CIA-18 | ubiquity: superm1 * r2179 mythbuntu-ubiquity/ (debian/changelog scripts/mythbuntu/apply-services): fix typo in mysql-server | 03:15 |
---|---|---|
twb | cjwatson: hale. | 10:56 |
cjwatson | dia duit | 10:57 |
twb | So I'm using just the kernel & initrd to do the unattended network install -- rather than using casper over nfs and the live installer. | 11:01 |
twb | Does the simple d-i installer support GTK, like Etch? | 11:02 |
cjwatson | no | 11:02 |
cjwatson | I spent quite a bit of energy on that, but eventually decided to go the ubiquity path instead | 11:02 |
cjwatson | I think GTK d-i needs a lot more effort put into plugins before it's sufficiently usable | 11:03 |
cjwatson | but for unattended installs it doesn't really seem like a big deal | 11:03 |
twb | On Ubuntu, or generally? | 11:03 |
twb | Yeah, I was more curious than anything. | 11:03 |
cjwatson | well, generally, but obviously Debian can do what it likes :-) | 11:03 |
cjwatson | the partitioner is the main problem | 11:04 |
twb | What's wrong with it? It seemed fine to- ah. | 11:04 |
cjwatson | it's pretty much just a straight conversion of the newt installer | 11:04 |
cjwatson | the only win seems to be that you get to use the mouse | 11:04 |
twb | I think that's a corrollary of trying to smush the partitioner into the UI-agnostic infrastructure | 11:04 |
cjwatson | only because nobody's written plugins :) | 11:05 |
cjwatson | cdebconf plugins let you make things UI-aware | 11:05 |
twb | Interesting. | 11:05 |
cjwatson | but it's hard work to do for the partitioner | 11:05 |
twb | I get the impression that people have had trouble with the Ubuntu live partitioner, though, too. Under Feisty. | 11:05 |
cjwatson | and the infrastructure for cdebconf plugins is unfortunately not quite finished (libd-i needs work) | 11:06 |
cjwatson | sure, ours is far from perfect either, I acknowledge | 11:06 |
cjwatson | don't get me wrong, I'm not doing a "rah rah Ubuntu rocks Debian sucks" kind of thing | 11:06 |
twb | No worries. | 11:06 |
twb | It's clear to me you're level-headed about it. | 11:07 |
cjwatson | partitioning UI is hard; even the dedicated implementations (gparted, qtparted) have major flaws | 11:07 |
cjwatson | I don't think anyone's done it right on Linux yet | 11:07 |
cjwatson | at least not in a graphical way; the text-only ones are pretty good within that medium | 11:08 |
twb | Especially if you introduce multiple label types, LVM, md, and trying to maybe not blow away existing partitions. | 11:08 |
cjwatson | right. we have a design we're working towards that should accommodate LVM and md, but (again, spot the theme) it's hard work | 11:08 |
twb | Just "blow it away and do something sensible" is working fine. | 11:09 |
cjwatson | and the flaws will probably only become apparent once it's in place | 11:09 |
cjwatson | (https://wiki.ubuntu.com/Ubiquity/AdvancedPartitionerRewrite) | 11:09 |
twb | What I'm keen on is for the whole metainit/upstart-sysv-compat stuff to clear up so we can actually have daemons using upstart properly | 11:10 |
cjwatson | yeah, I'm no expert on that though | 11:11 |
twb | I did some work on thttpd recently and it's distressing how bad the init.d scripts are, because they're currently written by people who aren't experts on init, but rather experts on <whatever daemon> | 11:11 |
cjwatson | twb: so from your comments in /msg earlier I gathered that there was some problem with putting the keymap stuff on the kernel command line? | 11:12 |
cjwatson | are you running over the length limit? | 11:13 |
twb | cjwatson: no, it's just inelegant | 11:13 |
cjwatson | if you're prepared to modify the initrd, you can use a /preseed.cfg file in there for the stuff that comes before bringing up the network | 11:14 |
cjwatson | or, as I say, kickstart works around it by bringing the network up early before fetching the preseed file | 11:14 |
cjwatson | er, before fetching the kickstart file | 11:14 |
twb | cjwatson: I was already wondering about that so I won't have to run an HTTP daemon | 11:14 |
twb | Using file= that is | 11:15 |
cjwatson | you can preseed everything in /preseed.cfg if you like, that's valid enough | 11:15 |
cjwatson | file= doesn't come right at the start | 11:15 |
cjwatson | it comes after keymap handling too | 11:15 |
twb | Hmm, so without arguments it will use /preseed.cfg? | 11:15 |
cjwatson | file= is basically for CD installs when you want to do file=/cdrom/... | 11:15 |
twb | Oh, I see. | 11:15 |
cjwatson | or to fetch stuff off the hard disk | 11:16 |
cjwatson | (but nevertheless after hardware detection runs) | 11:16 |
cjwatson | yes, /preseed.cfg in the initrd is used unconditionally - any file= or url= arguments supplement it | 11:16 |
twb | If it uses /preseed.cfg from the initrd without being told, can I tell it NOT to do that -- for the odd occasion when I want to cancel out preseeding and do a fully manual install | 11:16 |
twb | ...or would I have to have a separate initrd? | 11:16 |
cjwatson | you'd have to use a separate initrd | 11:17 |
twb | OK. | 11:17 |
cjwatson | interesting thought though, maybe a wishlist on the preseed package | 11:17 |
twb | An example would be that all my systems would be seeded with TZ=Australia/Melbourne, unless we're building a box to ship out to a large customer. | 11:18 |
twb | Although normally I guess we'd make them a new netboot entry with different initrd/seed anyway. | 11:19 |
