| CIA-8 | debian-installer: cjwatson * r868 ubuntu/ (build/config/powerpc.cfg debian/changelog): | 00:22 |
|---|---|---|
| CIA-8 | debian-installer: * Drop powerpc/cell subarchitecture at least for now, as the kernel no | 00:22 |
| CIA-8 | debian-installer: longer builds it. | 00:22 |
| CIA-8 | console-setup: cjwatson * r39 ubuntu/debian/changelog: releasing version 1.19ubuntu2 | 00:32 |
| evand | :q | 01:11 |
| evand | whoops | 01:11 |
| twb | I'm trying to do "d-i preseed/include string common.cfg" to put common bits (like the locale) across Gutsy and Etch in a separate .cfg file. | 01:44 |
| twb | d-i fails to fetch common.cfg, is it because after setting mirror/http/proxy it uses the proxy for the preseed files, too? | 01:44 |
| twb | newsflash! It does use the proxy | 01:47 |
| twb | Can I use d-i apt-setup/something to set the proxy *only* for apt-get, (well, and anna-install)? | 01:48 |
| twb | Or, can I set a no_proxy variable? | 01:48 |
| Le_Vert_ | hi :) | 09:21 |
| Le_Vert_ | okay :) | 10:10 |
| Le_Vert_ | I managed to start ubuntu 7.10 by pxe / nf | 10:10 |
| Le_Vert_ | nfs | 10:10 |
| Le_Vert_ | all what's needed is to disable network manager launch by dbus | 10:11 |
| Le_Vert_ | because it causes network connection reset | 10:11 |
| Le_Vert_ | and thus, break the / | 10:11 |
| Le_Vert_ | but there's still something I don't like | 10:18 |
| Le_Vert_ | with the livecd, I get a gnome session | 10:18 |
| Le_Vert_ | with the pxe/nfs boot, I get gdm login screen | 10:18 |
| Le_Vert_ | do you know what's the problem could be ? | 10:19 |
| cjwatson | are you sure it's set up a tmpfs correctly? | 10:20 |
| cjwatson | the code that sets up autologin doesn't look like it should care whether it's booting from CD or net | 10:20 |
| Le_Vert_ | tmpfs for what ? | 10:20 |
| cjwatson | but it will need to write to /etc/gdm/gdm.conf ... | 10:20 |
| cjwatson | as in, the root filesystem should have ended up as a unionfs containing the read-only squashfs plus a writable tmpfs | 10:21 |
| Le_Vert_ | yes... that's not the case | 10:21 |
| Le_Vert_ | here is how it works for now | 10:21 |
| Le_Vert_ | boot with root=/dev/nfs nfsroot=/path/to/squash/nfs/export ip=dhcp | 10:22 |
| Le_Vert_ | so casper is not used at all | 10:22 |
| Le_Vert_ | so now unionfs | 10:22 |
| cjwatson | well that won't work | 10:22 |
| cjwatson | you have to have casper involved :) | 10:22 |
| Le_Vert_ | I guess I'll have to do this by hand | 10:22 |
| cjwatson | it's not possible to boot an Ubuntu live session without either casper or a reinvention of casper | 10:23 |
| Le_Vert_ | mmmh | 10:23 |
| Le_Vert_ | so what's the right way ? | 10:23 |
| Le_Vert_ | find something a casper'ized initrd ? | 10:23 |
| cjwatson | well, I sketched it yesterday | 10:23 |
| cjwatson | what went wrong? | 10:23 |
| Le_Vert_ | dunno, I'm trying to figure out how to use casper for now :p | 10:24 |
| cjwatson | you should be using /casper/vmlinuz and /casper/initrd.gz from the live CD as the starting point | 10:24 |
| Le_Vert_ | replace ubuntu stock kernel by the livecd one | 10:25 |
| cjwatson | and pass NETBOOT=nfs NFSROOT=server:/path on the kernel command line | 10:25 |
| * Le_Vert_ 's giving a try | 10:25 | |
| cjwatson | what's to replace? they're the same | 10:25 |
| cjwatson | it's the initrd that's special | 10:25 |
| Le_Vert_ | yes and I'm not using the right one :) | 10:25 |
| Le_Vert_ | my kernel/initrd are not using casper at all | 10:25 |
| cjwatson | like I say, I'm sure the above won't be perfect and there'll be some things to fix, but it's the way to start | 10:25 |
| Le_Vert_ | so I'll just replace them by the livecd one and check if it works better | 10:25 |
