/srv/irclogs.ubuntu.com/2009/02/22/#ubuntu-devel.txt

macokirkland: hey can you take a look at bug 317895?00:23
ubottuLaunchpad bug 317895 in ecryptfs-utils "netboot newuser and ecryptfs fails to login" [High,Triaged] https://launchpad.net/bugs/31789500:23
sabdflcjwatson: http://thunk.org/tytso/blog/2009/02/20/aligning-filesystems-to-an-ssds-erase-block-size/ makes good installer fodder?00:45
ScottKIf there's an archive admin available, the kde4.2 backport is once again hung up on binary New.  I'd appreciate it if you would accept kde4bindings from Intrepid New.  It's got mixed Main/Universe binaries so I can't do it from the LP U/I.01:13
directhexkde 4.2 backport? sounds like a big job01:15
ScottKIt is.  sigh.01:17
directhexi didn't think major infrastructure was even eligible for backporting01:17
ScottKIt's several dozen packages that mostly have to get built in a certain order and several of which get stuck in New because they grew new binaries in Jaunty.01:18
ScottKKubuntu has historically had a more generous policy about post release updating.01:18
directhexfair enough01:18
ScottKAlso because KDE4 is such a rapibly moving target ....01:18
directhexanyway, good luck with the backport. i think it's time for sleepybyes01:18
ScottKdirecthex: JFTR if I'd dropped all the CIL bindings for the backport I'd have avoided a trip through New, but I didn't.01:20
HobbseeScottK: waved thru01:21
ScottKTHanks.01:21
directhexScottK, see, that wasn't as bad as all that, was it?01:22
* directhex hands Hobbsee a fiver01:22
Hobbseedirecthex: \o/01:25
directhexhmm... sounds like rhythmbox development is winding up01:29
directhexi wonder what that'll mean for ubuntu desktops01:29
Hobbseewould help if the power didn't trip out, too...01:33
HobbseeOK, that's actually accepted this time01:33
ScottKThanks.01:35
=== freeflyi1g is now known as freeflying
dtchenhunger_t: yes, currently at least bug 33227004:32
ubottuLaunchpad bug 332270 in udev "[jaunty] doesn't boot anymore after udev upgrade" [Critical,Incomplete] https://launchpad.net/bugs/33227004:32
pwnguinoh neat04:39
pwnguinfor some reason update-manager is blocking on a etc file in the embedded VT04:40
MrMacPlusany chance of getting the PC-BSD package manager ported to Ubuntu?04:51
MrMacPlus(it doesn't require internet)04:51
MrMacPlusthe only machines I have don't have an internet connection04:52
MrMacPlusnevermind04:54
macokirkland: i *think* that patch is doing the right thing.  when the user's not logged in, their ~ is 755 with visible filenames (as in release notes) but all files appear empty (even to root).  when the user is logged in, ~ becomes 700 and root can see into the files (which makes sense to me, since the fs is mounted). i just want to check with you that this is the correct behavior for fixing that bug.05:58
cjwatsonsabdfl: yeah, there was somebody on ubuntu-devel-discuss (I think?) a little while back who was interested in working on that09:22
sabdflcjwatson: would that logic best be encoded in the d-i partitioner09:23
sabdfl?09:23
cjwatsonsabdfl: https://lists.ubuntu.com/archives/ubuntu-devel-discuss/2009-January/006760.html09:24
cjwatsonsabdfl: and followup, where I provide some suggestions09:24
sabdflthanks colin09:24
cjwatsonlibparted is a good common place to put this kind of thing - d-i would just need to flip the switch, ideally. However from the look of tytso's blog it does seem that we'd have to pass some extra arguments to mkfs.ext4 and such09:25
cjwatsonbut we could stash that information in /var/lib/partman if libparted were doing the work of fishing it out from the kernel09:26
sabdflyes, it looks like there needs to be a level of intelligence in a variety of the tools09:26
sabdfldid daniel respond to your question about whether he's willing to do the work? we could connect him to ted tso09:26
cjwatsonsabdfl: haha, and in my work mailbox this morning there's a followup from Daniel J Blueman on parted-devel about this exact subject ;-)09:26
