd appreciate it. === davmor2 is now known as davmor2_away [09:55] persia: done, pushed [09:57] cjwatson, Thank you. Soon I'll have enough core-dev points :) [10:24] clock-setup: cjwatson * r199 clock-setup/ (debian/changelog finish-install.d/10clock-setup): [10:24] clock-setup: Apparently OpenSolaris keeps the hardware clock in local time [10:24] clock-setup: (surprisingly). Assume UTC=no if Solaris is detected (LP: #279065). [11:18] debian-installer: cjwatson * r971 ubuntu/ (3 files in 2 dirs): Move mainline architectures to 2.6.27-6 kernels. [11:28] cjwatson: Does the next ABI bump get "Stop it!" in a d-i commit message? :-) [11:29] I think I sorted the remaining blocker to --automatic for ubuntu-mid with passwd/allow-password-empty in user-setup. Please review lp:~persia/user-setup/ubuntu and lp:~persia/casper/trunk to turn it on. [11:29] StevenK: I can't be bothered :) [11:30] (and I'm not generally into silly commit messages anyway ...) [11:30] Haha [11:33] persia: user-setup is fine. I'm not sure I like that casper change; why can't ubuntu-mid just boot with automatic-ubiquity? [11:33] What's the casper change? [11:33] persia: (please use UNRELEASED in changelogs before they actually get uploaded though) [11:34] user-setup: cjwatson * r118 ubuntu/ (3 files in 2 dirs): merge from lp:~persia/user-setup/ubuntu [11:38] as it happens I'll just upload it now anyway though ;-) [11:38] user-setup: cjwatson * r119 ubuntu/debian/changelog: releasing version 1.20ubuntu8 [11:40] StevenK: http://paste.ubuntu.com/54955/ [11:40] Ohh, that change. [11:42] cjwatson: Putting automatic-ubiquity in the kernel command line will achieve the same thing? [11:45] persia: or do you not launch the installer straight away? === davmor2_away is now known as davmor2 [14:13] ubiquity: cjwatson * r2872 ubiquity/ (15 files in 10 dirs): [14:13] ubiquity: Fix intltool handling; po/Makefile.in.in was still coming from gettext. [14:13] ubiquity: Pass no-xml to IT_PROG_INTLTOOL so that we don't need to build-depend on [14:13] ubiquity: libxml-parser-perl. [14:19] oem-config: cjwatson * r535 oem-config/ (9 files in 5 dirs): [14:19] oem-config: Fix intltool handling; po/Makefile.in.in was still coming from gettext. [14:19] oem-config: Pass no-xml to IT_PROG_INTLTOOL so that we don't need to build-depend on [14:19] oem-config: libxml-parser-perl. [15:54] oem-config: cjwatson * r536 oem-config/debian/ (changelog init): [15:54] oem-config: Add a comment to /etc/init.d/oem-config to explain the purpose of the [15:54] oem-config: oem-config-firstboot check. [16:02] cjwatson: thank you for your response in bug 33649 to tricky1 ... s/he is troll and i was done arguing with him/her [16:02] you actually said you were going to follow up again and didn't ;-) [16:02] but no problem [16:02] I just came across it on the sponsorship queue [16:22] cjwatson: i think you can drop the ubuntu-main-sponsors from that bug [16:23] cjwatson: all of my changes related to that bug have been sponsored, and i believe that bug to be closed [16:23] yeah; done [16:24] cjwatson: there may be more issues related to that bug, but i'd like to see them opened as a clean, new bug, without the 200+ comments from the peanut gallery about all things wrong with Ubuntu [16:26] kirkland: speaking of which, do you think you could take a pass through the big list in bug 68308 and see if there's anything there we can pick up and fix? [16:26] cjwatson: sure, no problem... also, on my todo list, is to look at all the various raid bugs open and see if we can close some of them [16:26] cjwatson: i'm in #u-meeting at the moment, but thereafter [16:34] cjwatson: wow, i didn't even know that 68308 existed ... that was exactly what i was trying to solve with my intrepid spec/blueprint [17:20] cjwatson: it appears as though the DEBCONF_DEBUG=developer didn't provide any extra info in the /var/log/syslog or anything extra in /var/log/installer/ either - any ideas [17:20] is there a specific place in the kernel call line that it should be? [17:22] should produce a verbose /var/log/installer/debug [17:23] Fragadelic: in response to your