[12:12] evand: err, no it's not (waiting for approval) [12:12] https://launchpad.net/ubuntu/+source/ubiquity [12:13] as I'm getting kubuntu in the list, despite having the latest daily live in the drive [12:13] evand: that was the Alpha vs. Beta thing [12:13] (AIUI) [12:13] try wubi --debug [12:13] cjwatson: odd on the waiting for approval, my mail server must just be horribly behind [12:13] will do [12:14] michael_e_brown: the preseed file is read by casper; you should be able to see evidence of that in /var/cache/debconf/config.dat before starting ubiquity [12:14] (if you omit automatic-ubiquity) [12:14] are you using rev 301? [12:14] ok, yay. It is going now. (fails on partman, but that is where I have commented everything out) [12:14] michael_e_brown: yes, you need preseed/file= not preseed=; the latter is not valid [12:14] ah, so my cd is probably before the switch to beta [12:14] my mistake [12:16] cjwatson, since automatic-ubiquity is just being committed for this release, is it going to be stable enough for use in the Dell install? [12:16] And, since I just started looking at it, is there tiem enough to fix any bugs I might find? [12:18] michael_e_brown: it had better be, we're relying on it for other things [12:18] michael_e_brown: not for beta, but for final there ought to be [12:18] starting to use it now is the perfect way to get it stabilised [12:19] cjwatson, thanks. [12:19] cjwatson shall I disable wubi loopinstalltions for beta? [12:19] I'll probably be bugging you some more later, cjwatson. For now, I have it going to where I know it is looking at my stuff. [12:19] now I just have to sort out partitioning. [12:20] use a recipe for that [12:21] So, quick question: [12:21] One of the requests we have for the next version is for people to be able to keep their /home partition data. [12:21] xivulon: huh? no [12:22] But it is optional. I'll need to ask them. [12:22] michael_e_brown: we may well look into restructuring partitioning / ubiquity for that for hardy but it's unlikely to change for gutsy at this point [12:22] Are the debian docs on partitioning relevant, or where should I look? [12:23] What I was thinking of doing: partition myself in the pre scriptlet, and then just tell ubiquity to use the preconfigured partitions [12:23] https://help.ubuntu.com/7.04/installation-guide/i386/appendix-preseed.html [12:23] as a general rule (though not universally) we aim to be compatible with d-i preseeding [12:23] ah. I've been looking for something like that. [12:24] of course there are areas where it doesn't make sense [12:24] hasnt come up in any of my google searches. [12:24] I recommend strongly against doing partitioning yourself, unless you absolutely have to [12:24] at most, delete partitions you don't want and then tell ubiquity to partition the free space [12:25] Well, since we are running from /dev/sda2, I'll see what I can come up with. [12:25] free space might work. [12:25] But I'm trying to work in this requirement that they've given me that people would like to (optionally) re-use /home. [12:25] assuming, presumably, that it's a separate partition? [12:25] So, I need to come up with a way to ask the user, and so switch around. [12:26] cjwatson, not necessarily. With our current feisty install, [12:26] I was planning on just doing an 'rm' of everything except /home [12:26] you'd also have to disable sanity checks in the installer [12:26] cjwatson talking about /home, that's also a wubi requirement at this point [12:26] xivulon: not for gutsy it can't be [12:26] I'll need to disable /home backups then [12:27] michael_e_brown: it's hard to do this with unmodified Ubuntu I'm afraid [12:27] it's a bug we recognise, but is complicated to fix properly [12:27] to what extent are your requirements modified by being possible within Ubuntu? [12:27] Can I simply tell the installer to use an existing partition for certain mountpoints? [12:28] not with the current automatic partitioning recipe format, no [12:28] but if I dont care about automatic partitioning? [12:28] I saw in the docs that it is possible to specify partitioning [12:28] that's the only way to tell the installer to do anything with partitioning in a preseedable way [12:28] but havent read them enough yet [12:28] without writing code [12:28] I can write code [12:28] thats my job. [12:28] :) [12:28] I know, but it starts deviating from