| persia | lool: I picked Recommends because I thought most people *should* builde for armel :) Did you manage to fix the issue with the conflicting binfmt-misc handlers? | 17:51 |
|---|---|---|
| lool | persia: I think so | 18:20 |
| lool | persia: The issue was one present before the rename, but becoming serious after the rename | 18:20 |
| persia | Aha. I'm just installing the latest version, and will confirm then :) | 18:21 |
| lool | persia: The handler was registered on postinst/configure and unregistered in prerm/remove instead of postinst/install prerm/remove or postinst/configure and prerm/upgrade | 18:21 |
| lool | (Hope I'm making sense) | 18:21 |
| persia | Right. That would explain all sorts of things :) Sometimes I wish maintainer scripts weren't quite so complex, because more people might get them right the first time :) | 18:22 |
| lool | Sadly, I can't retrofix the qemu-arm-static packages, but I've added qemu-arm-static.postinst and qemu-kvm-extras-static.postinst configure snippets to remove the old name | 18:22 |
| lool | I didn't really test an old qemu-arm-static -> new world full upgrade though; if you do that, I'd love to hear the results | 18:23 |
| persia | I can't easily test the full upgrade, but I may be able to figure out a way. | 18:23 |
| persia | I can test broken -> fixed. | 18:23 |
| persia | Would a karmic -> lucid upgrade do it for the full upgrade case? | 18:24 |
| lool | Yes | 18:25 |
| lool | Or simply, removing everything, manually installing an old lucid qemu-arm-static and upgrading | 18:25 |
| persia | That's fairly trivial to do then: I can just install the karmic qemu-arm-static in a lucid chroot and dist-upgrade. | 18:26 |
| persia | Separately, I was thinking: would it make sense to use build-cross-chroot rather than build-arm-chroot, and then try to encourage enabling other architectures? | 18:27 |
| persia | I hear qemu powerpc is getting there, although sparc and ia64 probably need some work. | 18:27 |
| persia | My thought was that it would be nifty if developers could work on most architectures for most purposes from an amd64 install. | 18:28 |
| lool | persia: Ack, I don't like "-arm-" in "build-arm-chroot" | 18:28 |
| lool | I think qemu-debootstrap would be a better name for it | 18:28 |
| persia | I'd be happy with that name. | 18:29 |
| persia | Note that it's not currently command-line compatible with debootstrap. Specifically, `debootstrap --arch armel` works, but `build-arm-chroot --arch armel` doesn't. | 18:29 |
| persia | Both also work with "--arch=armel", but that's a separate case :) | 18:30 |
| persia | I've just upgraded from a broken system, and I still get "sudo: must be setuid root". | 18:30 |
| lool | or qemu-debootstrap-helper, but I would call it qemu-debootstrap and fix it to be compatible | 18:30 |
| lool | persia: Oh sudo is unrelated | 18:30 |
| persia | If you would, that would be great. | 18:30 |
| persia | Um, sudo is the problem I had, which ogra convinced me was related to the binfmt issue. | 18:31 |
| persia | What's the sudo issue? | 18:31 |
| lool | If you try running sudo with binfmt, the kernel will run qemu-arm-static which doesn't have the SUID bit set | 18:31 |
| lool | It can't work | 18:31 |
| lool | persia: Check whether you have one and only one arm entry in /var/lib/binfmt* and /proc/sys/filesystems/binfmt*/ | 18:32 |
| lool | The entry should be qemu-arm | 18:32 |
| lool | /var/lib/binfmts/qemu-arm and /proc/sys/fs/binfmt_misc/qemu-arm | 18:32 |
| persia | I do only have one of those. | 18:33 |
| lool | You have both and these are named qemu-arm though? | 18:34 |
| lool | If you do, then your system is correctly configured | 18:35 |
| persia | Right, but I still can't run sudo in my chroot, which was the original issue that brought the other to my attention. | 18:35 |