question last night, yes, we do include /var/cache/debconf in the squashfs - wouldn't work otherwise [17:23] there were only a few lines in /var/log/installer/debug [17:23] doesn't matter where DEBCONF_DEBUG=developer is, as long as it's not right up against another argument without space in between of course :) [17:23] 'cat /proc/cmdline' for me? [17:24] I found that out - I just loop mounted an official filesystem.squashfs and tried - would the *-old files maybe cause a problem? [17:24] no [17:24] we remove them in ours to save space, that's all [17:24] It shows up in the /proc/cmdline - I'm doing it again now in a virtual machine [17:25] oh, whoops, I was thinking of the alternate install CD and giving you completely wrong advice [17:25] instead of DEBCONF_DEBUG=developer, run ubiquity with the -d argument [17:25] sorry about that, I get confused sometimes [17:25] ah - ok - will try that now [17:25] no problem - I'm just glad you are willing to try to help me out [17:27] ok - its installing right now - will doing a tail -f /var/log/installer/debug casue any issues? [17:27] no [17:28] thats much better - lots of info in the debug now - lol [17:28] what are the other things that you can pass for troubleshooting - both cmdline for boot and ubiquity - couldn't find much info about casper or ubiquity online [17:32] stuck after get popularity-contest/participate and next line 1 false - hanging now - hopefully when it starts up again there will be something meaningful [17:32] unfortunately the Ubuntu installer team has not historically been all that great at documentation :( [17:33] thats fine - most folks aren't interested in it as long as it works - I'm interested in it since remastersys builds the livecd as close to original as I cna get it including using ubiquity [17:34] Fragadelic: there's https://wiki.ubuntu.com/DesktopCDOptions [17:34] all I do is simply disable a couple of the bottom scripts like adduser and autologin for backup mode [17:34] ok, your log as described suggests that in fact it's not getting stuck on debconf interaction, which had been my first hypothesis [17:35] could you post the output of 'ps auxf' on a pastebin from the point when it gets stuck? [17:37] will have to figure out how to get it from the vm - lol - vm won't allow me to copy and paste right now - have to find out why [17:37] TheMuso: bug 279288 - looks like dmraid needs to be activated only when explicitly asked for [17:37] scpfra scp? [17:37] oops [17:37] Fragadelic: scp? [17:39] ok - logged in from the vm - don't do irc much - how do I post to pastebin? [17:40] Fragadeli1: sudo apt-get install pastebinit [17:40] Fragadeli1: you can echo $foo | pastebinit [17:40] Fragadeli1: and it'll spew a url back at you [17:41] http://pastebin.com/f1019d4e5 [17:41] this is the debug file [17:42] kirkland: thank you very much [17:42] Fragadelic: that seems like the debug file from a successful installation [17:43] but take a look at the time from one step to the other at line 2956 [17:44] 6 minute hang - sometimes longer and sometimes it doesn't even finish - hangs indefinitely [17:44] oh, I see [17:44] removing popularity-contest and install flows along nicely [17:45] there is a perl call for debconf that seems to be where it hangs [17:45] I think you're misinterpreting that - I'd need the full output [17:45] ubiquity uses debconf throughout and there'll be a perl process running for debconf rather a lot really [17:46] the point you refer to happens to be where update-initramfs is run [17:46] ah ok - its just that removing popularity-contest makes the issue disappear [17:46] that's a very time-consuming process [17:46] update-initramfs on my system goes quickly [17:47] could take a long time if you're low on memory for some reason [17:47] I'm not questioning your observations, but the coincidence is very suspicious [17:47] never been 6 minutes and like I said if popularity-contest is removed it doesn't hang there - night and day diff [17:47] i hear you - but...I rebuild the initramfs during remastersys after making changes to casper so its ready for the livecd [17:48] to