| cjwatson | ah, well that would be a problem yes :) | 10:26 |
| cjwatson | casper isn't particularly deep magic or anything, but it does do stuff like creating a live session user, setting up autologin, disabling some stuff that doesn't make sense in a live session ... | 10:27 |
| cjwatson | if you're netbooting, casper will put a 'static' entry in /etc/network/interfaces, which should have the effect of disabling network-manager | 10:28 |
| cjwatson | s/static/manual | 10:28 |
| Le_Vert_ | that worked better with the ubuntu stock kernel :p | 10:30 |
| Le_Vert_ | initrd doesn't initiliaze the network card | 10:31 |
| cjwatson | my point is not that this will work directly but that once you get it working you will not have stupid niggling issues like network-manager and autologin | 10:32 |
| cjwatson | and again, the kernels are bit-for-bit identical | 10:32 |
| cjwatson | hmm, it seems to try to use busybox ipconfig | 10:33 |
| cjwatson | is the network device there at all and just not being DHCPed, or is it totally missing (e.g. relevant kernel module not loaded)? | 10:33 |
| Le_Vert_ | eth0: linkup | 10:33 |
| Le_Vert_ | then drop me to a busybox initrd shell | 10:34 |
| Le_Vert_ | (no /sbin/init) | 10:34 |
| Le_Vert_ | can't see and dhcp request | 10:34 |
| cjwatson | what happens if you do 'ipconfig eth0' from there? | 10:34 |
| cjwatson | actually, first, what's in /netboot.config? | 10:34 |
| Le_Vert_ | err that's strange | 10:34 |
| Le_Vert_ | ifconfig show me eth0 is okay | 10:34 |
| cjwatson | ipconfig | 10:34 |
| cjwatson | that was NOT a typo | 10:35 |
| Le_Vert_ | and even mount show nfsroot has been mounted to /root/cdrom | 10:35 |
| cjwatson | hmm, ok | 10:35 |
| cjwatson | /casper.log please? | 10:35 |
| Le_Vert_ | (and I can read /root/cdrom) | 10:35 |
| cjwatson | ok, so that's a good start | 10:35 |
| Le_Vert_ | full of no such file or directory | 10:36 |
| Le_Vert_ | (casper.log) | 10:36 |
| Le_Vert_ | got it :) | 10:41 |
| Le_Vert_ | got it | 10:42 |
| Le_Vert_ | !! | 10:42 |
| Le_Vert_ | :) | 10:42 |
| Le_Vert_ | You really want to export the iso mounted, not the squashfs image | 10:43 |
| cjwatson | sorry, had to clean up after the dog (don't ask) | 10:54 |
| cjwatson | yes, you're quite right - sorry about that, my mistake :-/ | 10:54 |
| Le_Vert_ | no problem | 11:04 |
| Le_Vert_ | you ware really nice to help me through this success :) | 11:04 |
| cjwatson | wow, did it just work after that? | 11:23 |
| cjwatson | I've created https://wiki.ubuntu.com/LiveCDNetboot as a starting point for documenting this | 11:24 |
| Le_Vert_ | yes cjwatson | 12:44 |
| Le_Vert_ | it does work :) | 12:44 |
| Le_Vert_ | I just made an initial pressed config based on evand one | 12:44 |
| Le_Vert_ | ps: no need to copy, mounting the iso as loop is enough | 12:45 |
| Le_Vert_ | you have to pass boot=casper too | 12:45 |
| Le_Vert_ | (kernel parameter) | 12:45 |
| Le_Vert_ | however ubiquity failed (partman broke) | 12:47 |
| cjwatson | ok, made those changes. I think for some people copying will be more convenient as then they won't have to arrange for the ISO to be loop-mounted on every boot | 12:47 |
| Le_Vert_ | are you sure ubiquity can be partially presseded ? | 12:48 |
| cjwatson | (direct loop-mounting is less worrying if squashfs isn't involved) | 12:48 |
| cjwatson | Le_Vert_: please show me (a) preseed file with passwords removed (b) /var/log/syslog | 12:48 |
| cjwatson | yes, I am sure it can be partially preseeded; but not all combinations will necessarily work :-) and you might be running into some other problem | 12:49 |
| Le_Vert_ | http://paste.ubuntu.com/3472/ | 12:49 |
