=== superm1|away is now known as superm1 [07:27] cjwatson, I've been looking over cdimage and was wondering if you had any experience running the daily cron using another server's pool instead of rsyncing it locally [07:27] or if that's even possible [07:48] I'm going to get some sleep, I'll try you tomorrow [09:43] cjwatson: What's the major fail with i386 on the cd builds do you know? [09:45] live only [10:35] acoc: it's not possible with debian-cd as it stands [10:36] davmor2: more detail please, I just got up [10:46] cjwatson: sorry afk [10:48] I386 desktop doesn't show up here http://people.ubuntu.com/~ubuntu-archive/livefs-build-logs/intrepid/ubuntu/20080926/ at all which is why I'm guessing it is using the 23rd image but because it's not showing up I don't know why? So I'm assuming a major failure of build being as amd64 is there... [10:48] daily alternate seems to be up to date though [10:51] looks like the livefs build machine got stuck on a lock [10:52] I've cleaned up antimony's side of it and asked our sysadmins to fix terranova [10:52] cjwatson: Okay cool just throw me when it didn't show up at all :) === cjwatson_ is now known as cjwatson === davmor2 is now known as davmor2_Lunch === davmor2_Lunch is now known as davmor2 [14:01] cjwatson: StevenK and I were looking at how to add the local installation pool to the ubuntu-mid images, and got a little lost in the debian-cd internals. Would you be able to provide any guidance? [14:02] I got very lost in debian-cd :-( [14:05] persia: do you have a separate archive already, or do you just want to throw in a few .debs? [14:05] err. I'm not actually sure what you mean, thinking about it [14:05] "local installation pool"? [14:06] cjwatson: Like pool and dists on the Live CD [14:06] On the live CD, there's a directory /pool/ in the VFAT, containing a local archive snapshot. We'd like to do the same sort of thing. [14:06] Where persia says VFAT, he means ISO9660 :-P [14:07] Indeed :) [14:07] is there any desire to add mythbuntu or other flavors that provide ubiquity live CD to wubi? [14:08] xivulon: I'm interested, but I'm not sure about the timing for the ubuntu-mobile flavour. [14:11] persia, as we discussed at UDS, I am a big fan of any flavor which can be used as a platform for providing a simplified desktop for young/elderly users [14:13] xivulon: There's one big outstanding bug to hit first, but as soon as that gets hit, I'd love to look at how to make it work with wubi. I don't have any windows though. [14:14] don't need one for that I can provide a script to loopinstall onto another ntfs/ext3 partition for testing purposes, also in vm, (which is my testing rig) [14:15] persia: ok, I'll take a look ... [14:15] now *I'm* lost in build-mobile*. [14:15] StevenK: Please provide guidance. [14:16] is there any tool to recursively copy a filesystem tree into a vfat image using mtools? [14:16] Not that I've found. [14:17] OK, I suppose that could be hacked up with find+xargs [14:17] cjwatson: Ignore build-mobile, it essentially does download-live-filesystems and publishing, build-mobile-img is the script that creates the vfat. I suspect you've already discovered this. [14:17] ok, um, so. I think the first thing you need to do is to bin build-mobile and integrate its functionality into build-image-set. [14:17] because you're going to need nearly all of build-image-set anyway in order to make this work [14:18] But build-image-set scares me [14:18] yes. but it's doing the work you need. [14:18] you need the whole sync-an-archive and call ./build_all.sh stuff [14:18] not to mention running germinate [14:18] ... We do? [14:18] yes, you do [14:19] that's what builds pool/ and dists/, ultimately [14:19] Oh, right [14:19] otherwise where are you going to get it from? [14:19] We call build-image-set for the dailies actually, we just drop out early. [14:20] where's that? [14:20] oh dear god [14:20] that's horrible. doesn't count. :) [14:20] persia, can mid be run as a desktop on standard PCs? [14:20] Oh yeah. We use it for setting up mail and then we jump to build-mobile. [14:21] xivulon: -mid might be a little funny, but ought work on newer machines. -mobile *definitely* can run as a desktop on regular PCs. [14:21] no, I meant using it properly [14:21] build-mobile isn't really all that different - you should just need to set IMAGE_TYPE=mobile or something and then key off that [14:21] cjwatson: Hm. Maybe I shouldn't have pointed that out. :-) [14:22] and you've already got $PROJECT set [14:22] persia: looks intersting will play with that in coming days [14:22] superm1: what are your thoughts to have mythbuntu