[00:05] xivulon, use -us -uc [00:06] avoids signing files that way [00:06] superm1 thx [00:07] np [07:38] hrm, mpt is missing. [07:38] cjwatson_: is this acceptable in your eyes: http://evalicious.com/resize-partition.png [07:39] I figured the names can be grabbed from os-prober and .disk/info respectively [07:39] color suggestions welcome (I'll probably grab the theme colors) [07:43] oh and I'll shrink the height down when I wire up the size request to the label height. [07:47] or perhaps I should replace Ubuntu 8.04 with "Install Target" or something along those lines. Hrm. [09:24] evand: looks good! [09:24] fishing it out of .disk/info if possible would be neat, maybe with a fallback for when it isn't possible === cjwatson_ is now known as cjwatson [09:44] cjwatson: I need some more help with this iscsi.. I can't figure out the easiest way to test it. If I shove the udeb into the initrd (using EXTRAS), it's already unpackad, so it seems anna-install doesn't do anything, but I need it to execute the postinst script for it to ask me the relevant questions.. How to do? [09:46] do you have an installer-menu-item field in control? [09:46] if so, you should be able to go back to the main menu and select it [09:47] otherwise, (a) you probably should, (b) try udpkg --configure [09:47] Ah, no I don't. [09:49] Just Installer-Menu-Item: something or X-&? [09:50] XB-&, it seems. [09:57] right [09:58] if you want tedious facts ;-) [10:00] *g* [10:00] Erm.. Ok, now I get: [10:00] Unable to set title for open-iscsi-udeb. [10:01] and shortly after: [10:01] Menu item 'open-iscsi-udeb' succeeded but requested to be left unconfigured. [10:01] I'm guessing this means something to you? [10:01] cjwatson: ^^ [10:06] that means it exited with status 10 [10:07] unable to set title means you forgot to include a debian-installer/open-iscsi-udeb/title template, but is otherwise harmless [10:07] exiting 10 usually means either you deliberately backed up or else you failed to handle some other error from debconf [10:07] Exited with status 10? I'm guessing something debconfy does this if you don't play nicely? I certainly neve explicitly exit 10. [10:08] code 10 from debconf means "question does not exist" [10:08] so perhaps you messed up your templates somehow [10:10] cjwatson: Ah, that's quite possible. [10:10] Yes, my templates are not even in the udeb. wtf.. [10:11] that would do it. missing dh_installdebconf? [10:12] Er, yes. I probably removed it when I got offended that it didn't want to install my .config because it was a udeb... I was quite tired at that point :) [10:13] well, it was correct [10:13] you don't use .config in udebs [10:13] Exactly. [10:13] I wasn't thinking straight. [10:13] :-) [10:13] I'm better now (I hope). [10:15] er... [10:15] Oh, right, it's in control, not data. Never mind. [10:16] With a bit of luck this'll be ready later today, so we can have a new installer tomorrow with shiny new kernels full of virtio love and iscsi magic. Woo! [10:17] kewl [10:18] maybe by then my mirror run will have stopped eating ALL MY BANDWIDTH [10:19] :) [10:20] Hmm... Where are templates stored in the installer? They're not in /var/lib/dpkg/info.. [10:20] And /var/cache/debconf doesn't exist. [10:23] they're in /var/lib/dpkg/info in the initrd, until /lib/debian-installer-startup.d/S30templates runs and loads them into the debconf db in /var/lib/cdebconf/ [10:23] S20templates I mean [10:23] ...at which point they'll disappear from /var/lib/dpkg? [10:23] S20templates explicitly removes them to save memory [10:24] templates files tend to be pretty big, what with translations and all [10:24] I was just about to mumble something about memory :) [10:57] Woo! It's asking me about iscsi now. [10:58] * soren does a victory dance [10:59] nice [11:01] Hm... It needs a few modules to be installed. Should I just add them to the core-storage-modules or should I be adding a new udeb for that? [11:02] are they iscsi-specific? [11:03] Yes. [11:05] So add another udeb to the kernel build for the icsi modules and make the open-iscsi-udeb depend on that? [11:06] I think so, yeah [11:07] One step forwards, two steps backwards. [11:30] *headdesk* Why, oh why, is the key combo for closing a kvm window and for switching to the syslog vt in the installer the same? [11:31] * soren <---- not bitter [11:43] cjwatson: Any reason why getpwuid would fail in the installer? [12:08] Odd.. open-iscsi doesn't care if your uid is 0. It cares about whether or not the username assigned to your uid is "root". [12:08] and for some reason getpwent(0) fails in the installer. [12:10] * soren -> lunch [12:18] might need to depend on libnss-files-udeb [12:18] the nss modules you need aren't necessarily there otherwise === ebel_ is now known as ebel [13:04] cjwatson: Ah! Yes, that helped significanly. [13:04] W00t! http://warma.dk/iscsi-installer.png [13:05] 14:04:23 < soren> cjwatson: Ah! Yes, that helped significanly. [13:05] 14:04:42 < soren> W00t! http://warma.dk/iscsi-installer.png [13:12] soren: nice! === cjwatson_ is now known as cjwatson [15:37] cjwatson: Final piece