| Le_Vert_ | (ps: you can add quiet slash as kernel parameters too ;)) | 12:49 |
| cjwatson | slash? | 12:49 |
| cjwatson | oh, splash | 12:49 |
| Le_Vert_ | this presseed file is supposed to do everything, except disk partitionning and user account creation | 12:49 |
| Le_Vert_ | (yes splash, sorry) | 12:50 |
| Le_Vert_ | (and I guess root should be exported with no_root_squash) | 12:50 |
| Le_Vert_ | (I didn't give a try without this option but I bet it will fail) | 12:50 |
| cjwatson | yeah | 12:51 |
| cjwatson | preseed file looks fine, need syslog | 12:52 |
| Le_Vert_ | yep but I can't get a shell :D | 12:52 |
| Le_Vert_ | I'm pretty sure ctrl+alt+fX in vmware is ctrl+fX or alt+fX | 12:53 |
| Le_Vert_ | but none drop me to a console tty | 12:53 |
| Le_Vert_ | :/ | 12:53 |
| cjwatson | ctrl-alt-shift-f1 | 12:53 |
| cjwatson | or ctrl-alt-space ctrl-alt-f1 | 12:53 |
| Le_Vert_ | err I'm so stupid sometimes | 12:54 |
| Le_Vert_ | syslog on the way.... | 12:54 |
| cjwatson | alternatively, cancelling the installer should drop you into a normal session | 12:54 |
| Le_Vert_ | it doesn't | 12:55 |
| Le_Vert_ | (probably because of auto-ubiquity kernel parameter) | 12:55 |
| cjwatson | for testing purposes I'd advise launching the installer manually from a shell | 12:56 |
| cjwatson | it'll get rather tedious otherwise | 12:56 |
| cjwatson | ubiquity --automatic | 12:56 |
| Le_Vert_ | http://paste.ubuntu.com/3474/ | 12:57 |
| Le_Vert_ | looks like being a floppy issue | 12:57 |
| Le_Vert_ | but none is defined in the vmware virtual computer | 12:58 |
| cjwatson | red herring | 12:58 |
| Le_Vert_ | ? | 12:58 |
| cjwatson | not a floppy issue, that just happens to be adjacent in the logs | 12:59 |
| cjwatson | hmm | 12:59 |
| Le_Vert_ | File "/usr/lib/ubiquity/ubiquity/frontend/gtk_ui.py", line 1122, in watch_debconf_fd_helper < fd_helper = floopy disk helper ? | 13:00 |
| cjwatson | could you run parted and see if it recognises the disk? | 13:00 |
| cjwatson | fd => file descriptor. Forget about floppies. They are not relevant to this. | 13:00 |
| Le_Vert_ | okay :) | 13:00 |
| cjwatson | (you might need to run parted as root to let it poke at partition tables) | 13:01 |
| Le_Vert_ | parted read the disk successfully | 13:02 |
| Le_Vert_ | it's a debian etch installation on it | 13:02 |
| Le_Vert_ | with ext3/swap/xfs filesystems | 13:02 |
| Le_Vert_ | fdisk read it too | 13:02 |
| cjwatson | ok, could you reboot without auto-ubiquity, run 'ubiquity --automatic --debug' from a shell, reproduce, and get /var/log/installer/debug? | 13:03 |
| Le_Vert_ | off course I do ;) | 13:03 |
| Le_Vert_ | minute... | 13:03 |
| Le_Vert_ | another feature request... | 13:04 |
| Le_Vert_ | could you please add console-data into the cd root ? | 13:04 |
| Le_Vert_ | that would be really great if azerty keyboard owner could use loadkeys :p | 13:04 |
| cjwatson | console-data is obsolete | 13:05 |
| cjwatson | (in Ubuntu) | 13:05 |
| cjwatson | bug 83487 is scheduled to be fixed in hardy | 13:05 |
| ubotu | Launchpad bug 83487 in console-setup "old-style loadkeys invocations no longer work" [Low,Confirmed] https://launchpad.net/bugs/83487 | 13:05 |
| Le_Vert_ | okay :) | 13:06 |
| cjwatson | I'd suggest putting console-setup/layoutcode=fr on the kernel command line for the time being | 13:06 |
| Le_Vert_ | okay :) | 13:07 |
| Le_Vert_ | starting ubiquity in debug mode.... | 13:07 |
| Le_Vert_ | http://www.le-vert.net/divers/debug | 13:10 |
| Le_Vert_ | no clue ? | 13:16 |
| cjwatson | evand: ^-- looks like the seen flag hack has some problems with partman - maybe we need to avoid questions being accidentally marked as seen when they're never presented to the user? | 13:17 |