sabdflincluding a ref to ted's blog?09:27
cjwatsonno, but it would fail to surprise me if he were prompted by the same thing09:29
cjwatsonthe coincidence is a bit much09:29
cjwatsonI'll keep the link and nudge things alon09:29
cjwatsong09:29
cjwatsonsabdfl: I've mentioned Ted's blog on parted-devel09:35
cjwatsonsabdfl: the bit about alignment of the first partition is interesting, given that we haven't historically created a /boot09:35
cjwatson(well, not by default anyway)09:35
sabdflhe's encoded a *lot* of "wisdom and best practice" into that blog09:36
cjwatsonI wonder whether the right answer is to create a /boot on SSDs, or whether it's to say screw MS-DOS compatibility and hope that not too many partitioning programs will whine at us if we 128KB-align the first partition09:37
sabdflcan we detect SSD's? and worse, can we determine things like which type of SSD it is?09:37
cjwatsonwe can detect SSDs with ATA IDENTIFY. I don't know about which type09:37
cjwatsonDaniel says: """09:38
cjwatsonI've checked into this, and since libparted sees the SATA block device09:38
cjwatsonas SCSI, it doesn't perform the expected ATA 'identify' command to09:38
cjwatsonfill out the 512 bytes of device info, of which (short) word 217 is09:38
cjwatsondevice RPM, defined to be 1 on newer compliant SSDs. The kernel uses09:38
cjwatsonthis word to detect if a device is an SSD or not, so I suggest we use09:38
cjwatsonthe same.09:38
cjwatson"""09:38
cjwatsonthat much is pretty trivial09:38
sabdflassuming we fix the libparted-sees-scsi bit09:39
=== hunger_t is now known as hunger
ogracjwatson, do we have a way to enforce that users create swap in partman ? (would be handy to not make nslu2 users shoot themselves in the foot, though i can just document that you need it indeed)12:14
=== kyselejsyrecek1 is now known as kyselejsyrecek
LordMetroidI just installed 8.10 and now 9.04 is soon to be released, I can't keep up with this high pace...12:58
LordMetroidHow can I get insight in what is being developed for each release?12:59
hungerLordMetroid: CHeck launchpad:-)13:03
LordMetroidI am hanging around on Launchpad, I even contributed some bug reports13:03
LordMetroidBut how do I know what people are working on?13:04
hungerLordMetroid: https://launchpad.net/ubuntu/intrepid/ is the "homepage" for intrepid. The most interesting section there wrt. planning is "Blueprints".13:04
hungerYou might want to replace intrepid with jaunty... Not that much planning is happening in intrepid nowadays:-)13:05
LordMetroidkarmic has no section of its own?13:10
StevenKLordMetroid: Not yet, no.13:40
gesera quick question: why is python3-minimal an essential package?13:41
directhexbecause ubuntu wuvs python?13:42
highvoltagegeser: are you wondering more about the '3' part or more about the 'python' part?13:42
geseryes13:42
geserthe 313:43
geserpython3 is still in universe and iirc doko waits on python 3.113:43
highvoltageyeah that is a bit weird13:43
directhexany rdeps?13:44
geserI just wondered why my jaunty pbuilder install python3-minimal just to find out it's because it's essential13:44
dokogeser: it's a bug and should be fixed14:01
Keybukmdeslaur: so you're getting this udev bug too?16:11
mdeslaurKeybuk: yes...not fun16:11
Keybukhave you worked around it yet, or can I pick on you for some debugging?16:12
Keybukit seems to be uniquely affecting people who use some magic combination of lvm, devmapper or mdadm16:13
mdeslaurKeybuk: well, I commented out the watch line, but I booted on an old kernel, so I can simply add it back in16:13
Keybukmdeslaur: do you have some time to do that now?16:13
Keybukor Monday (since today is Sunday and you might be busy :p)16:13
mdeslaurproblem is, I had a hell of a time getting my laptop to boot so I could make the change16:13
mdeslauryes, let me just save my work16:13
mdeslaurKeybuk: what would you like me to do?16:14
