[09:53] Argh, this stupid attempt to be helpful and not ask for a proxy is annoying! [09:54] Even when I seed the proxy setting in advance, it still tries to talk to archive.ubuntu.com :-/ === thom_ is now known as thom === cjwatson_ is now known as cjwatson [10:39] twb: I know. It's fixed in hardy now [10:39] (not that that helps you immediately) [11:35] hi, [11:36] ubuntu-installer can work on any debian based distro???? [11:42] hi, iam getting problem like this, [11:43] Traceback (most recent call last): [11:43] File "/usr/bin/ubiquity", line 166, in ? [11:43] main() [11:43] File "/usr/bin/ubiquity", line 163, in main [11:43] install() [11:43] File "/usr/bin/ubiquity", line 56, in install [11:43] wizard = ui.Wizard(distro) [11:43] File "/usr/lib/ubiquity/ubiquity/frontend/gtkui.py", line 177, in __init__ [11:43] self.customize_installer() [11:43] File "/usr/lib/ubiquity/ubiquity/frontend/gtkui.py", line 316, in [11:43] customize_installer [11:43] gtk.window_set_default_icon_from_file('/usr/share/pixmaps/' [11:43] the Ubuntu installer depends on modifications made in Ubuntu [11:43] GError: Unable to load image-loading module: [11:43] /usr/lib/gtk-2.0/2.4.0/loaders/libpixbufloader-png.so: [11:43] /usr/lib/gtk-2.0/2.4.0/loaders/libpixbufloader-png.so: cannot open [11:43] shared object file: No such file or directory [11:43] Traceback (most recent call last): [11:43] File "/usr/bin/ubiquity", line 166, in ? [11:43] that sounds like your GTK installation is broken [11:43] main() [11:43] File "/usr/bin/ubiquity", line 163, in main [11:43] install() [11:43] File "/usr/bin/ubiquity", line 56, in install [11:43] wizard = ui.Wizard(distro) [11:43] File "/usr/lib/ubiquity/ubiquity/frontend/gtkui.py", line 177, in __init__ [11:43] self.customize_installer() [11:43] please do not paste large chunks of text here in future; there are a number of pastebin services for that kind of thing [11:43] File "/usr/lib/ubiquity/ubiquity/frontend/gtkui.py", line 316, in [11:43] customize_installer [11:43] gtk.window_set_default_icon_from_file('/usr/share/pixmaps/' [11:43] GError: Unable to load image-loading module: [11:43] /usr/lib/gtk-2.0/2.4.0/loaders/libpixbufloader-png.so: [11:43] /usr/lib/gtk-2.0/2.4.0/loaders/libpixbufloader-png.so: cannot open [11:43] shared object file: No such file or directory [11:44] sorry!!!!!!!!how to rectify that one?? [11:44] I don't know, it depends on your system. It works fine in Ubuntu ... [11:45] if you don't know for yourself, file a bug with your vendor (pointing out that you know they don't support the Ubuntu installer, but that the problem is that PyGTK can't load PNGs) [11:46] 19:39 twb: I know. It's fixed in hardy now [11:46] cjwatson: awesome, that WILL certainly help in 6 months [11:47] cjwatson: I'm also nagging our netadm about changing our network so non-proxy packets get dropped immediately. [11:48] sridhar: welcome to dll hell. Instead of installing the .deb with dpkg -i, build it from source. Add a deb-src entry to sources.list then do apt-get --build source ubiquity, or whatever. [11:49] err [11:49] that won't have the slightest effect on his problem [11:49] Oh, on reading cjwatson's response I may have things wrnog [11:49] please do not do the above as it will just make things more complicated [11:50] cjwatson: is it saying that he's somehow got a version mismatch between Python's GTK bindings and libgtk? [11:50] ubiquity isn't linked to any particular version of gtk; if there is a problem of that kind, it's in pygtk, not in ubiquity [11:50] but ultimately, it's something his OS provider is best placed to resolve [11:51] I don't want to get into trying to fix somebody's pygtk installation [11:51] Fair enough [11:51] I'm used to #emacs where we happily stray from the topic in order to fix a problem :-) [11:52] hi, iam getting the live CD from Debian-live, and trying to make Live come install Cd, by using ubiquity. There iam getting the problem like that [11:53] sridhar: are you in #debian-live on OFTC? [11:53] twb: I don't object to that, but this is a case where it's pretty easy to bugger up your system, so we should be correct if doing so :-) [11:53] sridhar: odd that Debian Live's pygtk would be busted [11:54] cjwatson: he's probably using Sid [11:54] cjwatson: debian live rarely works at all, so mainly people have to use Sid :-/ [11:54] I expect that making ubiquity work on Debian will be a considerable effort [11:54] quite possibly a worthwhile one, but still considerable [11:55] Incidentally, when I tried to use Ubiquity the other day it looked like the partitioner didn't support md *or* lvm [11:55] it doesn't [11:55] That sounds pretty lame [11:55] there's a plan for the UI [11:56] but I was much more interested in getting plain old partitioning working well [11:56] rather than diving down that rabbit-hole [11:56] I would never deploy a system without LVM in this day and age [11:56] Perhaps I'm a weirdo. [11:56] I think it's perfectly reasonable for you to use the alternate or server CD, then :-) [11:56] Those use d-i, yes? [11:56] yes [11:57] install from live CD to LVM seems sort of niche to me [11:57] I'd like to support it eventually, but I don't think it's critical [11:57] The reason I was even trying it was because neither the CD drive nor the NIC were being detected [11:57] I'd love to be in a world where the desktop tools for LVM were good enough that we could deploy it by default [11:57] at the moment I think we'd just be making a load of inexperienced folks play around with the command-line LVM tools and I doubt much good would come of that [11:58] do you have a bug (or two) about the detection problems that I can look at? [11:58] Erm, sorry, no. [11:59] The hardware is in the office, and I was trying things haphazardly because I was in a hurry, and I didn't take notes. [11:59] But launchpad requires a GUI browser anyway so I don't use it. [11:59] I always reproduce any bugs I find under Debian and then report them to Debian, because it's so much easier. [11:59] launchpad works fine with w3m [11:59] Not for me. [11:59] I use it quite often [12:00] I can't log in with w3m [12:00] It always just takes me straight back to the login page [12:00] Non-logged-in-ness works. [12:01] *blink* you're right, they broke that recently it seems [12:01] It's been broken for months [12:01] #launchpad told me it was because they use cookies wrong, and they do this because otherwise everyone would have to go to www.launchpad.net instead of launchpad.net so they refused to fix it [12:01] anyhow, you can file Launchpad bugs by mail as long as you can GPG-sign the initial report [12:02] Yeah, I know [12:02] https://help.launchpad.net/BugTrackerEmailInterface [12:02] It's still a major pain compared to bts(1) and reportbug(1) [12:02] bug 59510, I see [12:03] bug 152706 has a clearer description [12:03] that actually sort of suggests that it may be fixable in w3m [12:03] depending on your value of "fix" [12:04] I do remember it was a SATA optical drive and an integratd e1000 NIC with an ICH9 southbridge. [12:04] sounds pretty mainstream [12:04] Yep [12:05] it was ironic -- I netbooted, and the initrd couldn't see the nic. [12:05] I CD booted, and it couldn't see the CD drive [12:06] OK, so using d-i, if I answer a question via the GUI can I then switch to vt2 and find out what setting and value it was? [12:07] yes, if you know the underlying question name [12:07] debconf-get can extract it [12:07] debconf-get-selections? [12:07] possibly easier, though: run the whole thing in debug mode [12:07] boot with DEBCONF_DEBUG=developer and syslog will have a trace of all debconf activity [12:07] no, not debconf-get-selections, but never mind, do the above :) [12:07] cjwatson: on the kernel command line? [12:07] right [12:07] Cool, thanks [12:08] you'll see an INPUT for everything that corresponds to a question being asked in the GUI [12:08] (some INPUT commands may not be asked, but everything that's asked will come from an INPUT command) [12:09] if you're using this for preseeding, that's mostly a reasonable way to do it, but beware that it will not work in the partitioner; see the documentation for that instead [12:10] OK, thanks. [12:11] * cjwatson notes that RFC2965 Obsoletes: 2109 [12:12] hmm, that may not be useful though [12:17] Regarding the proxy issue, our netadm writes: "It isn't the proxy server configuration. It is the firewall which doesnot understand TCP window scale options. If you can set net.ipv4.tcp_window_scaling = 0 either via the command line or in /etc/sysctl.conf then the problem should go away. A rebuild of the firewall is listed as part of the move which should fix this problem." [12:18] I don't suppose it's possible to do that via preseeding or kernel parameters? [12:19] Never mind, apparently he was confusing my complaint with an unrelated issue. [12:24] Argh, the busybox on the installer is real old [12:24] no grep -C [12:25] it's not that it's old, it's that it's reduced [12:25] that being the point :) [12:25] debian/config-udeb:# CONFIG_FEATURE_GREP_CONTEXT is not set [12:26] It has 1.1; my router has 1.4 :-( [12:26] I normally use nano -v's search [12:26] it is not relevant that it is old in this