| cjwatson | Le_Vert_: right, while partial preseeding should *in general* work, looks like it has some problems in this specific case | 13:17 |
| Le_Vert_ | sounds bad | 13:18 |
| Le_Vert_ | I may pressed the question as not seed ? | 13:18 |
| Le_Vert_ | s/seed/seen/ | 13:18 |
| cjwatson | no, it's getting marked that way due to internal actions | 13:18 |
| Le_Vert_ | but that internal process may be overriden by a user preseed ? | 13:20 |
| cjwatson | no | 13:22 |
| cjwatson | it's a bg | 13:23 |
| cjwatson | bug | 13:23 |
| cjwatson | ubiquity uses d-i as much as possible in order to reduce the amount of code duplication involved | 13:23 |
| cjwatson | but the UI models are quite different, and resolving those differences can be difficult in some places | 13:24 |
| cjwatson | this is one of those places | 13:24 |
| Le_Vert_ | okay.... | 13:24 |
| Le_Vert_ | so no preseed | 13:25 |
| Le_Vert_ | or all preseed | 13:25 |
| Le_Vert_ | could you please report this bug and add me as 'cc' ? | 13:27 |
| Le_Vert_ | gandalf@le-vert.net | 13:27 |
| cjwatson | I think things other than partitioning should work OK if partially preseeded | 13:29 |
| cjwatson | it's just that the partitioner is weird, as always :-/ | 13:29 |
| cjwatson | you're launchpad.net/~gandalf-le-vert? | 13:29 |
| Le_Vert_ | yes | 13:29 |
| Le_Vert_ | I never created an account there | 13:30 |
| Le_Vert_ | I think it got imported from debian | 13:30 |
| Le_Vert_ | cause I have some packages there :) | 13:30 |
| cjwatson | :-) | 13:30 |
| Le_Vert_ | i'm not really aware of the ubuntu specific stuff | 13:34 |
| Le_Vert_ | like launchpad and ubiquity | 13:34 |
| Le_Vert_ | that's why I think you'll be a really better bug reporter ;) | 13:34 |
| cjwatson | bug 182004 | 13:37 |
| ubotu | Launchpad bug 182004 in ubiquity "partitioner fails if partially preseeded due to seen flag madness" [Undecided,New] https://launchpad.net/bugs/182004 | 13:37 |
| Le_Vert_ | thanks :) | 13:57 |
| Le_Vert_ | I'll just have to wait that evand fix this bug :p | 13:57 |
| evand | hooray. I'll look into it after I release a new ubiquity. cjwatson, do you mind if I release a new oem-config as well? | 14:22 |
| cjwatson | go ahead | 14:27 |
| evand | ok | 14:27 |
| CIA-8 | ubiquity: evand * r2403 ubiquity/ (d-i/manifest debian/changelog): | 14:37 |
| CIA-8 | ubiquity: * Automatic update of included source packages: console-setup | 14:37 |
| CIA-8 | ubiquity: 1.19ubuntu2, partman-auto-loop 0ubuntu12. | 14:37 |
| Le_Vert_ | evand: do you think the fix will be backportable to 7.10 ? | 14:51 |
| evand | We don't backport new releases of ubiquity, but I suspect the patch will apply cleanly to 7.10. | 14:53 |
| Le_Vert_ | okay... | 14:56 |
| CIA-8 | ubiquity: evand * r2404 ubiquity/debian/changelog: releasing version 1.7.3 | 14:56 |
| Le_Vert_ | then I'll backport the fix to ubiquity package or just apply the diff against the ubiquity binary directly :) | 14:56 |
| Le_Vert_ | I'll see | 14:57 |
| evand | To clarify, we don't make new releases of ubiquity for older versions of Ubuntu because nothing updates the installer from the archive before running, nor do we release new CDs (LTS aside). | 14:58 |
| evand | But yes, I imagine it wont be hard to give you an updated deb. | 14:58 |
| evand | Sorry about the lack of clarity on that. | 14:58 |
| Le_Vert_ | that sounds great ;) | 14:59 |
| Le_Vert_ | any estimated ETA for the fix ? | 15:00 |
| Le_Vert_ | I'd go ahead with that project asap ;) | 15:00 |
| evand | honestly, no idea. I've read through your conversation with cjwatson, but I have not had a chance to take a hard look at the problem yet. | 15:00 |