cjwatson | timezone's handled much later, of course, so you could do that with URL preseeding or similar if you wanted | 11:24 |
twb | Good point -- I could have the preseed in the initrd ONLY set up the keyboard and suchlike, and then do the rest via HTTP | 11:25 |
cjwatson | right | 11:32 |
Nyct | hello there | 12:34 |
Nyct | anyone free to give me some advice on installing Ubuntu - first time user here | 12:34 |
xivulon | cjwatson, loopinstallation requires to override a few files in the target system (init.d scripts, initramfs-tools, and update-grub), what is a good way to do that? | 12:42 |
xivulon | I'd guess that the patched packages should hit the gutsy repo and I'd only need to override the packages inside the ISO. Is that ok? | 12:45 |
xivulon | I still need to go over the patches and clean them up by the way (probably this w/e). | 12:46 |
cjwatson | xivulon: gutsy isn't going to change now | 13:25 |
xivulon | cjwatson, what do you suggest? | 14:03 |
xivulon | I mean overriding the files is ugly but easy, the issue is of course when people upgrade such packages. | 14:05 |
cjwatson | xivulon: for init.d scripts and initramfs-tools, write them as separate scripts rather than changing existing ones; that should be possible for the most part | 14:26 |
cjwatson | xivulon: what needs to be changed in update-grub? | 14:26 |
xivulon | Hmm umountfs must skip /host (and /boot) whenever umounting /host will cause / to jump... | 14:28 |
xivulon | Not sure I can do that in a separate script | 14:28 |
xivulon | So basically if umounting mountpoint X you unmount device Y which also host root, then you have to skip | 14:29 |
xivulon | As for update-grub kopt is not set correctly in the case of a loop installation | 14:31 |
xivulon | http://codebrowse.launchpad.net/~ubuntu-installer/wubi/gutsy/annotate/ago%40nb-ago-20071019072229-l4szhanyv3n0erks?file_id=updategrub-20071009210021-2g4lnkrfrydx012e-3 | 14:31 |
xivulon | Then of course you need to take care of /host & co in umountroot | 14:37 |
xivulon | But that can be done in a separate script | 14:38 |
cjwatson | OK, I'm just telling you what the constraints are. It is not possible to change gutsy for things other than major regressions, security problems, or data loss bugs. | 14:40 |
cjwatson | so that's what you need to work around in order to support gutsy | 14:41 |
xivulon | cjwatson, what we used to do before, was to have an init.d script that would patch subsequent init.d scripts. That will work, but it's not the most elegant solution. | 14:54 |
xivulon | dpkg-divert will be used for update-grub | 15:04 |
xivulon | The plan is to use dpkg-divert for init.d/update-gurb/initramfs-tools within lupin-support, and override the ISO to provide an updated lupin-support.deb. Any issue with that? | 15:20 |
cjwatson | you can't use dpkg-divert for anything in /etc | 15:21 |
xivulon | I thought it was too easy | 15:21 |
cjwatson | it doesn't work at all for configuration files (not shipped in .debs), and it breaks in corner cases for conffiles (shipped in .debs) | 15:21 |
cjwatson | dpkg-divert is probably reasonable for the rest of it | 15:22 |
xivulon | For init.d I can use an early start/stop script to override umountfs | 15:22 |
xivulon | Okthat's all I needed to know, thx | 15:31 |
xivulon | cjwatson, I really think that /sbin/mount.ntfs should be skept within sendsigs. | 17:47 |
xivulon | users that have ntfs(-3g) mounted and reboot, will end up with the ntfs killed by sendsigs as opposed of being unmounted | 17:48 |
xivulon | this is not a lupin only featrue, and can result in data loss/fs corruption | 17:48 |
xivulon | now reading your comment in #151580 | 17:51 |
cjwatson | please make this sort of comment in bug reports rather than by IRC | 17:52 |
cjwatson | I find it very difficult to keep track of bugs raised on IRC and would prefer people not to do it | 17:52 |
xivulon | Sure, I had missed your previous comment. I think you have a good point | 17:52 |
cjwatson | and shifting between bug reports and IRC when the issue isn't immediately resolvable makes it extremely difficult to look back later and see why something was done | 17:53 |
joshk | evand: hey | 21:52 |
joshk | is there a way to get proof that the ubuntu boot procedure did or didn't successfully fetch the preseed file? | 21:52 |
evand | joshk: I *think* it writes stdout/stderr to /var/log/casper.log | 21:55 |
joshk | ah, darn. network is unreachable | 21:55 |
joshk | DHCP being weird, i guess | 21:56 |
joshk | that kinda sucks, though. | 21:56 |
evand | what do you mean? | 21:57 |
joshk | reconciling syslog and capser.log, network was available 13 seconds later | 21:57 |
evand | hrm, I could move that further down in the chain of events. | 21:59 |
evand | cjwatson: ^-- thoughts? | 21:59 |
joshk | don't you have all sorts of dbus noise to figure out whether dhcp client has gotten a lease yet? | 21:59 |
evand | hrm, indeed. I'll toy around with that later. | 22:02 |
evand | But first I need to head to the bank. | 22:02 |
joshk | money's important. | 22:02 |
evand | 'tis indeed :) | 22:02 |
joshk | is there a way to tell preseed to mount an arbitrary device for preseeding? | 22:38 |
joshk | say, /dev/cdrom1 | 22:38 |
joshk | alternatively, a way to specify an arbitrary command on the command line | 22:48 |
joshk | (isolinux command line) | 22:48 |
joshk | is there like a ubiquity/really_early_command that runs before preseeding that i can sepcify maybe? | 23:32 |
joshk | *specify | 23:32 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!