[00:00] is there anything mounted under chroot/proc? [00:00] Oh, derp. [00:01] llvm-* build-deps on binfmt-support. Of course. [00:01] Now I feel silly. [00:01] so umount-chroot is just broken, because it assumes that reversing the contents of /proc/mounts (which is somewhat random) will give a nicely ordered list [00:01] lamont: So, we just need smarter umount ordering, probably. [00:01] just loop [00:01] lamont: Walk tree depths. [00:01] it's saner than trying to order [00:01] (or loop) [00:01] and life is short. feel free to update the bug? [00:02] * lamont lets the panda fsck [00:02] 210GB disk... sigh [00:03] You know, screw sorting, I bet if you just inverted the list from the grep /proc/mounts, it would solve it. [00:03] And much more elegantly. [00:03] we already do that... [00:04] and that's probably because we needed it _that_ order [00:04] I don't see it inverted. [00:04] maybe that's just my brain messing with me [00:05] This is essentially what lp-buildd does: [00:05] root@shiva:~# grep oneiric-llvm /proc/mounts | cut -d\ -f2 | xargs -r -n 1 echo [00:05] /srv/chroot/oneiric-llvm/proc [00:05] /srv/chroot/oneiric-llvm/dev/pts [00:05] s/echo/umount/ for the real deal. [00:05] /srv/chroot/oneiric-llvm/proc/sys/fs/binfmt_misc [00:05] Invert that list, bug fixed. [00:06] for now. the ordering in /proc/mounts isn't exactly scientific [00:06] It's FIFO. [00:06] Sure, a package could mount -o=remount /proc in its postinst and mess with you, but... Eww? [00:07] (I guess you could keep the loop too, but still invert the list) [00:09] Anyhow, commented on the bug and the merge request. [00:10] Reversing the order would be more efficient, but the loop is probably a safe catch-all anyway. [00:13] lamont: The plus side is that there's no "underlying issue" or processes cleverly hiding from us, just a bit of umount PEBKAC, so I'm that much happier about your hack. It will DTRT. :P [00:13] heh [00:14] thanks for digging into it [00:14] in other news, chort.buildd will shortly be another panda builder [00:14] \o/ [00:16] so if the panda isn't at 115200, what speed should I try? [00:17] That's the only speed it should be... [00:18] that's sad [00:18] lamont: Actually, regarding the same bug. No complex heuristics are required. Piping a "sort -r" between the cut and xargs will DTRT. [00:19] lamont: Cause paths sort quite nicely. :P [00:19] for the non-pathological case, yes [00:19] Build me a case where it fails? [00:20] �� <-- I do not like it when hitting return several times gets me that.. [00:21] I'm having BBS flashbacks. [00:21] they're only slightly faster than some of them [00:22] for the real BBS experience, you want a beaglexm or a bbg3 [00:22] And a copy of LORD. [00:25] will that even run on arm? [00:26] Under freedos under qemu, sure! [00:26] Or something equally convoluted. [00:37] Well, time to go have a birthday. I'll be around here and there on the weekend for releaseish stuff. [00:50] infinity: happy birthday [00:51] Danke. [01:04] infinity: I just realized that serial console is somewhat overrated [01:53] until it isn't. :( [08:48] infinity: curl> I hadn't noticed it, but yeah, I'm probably biased. If you want to, that'd be great [08:48] infinity: on the umount thing, maybe umount -l would be helpful as well? [09:51] cjwatson: Already did overrides and accepted the two new udebs. [09:51] cjwatson: As for the umount thing, I'm leaving it in the hands of lamont and the launchpad team at this point, I've commented more than enough on the bug. :P [10:24] ta [12:45] Hello folks, I recently added a new bzr branch for a package (lightdm-theme) for our team, ~ubuntustudio-dev. We'd like to include it in oneiric, still. Does it need an FFe? [15:25] astraljava: Technically yes, but on matters of flavor branding we'll approve whatever you all want, so whoever is the ubuntu studio lead just needs to say it should go in. [15:31] ScottK: Thanks, just a minute ago I got a response on -devel, though, that I should file an FFe. It's here: http://pad.lv/852623 [15:48] ScottK: hi, ScottL here (ubuntu studio project lead) asking about getting a lightdm theme into the repos [15:48] ScottK: anything special we would need to do? [16:07] scott-work: Ack in the bug that you want it (I approved it for the release team based on that and then subscribe ubuntu-sponsors/find someone to upload it. [16:08] ScottK: aye === s1aden is now known as sladen [20:27] cjwatson: umount -l feels way too much like a total hack to me. [20:27] 18 arm builders are making a bit of a dent in the rebuild test