KeybukI'd like to find out what's causing these inotify events16:14
Keybukso firstly need to get booted up, so we can find out which block devices they're happening for16:15
Keybukie. run udevadm monitor16:15
mdeslaurKeybuk: I can't successfully boot _with_ the problem16:16
mdeslaurKeybuk: the best I can do is add the watch line back in, but boot with the 2.6.28-7 kernel16:17
mdeslauris that ok?16:17
Keybukdoes the kernel make a difference?16:18
geserKeybuk: I'm affected by the bug too (I downgraded udev in the end). I didn't get near a shell with the bugged udev (perhaps I didn't wait long enough) to be able to test something.16:18
mdeslaurKeybuk: I think it's juste because the old kernel has an old initrd16:19
Keybukoh, hmm16:19
Keybukok, let's just debug this from the other angle then16:19
Keybukcan you remember which filesystems you saw inotify events for?16:19
Keybukgeser: likewise?16:20
mdeslaurKeybuk: I think it was my home directory.../dev/mapper/defaultvg-home on /home16:20
mdeslaurKeybuk: but I'm not 100% sure16:20
Keybukis there anything special about that drive, other than using LVM?16:21
Keybukis it encrypted?16:21
Keybukdo you use encrypted drives or swap?16:21
mdeslaurKeybuk: I use encrypted swap, and have an ecryptfs ~/Private directory16:21
Keybukgeser: what about you?16:22
geserI use just simple raid + lvm for / and /home16:22
Keybukgeser: lvm-on-md-on-sda?16:23
mdeslaurKeybuk: I use LUKS for my encrypted swap16:23
geseryes, raid1 on sda+sdb and lvm on that raid16:23
Keybukinteresting16:24
Keybukmdeslaur: so you use md at all?16:24
mdeslaurKeybuk: nope16:24
Keybukmdeslaur: but you do use lvm?16:24
mdeslaurKeybuk: yes16:24
Keybukmdeslaur: cryptswap is done all inside the kernel, right?16:25
Keybukthere's no userspace helper that writes to the block device?16:25
mdeslaurKeybuk: yeah, I'm pretty sure it's kernel16:26
Keybukgeser: do you see inotify events for both underlying disks/16:26
mdeslaurKeybuk: of course, I get prompted for a password during boot up for my encrypted swap16:27
mdeslaurKeybuk: It's during usplash, from initr16:27
mdeslaurinitrd16:27
Keybukthe confusing thing is that none of this is supposed to generate inotify events16:28
Keybukcertainly not the IN_CLOSE_WRITE event that udev actually looks for16:29
geserKeybuk: I could upgrade udev again and check.16:30
Keybukgeser: you should be able to upgrade in-situ without rebooting16:30
Keybukand then trigger it16:31
gesereven better16:31
geserKeybuk: I've started udevadm monitor and restarted udev and no output from udevadm16:34
Keybukok16:34
Keybukrun udevadm trigger --subsystem-match=block16:34
mdeslaurKeybuk: I've done the same16:34
Keybuk(have a very niced root terminal somewhere :p)16:35
mdeslaurKeybuk: whoa, I'm getting a TON of events16:35
Keybukmdeslaur: are they repeating?16:35
mdeslaurKeybuk: yes16:36
Keybukif so, kill udevd from the other terminal and pastebin me the output ;)16:36
geserKeybuk: http://paste.ubuntu.com/121466/ (it didn't stop)16:36
mdeslaurKeybuk: https://pastebin.canonical.com/14086/16:37
Keybukgeser: kill udevd harder16:37
mdeslaurKeybuk: I didn't get the beginning16:38
geserKeybuk: that was before I killed udevd16:38
Keybukboth of you: can you run "find /sys/block" for me and pastebin that16:38
mdeslaurKeybuk: http://paste.ubuntu.com/121467/16:39
geserhttp://paste.ubuntu.com/121468/16:39
IntuitiveNippleKeybuk: Some info: If running with udev_log=debug (and a serial console) the problem isn't nearly as bad, although there are a huge number of "vol_id..." invocations early in the initrd - before netconsole  gets going. Without either "debug" or "info" log-levels, the disk will just thrash constantly and not reach a prompt for a very long time (hour?). This is with just a single LVM volume, not even mounted via fstab.16:47
IntuitiveNippleKeybuk: If you want to ssh in to observe/interact, I can do that16:48