case [12:26] OK. [12:27] I'm largely waiting for Debian to update since I don't want us to end up maintaining busybox ourselves :) [12:27] Understandable [12:31] Ah, I can find the line number with grep -n then print it with sed -n x,yp [12:31] Is netcfg/get_hostname asked before fetching the preseed file? [12:32] for netboot installations, yes [12:32] It seems to be, and that kinda makes sense if it wants to send that hostname out via preseed. [12:32] *via DHCP [12:32] OK. [12:32] well, also that it's part of bringing the network up in general [12:32] (it perhaps doesn't need to be, but right now it is) [12:33] Is there a way to avoid having all my machines end up with hostname=ubuntu? [12:33] set something else by DHCP [12:33] Well, I do. [12:33] So they will all be called dhcpNN, and the kernel option is ignored? [12:35] # Any hostname and domain names assigned from dhcp take precedence over [12:35] # values set here. However, setting the values still prevents the questions [12:35] # from being shown, even if values come from dhcp. [12:35] d-i netcfg/get_hostname string unassigned-hostname [12:35] d-i netcfg/get_domain string unassigned-domain [12:35] from https://help.ubuntu.com/7.04/installation-guide/i386/preseed-contents.html [12:35] Yes, but I don't believe everything I read. [12:36] That's for the preseed.cfg, which may not be the same as the kernel option [12:36] try things you read before asking, though :-) [12:36] Point. [12:36] Things are very slow to try, though, because of the proxy problem. [12:36] So while it's booting, I start asking. [12:37] kernel options get turned into preseeding; they're equivalent to /preseed.cfg in the initrd aside from minor matters of quoting [12:37] OK. [12:38] Can you only preseed d-i options via kernel parameters, or can you seed, say, tzdata/Zones/Australia? [12:38] (Just curious) [12:40] I don't know if it's documented, but 'tzdata:tzdata/Zones/Australia=value' [12:41] the bit before the colon should be the package being preseeded [12:41] Oooh! [12:41] if there's no colon, it's implicitly 'd-i' [12:41] I think that's undocumented; I've never seen that. [12:42] it's in the current upstream installation guide [12:42] and in fact it's in the gutsy guide too; may just not have been in time for feisty [12:43] I can't see it in the gutsy guide; what page number? [12:43] hang on while I install the PDF [12:43] Oh, 83 [12:44] ftp://twb.ath.cx/tmp/output.pdf is just Appendix B of the i386 Gutsy guide [12:44] that's the one [12:44] it's OK, I have it [12:52] cjwatson: How does the the alternate installer decide which kernel to install? [12:53] soren: base-installer/kernel/*.sh [12:53] cjwatson: Um.... Err.. [12:54] cjwatson: Ok, pretend I'm a complete idiot.. [12:54] cjwatson: I've unpacked the contents of the alternate CD.. [12:54] cjwatson: As per the instructions on the wiki. [12:54] find base-installer_*.udeb and look for /usr/lib/base-installer/kernel.sh in that [12:55] there's also preseeding for it if you just want to hit it with a hammer [12:55] Gah, phone.. [12:56] e.g. 'd-i base-installer/kernel/override-image string linux-server' [12:56] Ok. [12:56] I find that more appealing than reworking the base-installer udeb. [12:56] Does d-i support non-linux flavours of debian? [13:02] twb: there has been some initial work on it but it hasn't been completed [13:02] I did a fair amount of work on porting to the Hurd but got stuck on getting the damn thing to boot with something roughly equivalent to an initrd [13:04] Heh [13:04] Do you use syslinux to load the hurd? [13:04] much of the same work should apply to FreeBSD provided that somebody solves the same problem [13:05] goodness no, grub is the only supported way to boot the Hurd [13:05] chaining from the BIOS wasn't the problem, it was getting a ramdisk set up with writable / [13:05] or at least a good emulation thereof [13:05] maybe some day I'll come back to it [14:12] cjwatson: What was I supposed to put in the preseed file to make the installer accept unauthenticated stuff? [14:22] soren: 'd-i debian-installer/allow_unauthenticated boolean true' or debian-installer/allow_unauthenticated=true on the command line [14:23] cjwatson: Alright. Thanks! [14:37] cjwatson: Does that affect the installed system, by the way? [14:38] cjwatson: I.e. is it carried over into apt.conf? [14:38] I don't recall, and unfortunately don't have time to check at the moment [14:38] No problem. [14:50] Is the OEM mode supposed to prompt me to create a user at install time? [14:51] only the OEM customisation user [14:55] Ah, I see from the wiki that it creates