in wubi? [14:22] so you can make run-germinate check that to decide which seed collections to use, and germinate-to-tasks can pick out the right seeds from that and select the right set of packages [14:23] xivulon: Thanks. Please let me know if you need anything. [14:23] cjwatson: But it looks like all of the heavy lifting is done in build_all.sh, which is ... odd code [14:23] oh, and list-seeds would need to spit out the right set of top-level seeds [14:23] StevenK: don't worry about build_all.sh for now [14:23] ps is there any plan to do a livecd for gobuntu? [14:23] actually, it will largely just do what you tell it [14:23] xivulon: no, we aren't building Gobuntu any more [14:23] cjwatson: From my reading of it, it eats small children. [14:23] https://lists.ubuntu.com/archives/ubuntu-devel-announce/2008-June/000434.html [14:24] StevenK: honestly, most of cdimage is dedicated to setting up debian-cd so that you can just point it in the right direction and say "eat that *very specific* small child over there" [14:24] Heh [14:24] you are going to have to use it if you want pool/ and dists/ [14:24] otherwise you'll have to reinvent it [14:25] cjwatson: thx had missed that, didn't notice the "Free Software Only" option (also because I am mostly in automatic mode) [14:25] cjwatson: So, I can see where you're going, I can leave CDIMAGE_LIVE set and have download-live-filesystems run. Maybe then I have a check for CDIMAGE_MOBILE and run build-mobile-img [14:26] right, that's sort of what I was thinking [14:26] And as a bonus, we can have proper tracking of preseed files with debian-cd. [14:26] I never liked build-mobile being a separate script anyway :) [14:28] So I'm probably going to have to set IMAGE_TYPE, which I've sort of ignored up till now [14:31] cjwatson: Mind you, I like the new build-mobile a lot more than the old one [14:32] xivulon, they would have to be dedicating a significant amount of space to do so [14:32] xivulon, for recordings and such [14:32] i think if the frontend only mode were activated however, its doable [14:32] and quite interesting then [14:33] StevenK: I think you ought to. IMAGE_TYPE defines the "form factor" of the image, and should be used when you're creating a different type of object - d-i CD vs. live CD vs. DVD vs. VFAT mobile image [14:34] PROJECT is supposed to control the contents of the image (list of packages it contains, preseed files, particular live filesystem to fetch, etc.) [14:34] cjwatson: Yes, I'm guessing I'm also going to have to either do strange things to cron.daily-live or write cron.daily-mobile [14:34] cron.daily-mobile would be fine [14:34] the cron.* scripts are all just thin wrappers to encapsulate setting environment variables and stuff [14:34] superm1 would anything prevent users from storing stuff on ntfs directly? [14:35] users can allocate a "normal size" (~10GB) but still access rw all of the host disk (/host) [14:35] xivulon, performance would be questionable i think [14:35] xivulon, but otherwise i suppose not.. [14:36] hmm /host access is native ntfs no loopfile involved [14:36] cjwatson: Right, so aside from looking at how publish-daily works (meaning publish-mobile dies too), I need to do what to ./build_all.sh ? [14:37] xivulon, well at least implementing the frontend only portion would be more feasible to start and see how that goes [14:37] xivulon, additional questions would have to be asked once ubiquity took over though or asked during wubi [14:39] superm1: the options are a) we preseed those questions, b) we show them in wubi interface (unlikely and certainly not in intrepid), c) we ask them after reboot (not nice) [14:40] xivulon, well at this point there are a few that are "critical" to be asked - [14:40] It would be possible to show a few profiles in wubi and preseed from there [14:41] xivulon, the ones i'm referring to tell the machine the user/pass of the backend it connects to [14:41] and the hostname/database [14:42] i think it'd be fine to just ask them after reboot during --automatic mode though [14:42] I'd guess it might be passable to leave that blank and have a dialog after rebooting [14:42] yeah [14:42] xivulon, let me find the preseed that i was using for enabling --automatic this summer [14:43] We could have one profile for frontend only and one frontend + backend [14:43] evand/cjwatson any view on the above? [14:43] yeah, develop frontend first and then frontend+backend after proving frontend first works [14:44] for me it is only a matter of changing the configuration file, and add a new preseed template [14:44] there should be no change in code [14:44] well