Keybukmy guess is that mdadm and lvm (and probably just devmapper) are generating open/close events in the kernel16:53
IntuitiveNippleIt is very weird how with logging on, the timing difference allows the system to boot. Also very frustrating, because it is making it almost impossible to catch the activity!16:54
Keybukthe timing difference is just because udevd is much slower16:54
Keybukyou could replicate the same by adding a rule like16:54
KeybukSUBSYSTEM=="block", RUN+="/bin/sleep 5"16:54
IntuitiveNippleyeah, but my point is, the timing delay is preventing the problem... the disk doesn't thrash like it does when the system's being hit16:55
Keybukis it preventing it? or just slowing it?16:55
geserKeybuk: you need more testing? else I'd downgrade udev again so I can boot tomorrow16:56
Keybuka continuous loop of events is bad whatever16:56
Keybukgeser: I think I have enough information to try and put together a test install16:56
Keybukif I still can't replicate it, I'll come back to you16:56
IntuitiveNippleWith logging enabled, I can get to the root prompt in about 80 seconds. Without it the disk thrashes with no screen reports for a _long_ time... longest I left it was 30 minutes16:56
IntuitiveNippleKeybuk: I found as soon as I added lvm2 to the intrd it fired it off - remove lvm2 from initrd and it's fine16:57
Keybuksure, that's just the symptom though16:57
Keybukthe real problem is udev picking up events that seem to cause other events16:57
Keybukit'd be interesting to find out whether it's caused by a udev event16:57
Keybukor whether the kernel simply generates events over and over again anyway16:57
IntuitiveNippleI have a log here now (captured via screen from the serial console). I *think* it might be useful but I'm trawling through it to check16:58
IntuitiveNippleKeybuk: It's got a lot of this kind of thing: http://paste.ubuntu.com/121476/16:59
Keybukwhat I need to find out is whether the *next* event is caused by the previous one17:00
Keybukthat's possibly quite easy17:00
KeybukSUBSYSTEM=="block", ACTION=="change", OPTIONS+="last_rule" as a 00-xxx.rules file17:00
Keybukadd that to the system, and see if the eventS stop17:00
IntuitiveNippleOK I'll try that now. I've just attached this most recent log-file to the bug report for you.17:03
geserKeybuk: http://paste.ubuntu.com/121483/ with 00-xxx.rules (complete output, no looping)17:13
geserafter a udevadm trigger --subsystem-match=block17:14
cjwatsonogra: partman already complains if you don't create swap, AFAIK17:15
Keybukgeser: aha! fantastic17:19
KeybukI bet it's something silly17:19
lamontshould 'Running "grub-install (hd0)"...' take forever in a qemu box?17:20
IntuitiveNippleKeybuk: With the 00-xxx.rules the boot with udev_log="err" succeeded. About 60 seconds to the root prompt17:20
KeybukIntuitiveNipple: does udevadm monitor show continuous activity?17:20
IntuitiveNippleKeybuk: After the system reaches the root prompt, the 'storm' has gone away. So no, it shows 'normal' behaviour at that point.17:21
KeybukIntuitiveNipple: what about throughout the boot?17:21
IntuitiveNippleDuring the boot there's no way to run that manually since the storm kicks off the very moment udevadm starts.17:23
Keybukso you're saying that you still get the looped events during boot?17:23
Keybukor are you saying you simply don't know17:23
Keybukplease be accurate17:24
IntuitiveNippleWith udev_log="err" it is difficult to know what is going on. The console reaches about 6 seconds and then all console logging ceases, the hard disk begins thrashing continuously, and it never appears to get out of the initrd stage (I was hoping by leaving it that /var/log/udev might catch something)17:25
Keybuk"it" ?17:25
Keybukplease be specific about what you are doing17:25
KeybukI cannot read your mind]17:26
IntuitiveNippleWith udev_log="info/debug" the console continues to deliver kernel messages as well as the udev output, and it takes about 80 seconds to reach the recovery root prompt17:26
