=== BenC__ is now known as BenC === cpg is now known as cpg|away === Logan__ is now known as Logan_ [02:39] Anyone know how to fix this? E: Couldn't download packages: python-minimal python2.7-minimal sysvinit-utils [02:39] when running pbuilder create [02:40] oh nevermind [05:19] Good morning === cpg|away is now known as cpg [05:27] bdmurray: followed up to bug 1048059 [05:27] Launchpad bug 1048059 in udisks2 (Ubuntu) "Adding ACLs to /media/$user does not work" [Undecided,Incomplete] https://launchpad.net/bugs/1048059 === chilicuil is now known as chilicui1 [05:46] Riddell: btw, did you see my response to your apport-kde patch? it's still dubious to me === dendrobates is now known as dendro-afk [05:56] ev: apport-retrace crash> reproduced in a test case, and fixed in http://bazaar.launchpad.net/~apport-hackers/apport/trunk/revision/2514 [05:57] ev: you are running retracers out of trunk, I presume? so this should just work === chrisccoulson_ is now known as chrisccoulson [06:57] good morning === smb` is now known as smb === Sp4rKy_ is now known as Sp4rKy [07:22] Morning dholbach. [07:23] hey iulian [07:56] ev: quick question - I use the recoverable errors mechanism in software-center right now to learn about the failure that people get when installing stuff - I would like to break it down further, can I use a string like "resolve-failed", "no-policykit" etc for the error and send that as "Traceback" or does it need a real traceback, i.e. will I just have to make up a bunch of exception? the ration is that it current just sends a TransactionFaile [07:56] d which apparently put them all into the same bucket [08:05] ev: or can I have some other "signature" mechanism to help the thing that looks for the dupes to sort them into the right buckets? [08:07] mvo, while we're on that subject, can you explain the difference between what you're using RecoverableError for, and the built-in ? [08:08] * mpt wonders if it's actually that the latter isn't implemented yet [08:09] mpt: the s-c mechanism will kick in even if no apt/dpkg is touched, i.e. if there is no policykit authentication agent running in the session or if there is a dependency resolve failure [08:09] pitti: yeah I'm not entirely convinced by the apport change but it seemed to work [08:09] Riddell: but dropping the sys.exit() seems wrong, it always exits with 0 now? [08:09] mvo, does that mean it's possible for a single installation request to trigger both errors? === henrix_ is now known as henrix [08:11] (ev is afk atm) [08:11] pitti: I just copied what apport-gtk does, that has no sys.exit() [08:11] Riddell: hm; curious that the sys.exit() causes a problem [08:11] Riddell: what's the crash anyway? I do get an error message, but I get it no matter whether it calls sys.exit() or not [08:12] python3: Fatal IO error 2 (Datei oder Verzeichnis nicht gefunden) on X server :0.0. [08:12] Riddell: that's "(file or directory not found)" [08:12] is there any chance the poppler 0.20.5 that i released yesterday finds it way to quantal? it has lots of crash fixes that could be CVE classified if we issued CVEs :D [08:13] mpt: I need to double check - it could be [08:14] pitti: it's a crash somewhere in the python binding code, previously it ran sys.exit then it ran the qt mainloop which confused it [08:14] Riddell: the qt mainloop was dead code; sys.exit() doesn't return [08:14] apw: So, the desktop amd64 image should now boot with a signed kernel, assuming I got the scripting right at oh-god-o'clock last night, but the server image is a different story [08:15] apw: What do you think about having linux-signed emit a kernel-signed-image udeb? [08:15] Hello. ATI legacy driver is not working properly on toshiba l500. Unity is failing to start and screen resolution is broken. [08:17] cjwatson, well i guess i'd have to make it manually (the udeb), but i guess its feasable [08:18] pitti: hmm, and now I add it back the crash doesn't occur :( [08:19] apw: It'd basically be a vector for getting it into /dists/quantal/main/installer-amd64/current/images/ === mcclurmc_away is now known as mcclurmc [08:20] cjwatson, is there