a user called `oem' [14:56] Any idea if this user gets uid=1000? For purely aesthetic reasons I'd like the "normal" user to get that ID [14:56] the oem user is 29999, and your normal user will be 1000 [14:56] Cool. [15:09] What's an "Ubuntu Studio desktop"? [15:09] Is that basically GIMP and Inkscape? [15:30] twb: I believe it's more geared towards audio engineering. [15:31] ah, I stand corrected. Audio, graphics, and video. [15:38] Righto. [16:47] evand, at the current moment does migration assistant detect /home on existing linux installs? [16:47] evand, and copy that over if requested? [16:48] superm1: it copies individual items out of it, not the whole thing [16:48] provided it's either part of / or mounted using UUIDs [16:49] evand, well i'm considering proposing / attempting to implement a spec for hardy to properly import knoppmyth and mythdora existing installs into a new mythbuntu install. Before i can commit to it, i'm trying to gauge the difficulty (based on the existing code and your opinion of how difficult it should be). What do you think? [16:51] superm1: so to clarify, this would mean in those cases copying over /home/$USER for those accounts? [16:52] evand, well it would copy over items from /home, and probably a few items in /etc and a mysql database [16:52] How does the installer decide if it should install usplash? [16:53] copying /home/$USER in its entirety was listed in the previous m-a specification. I make no guarentees as to whether or not I'll have time to get to it in Hardy though as m-a isn't scheduled and lots of other things are :) [16:53] but /etc and mysql would be outside the scope of m-a [16:53] I should also note the preserve /home specification [16:54] that's what i was fearing (regarding the /etc and mysql) [16:55] mysql would probably require some sort of binding of /proc and /dev, a chroot run to start the sql server, and then grab the table it needs, and then stop sql, and un bind mount those areas [16:55] there are cases where usplash isn't installed? [16:56] oh, server, duh [16:56] evand: :) [16:56] evand: Yeah. Is it buried in the code somewhere or does it depend on the presence of the usplash packages or something entirely different? [16:57] soren: not entirely sure, I'd have to dig through d-i, which I can do unless cjwatson has a quick answer. [16:57] usplash is just done by seeds nowadays [16:57] it used to be more complex [16:57] desktop: * usplash [16:57] desktop: * usplash-theme-ubuntu [16:57] so any desktop install will get usplash [16:58] cjwatson: Ah, of course. [16:58] cjwatson: ... "'d-i pkgsel/include string usplash-theme-ubuntu" ought to do it, too, I guess. [16:59] yep [16:59] Cool. [16:59] Thanks again! [16:59] Ok, now I'm just blocking on lack of upstream bandwidth... /me sighs [16:59] Well, s/blocking/stalling/ [17:00] * evand adds that to the list of things that should've been obvious to him. :) [17:00] * soren too :) [17:01] bear in mind it took me two releases to reach this "obvious" conclusion [17:01] *g* [17:02] actually it did use to be harder due to how usplash generates its configuration file [17:02] which is still a bit of a mess - it does formally disallowed things with debconf [17:03] cjwatson, i was trying to investigate what would need to be done to get the mythbuntu disks be part of the normal build process, and was glancing through your ~cjwatson/cdimage/mainline branch. Can you point me to how these live filesystem that find-live-filesystem grabs are built? [17:04] superm1: the livecd-rootfs package [17:05] cjwatson, at a bzr branch somewhere i should look, or the direct package in the archives? [17:06] https://code.launchpad.net/~ubuntu-core-dev/livecd-rootfs/trunk [17:07] ah okay. debian/control and debian/copyright don't mention that in the package's source. [17:07] thanks [17:09] this probably would have been useful for us to look at a lot earlier before hand writing a script to do such things :) [17:09] I've added XS-Vcs-Bzr in bzr now [17:09] it was only released publicly in July [17:10] oh i see [17:20] cjwatson, glancing through the source on this, its fairly similar to how we were building the live filesystem used on our disks. Only exception is that several items that are supposed to be in /home/ubuntu (but not future users that are created) are copied over in our process. What is the best way to work around that? [17:30] we deal with /home/ubuntu in casper [17:30] doing it in the squashfs is a mistake IMO, because it means it gets copied to the installed system and has to be removed by hand [17:30] yeah i was realizing that shortly ago