Ubuntu at that point [12:28] cjwatson, what I meant earlier is that if there is no code for skipping mounted partitions in mount.d hence loopinstallations should not work at the moment [12:29] we would greatly prefer Dell's changes to Ubuntu to be rolled into Ubuntu [12:29] yes. [12:29] Agreed. [12:29] but, the debian preseed example file (http://www.debian.org/releases/etch/example-preseed.txt) [12:29] has manual partitioning. [12:30] xivulon: please leave it switched on. if it doesn't work we fix it. [12:30] Is that not applicable to ubunut? [12:30] look under d-i partman-auto/expert_recipe string [12:30] that's automatic partitioning, not manual partitioning [12:30] manual partitioning is where you have user interaction at run-time [12:30] cjwatson, that's the fix you sent me from your local branch [12:31] it's manually-configured automatic partitioning, but it's still automatic :) [12:31] Yes, that is what I want to do. [12:31] and yes, expert recipes are supported in Ubuntu [12:31] however there is no way to say in an expert recipe "use this partition that already exists" [12:31] Can I have a script in the preseed/early_command that pops up a GUI dialog [12:31] the syntax has not yet been established [12:31] ah. [12:31] hmm. [12:31] :( [12:32] preseed/early_command is d-i only [12:32] ouch. [12:32] ok. So I have one more question, then. not partitioning related. [12:32] I have to have a way to put up a BIG FAT WARNING (TM) that the user has to acknowledge before it continues. [12:33] Right now, the way it works is that there is a GRUB menu entry for reinstalling the system [12:33] cjwatson is there something equivalent to early_command hooks in the live CD? [12:33] and they choose that, and there is a script that displays the BFW and asks them to type "REINSTALL SYSTEM" or some such [12:33] there's no preseeding facility to do that, but you could modify any number of different places to do that - the boot loader, casper (via usplash and TEXT-URGENT and the like), ubiquity [12:34] xivulon: not at present [12:34] I need something equivalent for us [12:34] ok. Well, I'll dive into that tomorrow, then. [12:34] cjwatson, TheMuso had said to double check with you with what needs to be done beyond adding a task to mythbuntu.gutsy seeds branch, to make it an available task (eg so it shows up in synaptic as an option etc). What more needs to be done? [12:34] I'd be inclined to modify the ubiquity welcome screen [12:34] cjwatson, then I have some useless code in lupin, shouldn't do much harm anyway [12:34] Thanks for all the help, guys. I am certainly *MUCH* more happy now than three hours ago. [12:34] michael_e_brown: there's an intro.txt file that is displayed on ubiquity's first screen [12:34] /usr/share/ubiquity/intro.txt [12:34] I'd modify that and put your BFW there [12:35] though the automation facility might need to be changed to ensure that that is displayed [12:35] Ok. [12:35] evand's your man for figuring that out [12:35] time for me to go for today. I'll get partitioning sorted out, and then come back after I have more stuff working. [12:36] but... my aim here is to be able to burn /dev/sda2 onto a CD and drop on the desktop as the last step in the install process. :) [12:37] so, we boot in the factory with one set of kernel parameters which do a fully automatic install [12:37] xivulon: it's valid for d-i, but overkill - I've fixed it to use anna/choose_modules [12:37] and the user can either A) boot off their CD copy for a fully automatic install (except for BFW, of course) [12:37] or B) they can use the Live CD [12:38] anyways, thanks for the help. I'll talk later. === michael_e_brown [n=michael_@adsl-67-65-14-121.dsl.austtx.swbell.net] has left #ubuntu-installer ["Leaving"] [12:38] right, I do think ubiquity is a sane option for recovery CDs [12:38] argh [12:39] cjwatson, I was thinking (post beta) to support a live hook like "find_override_script" and "find_override_folder" the fist finds and run a shell scripts after squashfs is ready, the second makes the special folder specified available (so that the script can use the files in there). [12:40] I'd like to have a chance to influence the naming of such options. I accepted find_preseed and find_iso because you'd already done them, but they're not really in the standard style [12:41] "folder" is a very non-Unix term too :) [12:42] stat rosa