[04:00] <wcr2007> hi, everyone. I have one silly question. I failed to install 32bit ubuntu(8.04 desktop) on AMD phenom 8450 CPU. I have turned acpi and apic off. Can anyone has such expericnce tell me how to? Is there any kernel options to turn on or off? Thanks a lot.
[04:01] <wcr2007> this is the message I got:Ubuntu:    BUG: Int 6 : CR2 00000000       EDI 00000080 ESI 01000000 EBP c04cd440 ESP c0417e70       EBX c04cd440 EDX 00000006 ECX 00000046 EAX 00000036       err 00000000 EIP c0430559 CS 00000060  flg 00010082    Stack: xxxxxxxxx xxxxxxx  xxxxxxxx  xxxxxxxxx ........        Debian:         Unknown interrup or fault at EIP 00010096 00000060 c031f3f5
[09:17] <xivulon> cjwatson are UUID going to be supported by grub in Intrepid?
[09:52] <davmor2> xivulon: Now that ubiquity is fixed I might be able to test wubi again :)
[09:53] <xivulon> you mean in intrepid?
[09:53] <davmor2> yes
[09:53] <xivulon> I still have to bump the version and change isolist.ini...
[09:53] <xivulon> same for umenu
[09:53] <xivulon> I guess that as it is it will just install 8.04
[09:54] <xivulon> ^note for self
[09:54] <davmor2> ah okay no probs give me a ping when it's ready then and I'll test it out :)
[09:56] <xivulon> will do it today and ask evand to upload
[10:02] <xivulon> davmor2 I just noticed that the ntfs-3g patch in 204133 is still in proposed for hardy, would it be possible to test that in hardy and reply to pitti?
[10:03] <davmor2> I can look at it this afternoon I need to do some smoke tests this morning for Intrepid.
[10:04] <davmor2> What does pitti need to know?
[10:04] <xivulon> np, I'll test it myself tonight if I have time left after thewubi intrepid
[10:08] <xivulon> You need to install 1:1.2216-1ubuntu3 in hardy (proposed repo) then mount an ntfs partition with and without the syncio option (run chkdsk /r first so that you are sure it is clean), hammer it, copy files until it is full, delete files etc..., then check that all is good (run chucksums on the copied files), see if there is any relevant speed impact when using syncio, and see if there is any fs corruption (chkdsk)
[10:08] <xivulon> ^ntfs-3g 1.2216-1ubuntu3
[10:09] <cjwatson> xivulon: yes, although we're still working on some of the details
[10:10] <xivulon> cjwatson cool I'll need to notify grub4dos devs so that they can port the same setup
[10:12] <xivulon> well in fact that is half relevant since at the moment I use the grub find command to check whether the system is already installed or not...
[10:14] <xivulon> I'd guess that on my side it is more relevant to use UUID in place of find-iso/find-custom-installation-folder... not much to do with grub
[10:14]  * xivulon writes other note to self
[10:25] <xivulon> any preference for the syntax? preseed/file=UUID=blablablabla/path/to/preseed.cfg or preseed/file=/dev/disk/by-uuid/blablablabla/path/to/preseed.cfg or ???
[10:28] <cjwatson> haven't thought about it yet
[10:28] <cjwatson> though both of those look horrible
[10:28] <cjwatson> shouldn't mix device specifications with paths like that
[10:29] <cjwatson> I don't want to think about this right now though, sorry
[10:29] <xivulon> np feel free to edit https://wiki.ubuntu.com/WubiIntrepid, I'll pick it from there
[10:29] <cjwatson> however, my general advice is that paths should be pure Unix paths, and deal with device locations separately
[10:30] <xivulon> preseed/file-device=UUID=blablabla preseed/file=/path/to/preseed.cfg?
[10:31] <CIA-1> casper: cjwatson * r525 casper/ (debian/changelog scripts/casper-bottom/13swap): merge lp:~tormodvolden/casper/no-raid-swap
[10:51] <TheMuso> cjwatson: Do we have a functional casper branch again?
[10:51] <cjwatson> not if you don't already have a copy
[10:51] <cjwatson> it needs either (a) bzr fix to not mind ghosts on mainline (b) very very very careful import and fetch-ghosts of old baz branches
[10:52] <cjwatson> but for (b) you have to get the import exactly right - you have to be importing with the same history as the existing converted branches were imported with, sort of
[10:52] <cjwatson> otherwise you get sha-1 conflicts
[10:53] <TheMuso> Sounds somewhat nasty.