build a cd that ends up with a 3.6G squashfs file it takes less than 10 minutes total [17:48] can you put 'set -x' on the second line of /var/lib/dpkg/info/popularity-contest.postinst in the squashfs? [17:48] and then rerun with debugging in the same way [17:50] cjwatson: i'm going through that list of RAID bugs, i'm marking "Fix Released" for the ones I can't reproduce, and have a good reason to believe that the problem has been fixed by some recent work by myself, or others [17:50] I'll try that but I have to go back to work now :( - will do it after work and pastebin the debug again [17:51] I can't see anything in popularity-contest.postinst that should take significant time, so while I understand and am not questioning your comment, please understand that I'm thinking about possible secondary effects of removing popularity-contest as well as the obvious direct effect [17:51] cjwatson: inviting, of course, anyone subscribing to the bug to reopen if they can confirm the behavior in Intrepid [17:51] all very puzzling [17:51] kirkland: thanks, much appreciated [17:51] the only thing popularity-contest removes is the ubuntu-standard meta package placeholder [17:51] cjwatson: k, just checking if that's the proper response for bugs that were reported against something like Edgy [17:52] Fragadelic: at the package management system, certainly; but in theory its absence could cause ubiquity to skip some other step that it would otherwise have performed [17:52] Fragadelic: the reason I need the output of 'ps auxf' is to see whether popularity-contest.postinst is running at the time of the hang [17:52] Fragadelic: if it isn't, then that's pretty definite proof that popularity-contest itself is not at fault, but rather something near it [17:53] I can do that part real quick now if you like - how do I reset the entries cause if I remember correctly, ubiquity thinks its already been answered [17:53] kirkland: seems reasonable to me [17:53] Fragadelic: I'd rather not introduce further variables - start from a clean boot [17:53] ok [17:54] so DEBCONF_DEBUG is not needed right? [17:55] correct [17:55] thanks [17:57] running installer now - will pastebin the info you asked for shortly [18:01] http://pastebin.com/f5cf7af01 [18:02] huh, a zombie process [18:02] i guess - defunct - any ideas how to get around it? [18:02] or why it is happening [18:03] give me a minute! [18:03] i tried several different things - purgin popularity-contest - installing it but not configuring it,etc nothing seemd to work [18:03] thanks - no hurry [18:07] I really have to go now - take your time - I posted on the mailing list and maybe you can respond with what you find to that - tb6517@yahoo.com [18:07] thank you all very much for your time! [18:13] this really looks like the symptom you get when a postinst starts a daemon process that fails to close stdin [18:13] but popularity-contest doesn't do that [18:13] what changes have you made to your image? [18:20] cjwatson: would this break networking in d-i? http://pastebin.ubuntu.com/55062/ [18:20] or rather, do you think it could [18:21] I'm afraid I think it probably could [18:21] hw-detect calls update-dev and netcfg runs afterwards [18:21] indeed, that's what I suspected. [18:22] from poking through hw-detect [18:22] you could patch it in ubiquity [18:22] indeed, ugly, but it seems like the only other option. [18:22] ok, thanks for the advice! [18:22] np [19:40] cjwatson: what was the reasoning behind having the installer copy directly from the filesystem.squashfs as opposed to copying it from the working live system? I've always been curious about this [19:42] Fragadelic_: users tend to muck about in the working live system and expect a prestine copy of the OS to be installed to their disk. [19:42] lol - I always just install it and then muck later [19:43] did you ever try it the other way or was it always an executive decision to just install from the pristine file? [19:44] I can't speak to the research behind this as I wasn't working on ubiquity back then. cjwatson, as you rightly requested originally, would be able to speak on