in theory it should just "work" w/ the ubiquity changes i did this summer too [14:45] so if you can throw something together, i'll throw a windows vm together and give it a shot [14:45] http://paste.ubuntu.com/50877/ [14:45] so that should be all the possible questions that get asked [14:45] StevenK: you should do nothing at all to build_all.sh itself. Leave it alone. [14:46] xivulon, so of the mythbuntu specific ones, would need to preseed the advanced_install, install_type, and the mythbuntu summary page [14:46] cool will play with that today, will try to setup a backend first (yesterday attempts failed as my only desktop has only tv out which did not work...) [14:46] the rest would be optional [14:46] StevenK: the only things you should need to edit are tools/boot/intrepid/boot-* (either to make the image bootable in whatever's the appropriate way, or to skip it and let build-mobile-img do it), tools/add_live_filesystems (to add mobile instead), and perhaps preseed files in data/intrepid/preseed/ [14:47] xivulon, oh the proprietary driver installation stuff is fixed in ubiquity trunk, but broken in our alpha6 disk [14:47] if that's where things went wrong [14:47] (if you didnt see it during install, that's a different problem) [14:47] yep I thought that was the issue, it's an nvidia video with tv out [14:47] xivulon: I have too much to do the week before beta to help with adding new features which should have landed several weeks ago in order to be ready for 8.10 ... [14:48] xivulon, if this just "works" without too much effort, then i'm for it and will get some people on my team to help test it, but if we run into a handful of problems, lets defer to jaunty okay? [14:49] cjwatson I appreciate that, but as mentioned this should only involve a change of configuration files, and we would go through the feature freeze exception approval anyway [14:49] superm1 that is what I had in mind [14:49] xivulon: Don't do an FFe just on account of ubuntu-mobile support. Many of the target devices come with linux anyway. [14:49] xivulon: superm1 is much better placed to figure out what to do with mythbuntu than I am. I'd have to educate myself about it all from scratch [14:50] cjwatson: build-mobile-img already deals with preseed files [14:50] cjwatson, yeah i'll work with xivulon to look at this [14:50] StevenK: ideally, those ought to move into data/intrepid/preseed/, IMO [14:50] the way it is now is just really awkward NIHJ [14:50] NIH [14:51] I can understand why you did it that way, but if you're reengineering things anyway, might as well bring it into sync [14:51] I did it that way because build-image-set and friends made me go "OMG, my head is exploding" [14:51] StevenK: I'd rather have the preseed files live someone accessible. [14:52] cjwatson: So if build_all.sh is going to build the filesystem, what needs to be changed so it knows how? [14:53] I suspect making it actually build the entire filesystem is a little optimistic for now. However, you could make it build the filesystem tree and then bundle that into VFAT in build-mobile-img [14:53] That works [14:55] edit the bin-images target in debian-cd/Makefile, and stick an extra case at the top of the if [ "$(DOJIGDO)" = "0" ] block to cover the "don't build an ISO at all" case [14:56] * StevenK is currently trying to determine where the checkout of debian-cd lives [14:56] BTW, what are you planning to put in pool/? [14:56] sftp://antimony/srv/cdimage.ubuntu.com/bzr/debian-cd/ [14:57] I was hoping persia could tell me. [14:57] * StevenK is also thinking that IMAGE_TYPE=mobile doesn't appeal [14:57] Since that turns up in the published path [14:58] I wanted to put in the DKMS-enablement stuff and some alternate networking for people with devices where the network didn't work out of the box (like mine). [14:58] published path? [14:58] cjwatson: http://cdimage.u.c/.... [14:58] oh, just special-case publishing *shrug* [14:59] the primary purpose of IMAGE_TYPE is to control form factor; publishing URLs are secondary and it's OK to special-case those [15:00] Oh my god, my eyes. [15:01] Who wrote this? [15:01] Oh, Raphaƫl. That explains so much [15:01] err, yeah, don't look at the Makefile too hard [15:01] various people have had a go at rewriting bits, and I think it's more sane in Debian now, but I can't face the merge [15:02] cjwatson: I think you'd need to be *completly* smashed to start it ... [15:04] persia, if these are "extra" modules that don't conflict with anything, it shouldnt be too late to add them to the kernel though i'd think? [15:04] I haven't been that drunk for some time [15:04] cjwatson: Reminds me of code at $OLD_WORK. "# eyes closed now. <40 lines or so of evil TeX> # okay, you can open your eyes. If you can read this, you were cheating!" [15:04] heh [15:05] * StevenK re-reads cjwatson's comment, trying to apply it to the code [15:05] superm1: For modules, I'm working with kernel devs, but I'm thinking of the same stuff that comes from the ship-live seed. [15:06] persia, so these would be NEW packages then? [15:06] persia: right, this is the sort of thing that the live CD does [15:07] cjwatson: I'm guessing I want ifeq ($(CDIMAGE_MOBILE),1) and else and endif around that block, but that breaks the like 50 lines that are \'d [15:07] so the live CD is a sensible model [15:07] StevenK: no, use if in shell [15:07] if [ "$(CDIMAGE_MOBILE)" = 1 ]; then ... [15:07] superm1: No. Just existing packages on the CD. For the modules, I've been told that it's just a matter of understanding why they aren't loaded. [15:08] cjwatson: If I have to follow the coding style, I may claw my eyes out in sheer self-defense. :-P [15:09] I want else; \ or else \ ? [15:09] I hardly ever collapse if's to one-line [15:11] ok, you can stop going on about how awful everything is now. :) [15:11] if [ ... ]; then \ [15:11] ... [15:11] elif [ ... ]; then \ [15:12] cjwatson: So, in that bit, I want to make my filesystem tree under $(BDIR)/CD$$n ? [15:12] err, I think you misunderstood me [15:12] debian-cd will make the filesystem tree under $(BDIR)/CD$$n [15:12] in bin-images, you just have to arrange *not* to call mkisofs - you don't have to build anything [15:16] cjwatson, do you happen to know about how much space would be required for a single dist pool sync [15:17] acoc: for which components and which architectures? [15:17] for i386 and sorry what were components again [15:18] main, restricted, universe, multiverse [15:18] just main [15:18] cjwatson: Oh, right [15:19] cjwatson: Do I also want the $(jidgo_cleanup) bit? [15:20] StevenK: yeah, you might need to guard that too [15:21] I can't remember, check what the called script does [15:22] acoc: a bit over 6GB [15:22] cjwatson: oh yeah that's not bad at all thanks [15:23] cjwatson: Meh, it's dealing with .jigdo, don't care [15:32] cjwatson: Is there anything I need to do in that Makefile? [15:33] Er. Anything *else* [15:35] not that I can think of right now, but consider this an iterative process ;-) [15:35] * StevenK grins [15:36] preseed file added, too [15:36] speaking of iterative processes, I've added branches for bugs 274781 and 274785, if there's any chance of a review. This is more lpia bits for grub-installer and ubiquity. === superm1 is now known as superm1|away [16:37] evand: hiya, i tried usb-creator, and it doesn't appear to be working for me [16:37] persia: you don't seem to have pushed to the branch for 274781 properly [16:37] evand: i should be able to point kvm to it, and test it that way, right? [16:37] kirkland: Did it crash or is it just not booting? [16:37] evand: not booting [16:38] evand: i got the install to succeed [16:38] kirkland: server CD? [16:38] cjwatson: no, kubuntu-desktop-amd64 [16:38] kirkland: That;s probably my recently discovered bug of it not properly setting the boot flag. [16:38] ok, I'll shut up :) [16:38] evand: i was just about to try and change that in fdisk [16:38] kirkland: fdisk /dev/whatever, a, partition number, w; done [16:38] evand: thought i'd ask first [16:38] evand: yeah, i just wanted to check [16:38] cjwatson: Sorry. repushing now/ [16:39] cjwatson: :-) i suppose i should test the server [16:39] Speaking of which, anyone know a reliable way of dermining the partition number that a block device represents, given the knowledge of it and its parent? [16:39] i'm just not a kubuntu user, but i wanted to see what all the kde4 hype was about [16:39] kirkland: won't work yet, bug 234185 [16:39] ah [16:39] evand: /sys/block? === superm1|away is now known as superm1 [16:40] evand: cool, that fixed it (bootable partition) [16:40] evand: very nice [16:42] cjwatson: can you be more specific? I thought the device's minor number corresponded to the partition number, but apparently I'm very wrong. [16:42] it should do, sure. I'm not sure I understood your question [16:44] I need to be able to pass parted a partition number to mark as bootable, but the utility works in device nodes, so I need to go from "/dev/sdb1" to "1". [16:45] I could take the difference of /dev/sdb and /dev/sdb1, given the knowledge of both of them, but that seems hackish. [16:45] see how grub-installer does it [16:45] it is hackish :-) [16:45] ah, good point. Will do [16:45] hah [16:45] or, hang on [16:46] hmm, no, /sys/block doesn't have it [16:46] you pretty much just have to subtract the parent device name, but note that some devices use a leading "p" before the partition number [16:47] evand: lilo-installer might involve less bonkers code to copy from [16:47] wow that's a lot of sed [16:48] but if you are using grub-installer, it's the make_active_partition function [16:48] ok, thanks [16:50] evand: are you going to be able to fix the grey partition thing by beta? [16:50] I was hoping to do an upload today assuming I can fix 182004 [16:51] cjwatson: I'm not sure I can fix it by end of business today, but I can surely have it done by the end of the weekend. [16:51] is it just a change of some constant somewhere? [16:51] oh, sorry, I'm mixing bugs. [16:51] That I can fix quickly [16:51] and yes, it is [16:51] ok, cool [16:51] on it now [16:52] (thought you were talking about the rendering issues previously) [16:52] ah, no [16:53] Which is the rendering bug? [16:56] http://launchpadlibrarian.net/17699770/install1.png [16:57] from https://bugs.launchpad.net/ubuntu/+source/ubiquity/+bug/271512 and possibly others [16:58] Oh good. That was one of the more minor things on my list, and I was hoping it wasn't -mid specific. Thanks for the reference. [17:04] evand: okay, now i'm trying to boot from the usb stick on real hardware, and i'm stuck at the SYSLINUX bootloader boot: prompt [17:05] what happens if you press enter? [17:05] Could not find kernel image: linux [17:05] i've dorked around with bios, trying to rearrange the drives [17:05] i suspect it's booting off of one drive, looking for a kernel on another? [17:05] ...curious [17:06] cjwatson: LP doesn't seem to want to display the branch despite a refresh. I've attached a diff to 274781 [17:06] hrm [17:07] Differences between syslinux.cfg and isolinux.cfg ? [17:10] grub-installer: cjwatson * r747 ubuntu/ (debian/changelog debian/control grub-installer): Add support for lpia (LP: #274781) [17:10] kirkland: hrm, I cannot reproduce this. Are /syslinux and /syslinux.cfg present on the root of the disk? [17:11] evand: yes, and yes [17:11] evand: i strongly suspect something funny in my bios [17:11] evand: unfortunately, i overwrote the working DSL i had installed on this device :-) [17:13] kirkland: Could you try running usb-creator with -s? [17:13] evand: sure, i'll dork around with it a bit more [17:19] ubiquity: evand * r2851 ubiquity/ (aclocal.m4 configure configure.ac): Bump to 1.9.20 [17:20] maybe 1.10.0 since we're approaching beta? (historically I bumped minor just before beta) [17:20] ah, will do [17:21] ubiquity: evand * r2852 ubiquity/ (configure configure.ac): Bump to 1.10.0 [17:23] ubiquity: evand * r2853 ubiquity/ (debian/changelog ubiquity/frontend/gtk_ui.py): [17:23] ubiquity: Use a normal color for the partition bar when use entire disk is [17:23] ubiquity: selected (LP: #273271). [17:23] Hooray, usb-creator is in main. [17:26] evand: That can be fixed :-P [17:27] heh [17:28] evand: ooh [17:32] * evand will be quite happy when bug 232429 is fixed (auto marking fix committed). [17:38] that's more about projects than source packages, I think, though I added a comment about the latter [17:40] ah, thanks [17:52] tasksel: cjwatson * r1374 ubuntu/ (debian/changelog ubuntu-seeds.pl): [17:52] tasksel: Work around a bug when checking out ubuntustudio seeds into a [17:52] tasksel: repository. [17:59] ubiquity: cjwatson * r2854 ubiquity/ (debian/changelog ubiquity/components/partman.py): [17:59] ubiquity: Don't mark partman questions as seen when we expect them to be asked [17:59] ubiquity: again (LP: #182004). [18:03] \o/ [18:03] ubiquity: cjwatson * r2855 ubiquity/ (debian/changelog gui/glade/ubiquity.glade): Set GtkAdjustment page_size to zero in Glade files too. [18:05] tasksel: cjwatson * r1375 ubuntu/ (5 files in 3 dirs): [18:05] tasksel: * Build mobile-* tasks. [18:05] tasksel: * Update Ubuntu tasks from seeds, adding mobile-mid, mobile-mobile (ahem), [18:05] tasksel: and virt-host tasks. [18:06] tasksel: cjwatson * r1376 ubuntu/debian/changelog: releasing version 2.73ubuntu9 [18:16] ubiquity: cjwatson * r2856 ubiquity/ (debian/changelog scripts/install.py): merge from lp:~persia/ubiquity/lpia-grub [18:17] persia: BTW if you use debcommit then bzr will know about the bug-fix metadata [18:17] cjwatson: Thank you. That ought sort everything for the -mid installs. [18:18] Oh. That saves the frustratingly annoying duplicate typing. Thank you. [19:04] evand: (feel free to upload ubiquity at some point, I'm done for today and probably happy with it for beta) [19:04] cjwatson: noted, will do