somewhere i can read about what has to be in a udeb [08:20] Just ship /boot/vmlinuz-* [08:20] * apw pokes the nest to see what happens [08:20] There's doc/devel/modules.txt in the d-i source tree but it's probably not hugely useful for kernel stuff [08:21] Put Package-Type: udeb, Section: debian-installer, Priority: extra in the control stanza [08:21] Call it kernel-signed-image-$(ABI)-di [08:21] Should basically be it [08:21] ok ... sounds doable ... [08:31] ev: aha! DuplicatesSignature is the one I can use I guess - I will try that nw [08:32] cjwatson, http://people.canonical.com/~apw/misc/linux-signed-image-3.5.0-17-generic-di_3.5.0-17.28+signed2_amd64.udeb does that look like the sort of thing? [08:33] apw: linux-signed-image -> kernel-signed-image please to match the kernel-image udeb [08:33] apw: Except that, for whatever reason, the naming con... yeah. [08:33] Fairly historical but it would help my sanity some [08:33] yeah no problem [08:34] so dh seems to grok udebs, i am slightly supprised [08:34] Yeah, ages ago :) [08:34] debhelper and debian-installer both sprung from the same bearded mind, it would be odd if they didn't get along. [08:34] Most of the d-i packages tree is dh(1) with rather few overrides these days [08:35] apw: Possibly add "Provides: kernel-signed-image" for reasons to do with the d-i build system I haven't had enough coffee to explain properly :-) [08:36] cjwatson, i am in your hands, whatever you need :) [08:36] (I can't remember if that's strictly needed but it would be regular and wouldn't hurt) [08:36] apw: Everything else looks fine [08:36] Though obviously the proof will be in actually managing to construct images using it [08:37] But probably simplest to do that with it in the archive - I see no other structural problems [08:38] cjwatson: Were you going to try to make netboot work with SB too, or is that way out of scope for the next N hours before you give up on quantal? :P [08:38] infinity: This should handle netboot as a side-effect [08:38] With any luck [08:38] Cute. [08:38] infinity: Well, the netboot mini.iso [08:38] please don't break it :) [08:38] Yeah, I was thinking actual netboot. I assume SB also diddles with PXE? [08:39] infinity: Actual netboot will take longer - I have a to-do item to test GRUB efinet [08:39] infinity: UEFI netbooting is a kind of separate thing with its own protocol stack [08:39] This doesn't surprised me. [08:39] Nor surprise. [08:39] I've only stuck my toe in it so far [08:39] But it doesn't also provide a PXE emulation layer? [08:39] cjwatson: How well standardised to you expect it to be? [08:40] Oh, but maybe enabling such a beast wouldn't require signed images. [08:40] infinity,Daviey: no idea to either of those, would have to read the spec in detail [08:40] which I do not have time for right now :) [08:40] Daviey: It's bound to be more standardised than PXE. [08:40] Daviey, that sounds like a server thing, i think i'd ask Daviey about it :) [08:40] My breakfast is more standardised than PXE. [08:40] Daviey: Which is basically just "the world attempting to mimick Intel's behaviour, badly". [08:40] apw: he's a joker, who doesn't know jack. [08:41] Daviey, we'll have to introduce him, jack is a nice fellow [08:41] infinity: with arm doing it worse :) [08:41] cjwatson: Anyhoo. Sounds like another d-i upload in the offing, there's an armadaxp ABI bump building, if you'd like to do that too. === henrix is now known as henrix_ [08:41] infinity: So, if we're lucky, it works in the signed images, with some configuration yet to be determined. If we're unlucky, R + quantal-updates. [08:42] infinity: Yeah, was on my list. [08:42] Daviey: u-boot's PXE emulation was a last-minute hacking "here, have this, cause you're a bunch of whiney gits who don't think that bootp+tftp is cool enough for you", don't complain about it. :P [08:42] Daviey: Real UNIX admins <3 bootp+tftp. [08:42] infinity, it was a simpler time === henrix_ is now known as henrix [08:42] infinity: I was one of the whiney poeple that complained the pxelinux implementation didn't respect enough of the pxelinux standard :) [08:42] infinity: yeah I set up my wndr3800 router for that, it points netboot to my boot server :-) [08:43] Daviey: It's almost kinda halfway functional now, thanks to all of the bugs/whining from your team. [08:43] netboot ftw [08:43] Daviey: Nowhere near as feature-complete as actual PXE (and never will be), but it seems to mostly DTRT now. [08:48] cjwatson, ok if you are happy i'll chuck this in the queue [08:49] apw: Yep [08:50] apw: Do you think there's any hope at all of bug 1065263 getting in, or are we toast at this point? [08:50] Launchpad bug 1065263 in linux (Ubuntu) "wrong stride for efifb on some systems" [High,Triaged] https://launchpad.net/bugs/1065263 [08:50] * apw will look at the fixes there next [08:50] Basically hunting for ways in which we might manage to QA the server image on SB [08:51] Though we may end up waiting for stgraber to unpack on Friday anyway [08:52] cjwatson: I didn't think Server/SB was a requirement for 12.10? Is it compelling to squeeze it in? [08:53] I'd been given to understand that some level of SB support was a requirement across the board for 12.10 [08:53] Daviey: feel free to clarify with Rick [08:53] Plus, it should be nearly there anyway [08:54] (Except netboot, which is an unknown quantity) [08:54] cjwatson: I'm not pushing against it.. I thought when we discussed it before, it was a Destkop deliverable for 12.10... :) [08:54] I guess I've run into bureaucracy again. [08:54] * cjwatson goes for breakfast and coffee [08:54] But weehaaa, value added :) [08:58] mpt: I fixed that it would also report dpkg errors now, thanks for this hint [08:59] mvo, you mean that before it would, but now it won't? [08:59] mpt: yes [08:59] ok [09:01] Daviey: it's just, things I don't want to be doing a week before release rather than fixing RC bugs: [09:01] * arguing about whether something is a required deliverable or not [09:01] if somebody else wants to do that and tell me the answer, awesome [09:02] cjwatson: That is what i was /trying to help with. If it's something that isn't needed, perhaps time is best spent elsewhere? [09:02] and you may well be right; but OTOH I'd be surprised if the efifb bug didn't show up in some desktop use cases too [09:02] (shrug* [09:02] I really like to get the server installer working because it's a lot easier to debug than ubiquity, too [09:03] Plus don't like leaving jobs incomplete [09:03] right! === tkamppeter_ is now known as tkamppeter [09:38] unity-2d seems to mess with drag and drop within my Qt application. Anyone know how to prevent/fix this? === doko_ is now known as doko [09:39] pitti: would you care if I point some gnome/gtk/vala universe issues? [09:39] doko: I'd rather you send them #ubuntu-desktop-wards, TBH [09:40] I can look into an FTBFS or two, but I'm already doing more disto work than intended [09:40] doko: (pitti isn't desktop team any more) [09:40] cjwatson, but vala specialist ;) [09:40] steep career -- touch vala once, and you are a specialist :) [09:41] If somebody could fix avant-window-navigator, I'm really reluctant to remove that since I get the impression it's widely used [09:41] That'd probably be one of the more valuable things to do [09:42] uh, I remember looking at that at the beginning of precise already [09:42] cjwatson, I don't see the ftbfs ... [09:43] It doesn't fail right now, but it requires a version of vala that's slated for removal [09:43] It's on the ubuntu-archive bug list [09:44] my d'n'd problem is prevented when the dock is not hidden. Turned of auto-hide for now. But supprised there is no bugfix [09:45] cjwatson: UEFI bricks Samsung laptop bug > I've attempted to identify x86 UEFI contacts at Samsung by pinging Samsung folks working on UEFI for ARM, but no luck so far; we're now passing the bug to their management in case they know someone but Samsung is so big... [09:50] Thanks [10:00] sse2 isn't default, even for amd64??? [10:03] hmm, seems so === amitk is now known as amitk_ === amitk_ is now known as amitk__ [10:18] what's the command to know which packaging branch a package has? [10:26] pitti: btw, I've commented on your unittest blog