[10:57] <cjwatson> it's absolutely foul
[10:58] <CIA-1> casper: cjwatson * r526 casper/debian/changelog: merge lp:~ubuntu-installer/casper/hardy.proposed (purely changelog)
[13:08] <cr3_> cjwatson: ping, I've been encountering a problem recently with netinstalls of intrepid where the retrieval of base packages over http seems to be timing out and I get prompted to Retry. does that ring a bell?
[13:08] <cjwatson> cr3_: no, it doesn't
[13:09] <cr3_> cjwatson: could you suggest a place to look to diagnose the problem?
[13:10] <cjwatson> you could start by putting the syslog somewhere we can see
[13:15] <cr3_> cjwatson: http://people.ubuntu.com/~cr3/intrepid/syslog from /var/log/installer/syslog
[13:16] <cr3_> cjwatson: looking for "retry" seems to return a few lines. by the way, this only happens sometimes: might not happen at all or might happen a few times
[13:41] <cjwatson> cr3_: seems to be something odd wrong with the fetch-url helper
[13:41] <cjwatson> Jul 30 11:19:58 main-menu[2870]: (process:6216): sed: can't open '/dev/fd/4': No such file or directory
[13:41] <cjwatson> all that stuff looks implicated
[13:46] <cr3_> cjwatson: I just did the exact same install on the same machine and I wasn't prompted to retry this time
[13:47] <CIA-1> debian-installer-utils: cjwatson * r660 ubuntu/ (README.wget404 debian/changelog fetch-url-methods/http):
[13:47] <CIA-1> debian-installer-utils: Use /proc/self/fd/4 rather than /dev/fd/4 in fetch-url-methods/http; the
[13:47] <CIA-1> debian-installer-utils: latter doesn't seem to exist in d-i.
[13:47] <cjwatson> ^- might be that, hard to say for sure though if it's transient
[13:47] <cjwatson> your mirror connection might just be dodgy somehow
[13:48] <cjwatson> I think that that error only happens if something 404s so you could check the server logs for suspicious stuff
[13:49] <CIA-1> debian-installer-utils: cjwatson * r661 ubuntu/debian/changelog: releasing version 1.59ubuntu2
[15:19] <crispin> cjwatson: I'm trying to preseed netcfg/no_interfaces (using "d-i netcfg/no_interfaces error"), and that isn't working - is it possible to preseed that ?
[15:20] <cjwatson> I don't see why it shouldn't be; netcfg doesn't do anything unusual with that
[15:20] <cjwatson> hit it with DEBCONF_DEBUG=developer and see what's actually going on
[15:22] <crispin> that goes on the kernel cmdline I assume
[15:24] <cjwatson> yes
[15:24] <cjwatson> results in a debconf protocol trace in /var/log/syslog
[15:27] <crispin> hmm, "netcfg/no_interfaces doesn't exist"
[15:27] <crispin> I guess that is why it isn't working ...
[15:27] <cjwatson> nah, that's just how preseeding works
[15:27] <crispin> yeah, just spotted that :-)
[15:27] <cjwatson> the question doesn't exist when you preseed stuff because all the templates haven't been loaded yet
[15:28] <crispin> yeah, of course!
[15:29] <crispin> the trace ends with "--> INPUT high netcfg/no_interfaces", "<-- 0 question will be asked", "--> GO"
[15:34] <cjwatson> is this a netboot image or a cdrom image?
[15:34] <crispin> cdrom image
[15:34] <cjwatson> I'm stumped then. can I see the syslog?
[15:36] <crispin> give me a minute, I need to finish the install, and then get the syslog off
[15:37] <crispin> presumably I could avoid the question as well by setting the priority to critical?
[15:40] <cjwatson> yes
[15:41] <crispin> damn it my installer is too automatic, it shutdown automatically - lets try that again!
[15:44] <evand> cjwatson: regarding Mario's crash from the other day, would it be reasonable to check that partitions are marked to be formatted before considering them busy in 01unmount_busy, given that somehow the install succeeds without being able to re-read the partition table?  Or should they hack around it by removing partitions from mtab or removing the check altogether?
[15:45] <evand> hrm, let me know if that got cut off.
[15:45] <cjwatson> evand: it's not whether they're formatted, it's whether there's any change to the structure of the partition table itself
[15:46] <cjwatson> if the list of devices and their extents is the same before and after, you might get away with it
[15:46] <cjwatson> s/devices/partitions/
[15:47] <evand> ah, indeed.  Somehow I didn't see beyond the simple case of reformatting partitions.