pristina... you know that one already... [12:42] they're directories in Unix, though if I understand you correctly you're talking more about mountpoints [12:42] (they used to be directories in Windows too until Microsoft decided for some reason that that was a scary term) [12:43] superm1_: I need to roll it into tasksel post-beta [12:43] cjwatson, okay i assumed it was something like that. [12:43] cjwatson, i'll add the task early next week and we can discuss after beta then [12:44] superm1_: depending on how the seed is written (I haven't looked yet), it may need small syntactic changes there) [12:44] s/)$// [12:49] find_override_files [12:50] parted_server doesn't want to die [12:51] it only dies if partman is shut down properly [12:51] if not, kill it and rm -rf /var/lib/partman /var/run/parted_server.pid so that it's reinitialised properly on the next run [12:51] and remember to losetup -d any loop-mounted devices and umount /host [12:52] hmm partman is ignoring kill -9 (not to mention 15) [12:56] and as mentioned tabcompletion freezes the shell [12:56] look in dmesg for unionfs bugs [12:56] looks like your kernel has fallen over [12:56] there are plenty of unionfs stuff in there [12:57] not sure whether they turned up debugging messages or if those are errors [12:57] but thing seem quite hosed, so I'd go for the second one [12:58] dmesg | grep BUG [12:58] if that matches then you're hosed, reboot [12:58] nothing [12:58] dmesg | grep -i oops [12:58] nope [12:58] dunno then [12:59] shell hanging on tab-completion sounds like it's bust anyway [12:59] I am sooo lucky [12:59] so even if it's unlogged I'd give it up [01:00] cjwatson going to reboot and try again, what's the plan with autopartition-loop, I don't think it's of much use without the patch we discussed last time [01:01] it works fine in the mode I care about :) [01:01] (IME) [01:02] http://www.fsl.cs.sunysb.edu/pipermail/unionfs/2007-September/005490.html [01:05] hmm jwatson, I don't preseed loop files in Use-CD mode or Read-Only mode [01:05] cjwatson [01:06] shall I change that? [01:07] also there is no "automatic-ubiquity" in either mode [01:07] that's great news! [01:08] ./wubi/installer/make_menu_lst.nsh:11: strcpy $automatic "automatic-ubiquity" [01:08] oh, that's if not read-only [01:09] automatic-ubiquity should be used in use-cd mode but not in read-only mode [01:09] read-only is just "boot into CD without having to reconfigure BIOS", right? [01:09] the name is confusing [01:10] by "use cd" I mean boot using the cd [01:10] by "read only" I mean boot from hd "installation" (but in read only mode). [01:10] err [01:11] I guess what you want is "install from CD" [01:11] I'm sorry, but I find these names incomprehensible [01:11] perhaps we can find better names for them :) [01:11] open to suggestions [01:11] even after you've explained them I'm not confident I know what they do [01:11] basically though the options are: [01:11] 1. Install using preseeding using physical CD [01:12] 2. Boot physical CD without running Ubiquity (no preseeding also) [01:12] 2 is "read-only mode"? [01:12] 3. Do a frugal installation (copy the CD to HD) and boot that [01:12] (3 is read only mode) [01:12] which one is "use cd"? [01:13] 2. is "use cd" [01:13] 1. is missing [01:13] ok [01:13] all others "4GB+" are loopinstallations using HD [01:13] 1 and 3 should have automatic-ubiquity set [01:14] 2 should not, and it shouldn't have the preseeding arguments either [01:14] Hmm 3 is not so much for installing, but for actually using when you only have 1GB [01:14] ah [01:14] If you set automatic ubiquity in 3. it becomes the same as 4GB+ [01:14] ok, then I guess only 1 should have automatic-ubiquity set [01:16] 1. is not there yet, also because with current interface you cannot choose the loopinstallation size, and the installation type [01:16] 1 needs to be the default we present on Ubuntu CDs [01:18] But does the user have to select the installation size? [01:18] I'm fine with the thing that defaults to some proportion of the available free space [01:21] Basically the difference would be that if you now use "4GB" I extract the ISO to HD, and use find_preseed. And you do not want that [01:21] when a CD is detected [01:21] find_preseed is fine, find_iso isn't [01:21] I meant find_iso [01:21] particularly since extracting the ISO to the hard disk doesn't work at the moment