KeybukI don't care about how long anything takes17:26
Keybukthat is irrelevant17:26
Keybukare you seeing the same change event for the same block device repeated over and over?17:26
geserIntuitiveNipple: did you also add that additional rule to the initrd?17:26
IntuitiveNippleKeybuk: The log messages blur the console. The log-files I attached to the bug from serial are hard to parse since there's so much in them. I've not spent a massive amount of time analysing every line of one so far17:29
IntuitiveNipplegeser: Yes, and it allowed the system to boot but problems with /dev/disk/by-uuid/ entries missing - obviously :)17:30
Keybukok, so you see continous log messages17:30
Keybukputting in that rule I asked, do you _STILL see continuous log messages17:30
Keybukor do they stop?17:30
IntuitiveNippleWith the 00-xxx rule *and* udev_log="debug" there seemed to be the same amount of log messages.17:31
Keybukcontinuously?17:31
Keybukie. the log messages never ever stop?17:31
IntuitiveNippleNo. Whenever I use udev_log="debug|info" the udev messages stop just before the root login is reached17:31
Keybukwell, that doesn't make sense in of itself17:35
Keybukchanging udev_log shouldn't change anything17:36
Keybukare you, honestly, telling me that changing udev_log makes this bug appear and go away?17:36
IntuitiveNippleYes.17:36
Keybukie. if you set udev_log to a particular value, your machine boots as normal, and udevadm monitor shows nothing if you run it while booted17:36
Keybukeven with the OPTIONS+="watch" rules all in place?17:36
IntuitiveNippleI've seen it before, with the kernel, debugging suspend/resume. Just adding additional logging added enough time delay to cure the problem17:37
IntuitiveNipplecorrect17:37
IntuitiveNippleThis test-bed has the default install on, I've not messed with it17:37
Keybukiiiinteresting17:38
Keybukyou know what this suggests?17:38
Keybukthis suggests that the inotify event is coming from one of the programs that udev runs17:38
Keybukand the logging is enough to slow down the call of inotify_add_watch such that the run rule finishes before it17:38
Keybukgeser's debugging suggests that udev is indeed chasing its own tail17:39
IntuitiveNippleThe thing I keep noticing is vol_id17:39
IntuitiveNippleIt feels as if it is triggering itself17:39
Keybukvol_id certainly shouldn't17:39
Keybukextras/volume_id/vol_id.c:fd = open(node, O_RDONLY);17:40
Keybukit opens the block device RDONLY17:40
IntuitiveNippleLooking through that log, I see a lot of this kind of thing, too: udevd-event[1569]: '/sbin/modprobe' (stderr) 'FATAL: Module acpi:LNXTHERM: not found.'17:41
Keybukthat's all normal17:41
IntuitiveNippleokay17:41
Keybukall we're looking for is something that udev runs that opens the block device for _writing_17:42
=== iulian is now known as Guest46483
IntuitiveNippleI'm reading every line in the log now, but its slow going17:44
IntuitiveNippleThere are a *lot* of these: udevd[835]: seq 729 forked, pid [914], 'add' 'acpi', 0 seconds old17:46
geserKeybuk: it's 65-mdadm.vol_id.rules what's causing this17:47
IntuitiveNipplegeser: they're not installed on this test-bed17:47
Keybukhttp://people.ubuntu.com/~scott/udev.inotify.patch17:48
Keybukcould you try applying that to udev's source and rebuilding17:48
IntuitiveNippleI think I've noticed something. The rules are firing like mad when ACPI and PCI are creating devices17:48
KeybukIntuitiveNipple: *shrug* you'd expect to see a lot of udev events then anyway17:49
geserKeybuk: when I comment out IMPORT{program}="vol_id --export $tempnode" from 65-mdadm.vol_id.rules the looping stops17:50
Keybukgeser: right17:50
Keybukthat's what will cause mdadm to be run though17:51
Keybukso commenting out the mdadm --incremental line has the same effect, no?17:51
