=== superm1 is now known as superm1|away === superm1|away is now known as superm1 [03:03] cjwatson: Thanks for the adjustment. I'll poke when I hear about lpia ABI bumps. [04:39] kirkland: Did you get a sponsor for your initramfs-tools fix? [04:42] kirkland: Never mind, seems you did. [08:14] Can anyone answer a few questions on how to run Ubiquity, if possible, as a normal application. I ran ./configure && make, but it only generated site libraries, and didn't spit out anything that was directly executable AFAIK [08:21] cjwatson in 268123 as mentioned yesterday mount -t ntfs does not work yet (no such device), mount.ntfs is ok [08:22] I was expecting mount -t ntfs to call mount.ntfs but it does not seem to be the case [08:22] NCommande: Have you looked in ./bin/ [08:23] NCommande: run debuild in the top directory, then install the deb. However, note that ubiquity is intended to run on the desktop CD and will break horribly without the presence of /rofs. [08:24] me goes out [08:24] evand: Is it a bug that it fails to complete the install with an empty blacklist set? [08:25] I can't imagine a use case for an empty blacklist, but thought the error was odd. [08:26] ugh, yeah, if it crashes in the blacklist code then it's my fault and should be filed as such ;) [08:26] evand: OK. I'll pinpoint it and open a bug. I worked around it by creating the correct manifests, but I'll try to link a branch with a naive fix to the bug. [08:28] thanks, much appreciated [08:38] evand: /rofs? [08:40] NCommande: It's one of the mount points on a live image. [08:40] Specifically the one of the pristine squashfs, without overlay, etc. [08:43] Oh I see [08:43] * NCommande rights now just wants to make sure the GUI code he wrote works correctly, and then worry about doing a real install [08:44] The only way to test it is from the live environment. I personally test under KVM, as I find that easiest. [08:44] Are you building a Xubuntu-specific UI? [08:44] No, Kubuntu [08:44] Qt -> KDE4 [08:45] For some reason, I said I'd look into it when I was made a Kubuntu member >.>; [08:50] So, test proceudre: [08:50] first, boot the latest livecd [08:50] next download (from somewhere: maybe ssh to your workstation) the updated ubiquity. [08:50] Install the updated version. [08:51] run ubiquity (I use ubiquity -d so I can see the logs) [08:51] Sounds like fun [08:51] :-P [08:51] * NCommande is going to be abusing his PPA it seems [08:51] WIth practice, you can get a cycle down to 5-10 minutes. [08:52] sounds painful :-P [08:52] Maybe I can setup build-tools and friends on the liveCD [08:52] Well, it just adds incentive to get the changes right the first time. [08:53] I've not had much success with running ubiquity twice in the same session. Seems that when ubiquity crashes, you need to manually remove hal locks, etc. Easier to just reboot. [08:53] and it just gets better and better [08:53] oh well [08:53] It shall be a learning experiment none the less [08:54] I often hack on ubiquity from the live cd [08:54] I figuring that might be possible [08:54] evand: You don't encounter issues with HAL? [08:54] What issues? [08:54] Might just be my images (which are full of other sorts of problems) [08:55] Well, if, for example, I do something that causes ubiquity to crash, running it again it complains about lockfiles. [08:55] hrm [08:55] I can probably reproduce it if you need: I figured it fell well into the "don't do that then" category. [08:55] you may have to kill stray processes [08:56] That might be it. [08:56] For me, rebooting was the fast way to do that :) [08:56] ps aux | grep part -- see if partman is still spinning in the background [08:56] heh [08:57] * NCommande just sets up his Bazaar branch with his source [08:57] Any reason why the Bazaar repo uses such an old Knit frmat? [08:57] *format [08:57] evand: Do you prefer conditional tests before execution, or exception processing when errors are encountered? [08:57] it's an old repo and no one has bothered to update it [08:58] persia: I'm not picky. Whatever best suits the situation. [08:58] so development isn't done