post. I've got an answer that I think you might like. [10:27] pitti: i have a patch for 0007-add-lightdm-support.patch in accountsservice that you originally did (or so says http://anonscm.debian.org/gitweb/?p=collab-maint/accountsservice.git;a=history;f=debian/patches/0007-add-lightdm-support.patch;h=da802bd00494292baa5c512adcb550a6f7b3b08b;hb=HEAD ) what do i do with it? [10:28] jml: hey, thanks; vila was just pointing out testtools as well, thanks! [10:29] jml: I'll mention this when I bring this up with upstream, whether they like that additional module [10:29] jml: it's indeed a lot more elegant [10:30] pitti: thanks. all that is pretty much lifeless's design. now that I've been using it, it's very hard to go back. [10:30] tsdgeos: it's from mterry; I suggest you attach it to a bug, and then we assign that to Mike? [10:30] ok, works for me [10:30] tsdgeos: thanks! [10:30] tx [10:30] pitti: making things out of composable pieces turns out to be a great idea [10:31] jml: indeed; and it would have been surprised if I were the first one to run into this [10:31] jml: *want in py3 by default* [10:32] *sigh* [10:32] I've had difficulty getting stuff into Python [10:32] but very much support it in principle, and we've made sure the licenses are all good for that. [10:36] I guess that's mostly because I've been lazy, and not super willing to engage on python-ideas. [10:36] jml: talk to barry and jobs done ;-) [10:37] jml: oh, http://pkgs.fedoraproject.org/cgit/python-testtools.git/ [10:37] jml: so it's in Debian, Ubuntu, Fedora -- good 'nuff :) [10:37] jml: (but would still be nice, of course) [10:37] pitti: oh neat. [10:38] jml: they don't seem to build it for py3, though [10:39] mpt: the graph is fixed: http://poppy-dev.local/ [10:39] pitti: a shame. [10:39] pitti: because we take Python 3 support pretty seriously [10:40] jml: yeah, and I write pretty much everything in py3 these days, so let's see how much that hurts with getting tests upstream ;) [10:40] well, I still need to change the calculation to make "by 12.04 standards" be "all collected" - "recoverable problems", but the 5th is back to normal [10:40] pitti: good grief, I think you're the first person I've heard say that. [10:41] jml: really? we have made great progress in py3-ification in quantal [10:41] ev, brilliant [10:41] jml: in a default install we are down to two packages that use py2 still, I think (systme-config-printer and ubuntuone) [10:42] pitti: I guess most of the people I talk to don't target quantal for their work. [10:42] If launchpadlib were ported I'd use py3 for everything. === MacSlow is now known as MacSlow|lunch [10:43] cjwatson: oh. hmm. interesting. I've got a half-formed plan to burn launchpadlib and have some kind of phoenix rise from its ashes. [10:44] * cjwatson dislikes non-incremental plans :-) [10:45] cjwatson: yeah, me too. [10:45] cjwatson: but there seems to be a fad for porting things to new APIs and languages for incremental benefit [10:46] infinity, cjwatson: did you give back google-glog this morning? failed again, but builds locally ;-/ [10:46] doko: no [10:46] can't we just have it in a non-python language ? [10:46] Daviey: like Python 3? :P [10:46] ev, it won't matter once "by 12.04 standards" is dotted, but for now it might be clearer if "all collected" was layered on top instead of underneath. [10:47] jml: No, like erlang.. or soemthing [10:47] * cjwatson looks sidelong at Daviey [10:47] cjwatson: more seriously, writing testable code that performs well with launchpadlib is way trickier than it has to be, and I think that's a direct result of the Python API making network calls appear transparent. [10:47] jml: I think what concerns me is the ordering of your plan. [10:48] pitti: if you want to help with systme-config-printer ... there's still a dependency which needs conversion [10:48] cjwatson: ashes first, phoenix second? [10:48] Something new's all well and good, but launchpadlib is production code and we can't afford to have it die before any replacement is ready. [10:48] cjwatson: oh yeah, sure, I didn't actually mean kill launchpadlib. I don't know