ramviI'm customizing Ubuntu. What are possible reasons for vntwsusb not showing up in lsmod? 1. sudo depmod -a vntwusb  2. lsmod . It's also added to /etc/modules17:52
Keybukhow does one set up encrypted swap?18:00
Keybukdo I select "physical volume for encryption" then put the swap in that?18:00
lamontis the xml that libvirt wants to see documented anywhere?18:03
lamontmeh. so it is18:03
cjwatsonKeybuk: yes18:04
cjwatsonKeybuk: (or put a physical volume for LVM in that, and put swap on LVM)18:05
cjwatsonthe reason you might do the latter is that it is much less faff to have just a single encrypted block device18:05
KeybukI've gone for a full spectrum attack18:07
Keybuk/dev/sda1 + /dev/sdb1 -> /dev/md0 -> ext3 /boot18:07
siretartthe other approach that many people seem to use is to avoid LVM, and just create a crypted block device and use /dev/random as key.18:07
Keybuk/dev/sda2 + /dev/sdb2 -> /dev/md_crypt1 -> swap18:07
Keybuk/dev/sda3 + /dev/sdb2 -> /dev/md1 -> lvm ubuntu -> root -> ext3 /18:07
Keybukso at least one of these should show up the bug :p18:08
Keybukcjwatson: about half a dozen people have reported that udev seems to chase its own tail with inotify watches18:16
Keybukie. it makes the node, sets the inotify watch, and then runs some things that write to the block device, so loops doing it over and over18:17
* lamont kills trackerd with 2+hours CPU time, wonders if it will ever be interesting enough to let it complete a run18:18
IntuitiveNipplelamont: That's a bug - it writes its IntegrityCheck=1 value into the database back to front :)18:20
lamonthuh?18:21
alex-weejjaunty broke some kind of very important library i think18:21
alex-weeja few weeks ago18:21
IntuitiveNipplelamont: https://bugs.edge.launchpad.net/ubuntu/+source/tracker/+bug/32430018:21
ubottuUbuntu bug 324300 in tracker "Incorrect parameter order for SetOption" [Unknown,Fix released]18:21
alex-weeji am trying to use apt to upgrade my system from the live cd18:21
alex-weejbut chrooting doesn't work because i can't run a shell or any other command due to said breakage18:22
alex-weejso i need a way of telling dpkg to work on files in /some/mountpoint18:22
lamonttrackerd uses sqlite?18:22
lamontZOMFG18:22
lamont /dev/md4             968864600 386122036 572982160  41% /home18:23
cjwatsonKeybuk: yeah, I saw - being called away now though, I'm afraid18:26
geserKeybuk: tested your patch, no improvement (it got only harder to stop it, had to use the init-script)18:28
Mamarokmdz: you around?18:33
lamontIntuitiveNipple: I've always assumed it just meant I needed to clean up /home :)18:34
IntuitiveNipplelamont: Well I thought it was just all my Evolution imap4 accounts and the number of mailing lists I'm subscribed to18:35
lamontthen again, apt-get remove --purge tracker; killall -9 trackerd seems to fix it for me quite nicely18:35
IntuitiveNipplelamont: upstream have completely reworked the code now, and so I think we'll be left to patch what we have ourselves18:36
lamontthough sometimes I wonder what I'm missing by doing that18:36
IntuitiveNippleI found google-desktop was better than trackerd18:36
lamontI choose not to share my desktop with google18:36
MithrandirI choose not to share my desktop with tracker.18:37
lamontthen again, on the rare times that I need to find something, uh... find works well enough for me18:37
lamontMithrandir: what exactly does tracker purport to do?18:37
directhexehm... track things?18:37
lamontduh18:37
IntuitiveNippleindex files via metadata etc18:37
Mithrandirlamont: apart from filling up your home partition and burn in your CPU?18:37
Mithrandirlamont: dunno, I've taken to just tossing it out.18:38
lamontMithrandir: well, yeah... /me is searching for any redeeming value in said burnage18:38
lamontMithrandir: I mean, not all of the crack in the desktop is bad...18:40
lamont(and most of it isn't crack at all - fun distinctions to make)18:40
Mithrandirlamont: I think it's supposed to be a search engine.  I tend to just use folders for organising stuff.18:41