[15:50] <crispin> cjwatson: the syslog is at http://patches.theflowerdays.com/syslog
[15:51] <crispin> and the preseed used is http://patches.theflowerdays.com/preseed
[15:52] <crispin> "Error questions are always shown even if they're asked at a low priority or have already been seen." is that the reason ?
[15:53] <crispin> debconf/Debconf/ConfModule.pm
[15:53] <cjwatson> debconf => not relevant
[15:53] <cjwatson> it's cdebconf in d-i
[15:53] <cjwatson> although as it happens, yeah, that would be the reason - sorry, should have noticed that
[15:57] <crispin> hmm, oh well, thanks
[15:57] <cjwatson> that's sort of awkward, let me see if I can find a way round it
[15:58] <cjwatson> is it not an error for you if there are no network interfaces? I'd have thought ZXTM would sort of like one
[15:58] <crispin> cjwatson: no, its no error at all - it happens because I am installing in VMware and the installer kernel doesn't know about the "Enhanced VMXnet" network driver
[15:58] <crispin> once everything installs and reboots the driver will be there
[15:59] <crispin> I'm only using that driver because the standard vmware e1000 driver fails to network boot :-(
[16:00] <crispin> ahh, I could probably just ignore netcfg entirely, let me try that :-)
[16:00] <cjwatson> you could have a preseed/early_command that blanks out /var/lib/dpkg/info/netcfg.postinst
[16:00] <cjwatson> as a last resort
[16:01] <crispin> hmm, yeah, just ignoring netcfg won't work as lots of things require it
[16:05] <crispin> btw, I thought I could preseed that because of https://bugs.launchpad.net/ubuntu/+source/casper/+bug/12482  - where you seemed to have success doing it! (although over 3 years ago ...)
[16:06] <cjwatson> yes, the same thought had occurred to me :-) it may be that netcfg/no_interfaces had a different type back then
[16:08] <cjwatson> or it may be that I just never tested that :-/
[16:08] <cjwatson> it seems to have had Type: error since revision 1
[16:10] <crispin> hmm, early_command is too early, netcfg hasn't been unpacked yet, I could totally hack things by making it fork and spin in the background until netcfg is unpacked and then blank it out
[16:10] <crispin> or maybe I'll just make vmware use a driver the installer understands :-)
[16:11] <cjwatson> or have early_command append something to the end of /var/lib/dpkg/info/load-cdrom.postinst that then does what you actually want ...
[16:11] <cjwatson> having the installer actually understand the driver is probably easier :)
[16:11] <crispin> lol, that is a superb hack :-)
[16:12] <cjwatson> I sometimes think the installer's sanity has suffered slightly from the number of crazy hacks that are *possible* (but not easy)
[16:13] <crispin> I do most of my the stuff we want to do (install extra packages etc) in the late_command so I'm well used to hacking things :-)
[16:34] <crispin> cjwatson: many thanks, that works perfectly (appening onto load-cdrom or load-iso depending on the install method)
[16:34] <cjwatson> cool
[16:35] <crispin> I can now do totally automatic, remove installs inside vmware :-)
[16:36] <crispin> s/remove/remote/
[19:15] <kirkland> evand: how do I run anna-install on a local file?
[20:09] <evand> (explained offline)
[21:37] <udit99> i have a question regarding overriding dependencies when installing packages..as part of my install process I want to prevent certain packages from being installed
[21:38] <udit99> How can I ensure that they dont get installed as a dependency from another package
[21:39] <udit99> I was thinking of using the overrides files but Im not sure if thats the right direction
[22:42] <cjwatson> udit99: the only ways that won't cause problems later are to modify the packages to remove dependencies, or to install fake packages to satisfy those dependencies (you can use equivs to generate them)
[22:42] <cjwatson> if you override the dependencies in other ways, apt will refuse to operate
[23:14] <kirkland> cjwatson: hiya, i've dropped to a shell on an intrepid installer, and installed the open-iscsi-udeb...  i have a working iscsi target that I can get to from other VM's on the same virtual network
[23:14] <kirkland> cjwatson: stracing the iscsiadm command in the install shell, i'm getting an ECONNREFUSED error
[23:14] <cjwatson> meeting atm
[23:15] <kirkland> cjwatson: k, np
[23:20] <hye> cjwatson: do you mean modifying the control file from packages to remove dependencies?
[23:20] <cjwatson> yes, or as I said creating new packages to satisfy them
[23:21] <hye> i see
[23:21] <hye> i like the second idea
[23:21] <hye> thanks