| persia | Any ideas? | 18:35 |
| lool | persia: I don't think that can work | 18:36 |
| lool | persia: Just avoid using sudo | 18:36 |
| lool | chroot as root and continue running stuff as root | 18:36 |
| persia | Why can't it work? Works for all my other chroots. | 18:37 |
| lool | Because the kernel runs a single binary for all binaries you run | 18:37 |
| lool | You could make qemu-arm-static suid root, but that would be a huge security hole | 18:37 |
| persia | Ah, I get it. I swear it worked on Tuesday, but maybe I did something else. | 18:38 |
| persia | It's probably worth noting that in a README somewhere, so people don't file bugs. | 18:38 |
| lool | persia: You might be able to run sudo as root | 18:38 |
| persia | That's probably it. | 18:39 |
| persia | my chroots were in a very primitive state on Tuesday. | 18:39 |
| persia | Anyway, I have to head off for a bit. Thanks a lot for explaining my issue. | 18:39 |
| lool | np, do feel free to pass the info to ogra :-) | 18:39 |
| persia | ogra: ^^ | 18:40 |
| * ogra reads it ... i'm just in fuzzy jetlag state | 18:40 | |
| lool | Ok, I give up trying to get virtio-pci to work under versatile | 18:41 |
| lool | It seems versatile has a very specific IRQ handling which might mess this up anyway | 18:41 |
| ogra | lool, did andy talk to you ? seems your patch didnt build at all | 18:49 |
| ogra | for now i asked him to just change the two config options you mentioned in the bug | 18:50 |
| lool | ogra: ? | 18:59 |
| lool | ogra: I talked to apw on #ubuntu-kernel; the kernel certainly built for me, as well as the modules | 18:59 |
| lool | I didn't build .debs though, so there might be some ABI issues | 18:59 |
| ogra | he said it exploded in his face on the first modules it tried to build | 19:00 |
| ogra | and said your patch changed more than just the two config options | 19:00 |
| lool | ogra: Which patch? | 19:06 |
| lool | Anyway, this is out of date, he merged my work AFAIK | 19:07 |
| ogra | i think he said you gave him a git tree or something ? | 19:07 |
| ogra | well, he said he didnt when we talked last night before i flew out | 19:07 |
| ogra | since it failed to build | 19:07 |
| lool | I attached a patch to the bug and the first commits of the git tree were to fix the lucid tree and then my patch | 19:08 |
| lool | Followed by large config updates in a separate patch | 19:08 |
| lool | That built for me, and he merged it | 19:08 |
| ogra | weird, he said it didnt | 19:08 |
| lool | He then decided to merge the updates in a single patch | 19:08 |
| lool | Check the tree yourself | 19:08 |
| lool | I certainly saw it merged | 19:08 |
| ogra | intresting | 19:09 |
| ogra | i only see the binutils.bin fix in the last upload | 19:10 |
| lool | It's in the git tree, not in the source package | 19:10 |
| ogra | right, because it made the package ftbfs according to him | 19:11 |
| persia | lool: Maybe it's because I'm in a chroot, but installing karmic qemu-arm-static in my lucid chroot creates double entries for me in /proc/sys/fs/binfmt_misc/*. dist-upgrading fixes it. | 19:21 |
| lool | persia: Yes, if you have it in the host, you'll see the entry in the chroot | 19:26 |
| lool | However if you add it in the chroot, it wont be visible in the host | 19:26 |
| lool | Can't believe lucid kexec-tools doesn't allow armv7l... | 19:26 |
| persia | Well, kinda. That was my experience for /var/lib/bin... but I seem to have the same /proc in and out of the chroot. | 19:26 |
| ogra | lool, huh ? we fixed that in jaunty ? | 19:26 |
| persia | NCommander was saying that kexec() hung his hardware. | 19:27 |
| * ogra clearly remembers he added a patch to kexec-tools | 19:27 | |
| ogra | hmm, but that might have been before the disastrous repackaging that lieb did | 19:27 |
| persia | You might want to reinvestigate | 19:28 |