there? [08:58] If it is updated, it probably oughtn't be updated to be newer than that supported by bzr 1.3.1 [08:58] NCommande: Development is rarely done in a public repo: it's a means to share the code. [08:59] NCommande: sorry, bad explanation. ~ubuntu-installer/ubiquity/trunk is in fact where the development is done, it's just been around for a long time. [08:59] persia: any idea how I go about doing that? [08:59] bzr upgrade ? [09:00] persia: I figured as much, just wondering [09:00] any special rules I need to know like about copyright assignment or such on hacking on ubuquity [09:00] evand: I think you need to use scp, but I've never done it. The guys in #launchpad have the answer. [09:00] persia: no, it should work remotely [09:00] http://wiki.ubuntu.com/InstallerDevelopment - I don't think there's any copyright assignment policy in place, but best to check with cjwatson on that. [09:01] arg [09:01] virtualbox doesn't work on current intrepid [09:01] kvm is pretty awesome, I'd suggest giving it a go if your processor supports it [09:02] I have the hack installed for my computers BIOS to enable VTx [09:02] But I haven't got a clue on how to get started with kvm [09:03] `kvm -hda somefile.img -m 768 -no-quit -k en-us` or some variation of is what I generally use [09:03] where somefile.img is generated by qemu-img create [09:03] I'll give it a spin [09:03] I just need to install kvm, right? [09:04] https://wiki.ubuntu.com/kvm [09:04] should be a good starting point [09:04] ah, I stand corrected: https://help.ubuntu.com/community/KVM [09:15] Unless I need to test sound in a VM, I use libvirt. [09:15] Very useful to manage multiple vms. [09:16] cjwatson: any objection to upgrading the repository format for ~ubuntu-installer/ubiquity/trunk? [09:26] evand: none, I just haven't done it because bzr upgrade sftp:// takes bloody forever [09:26] cjwatson: also, can you add usb-creator commits to the #ubuntu-installer CIA bot? I've added the project to cia.vc. [09:26] hahaha, ok [09:26] I would expect it to take hours, for the duration of which you don't get to commit to the branch [09:27] hm [09:28] I'll plan to do it this evening (EST), and if anyone feels that I should hold off so they can commit, please let me know beforehand. [09:29] This evening being real soon now, or in 18-20 hours? [09:29] 18-20 hours [09:30] well, 16 [09:30] heh. It's kinda late there :) [09:30] Just out of curiosity, is it not possible to upgrade via bzr+ssh? [09:31] early, really :) [09:31] not yet [09:31] ok, noted [09:31] or at least it doesn't seem to work all that well [09:31] I believe that bzr 1.7 will add that [09:31] cool [09:32] but at the moment, what bzr ends up doing is transferring every single file under .bzr back and forward several times, producing backup copies as well as for the actual upgrade [09:32] which is why it's so incredibly slow [09:32] each of those has to wait for round-trip times as well I think, it doesn't seem pipelinable as it stands [09:32] Is it faster if run from somewhere network-near? [09:32] somewhat [09:32] I think "faster" is the wrong word. "Less painfully slow". [09:32] lol [09:33] OK. Just wondering if it's worth asking for a run from either a machine in the DC or Ubuntuwire, rather than on an overseas link. [09:33] so I've upgraded a few installer repositories as they annoy me, but preferentially the smaller ones [09:33] you could ask #launchpad to upgrade it locally [09:34] mm, I think I'll try for that, assuming they can accomodate the time or stick it in cron. [09:34] I swear, cia.vc is run off a toaster oven. [09:34] locally, it should be fast enough that we don't have to worry about timing [09:35] I'd expect a minute or so at most [09:35] ah, good point [09:35] cia.vc> annoying isn't it? [09:35] I'm *trying* to save the bot change ... [09:35] hahaha, yeah it took me a good 10 minutes to just create the project [09:35] TheMuso: I'd use libvirt except that virt-manager's UI annoys me so much [09:36] cjwatson: Ah, I use cli to start it. :) [09:36] ubiquity doesn't have any ghost revisions by any chance? :) [09:36] no :) [09:36] grumble, lpia d-i failed to build [09:37] I know why it failed on lpia, and it's being sorted. [09:37] Oh, and I've pushed some changes to casper trunk on launchpad, but launchpad is not reflecting them in the UI. [09:37] Haven't looked at ia64 [09:37] TheMuso: yeah, it won't [09:37] persia: sorted how? [09:37] Is casper's bzr fixed? [09:37] cjwatson: Fair enough. [09:37] evand: No. [09:37] you've always been able to push to it, somewhat [09:37] fun [09:38] I've fixed xivulon's mount -t ntfs thing, I think [09:38] cjwatson: Actually shipping the additional d-i modules. The problem was http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-intrepid-lpia.git;a=commitdiff;h=5041792ac41d6c5425598a4995b20370e7a79bb1 [09:38] (enabled mount helpers in busybox) [09:38] ah, thanks for taking care of that [09:38] persia: oh dear goodness. When will the kernel team learn that they need to tell us about this shit? [09:38] persia: is the fix to revert that commit? [09:38] It's being reverted. [09:39] excellent [09:39] feel free to whack the rebuild button on d-i when it is [09:39] I can't, but I'll deputise someone. [09:39] persia: ia64 is because linux-ports fails to build on ia64, and has done for months [09:39] OK. That needs someone who cares about the ia64 kernel then. [09:41] Also, while reverting the commit, the lpia kernel will probably have an ABI bump. I've been promised this process will include an email to the installer team, as I suspect it will require a change to build/config/lpia.cfg. Will it save any time to prepare that branch, or is it just as easy to process from the mail? [09:50] it won't save any time to prepare the branch in advance [09:50] it's a sed [09:55] cjwatson any hint on previous posts? [09:55] 09:38 I've fixed xivulon's mount -t ntfs thing, I think [09:55] 09:38 (enabled mount helpers in busybox) [09:57] ah great, thanks a bunch, will resume tests tonight and let you know (davmor2 if you are around give it a go once the new busybox is up) [10:05] Unless anybody has some pressing casper changes they need to get in, I'd like to do an upload of casper to unbreak some accessibility profiles. [10:05] nothing from me [10:14] TheMuso, on a different subject, how are things with panic messages? [10:18] xivulon: Should be fixed for intrepid, but I have been so busy the past months with intrepid work, that I haven't yet done an SRU for hardy. [11:18] evand: which process did I have to kill to get ubiquity to work twice in the install? [11:19] NCommander: start by looking for partman. [11:20] I just had to kill hal-lock and it started [11:20] After that, start looking for anything with a higher process number than used by ubiquity. [11:20] Are you sure rebooting isn't easier? [11:21] as evand said, running ubiquity twice in one session should not normally be a problem [11:22] I've never heard of hal-lock sticking around before [11:22] sometimes parted_server sticks around, but that's all [11:22] I often run it successively - saves a lot of time when editing stuff on the fly [11:25] Maybe this is a new issue? I always saw hal-lock sticking around. [11:25] (well, since my images stopped being so broken that ubiquity couldn't run) [11:40] Happy Mid-Autumn Festival, everyone! 中秋节快乐！ :-) === NCommander is now known as mcasadevall === mcasadevall is now known as sonicmctails === sonicmctails is now known as NCommander === NCommander is now known as NCommander_ === NCommander_ is now known as mcasadevall_ === mcasadevall_ is now known as NCommander [13:30] TheMuso: kees did :-) thanks!!! [14:27] cjwatson: hi there, do you have a couple of minutes? I'm trying to implement your suggestions for https://bugs.launchpad.net/ubuntu/+source/partman-target/+bug/268580 [14:27] cjwatson: i want to knock this out ASAP because of UI freeze :-/ [14:28] kirkland: yes [14:28] cjwatson: sweet ... okay, the templates file.... [14:29] cjwatson: should i create an mdadm-udeb.templates with just the boot_degraded question? [14:29] cjwatson: ie, is that the proper method for getting the question into the udeb? [14:29] you already need