| Le_Vert_ | just boot the ubiquity with the preseed file attached, and the bug will appear :) | 15:02 |
| evand | heh, indeed | 15:03 |
| Le_Vert_ | anyway, thanks a lot to all d-i team for your help :) | 15:03 |
| evand | you're welcome | 15:05 |
| CIA-8 | ubiquity: evand * r2405 ubiquity/ (configure configure.ac): bump to 1.7.4 | 15:42 |
| CIA-8 | oem-config: evand * r394 oem-config/ (aclocal.m4 d-i/manifest debian/changelog): | 16:00 |
| CIA-8 | oem-config: * Automatic update of included source packages: console-setup | 16:00 |
| CIA-8 | oem-config: 1.19ubuntu2. | 16:00 |
| CIA-8 | oem-config: evand * r395 oem-config/debian/changelog: releasing version 1.25 | 16:12 |
| evand | uh oh | 16:23 |
| cjwatson | ? | 16:25 |
| evand | cjwatson: I missed that configure{|.ac} was at 1.24 and didn't bump it to 1.25 with that upload. I'm not sure what the correct path is here. Should I just upload 1.26 with the correct versioning? | 16:25 |
| cjwatson | it's not critically important | 16:25 |
| cjwatson | just bump it in bzr | 16:25 |
| evand | ok | 16:25 |
| cjwatson | I imagine nobody else will notice | 16:25 |
| cjwatson | it just affects --help and --version, possibly even only for ./configure | 16:26 |
| evand | ah, fantastic | 16:26 |
| cjwatson | I'd actually prefer autoconf to suck it out of debian/changelog automatically, and I did try to implement that once, but it wasn't working very well | 16:26 |
| cjwatson | I might try it again at some point | 16:26 |
| evand | hm, that would be quite neat | 16:31 |
| CIA-8 | oem-config: evand * r396 oem-config/ (configure configure.ac): Bump to 1.26 | 16:34 |
| xivulon | evand hi | 17:44 |
| xivulon | evand hi | 17:44 |
| evand | hi xivulon | 17:44 |
| xivulon | did you have any chance to start tackling the patches? | 17:44 |
| evand | xivulon: I looked them over and they appear to be fine. I'll start merging them in today. | 17:45 |
| xivulon | I'll try to be on chat/jabber in case you have question | 17:46 |
| xivulon | s | 17:46 |
| xivulon | a few upstream patches are duplicated in lupin-support | 17:47 |
| evand | ok. Worst case scenario, if I can't reach you and have a question, I'll just email you and hold off on that particular patch. | 17:47 |
| evand | oh, you needed this by the weekend though, right? | 17:47 |
| xivulon | that will need to be cleaned up, after the first round of marges | 17:47 |
| xivulon | ideally yes, so I can try to play with wubi over the w/e | 17:47 |
| evand | ok | 17:48 |
| xivulon | thanks a lot | 17:48 |
| evand | you're welcome | 17:49 |
| xivulon | To make things easier you can work in this order | 17:53 |
| xivulon | 1) merge ~ago/partman-auto-loop/lupin-support with partman-auto-loop | 17:53 |
| xivulon | that takes care of 4 separate bug reports | 17:53 |
| xivulon | 2) use the new lupin/hardy branch (no need to spend too much time over lupin-support since that will be put on diet) | 17:54 |
| xivulon | that affects 2 packages lupin/casper and lupin-support | 17:55 |
| xivulon | 3) grub patches: 181658 175772 | 17:57 |
| xivulon | 4) init.d patches 151579 181669 | 17:58 |
| xivulon | 5) poke mjg59 over 176112 | 17:59 |
| xivulon | 6) poke steve over 140458 | 18:00 |
| cjwatson | ROOTFSTYPE and LOOPFSTYPE would be set if those options were passed. | 18:00 |
| cjwatson | but it seems reasonable to do the pidofs regardless. I've reassigned that bug to the correct package. | 18:00 |
| xivulon | thanks | 18:01 |
| cjwatson | I really doubt 176112 is going to happen. I think you get to live with suspend-to-ram not working | 18:01 |
| cjwatson | at least I wouldn't rely on it being fixed | 18:01 |