when the CD's locked :) [01:22] true [01:22] shall we keep "read only" mode if a CD is detected? [01:22] There is little point IMO [01:23] 2 seems more useful than 3, and again has the advantage of working while the CD is locked [01:23] Does ejecting the physical CD work by the way? [01:23] I haven't noticed it working [01:23] but I haven't tested today [01:24] I didn't touched so if it didn't work yesterday it wouldn't work today either, will need to burn a CD to play with [01:24] Ok I am going to submit the required changes [01:25] have to log off [01:25] thanks [01:25] np [01:25] I'll do a build tonight, I'll wait for your feedback [01:25] not sure I'll be able to look at eject === xivulon [n=ubuntu@87-194-85-156.bethere.co.uk] has left #ubuntu-installer [] [01:45] evand: do you have a system known to exhibit unionfs problems? [01:46] cjwatson: yes [01:46] need logs? [01:46] evand: could you pipe up on #ubuntu-devel? [01:47] cjwatson: surely === xivulon [n=ago@87-194-85-156.bethere.co.uk] has joined #ubuntu-installer [02:16] cjwatson I pushed the required changes [02:17] and uploaded rev 304 binary [02:17] but did not test it yet [02:17] ok, I'm trying to get testing of a possible unionfs fix at the moment [02:18] basically now cd2iso is never called, so if a CD is detected, and you select say 4GB you have to remove the CD, reboot, then reinsert the CD [02:18] when requested to do so [02:18] probably ok for beta [02:18] eject should be fixed though, otherwise many people will boot off CD and will be quite confused [02:18] argh, I need about a week's sleep when this is done [02:19] same here [02:19] eject probably isn't working because wubi is running off the CD [02:19] and thus it's locked [02:19] I'm inclined to release-note it for beta if nobody can think of a good workaround [02:20] my suggestion for post-beta is to figure out a way for the CD bootloader to detect that we're in the middle of a wubi installation and boot straight through to the hard disk [02:20] we should have the technology to do that [02:20] simple workaround is to write "Please remove any CD/DVD and reboot to complete the installation" [02:20] that works too [02:21] cjwatson, for that you'd simply have to set find_preseed=XYZ [02:21] if it fails, you go on with normal boot [02:22] if it doesn't fail, it means you are in a wubi installation [02:22] that doesn't work, find_preseed isn't passed to the CD bootloader [02:22] and it isn't acceptable just to try it - we don't know what to try [02:22] hmm to see if we are in the middle of a loopinstallation you will have to scan the HD for some file [02:23] yep [02:23] that's what I meant [02:23] which is what find_preseed does [02:23] but it's in the wrong place [02:23] we need to know this in the CD bootloader before even showing the menu [02:23] I see [02:23] don't know much about isolinux workings [02:24] but IMO the code should replicate find_preseed [02:24] or better: it should try to find ubuntu/install/boot/grub/menu.lst and use the boot options in there [02:26] does the initrd take a decade to load for anyone else? [02:26] I don't want to design it now, it's just a thought [02:30] rev 305, changed the reboot message to mention ejecting CD/DVD [02:30] going to reboot to test [02:32] cjwatson, to be clear, "Use CD" works as before, "Read Only" works as before, 4GB+ whith CD inserted does not use find_iso, 4GB+ without CD uses find_iso [02:32] that sounds good to me [02:32] rebooting === xivulon need an XP VM image [02:33] xivulon: wait [02:33] xivulon: could you put a build of that up on /devel/minefields/ ? [02:34] we have an updated ubiquity in the archive now, so I'd like to roll updated CD images [02:34] the build 305 is up [02:34] but I have not tested it [02:35] oh, the apache directory listing is just lying to me about the timestamp [02:35] ok [02:35] gonna upload again just in case [02:36] evand: rebuilding images now [02:36] done [02:36] thanks [02:36] cjwatson: ok [02:36] going to reboot will let you know if anything went sour [02:36] I'm still trying to get unionfs to hate me === xivulon [n=ago@87-194-85-156.bethere.co.uk] has left #ubuntu-installer [] [02:36] normally it is so easy ... [02:37] it apparently has to be difficult, right up to the end [02:37] we should just go back to one of those really old COW filesystems, the kind that only work with 2.4 === xivulon [n=ago@87-194-85-156.bethere.co.uk] has joined #ubuntu-installer [02:41] hmm trying to download the file [02:41] I have 1 thing to fix [02:42] it's gonna be quick === xivulon [n=ago@87-194-85-156.bethere.co.uk] has left #ubuntu-installer [] === xivulon [n=ago@87-194-85-156.bethere.co.uk] has joined #ubuntu-installer [02:47] cjwatson, evand, uploaded rev 306 please use that [02:48] going to reboot to test again (it's becoming a pain) === xivulon [n=ago@87-194-85-156.bethere.co.uk] has left #ubuntu-installer [] === xivulon [n=ago@87-194-85-156.bethere.co.uk] has joined #ubuntu-installer [02:52] xivulon: the build hasn't got that far yet, so it'll use r306 [02:52] ubiquity: superm1 * r2157 mythbuntu-ubiquity/ (185 files in 15 dirs): [02:52] ubiquity: add old pending merges, and temporary changes that will be merged back [02:52] ubiquity: to main branch soon [02:53] whoops didn't realize bzr-cia was still enabled for that branch. not really necessary anymore i guess. [02:54] it's not a problem [02:56] cjwatson current build on website seems fine [02:57] in the rush I might have forgotten to up the revision [02:57] cool [02:57] no worries [02:57] can't that be generated automatically from bzr revno? [02:57] seems a hassle to have to bump it all the time [02:57] no rev 306 is fine [02:58] Yeah I should have a small script to automate that, not sure whether that should be part of make [02:59] curiously that vmware machine isn't hard locking, the mouse just doesn't work and if you try to read syslog the VT locks [03:09] cjwatson: getting an invalid module format on that unionfs.ko. Do I need a specific kernel version? [03:09] I'll continue investigating after I go grab something to eat. Back later tonight. [03:11] ubiquity: superm1 * r2158 mythbuntu-ubiquity/ (11 files in 7 dirs): add support for xmltv during installation [03:11] evand: are you on amd64? [03:12] that's built for i386 [03:12] oh, wow. [03:12] I'll start pulling down an i386 CD [03:12] new daily-live CDs up [03:12] evand: it might be quicker if I gave you the diff to apply to l-u-m [03:13] and you can do 'debian/rules build flavours=generic' yourself [03:13] well, I've gotta go get something to eat anyway. So I might as well have the computer do something in my absence [03:13] will grab that and resume testing tomorrow [03:13] evand: http://people.ubuntu.com/~cjwatson/tmp/unionfs-2.1.3-2.1.4.diff [03:14] going to sleep, anything you need? [03:14] need? sleep [03:15] :P [03:15] night === xivulon [n=ago@87-194-85-156.bethere.co.uk] has left #ubuntu-installer [] [03:27] cjwatson: it seems to work here, but I wasn't able to reproduce the unionfs bugs to have something to test against [03:27] no unionfs bugs thusfar === evand <- dinner [05:23] with further testing it still appears to be stable, but I haven't tried a remastered cd with wubi install yet, which I'll do now. [05:28] I had success reports from mjg59 and amitk [05:28] kyle has uploaded it [05:28] oh, great! [05:28] I'm working on unbreaking bits of wubi now [05:28] anything I can do to help at the moment? [05:28] beat me to it :-) [05:28] it's grotty partman stuff [05:28] heh, ok [05:29] I committed a lupin fix earlier [05:29] where is the install stopping for you currently? I don't get past 5% on the install itself, curious if you're running into the same issue. [05:29] ok [05:30] I think partman-basicfilesystems needs to be changed to redirect mount 3<&- [05:30] which is why log-output is hanging I believe [05:30] need to verify that though [05:30] ok [05:30] and partman-auto-loop needs to be changed to autoaccept the partitioning [05:32] hmm, or maybe not [06:18] evand: hmm, same thing for me now [06:18] and I switched to vt1 and now can't switch back to vt7 [06:19] I think it was a valiant attempt but maybe it's time to give up on wubi for beta [06:19] what do you think? [06:20] arr, I'm not sure. I want to fight to the end, but we're getting quite close. [06:20] hah, that made no sense [06:21] that hang we're both seeing seems pretty hard to tackle at short notice [06:21] well, at least someone may benefit from the automation work [06:21] oh, that will be generally applicable [06:22] but I have the feeling that if we don't get wubi in, someone will clearly beat us to the same feature before we have another release in 6 months [06:22] it just takes a while for the people using it to show up, because