an mdadm.templates, right? don't duplicate it [14:29] cjwatson: right, the question is already in there [14:29] cjwatson: it's used for dpkg-reconfigure mdadm in a running system [14:30] cjwatson: works there now... just want to add the question into the installer [14:30] kirkland: is there anything else in mdadm.templates that would need to be filtered out of the udeb? [14:30] cjwatson: autocheck, mail_to, start_daemon [14:31] cjwatson: i suppose those might be nice to actually ask, if someone is installing raid [14:31] then I'd recommend putting the boot_degraded question in a separate templates file; it doesn't matter what it's called [14:31] like debian/mdadm-boot-degraded.templates or something [14:31] cjwatson: but raid installs then become somewhat significantly more conversational [14:31] don't make them more conversational :) [14:31] cjwatson: agreed [14:31] cjwatson: okay, done. [14:32] then add that new .templates file to debian/po/POTFILES.in [14:32] then after dh_installdebconf in debian/rules, add this: [14:32] echo >> debian/mdadm/DEBIAN/templates [14:32] po2debconf debian/mdadm-boot-degraded.templates >> debian/mdadm/DEBIAN/templates [14:33] po2debconf debian/mdadm-boot-degraded.templates > debian/mdadm-udeb/DEBIAN/templates [14:33] something like that anyway [14:33] as you can see you can just build up the templates file in the package's control area dynamically [14:33] cjwatson: sounds good [14:33] cjwatson: okay, so do i add the check.d directory to the debian/ dir, or the top level? [14:33] an alternative approach would be to put that question only in mdadm-udeb.templates, and append that to debian/mdadm/DEBIAN/templates in much the same way [14:34] I think the above might be clearer in this case though [14:34] it doesn't matter. convention is the top level but it isn't mandatory [14:34] if you put it at the top level, then you can use the install-rc script from other partman-* packages to install it [14:34] cjwatson: that does sound better to me, the mdadm-udeb.templates [14:34] (without modifying it) [14:35] kirkland: ok, then you can just echo >> debian/mdadm/DEBIAN/templates; po2debconf debian/mdadm-udeb.templates >> debian/mdadm/DEBIAN/templates [14:35] to make sure it appears in both without having to duplicate it in the source [14:36] just keep debian/po/POTFILES.in up to date, and remember to run debconf-updatepo after changing the .templates [14:36] cjwatson: its in the clean block [14:36] was it there already from the current package? [14:36] cjwatson: yup [14:36] cjwatson: last step of "clean" in the rules [14:37] I generally prefer not to have it in clean myself, and to have the maintainer do it; saves on spurious timestamp diffs [14:37] but if it's there already, leave it [14:37] cjwatson: k [14:37] cjwatson: okay, so i put a check.d in the top level of mdadm source [14:37] cjwatson: there's a _numbers file [14:38] cjwatson: suggested value for, "05 root_on_raid" ... ? [14:38] cjwatson: this is an ordering, right? i'm the only one [14:38] cjwatson: so just leave some room for before and after? [14:39] kirkland: hmm, I don't think there's any particular sequencing in check.d. I'd put it with crypto if I were you, which would make it 07 [14:40] cjwatson: 07 root_on_raid, okay [14:40] it's not a big deal though, just controls which check will blow up first if multiple things are wrong *shrug* [14:40] cjwatson: i called the script, root_on_raid [14:40] sure, that's fine [14:40] cjwatson: seemed to read sort of like "proper_mountpoints", etc. [14:41] cjwatson: in that script, i borrowed some code from proper_mountpoints [14:41] cjwatson: basically looking if / or /boot is on a device named /dev/md* [14:41] cjwatson: i had: db_input critical mdadm/boot_degraded || true [14:41] cjwatson: i think you suggested: [14:42] db_get mdadm/boot_degraded [14:42] echo mdadm mdadm/boot_degraded boolean "$RET" | debconf-set-selections [14:42] right, that's to go after the db_go [14:42] just transfers the setting into /target [14:42] if [ "$mp" = "/" ] || [ "$mp" = "/boot" ]; then [14:42] db_input critical mdadm/boot_degraded || true [14:42] db_go || true [14:42] db_get mdadm/boot_degraded [14:42] echo mdadm mdadm/boot_degraded boolean "$RET" | debconf-set-selections [14:42] break [14:42] fi [14:42] or, more strictly, queues it up to be transferred into /target once /target exists [14:43] yep [14:43] remind me how you're checking that the mountpoint in question is on RAID? [14:43] cjwatson: okay, now what do I need to do to hook check.d/* to run? [14:44] 'debian/install-rc check.d' in the install target in debian/rules; see any partman-* package [14:44] cjwatson: http://pastebin.ubuntu.com/46227/ [14:44] cjwatson: ^^^ untested [14:44] cjwatson: okay, i'll grab it from partman-target, where I got most of this [14:44] looks plausible [14:44] oh, hmm [14:45] cjwatson: k [14:45] I think you might have to consume the rest of the output from the fstab.d scripts otherwise stuff will go wrong [14:46] so found=0 up top, found=1 instead of break, and if [ "$found" = 1 ]; then continue; fi at the top of the while loop [14:46] cjwatson: good suggestion, i'll do that [14:46] just to avoid stuff getting SIGPIPE. I can't remember whether that matters in this case but it certainly can do in other contexts [14:46] cjwatson: that'll be more clear [14:46] specifically when you're reading from parted_server, things go badly wrong if you don't consume all the output [14:47] cjwatson: okay, i'll need to install the check script into partman from mdadm [14:47] mdadm-udeb, rather [14:48] cjwatson: gotcha [14:49] cjwatson: so should the install-rc bit be in a udeb-specific section of the rules? [14:49] oh, maybe PACKAGE=mdadm-udeb debian/install-rc check.d [14:49] that'll force it to be copied into the right package [14:50] * kirkland puts that neat trick in his back pocket [14:51] cjwatson: okay, let me get a debdiff for spot-inspection [14:51] cjwatson: to test this, I'll build the udeb, and scp it over to a running installation kvm, udpkg -i the udeb, and restart the partitioning? [14:52] right, or even just udpkg -i before partitioning starts [14:52] mdadm-udeb doesn't have a postinst, right? [14:54] cjwatson: it does not [14:56] ok, that's good. If it did, you'd have to use udpkg --unpack as -i would run the postinst immediately. [14:57] cjwatson: untested debdiff looks like: http://pastebin.ubuntu.com/46231/ [14:58] cjwatson: more changelogging to come [14:59] kirkland: looks plausible. remember debian/po/POTFILES.in though [15:00] cjwatson: right, okay, i manually pruned the generated .po changes [15:01] cjwatson: and must have accidentally pruned that too [15:01] ok [15:01] cjwatson: okay, i'll go test this [15:01] cjwatson: thanks for your help [15:09] cjwatson: okay, i that db_input action did not fire... at what point in the partitioning would you have expected it? [15:09] after selecting finish [15:10] you can debug it by sticking set -x at the top [15:10] oh, make sure it's executable too! [15:10] you might need to put chmod +x in debian/rules [15:10] partman-* packages don't need that because they're all native, but the executable bit isn't preserved on files added in the Debian diff [15:11] (this is perhaps not obvious) [15:11] cjwatson: i bet that's it [15:11] cjwatson: i should have remembered that [15:55] cjwatson: i know your attention is elsewhere now, so just answer this question in time... I am certain that the code path containing "db_input critical mdadm/boot_degraded" is getting executed... but the question is not making it to the installer screen... ideas? [15:56] cjwatson: perhaps the question didn't make it into the debconf db? how can I check that? [15:57] oh [15:57] of course, you're running it in a while loop with stdin coming from a pipe [15:57] debconf uses stdio .. [15:58] have a look at the must_be_on_root function in partman-target's proper_mountpoints - you need to follow that model [15:58] i.e. have the while loop output a list of things to act on (or nothing), and then iterate over that in a for loop that *doesn't* have piped stdin [15:58] this is considered a wart in debconf [16:00] cjwatson: fooey :-) okay. [16:01] archive-copier