of the puzzle: The finish-install bit. This is where I'm supposed to install the iscsi package into the target system and copy over the configuration, correct? [15:44] you'd usually want to install it a bit earlier than that [15:44] you can call apt-install at (almost) any point, as long as it's before the CD is unmounted [15:44] (phone) [15:46] Ok. I thought that was exactly the sort of thing finish-install was for. [15:46] Well, I can't do it until there's a filesystem to install it onto, so I need to hook into something that runs after partman. [15:46] (Just get back to me when you're off the phone. No rush) [16:20] soren: do you need to put the configuration in place before or after installing the package? [16:20] soren: usually it's best to ask for the package to be installed as part of pkgsel, which involves running apt-install in a post-base-installer.d hook [16:20] cjwatson: Doesn't matter, I guess. [16:20] I see. [16:21] "apt-install foo" just queues foo for isntallation later? [16:21] soren: if you can arrange it, the most elegant way tends to be to create the configuration in post-base-installer.d (when /target exists and has a base system in it) and queue it for installation with apt-install [16:21] actually, wait a moment [16:21] yeah, that's right [16:21] I create the configuration before partman runs. [16:21] I just need to copy it over to the target system at some point. [16:21] apt-install queues for installation if it's running before apt-setup, and installs it immediately if it's running after apt-setup [16:21] Ah, ok. [16:22] the benefit of queueing if possible is that it's faster and you get to piggyback on pkgsel's progress bar [16:22] (saves starting up apt yet another time) [16:22] Ok.. [16:23] So /usr/lib/post-base-installer.d/open-iscsi, I guess? [16:23] it is possible to install stuff from early finish-install.d hooks, but I prefer not to if possible [16:23] Any advice on the numbering? [16:23] doesn't make much difference [16:23] Ok. [16:24] hmm, let me just see how other similar things do it [16:24] notwithstanding the above it might be better to be consistent :-) [16:24] I just noticed that the ordering in finish-install could be very important. [16:24] finish-install definitely, post-base-installer not really [16:24] That too, yes :) [16:24] Ok. [16:26] hmm [16:26] This has been an interesting exercise. This is my first package to a) be a udeb and b) use debconf. Not to mention all the installer integration stuff. [16:26] the nearest thing I can find to standard practice is actually to queue the package for installation with apt-install in a suitable partman hook (basically once the user has committed to using it) and create configuration in finish-install [16:27] it's not really hugely important where you put the configuration in place, so meh, whatever [16:27] one of the problems with flexibility is that you can use it. :-) [16:28] * soren chuckles [16:30] So in finish-install, it should be <20, or thereabouts? [16:32] apt-install must be <15 [16:32] copying configuration to the target must be <95 [16:33] look at what's there and pick the thing that seems closest in spirit [16:33] Will do. [16:36] cjwatson diff of iso-scan is here http://paste.ubuntu-nl.org/55041/ [16:36] did not test it though [16:37] thanks, I saw [16:37] would be better as a bug if you don't want it lost :) [16:37] Bug #189630 [16:37] Launchpad bug 189630 in lupin "d-i/hdmedia should have hooks similar to the ones in lupin/casper" [Undecided,New] https://launchpad.net/bugs/189630 [16:37] will add iso-scan to it [16:39] https://bugs.launchpad.net/ubuntu/+source/iso-scan/ or https://bugs.launchpad.net/iso-scan/ ? [16:39] the former [16:40] done [16:40] please feel free to comment on custom-installation/preseed part [16:41] in casper basically I use a custom-installation folder that contains preseed + override files (if required) + hook scripts [16:42] at the moment I do not require much more than preseed, but in the past hooks and overrides were handy [16:44] would like to add something like that to d-i hdmedia too [16:56] in the patch above I noticed they were using find | grep -i as opposed to find -iname. Is that because the version of find is crippled? I assumed it supported maxdepth [17:23] re moving wubi from nsis to python and friends> with feature freeze next friday, I don't think it'd be a good idea. NSIS may be losing its maintainer, but it seems stable in its current form. [17:24] now that's not to say you can't work on that in a separate branch with the intention of using it for the Wubi UI in Hardy+1 [17:24] I agree with that, we can use python in next cycle, also for winfoss [17:25] That will bring up space requirement by about ~ 2MB [17:25] In my experiments, using pywin.mfc + py2exe/pyinstaller + upx, I can get away with 1.5-2.5MB for a full gui executable [17:26] not terrible. [17:27] also it will be inpractical to use native nsis package for building wubi this time around [17:27] that's ok [17:27] I think our current setup works fine for this release [17:27] since it does not support System