Mithrandir(and most of the interesting bits where I want to search aren't on my desktop.)18:41
lamontyeah18:41
lamontas in it just looks in ~/Desktop?18:41
lamontthat'd be a waste18:41
Mithrandirnah, as in ~18:41
lamontok18:42
Mithrandirstill, I regularly use a bunch of different machines, and things like my email is not on this machine.18:42
lamontand yeah, the habits of actually organizing stuff that have grown over the past uh decade plus mean I don't really need something to find stuff for me18:42
MithrandirI tend to have duplicate copies of stuff, but I generally find what I'm looking for18:43
Mithrandirelse I have find and grep18:43
lamonthrm... I wonder how to bludgeon libvirt into putting the devices on the same bridge as the physical eth018:48
alex-weejif i can only run executables like this: /lib/ld-linux-x86-64.so.2 /bin/bash18:52
alex-weejthen what do i need to do fix it?18:53
Keybukgeser: sure?  my patch doesn't compile <g>18:56
geserKeybuk: it applied here, I build successfully a new source package and my pbuilder had no problems to build the deb18:57
KeybukI think you're all going mad ;)19:00
KeybukI just put together the most complicated setup I could think of, and still can't replicate the bug19:00
alex-weejhelp me please, my system is broken and i'm sure someone knows the quick answer19:01
alex-weeji can ONLY execute binaries when run under /lib/ld.so19:01
ScottKalex-weej: You really need to ask for help in #ubuntu or #ubuntu+1 if you are running Jaunty.19:01
alex-weejeveryone there is just like "how do i get compiz to spin?"19:02
ScottKThat doesn't make this a help channel.19:02
alex-weejok let me rephrase my problem19:02
alex-weejapt HOSED my system. there is a bug somewhere. i am trying to find out what it has broken so it doesn't happen again19:02
Keybukalex-weej: #ubuntu19:02
Keybukthis channel is for "I'm patching APT to ..."19:03
alex-weeji'm patching apt to not break my linker19:03
alex-weejbut i need to know what the difference is between me manually running /lib/ld exec args and just asking the kernel to execute exec args19:03
Keybukdepends how the program you're execing is compiled19:04
Keybukwhat linker is the program compiled to use?19:04
alex-weejwell i would have thought /lib/ld-linux-x86-64.so.219:05
alex-weejbut apparently not, as nothing works19:05
Keybukdon't think, look19:05
Keybukif you don't know how, you're in the wrong channel ;)19:05
alex-weej"the program" is anything ubuntu has installed on my system, even /bin/true19:05
alex-weejKeybuk: you obviously know the answer and are just torturing me19:07
alex-weej:(19:07
alex-weejubuntu@ubuntu:~$ sudo chroot /mnt /lib/ld-linux-x86-64.so.2 /bin/bash /usr/bin/ldd /bin/true19:08
alex-weejnot a dynamic executable19:08
Mithrandiralex-weej: is /mnt mounted noexec?19:10
alex-weejMithrandir: no. if it was, i wouldn't be able to run the linker ;)19:10
Mithrandirif you just cd to /mnt and run bin/bash, does that work correctly?19:11
alex-weejyes, but with the wrong libs obviously19:11
alex-weejwait -- ...19:12
alex-weejoh no, yes, ignore that. yes it works19:12
Mithrandirnaturally.  Try copying /mnt/lib into /mnt/lib.old or something and copy /lib/ld-linux-x86-64.so.2 into /mnt/lib and ditto for glibc, then chroot /mnt true and see if that works?19:12
geserKeybuk: I'm used to bugs which only a handful person can reproduce, in intrepid I had a kernel problem where only kees was affected too19:12
alex-weejMithrandir: glibc is just libc.so.6 ?19:15
Mithrandiralex-weej: the file that points to, yes.19:15
alex-weejlibc-2.8.90.so19:15
alex-weejyep got it, no luck19:15
Mithrandircan you strace it?19:16
Keybukgeser: are you around tomorrow?19:17
geseryes, starting from 10 UTC or so19:17
alex-weejMithrandir: yeah. do you want me to reinstate the real lib folder?19:19
Mithrandiralex-weej: either one which fails should tell us something19:19