used to have a supremely ugly solution to this involving redirecting to fd 9 [16:01] fortunately that abomination is long since dead [16:01] that's obfuscation ;-) [16:04] http://bazaar.launchpad.net/~kamion/archive-copier/mainline/annotate/170?file_id=x_Colin_Watson__Tue_Aug_10_16%3A59%3A47_2004_7462.3 lines 187-194 if you care [16:04] evand, how scriptable is this live usb creator going to be? Will it be possible to call it outside the live environment? Eg on normal ubuntu system w/ the cdimage as an argument? [16:07] superm1: I eventually intend to write a console frontend, though I can probably get away with accepting arguments for the source, target, and persistence size now with the GTK UI [16:07] however it does not download ISO files at the moment, and that's quite a large feature to add past FF. [16:08] oh [16:08] sorry, I misread [16:08] I thought you were suggesting that cdimage.ubuntu.com was the argument. [16:08] evand, nah, well I ask because one of the things I'm going to be throwing together this month is something that will convert the recovery media ISO that we generate from factory over to a USB image. I started to put it together and came across the existence of this project that you are were trying to get ready for main [16:09] superm1: it's in bzr if you'd like to have a look at the code to see if it's applicable to your project. [16:09] lp:~ubuntu-installer/usb-creator/trunk [16:10] evand, okay thanks, i'll see. [16:10] Anytime [16:10] evand, as it stands from my briefly reading, it looked like it had to be called while booted into the live cd. is that the case? [16:11] Not at all. It only requires a CD or ISO to use. [16:11] Nothing extra aside from the obvious (HAL, D-BUS, GTK) [16:11] okay then this may be a lot more feasible. I'll take a look. [16:12] given that you already mentioned HAL, probably a lot more intelligent than what I was starting to do w/ parsing /sys/block for devices and requiring the user to refresh [16:25] heh, yikes [16:26] yeah, this will automatically update the UI and its options based on the state of the drives, as well as automatically add new CDs and USB disks. [16:34] debian-installer: cjwatson * r962 ubuntu/ (build/config/lpia.cfg debian/changelog): Move lpia to 2.6.27-3 kernels. [16:39] cjwatson: Please don't upload that yet: the binaries haven't published, so it will FTBFS (or I suppose we can press the button later) [16:39] yep, I checked [16:39] just getting it into bzr for the meantime [16:40] OK. I'm watching the build, and will let you know when it's published. [16:42] cjwatson: dude, i'm definitely outside of the stdio pipe, mirrored must_be_on_root operation, definitely hitting the db_input critical call, but still not seeing the prompt [16:42] kirkland: DEBCONF_DEBUG=developer may help you [16:42] persia: tell me when it hits NEW and I'll shove it through [16:43] cjwatson: That's even better, as our archive-admin went to bed :) [16:43] cjwatson: http://pastebin.ubuntu.com/46280/ fyi [16:50] is there any ssl support in the busybox environment? libssl, stunnel, etc.? [16:52] doubt it [16:52] oh, actually we must ship a libssl udeb [16:52] libcrypto0.9.8-udeb IIRC, basically just for use by openssh-*-udeb [16:54] cjwatson: hm, not quite libssl, so I probably won't be able to perform https connections [16:54] no - why do you need that? [16:55] (busybox wget doesn't have https support anyway) [16:55] cjwatson: I wanted to submit test results during the installation process, but I guess I could support both secured and unsecured connections [16:56] if I wanted to submit to Launchpad, I wonder if that would be a feasible request [16:57] cjwatson: interesting, when i run ./07root_on_raid manually, it does-the-right-thing, but not when it's called from partman [17:04] cjwatson: how would I verify that the mdadm/boot_degraded question actually made it into the debconf database? [17:04] cjwatson: actually, scratch that... it's there when I run it by hand [17:05] let me get back to you after this meeting [17:05] cjwatson: no prob [17:05] cjwatson: fwiw, i have 2.5 hours, then I leave