| ogra | it wont apply anymore, the packaging lieb did isnt even remotely related to the debian packaging | 19:28 |
| ogra | he just took the latest upstream source and ran dh-make on it | 19:29 |
| persia | Can we just purge dh-make already? | 19:29 |
| ogra | and then copied all patches into debian/patches in case someone wants to re-apply (which indeed wouldnt work at all) | 19:30 |
| ogra | (IIRC) | 19:30 |
| persia | Right. That deserves an investigation to reclose n bugs. | 19:30 |
| lool | kexec-tools 2.0.1 fixes it though | 19:31 |
| ogra | did you reverse the mess ? | 19:32 |
| lool | I'm just reading the source | 19:32 |
| lool | lp:debian/kexec-tools and lp:ubuntu/kexec-tools don't relate at all, yeah | 19:33 |
| lool | haha I'm the one who uploaded the kexec-tools patch in jaunty | 19:33 |
| lool | actually I uploaded it to unstable | 19:34 |
| lool | http://paste.ubuntu.com/371141/ | 19:34 |
| lool | I pushed the new upstream release | 20:13 |
| ogra | great | 20:13 |
| ogra | to sad kexec is still broken in all supported kernels | 20:13 |
| ogra | (we tested it last week) | 20:13 |
| lool | How did it run last week? | 20:15 |
| ogra | quit good progress on the specs | 20:15 |
| ogra | we actually managed to get the workitem tracker to show green under the trendline :) | 20:16 |
| * ogra only has 4 items left in total and only tw in rootstock | 20:16 | |
| ogra | *two | 20:16 |
| ogra | the marvell issue was kind of eating the first half of the week though | 20:17 |
| lool | Let me reword my question: how could you test kexec last week since it doesn't run for utsname == armv7l? | 20:18 |
| ogra | hmm, ask NCommander :P he did the tests on both arches | 20:19 |
| ogra | i only saw it hanging when looking over his sholder | 20:19 |
| ogra | i guess he has to retest then tomorrow | 20:19 |
| ogra | thanks for pointing that out, i was expecting he had checked that before | 20:20 |
| ogra | ah ... | 20:21 |
| * ogra feels relief that jackd gets removed with the most recent upgrade | 20:21 | |
| ogra | and off to reboot ... | 20:21 |
| ogra | hmm, my boot got slower http://people.canonical.com/~ogra/osiris-lucid-20100207-3.png | 20:26 |
| persia | That's expected. Remember that the target is 10 seconds, not 8. | 20:30 |
| ogra | well, my system is usually a few seconds faster than the average HW | 20:30 |
| lool | New kexec still doesn't take armv7l, odd | 20:45 |
| lool | Bah I thought it was in 2.0.1 because Debian had 2.0.1 and had it, but it's not | 20:48 |
| lool | (and upstream had it too) | 20:48 |
| lool | it was committed after 2.0.1 | 20:48 |
| ogra | smells like another upload :) | 20:53 |
| lool | Ok, at least the new kexec crashes instead of refusing to run | 21:12 |
| lool | After the last fix, it does *something* | 21:31 |
| lool | "Starting new kernel | 21:31 |
| persia | That was the behaviour NCommander reported last week. At which point it hung. | 21:31 |
| lool | I see 100% CPU consumption from qemu | 21:32 |
| persia | Matches the previous report :) | 21:32 |
| lool | oddly, the versatile .deb build fails for me with /home/lool/qemu-versatile/linux-qemu/drivers/scsi/advansys.c:8352: error: implicit declaration of function 'dma_cache_sync' | 21:32 |
| lool | But only the debian way | 21:32 |
| lool | with the upstream make zImage and make modules, it builds | 21:33 |
| lool | I guess it's just not fatal for the upstream build, hmm | 21:34 |
| lool | I filed LP #518567 on the kexec issue | 22:01 |
| ubot4 | Launchpad bug 518567 in linux (Ubuntu) (and 1 other project) "armel/versatile: Can't kexec (affects: 1)" [Undecided,New] https://launchpad.net/bugs/518567 | 22:01 |
| lool | Got further by using a serial console | 22:01 |
| lool | Sent versatile fixes to kernel-team@ list | 22:19 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!