alex-weeji did tr this first but couldn't figure it out19:19
alex-weejMithrandir: actually, do you want me to strace the chroot process or trying to exec a process within the chroot?19:20
Mithrandiralex-weej: you need to strace the chroot call.19:21
Mithrandirso sudo strace chroot /mnt /bin/true19:21
Mithrandirthen pastebin that.19:21
alex-weejok the other one is interesting though, i'll show you both19:21
alex-weejMithrandir: http://rafb.net/p/N1GgmP21.html is what you asked for, and http://rafb.net/p/wVWnv565.html is the one within the chroot (running strace via ld and trying to execute "bash")19:24
Mithrandiralex-weej: can you try using true rather than bash?19:25
Mithrandiralex-weej: and can you do strings /mnt/bin/true | grep ^/lib ?19:25
alex-weejMithrandir: sure19:35
alex-weejMithrandir: http://rafb.net/p/UkEibD20.html and http://rafb.net/p/xTwDG588.html19:37
alex-weejalso true has /lib64/ld-linux-x86-64.so.2 in it, that's it19:38
=== ssweeny_ is now known as ssweeny
Mithrandirand that file exists and is executable and all, else nothing would have worked.19:38
alex-weejMithrandir: indeed, there's no /lib64 in /mnt19:38
Mithrandirah19:38
Mithrandirthat'd explain it19:38
alex-weejso it looks like coreutils was upgraded or something and they changed to /lib6419:38
Mithrandirjust make it a symlink to /lib19:39
Mithrandirno, it's always been lib6419:39
alex-weejthis system used to work...19:39
Mithrandiryes, so something ate the symlink.19:39
alex-weejMithrandir: ok now i can just chroot and indeed it seems to work fine19:40
alex-weejis strings the only way to check what linker a bin is supposed to use?19:40
Mithrandirit's at least an easy way to do so. :-)19:41
alex-weejubuntu@ubuntu:/$ dpkg -S /lib6419:41
alex-weejlibc6: /lib6419:41
alex-weejmaybe a bug in libc6 packaging?19:41
Mithrandirit could be.19:41
alex-weejok i'll get uptodate and see what happens19:42
alex-weejthe fact that when i booted it just says "/sbin/init - no such file or directory" is a bug though i think19:42
alex-weejthe missing linker is causing that error message and it's getting misrepresented by the looks of things19:43
lhoerstenwhenever I log out of gnome, my screen turns black (backlight still on) and my computer becomes unresponsive. ctrl + alt + backspace/delete do nothing. I don't see a cursor in the top left either. Anyone have any ideas how to debug this?20:33
lhoerstenseems launchpad has nothing similar20:34
IntuitiveNippleKeybuk: I've managed to catch it in the act and log it.20:42
lamontwtf do I get EPERM from a write to a connected UDP socket on localhost?20:42
lamont*giggle*20:44
=== asac_ is now known as asac
mdeslaurKeybuk: you can't reproduce it?21:52
mdeslaurKeybuk: here's the script I used to create my encrypted swap: http://paste.ubuntu.com/121574/21:55
hungerKeybuk: I have that issue without any encryption. Just plain LVM.22:01
hungerKeybuk: I do not even have / on LVM... that is on boring plain partition.22:03
cjwatsonKeybuk: you reassigned bug 328045 to partman-base - do you have any help to offer on it?22:26
ubottuLaunchpad bug 328045 in partman-base "[jaunty] Jaunty CD installation problem during harddisk partitioning" [Undecided,New] https://launchpad.net/bugs/32804522:26
cjwatsonstgraber: any progress on bug 328097?22:27
ubottuLaunchpad bug 328097 in partman-auto "preseeding partitionning isn't working anymore hardy 8.04.2" [Undecided,Incomplete] https://launchpad.net/bugs/32809722:27
stgrabercjwatson: haven't had much time to look at it recently, I should be less busy next week now that we're feature-frozen so I hope to have some time to debug it22:35
cjwatsonstgraber: thanks22:40
=== Daviey_ is now known as Daviey
LaserJockcjwatson: with the latest DVD the preseeding is gone by d-i complains about the ubuntu seed file being gone22:47
=== rgreening_ is now known as rgreening

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!