that. [19:49] ok thanks - I decided to start work on a simple installer based on one that I worked on for the Vector Linux livecd - trying to make it both cli and gui so it can be used with a live server as well [19:50] does anyone know if all the ubuntu kernels are capable of running in live mode - ie - have squashfs and loop? [21:00] ubiquity: evand * r2873 ubiquity/ (debian/changelog scripts/install.py): [21:00] ubiquity: Do not remove the files for packages that will break other packages when [21:00] ubiquity: removed (LP: #276657). [21:04] That's probably poorly titled, but I could not think of a better way of describing what it is doing. [21:04] Suggestions welcome. [21:20] I'm getting an error when trying to netinstall hardy because I request to pkgsel/include openssh-server but the client and the server versions don't match: openssh-server: Depends: openssh-client (= 1:4.7p1-8ubuntu1) but 1:4.7p1-8ubuntu1.2 is to be installed [21:21] cr3: pretty sure I do that. what are you using as a repo? [21:24] CarlFK: at first, I use the alternate image made available through apache. Then, I create a sources.list.apt-setup file containing: deb $archive_url $release main universe multiverse [21:24] CarlFK: where $archive_url is http://archive.ubuntu.com/ubuntu and $release is hardy [21:28] http://dpaste.com/83040/ [21:28] CarlFK: I moved my request for openssh-server from pkgsel/include to late_command string apt-install openssh-server [21:29] I run apt-cache on a local box so that I only dl the .deb's once - but if I am having troubles I use http://archive.ubuntu.com/ubuntu [21:37] cr3: are you using the 8.04.1 cd? [21:37] I bet the ssh-key thing is the root of your problem [21:39] CarlFK: yep, I'm using 8.04.1. maybe I could add: d-i debian-installer/allow_unauthenticated string true [21:43] i never needed that - your setup sounds like what I did before I used apt-cache [21:43] use the .deb on the cd as a local repo, then apt-get upgrade later [21:52] CarlFK: darn, the apt-install in the late command returns the same error message [21:53] wait... pretty sure you don't want that anyway [21:53] that will install it to the installer's environment (ram), not the disk you are installing to [21:54] CarlFK: what, the allow_unauthenticated thing? [21:54] i think [21:54] CarlFK: the apt-install should install to the disk, hence the example file which uses that for zsh [21:55] that sounds reasonable [21:56] how about skipping it and installing after boot? [21:57] do you plan on doing this once ibex is released? [21:57] where this=hardy [22:04] cjwatson: No changes at all to the image - it was a fresh install and then I just remastered right afterwards. It was one of my test of remastersys to make sure it worked properly with 8.04 [22:07] here is my remastersys script that does all the work - obviously the version I used didn't have the remove portion this one does. http://pastebin.com/f78cff54b [22:29] cjwatson: This is the problem with the metadata approach. A user may have all BIOS options turned off, but the metadata is still present. Unfortunately I can't tink of a way to only activate dmraid when asked for without having to deal with things like adding kernel command-line parameters, which to me seem not very user-friendly. [22:30] evand, oooh yuck. we just got the beta DVD into our factory install. "NotImplementedError: ubiqutiy.frontend.noninteractive.Wizard does not implement set_auto_login" [22:36] argh, fixing now. [22:38] evand, after you do, can you do a run through noninteractive and make sure nothing else is standing out? it unfortunately causes a ton of churn here every time we rev these DVD images for small installer things that break [22:41] absolutely [22:47] ubiquity: evand * r2874 ubiquity/ (debian/changelog ubiquity/frontend/noninteractive.py): [22:47] ubiquity: Fix a crash in the noninteractive frontend due to missing auto_login [22:47] ubiquity: functions. [22:47] ^ I'll test that in a bit. [23:21] evand: hey, today's server build isn't detecting the network or hard disks... known issue? [23:30] yikes, not known to me [23:40] evand: ugh