plugin calls (low level dll calls) and that requires a C wrapper for each time I use that [17:28] evand if you manage to build it, I'd rather have you generate the builds [17:28] I do not feel confident in having those being downloaded from a public server [17:28] what do you mean by public server? [17:28] sourceforge [17:30] I and the old wubi contributors have ssh access, but if any of our passwords has been compromised that would be a big issue [17:30] You can't rotate the passwords? [17:31] for me yes [17:31] if you're really concerned, we can move the stable wubi release that gets pulled in to http://people.ubuntu.com/~evand [17:32] I like that better [17:32] pulled into the live cd* [17:32] ok [17:32] I'll take care of that later today [17:35] I was also working on a much overdue wubi code cleanup (including makefile reorganization) but I do not think I will have that by friday [17:35] that should not affect functionality, but would make the life of others accessing the code much easier [17:37] that becomes less urgent though considering that I will probably not use native makensis for the time being and that I might recode in python later on [17:38] mmm, I imagine it will have to wait until the next release if it's not ready by Friday. This being a LTS release I imagine that slangasek will not appreciate freeze exception requests. [17:38] ah, indeed. [17:39] xivulon: thoughts on Henrik's email and follow up on the wiki page? Is coding proposal 2 something you can take care of? [17:39] evand: hiya, you might know this: what's the name of the command to grab all the packages from a system which can be piped to another command so that the same packages can be installed on another machine? [17:40] I usually use dpkg-query for such things. I'm not sure if there's a more elegant solution though. [17:40] dpkg --get-selections? [17:42] cr3: dpkg-query -W -f='${Package}\n' [17:43] evand I haven't tried that with nsis and nsDialogs, if I can hook mouseover and click events should be doable, will need suitable artwork though [17:44] I might not be able to complete it by friday though, or maybe I will be able to do only a rougher version (no graphics) [17:44] dpkg-query -W -f='${Package}\n' | tr '\n' ' ' | sudo xargs apt-get install [17:44] xivulon: that's what I was looking for, thanks! [17:44] would be both ends [17:45] xivulon: ok [17:45] I am also looking at the buglist now [17:45] translation staff I can postpone correct? [17:45] I will need to change translation strings a bit [17:46] I thought wubi had i18n support, no? [17:46] yes but need to update some translations [17:47] so far someone else was taking care of that, now it's just me [17:47] and have to hook to rosetta [17:48] so by postpone do you mean postpone until after FF or until the next Ubuntu release? [17:48] after FF [17:49] that will mostly involve updating PO files and changing strings around [17:49] will try to do as much as possible before ff though [17:49] I *think* that's ok as the language pack deadline is April 17th. [17:50] that was my understanding too [17:50] but on top of importing po files I may also change some strings/variables internally [17:50] that too requires a cleanup [17:51] I'd make as much of an effort to minimize the size of the delta post-FF. Obviously do what you have to wrt to i18n updates, but the less code that gets changed, the less chance there is of a bug appearing in the 11th hour. [17:52] of course [18:11] I am going through my list should be doable by friday, would need some help on #189630, I will submit a patch for custom-installation-scan mimicking iso-scan [18:21] debian-installer: cjwatson * r873 ubuntu/ (build/config/common debian/changelog): [18:21] debian-installer: * build/config/common: Update for Hardy (oops; fortunately this only [18:21] debian-installer: mattered for manual builds). [18:24] evand on your side (tag wubi) any item that seem problematic? [18:25] cjwatson re metalinks, even if the feature is not implemented yet, we'd need to agree metalink urls soon since they will be hardcoded [18:26] in an ini file at least, but it will be annoying changing that before releases to reflect metalink url name changes [18:26] I might point to some symlinks/redirectable urls as well... [21:09] ubiquity: evand * r2439 ubiquity/ubiquity/components/summary.py: You probably don't want to install grub to a swap partition. [21:30] ubiquity: evand * r2440 ubiquity/ (bin/ubiquity-dm debian/changelog): * Updated the location of gnome-settings-daemon in ubiquity-dm. [21:33] oem-config: evand * r408 oem-config/ (debian/changelog oem-config-dm): * Updated the location of gnome-settings-daemon (LP: #190029). [21:34] oem-config: evand * r409 oem-config/ (aclocal.m4 configure configure.ac): Forgot to bump to 1.27 [21:36] oem-config: evand * r410 oem-config/ (d-i/manifest debian/changelog): [21:36] oem-config: * Automatic update of included source packages: console-setup [21:36] oem-config: 1.21ubuntu2. [21:43] oem-config: evand * r411 oem-config/debian/changelog: releasing version 1.27 === cjwatson_ is now known as cjwatson