the office to catch a flight [17:54] evand, you i'm not sure if it was intended just because of your development process, but you have the glade file referenced directly within the tree, so when a deb get's built it's not usable [17:56] argh [17:56] I keep doing that by accident [17:57] might be best to set up a variable at the top perhaps [17:58] so glancing through, the app has the expectation that the drive you are going to use is already formated, so i take it you've decided that is a fair assumption? [17:59] ... and mounted (I filed a bug about that) [18:00] it's not a fair assumption, it just doesn't handle that case yet. [18:00] ah okay [18:02] usb-creator: evand * r10 usb-creator/usbcreator/ (backend.py gtk_frontend.py): Fixed paths. [18:02] * evand pets CIA-50 [18:02] something else that you might want to consider doing, when you copy over the contents of the ISO to the flash drive, you might want to regenerate the UUID in .disk and the initramfs. so as to prevent confusion/slowness if someone happens to have a cd in the drive with the same .disk etc [18:02] oh good, it worked [18:03] superm1: ah, good call. Noted [18:03] evand, in casper there's a script that I submitted sometime during hardy, a variant of it is what we've been using when making recovery media [18:04] ok [18:04] (for the exact same type of issue, where the recovery partition is always present when you put in the CD) [18:07] kirkland: did you try DEBCONF_DEBUG=developer? can I see the output from it? [18:07] (boot with that on the kernel command line) [18:08] kirkland: oh and FWIW you don't need the for loop, just do case $prompt in ... esac [18:08] (case's variable expansion semantics DTRT there) [18:12] cjwatson: i tried the DEBUG thing, where does the output land? [18:12] /var/log/syslog [18:12] you'll get a BIG trace [18:12] it's a hammer, but an exceedingly useful one [18:13] cjwatson: one sec, i'll grab it [18:14] cjwatson: linux-lpia hit new. Please forgive the missing Architecture:all packages - it's a known build issue, but the fix isn't critical for either the alternate CD or the live image (I think). [18:15] cjwatson: okay, db_input not found ... do I suppose I need to source something to get the debconf functions? [18:16] cjwatson: none of the other scripts in there have anything exotic sourced, just the /lib/partman/lib/base.sh [18:21] cjwatson: http://pastebin.ubuntu.com/46313/ [18:24] kirkland: /lib/partman/lib/base.sh sources the debconf confmodule as the first thing it does [18:24] cjwatson: good [18:24] kirkland: can I see the *exact* script you're currently using? [18:24] cjwatson: when I sh -x the script [18:25] I have http://pastebin.ubuntu.com/46280/ but it's not that [18:25] cjwatson: I get: RET=10 "mdadm/boot_degraded" doesn't exist [18:25] cjwatson: sure [18:25] don't run any of these scripts from the shell prompt [18:25] it will not work properly and it will confuse you [18:25] you need to edit set -x into them and have partman run them [18:26] cjwatson: http://pastebin.ubuntu.com/46319/ [18:26] DEBCONF_DEBUG=developer will not work in the script; you need to boot with it on the kernel command line [18:26] it's checked by the frontend, not the confmodule [18:27] cjwatson: didn't grok that, sorry [18:28] this is really weird, I can't figure it out [18:28] cjwatson: let me post the full debdiff, just in case [18:29] could you explain to me exactly what you did? did you scp the udeb into the installer? [18:29] cjwatson: yes [18:29] cjwatson: udpkg -i *udeb [18:29] ok, I'd like the debdiff, sure [18:29] cjwatson: then ran the partitioning [18:30] if possible on non-pastebin so I can check for funny characters [18:30] cjwatson: this one has the po poo : http://pastebin.ubuntu.com/46321/ [18:31] where is "date" coming from? it's not in your debdiff [18:32] cjwatson: after i install my debdiff, then i have been doing "active development" in the install filesystem [18:32] cjwatson: by manually editing 07raid_on_root [18:32] cjwatson: backing out to the partman screen that has "Finish" on it [18:32] cjwatson: and then I rerun finish [18:33]