CIA-7 | debian-installer: adconrad * r1716 ubuntu/ (build/config/armel/omap4.cfg debian/changelog): Move omap4 kernels to 3.4.0-204. | 00:44 |
---|---|---|
arthurlutz | hi | 07:41 |
arthurlutz | am having problems with casper on a livecd/installer generated by remastersys | 07:41 |
arthurlutz | there seems to be missing /dev/urandom and /dev/shm when I reach 25adduser | 07:42 |
arthurlutz | any hints or errors you can think of ? | 07:43 |
ogra_ | arthurlutz, nobody here touches remastersys. have a try at their bugtracker or their IRC channel | 08:26 |
arthurlutz | ogra_: sure, but this seems to be a problem with casper, not remastersys | 08:28 |
arthurlutz | /root/dev/ seems to not be mounted | 08:28 |
ogra_ | well, do you see it on the official livecd image from that ubuntu version ? | 08:29 |
arthurlutz | when I do break=bottom it should ? | 08:29 |
ogra_ | tools like remastersysw tend to apply hacks and diffs to the build process ... | 08:29 |
ogra_ | err s/build/boot/ | 08:30 |
ogra_ | no i mean if you boot a normal ubuntu live image of the release you are trying to remaster, do you see casper errors ? | 08:30 |
ogra_ | if not, thats definitely caused by remastersys or the process it uses to roll/modify the image | 08:31 |
ogra_ | (and i'm pretty sure for all released ubuntu versions we dont have such errors) | 08:31 |
arthurlutz | ogra_: ok | 08:33 |
arthurlutz | ogra_: thanks for the answer | 08:33 |
arthurlutz | I pass break=bottom to grub and in initramfs I find something a bit strange, can you tell me if it is correct or not ? | 12:26 |
arthurlutz | mount tells me that | 12:26 |
arthurlutz | /cow on /root type overlayfs | 12:27 |
arthurlutz | and also | 12:27 |
arthurlutz | /cow on /root/dev type overlayfs | 12:27 |
arthurlutz | and there is nothing in /root/dev/ | 12:27 |
arthurlutz | anyone here with some comprehensive documentation for overlayfs ? | 13:05 |
arthurlutz | when I chroot in /root | 15:04 |
arthurlutz | mount says /dev on /dev type non (rw,bind) | 15:05 |
arthurlutz | using debug in grub, I get "sh : 1 : qsd not found " | 15:33 |
arthurlutz | and above the cmd is "blkid -o value -s UUID /qev/sr0" | 15:37 |
arthurlutz | scratch the above, vim and a weird vnc keyboard did that... | 15:38 |
CIA-7 | ubiquity: cjwatson * r5541 trunk/ (bin/ubiquity-dm debian/changelog): merge lp:~kentb/ubuntu/quantal/ubiquity/add-custom-dm-scripts | 15:44 |
arthurlutz | forcing aufs instead of overlay to see if it works better... | 16:48 |
CIA-7 | debian-installer: adconrad * r1717 ubuntu/ (6 files in 2 dirs): Move master kernels to 3.5.0-4. | 19:41 |
stgraber | cjwatson: weird... if I believe my automated testing (that I finally ported to quantal), dual stack in d-i is busted. Though netcfg hasn't changed and the tests pass fine on precise... | 21:05 |
* stgraber investigates | 21:05 | |
stgraber | (all the ipv4 and ipv6 tests pass individually, it's just when running in a dual-stack environment that it fails) | 21:06 |
cjwatson | busybox busted maybe? | 21:06 |
cjwatson | or glibc, god help us | 21:06 |
stgraber | that or I messed up something in the new isc-dhcp that somehow breaks ipv6 connectivity when calling the dhclient script | 21:07 |
stgraber | kind of hoping for that one already ;) | 21:07 |
stgraber | busybox or glibc sounds a lot less fun to debug | 21:08 |
stgraber | hmm, that's odd | 21:10 |
cjwatson | ah, well, we knew the netcfg / dhcp client interface was fragile as all hell ... | 21:10 |
stgraber | ~ # ls /proc/sys/net/ipv4/conf/ | 21:11 |
stgraber | all default eth0 lo sit0 | 21:11 |
stgraber | ~ # ls /proc/sys/net/ipv6/conf/ | 21:11 |
stgraber | all default lo sit0 | 21:11 |
cjwatson | at one point I remember having to fix one of the dhclient scripts to stop bringing the entire interface down (both families) every time it was called | 21:11 |
cjwatson | smells a bit like that | 21:11 |
stgraber | well, that wouldn't explain why eth0 would vanish from ipv6/conf in proc would it? | 21:12 |
cjwatson | er, I forget :) | 21:12 |
stgraber | cjwatson: found it | 21:38 |
stgraber | addr flush triggers the bug | 21:38 |
stgraber | cjwatson: http://paste.ubuntu.com/1085255/ | 21:39 |
cjwatson | does addr -4 flush trigger it? | 21:43 |
cjwatson | since I think that's the fix I remember applying to some dhclient script | 21:43 |
stgraber | -4 is safe indeed | 21:44 |
cjwatson | I mean I guess we might need -6 flush occasionally ... | 21:44 |
stgraber | -6 is the one breaking the world indeed... | 21:45 |
stgraber | kind of surprised nobody noticed it already, it's not like I'm running a 3.5 kernel or anything fancy, I'm running the stable 12.04 kernel! | 21:46 |
stgraber | reproduced on 3.5 though | 21:48 |
stgraber | filed bug 1023174 | 21:55 |
ubot2 | Launchpad bug 1023174 in linux ""ip -6 addr flush" flushes much more than just the addresses" [Undecided,New] https://launchpad.net/bugs/1023174 | 21:55 |
stgraber | will workaround in isc-dhcp for now | 21:55 |
stgraber | cjwatson: test suite now passes with an hardcoded "-4" argument, though I'll make sure this bug is escalated to the kernel team because isc-dhcp definitely calls "ip -6 addr flush" in some ipv6 cases | 22:09 |
stgraber | we apparently don't hit these in NetworkManager or netcfg but I'm guessing someone using ifupdown directly probably would | 22:09 |
stgraber | isc-dhcp uploaded, will do a d-i upload a bit later to get new netboot images built (as that's what the automated testing is using) | 22:18 |
stgraber | I was kind of hoping to finish that open-iscsi merge today but apparently I'll need to update hw-detect for that too, currently getting a nice d-i red screen when running with the new udeb | 22:54 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!