| xivulon | :( | 18:01 |
| xivulon | my guess is that there is a similar situation to the one we fixed with sendsigs.omit | 18:02 |
| xivulon | whereby fuse/ntfs-3g are stopped too early | 18:03 |
| cjwatson | suspending to RAM doesn't unmount filesystems or kill processes | 18:03 |
| xivulon | do not know how it works internally, but I guess it must somehow freeze current processes to get a memory snapshot | 18:04 |
| cjwatson | (well, it kills *some* processes like dhclient, but not these ones) | 18:05 |
| cjwatson | suspend-to-RAM doesn't take a memory snapshot | 18:08 |
| cjwatson | it turns off devices and puts the CPU into a low-power state | 18:08 |
| cjwatson | you're thinking of suspend-to-disk (a.k.a. hibernate) | 18:08 |
| xivulon | you are right | 18:10 |
| xivulon | you are right | 18:10 |
| evand | which mjg59 seems to think is broken anyway | 18:11 |
| xivulon | I thought I remembered something matthew said about the "ordering" with which things were processed | 18:12 |
| xivulon | hibernation has additional problems, in that on top of fuse, has to work with swap on a file | 18:12 |
| evand | cjwatson: Do you have any thoughts on bug 177868? Specifically, I'm concerned about having to pull in a format script for every filesystem we want to work with. Then again, I don't see that being much of a problem as we only really care about ext3. | 19:31 |
| ubotu | Launchpad bug 177868 in wubi "When loopfiles are used mkfs has to target the file and not the containing device" [Low,Fix committed] https://launchpad.net/bugs/177868 | 19:31 |
| cjwatson | evand: I think unfortunately we are going to have to modify all the partman-* filesystem implementations; it's annoying but hard to avoid | 19:36 |
| cjwatson | evand: the patch there is wrong though | 19:37 |
| cjwatson | evand: there's no reason to run losetup; the loop device is saved in the 'loop' file mentioned there precisely so that things that partman-ext3 can get at it | 19:37 |
| cjwatson | so it should just check for the loop file and use its contents | 19:37 |
| evand | yeah, I thought the first bit might be an option, but I wasn't sure if you wanted to keep them clean of this. | 19:38 |
| evand | ok | 19:38 |
| cjwatson | preferably, it should arrange to set $device to the loop device name so that you don't have to duplicate the mkfs.ext3 code | 19:38 |
| cjwatson | I *want* to keep them clean of this, but I'm not sure we can :-/ | 19:39 |
| evand | heh | 19:39 |
| evand | indeed | 19:39 |
| cjwatson | having the filesystem implementations out of sync in that kind of way is bad mojo | 19:39 |
| evand | yeah, I figured as much | 19:39 |
| evand | cjwatson: if you have a free moment, can you give this a quick look over: http://pastebin.ubuntu.com/3488/ . Always forcing isn't going to break some case that I haven't considered, will it? | 21:19 |
| evand | hm, that's a silly question. Nevermind then. | 21:21 |
| cjwatson | evand: I *think* that's OK, but if you're worried, you could do force=-F inside the first if and then pass $force (without double-quotes) to mkfs.ext3 | 21:25 |
| evand | ah, good point | 21:26 |
| evand | ugh, I just made a mess by not realizing that we had synced the last few versions of partman-ext3 | 21:33 |
| * evand goes to unbreak things | 21:33 | |
| xivulon | do you prefer that I clean-up lupin-support now? | 21:52 |
| xivulon | since it contains some of the upstream-patches | 21:53 |
| evand | sure, if you so desire. | 21:56 |
| xivulon | done but cannot connect to launchpad | 22:43 |
| ago | evand lupin/hardy rev 79 | 23:02 |
| evand | ok, thanks | 23:18 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!