the people who care about that sort of thing tend to be more conservative [06:22] indeed [06:22] I just don't think we can do it for beta. If we can figure it out immediately post-beta, I'd be willing to consider it ... [06:23] I think it will take too much of our energy and consume extra testing resources that we don't have at this point [06:23] indeed [06:23] I think you're right [06:24] I'm not thrilled since we've put so much effort into it, but sunk costs [06:25] I suppose one way to tackle the hang would be to install openssh-server first, ssh in from another host, and strace the whole thing with output going to the terminal [06:25] then it *might* be possible to see what kills it [06:26] the hang locks up everything for you? [06:26] for me it's just the mouse and syslog [06:26] locked vt-switching which made it hard to tell about anything else [06:27] huh, I could switch VTs [06:27] you may stand a chance of debugging it then :) [06:27] what do I need to strace? [06:28] check the process listing first [06:28] for me it looked like mkfs.ext3 was running - does that match you? [06:28] check free memory, see if it's very low [06:28] then strace whatever the deepest process is [06:28] will do [06:29] I'll leave you to it, I think I need to fall over now [06:29] yeah, I was going to suggest it but 32,000 people already have. [06:29] I must say, I am quite impressed with your determination [06:34] cjwatson, 32,001 people have now, get some rest, i remember seeing you on when i first signed on earlier today :) [06:35] heh [06:35] evand: just committing the various things I've got so far from wubi testing [06:35] cjwatson: thanks, very much appreciated [06:43] evand: I won't upload them (at least for now), but you should find changes to lupin, wubi, and partman-basicfilesystems in their respective bzr branches [06:43] evand: the partman-basicfilesystems change, and indeed our problem, seems related to bug 134710 [06:44] evand: as a parting thought, I wonder if you might be able to work around it by blatting /bin/log-output and making it a symlink to true [06:44] er, except not a symlink to true, that won't work [06:44] but something that ignores its options and execs the command it's given, anyway [06:45] just a thought, may be wildly inaccurate [06:46] I follow your thought process. I'll give that a shot. [06:46] alternatively change ntfs-3g to close all its file descriptors when daemonisin [06:46] g [06:47] that might also help matters [06:47] definitely seems to be related to ntfs-3g one way or the other === evand shakes fist at nvidia VT bug [07:48] cjwatson: I don't think log-output is the issue. I just replaced calls to it in 50format_ext3 with just the commands and it's still getting caught up in the same place. [07:48] still investigating [07:49] oh wait, I didn't go high enough up. Hrmm, ignore what I said for right now. === superm1 [n=superm1@ubuntu/member/superm1] has joined #ubuntu-installer [08:20] ok yeah, it doesn't appear to be log-output. I replaced it with a script that stripped -t and it's argument and --pass-stdout, then ran the rest, and it's still locking in the same place [08:45] I can't call strace on mkfs.ext3 quick enough. It locks as I'm calling strace. [08:47] even via ssh? [08:47] superm1: indeed [08:48] I'm going to give it another go with a while [ true] ; do; strace -p$(pidof mkfs.ext3); done [08:48] evand, perhaps can you inject your call to strace into the source where mkfs.ext3 is called? [08:49] so that its immediately straced [08:49] hrmm [08:49] and then from your ssh session keep a tail -f on the log files [08:49] that's the problem though, I'd need to quickly tail the files [08:49] my window is very small [08:49] only a few seconds at best [08:50] oh yikes. [08:50] well however [08:50] when you are launching with strace, the process does take a bit longer to run [08:50] so your window will lengthen a bit [08:51] you can also consider injecting a sleep 10 or so into the source before mkfs.ext3 actually gets called [08:51] and I could create the log file beforehand and just tail it then. Hrm. [08:51] to lengthen that window [08:51] indeed [09:00] ubiquity: superm1 * r2159 mythbuntu-ubiquity/debian/ (81 files in 2 dirs): update text for myth_button [09:04] ubiquity: superm1 * r2160 mythbuntu-ubiquity/debian/ (80 files in 2 dirs): update text for mythweb_button [09:08]