how I would even do that or what it would mean. [10:51] I guess a new release of launchpadlib that totally broke API compatibility would do the trick. I don't know why anyone would do that though. === cpg is now known as cpg|away [11:15] mpt: fixed [11:16] ev: hey Evan, how are you? did you see my apport ping this morning? [11:16] splendid [11:16] pitti: I didn't until you mentioned it just now [11:16] wonderful! [11:16] thanks so much [11:19] no worries [11:30] hi. init: mounted-proc main process (1268) terminated with status 1 [11:30] anyone? [11:34] jml, pitti talk to me! we're starting to look for fun projects for py3.4 :) [11:34] got that w/ 11.10 12.04.1 and 12.10 beta2 [11:34] btw, ubuntu-core [11:35] barry: put testtools & fixtures into core Python. [11:35] barry: also, if you could make function definition an expression that'd be great too. [11:37] barry: the bit of testtools under discussion here is the 'details' API, which allows attaching arbitrary content to test results. [11:37] jml: the first two would be great. the last would be cool too, but probably harder to get through the gauntlet. but that's the kind of big ideas we're looking for. have you seen the recent discussion on generator based concurrency? we have a group of 4 dc pythonistas ready to sprint on big fun stuff [11:38] barry: no, I haven't. I assume dash's post was relevant though [11:40] barry: oh, actually, if you want bonkers crazy big ideas, I've got two more: [11:40] barry: persistent (i.e. immutable) data structures ala Clojure. (esp for dicts) [11:42] barry: do something about the Python module system: http://washort.twistedmatrix.com/2011/01/modules-in-python-good-bad-and-ugly.html (see also follow-ups) [11:45] jml: that's a very nice blog post. current thinking is generator-based with 'yield from' being an enabling syntax in 3.3. the examples i've seen so far are (imho) much easier to read and reason about than deferred/callback style. [11:48] barry: so, with iterCallbacks at least, you can't easily express "do a & b concurrently, and when they are both done, do c" [11:49] barry: and that's pretty important. [11:52] the module post is interesting. i think 3.3's switch to importlib provides a great opportunity for more experimentation. there is almost no built-in magic any more (the sys caches are still there, but we have thoughts about how to get rid of some if not all of them. to radical for 3.3, but maybe we can do something about them in 3.4) [11:52] jml: these are all great ideas. exactly the kind of thing we're collecting [11:53] barry: glad you think so. :) [11:54] jbicha, in June you synced google-glog, dropped ubuntu changes, and didn't look at the build failure :-/ [11:54] barry: re immutable data structures, that can be done easily enough without language changes, but to do it fast probably requires CPython extensions. [11:55] jml: do you have a reference for that? [11:55] now looking at the arm build failures too [11:55] barry: and, also, you wouldn't believe how many things there are on PyPI that implement dict helpers. [11:55] barry: reference for what exactly? [11:56] clojure's immutable data structures [11:58] barry: oh, umm, not really. it's a btree that shares structure. I think it's actually a solution to an exercise in Cormen's "Intro to Algorithms". https://github.com/clojure/clojure/blob/master/src/jvm/clojure/lang/PersistentHashMap.java [11:59] barry: and there's an explanation of how it works in _The Joy of Clojure_. === dendro-afk is now known as dendrobates [12:00] jml: cool [12:00] barry: actually, iirc, dash wrote a pure Python implementation. [12:04] jtaylor, Riddell: could you have a look at the komposer ftbfs on i386? === MacSlow|lunch is now known as MacSlow [12:08] doko: despite its name, there's nothing KDE in that package :) [12:08] ohh :-/ [12:09] You probably want micahg or someone that knows mozilla stuff. [12:16] anyone here use ubuntu-core armv7hfp? [12:20] ericbutters, some do, its a cheap way of getting a chroot (not of much use for anything else since many bits are missing for a proper rootfs though) [12:21] where can i find xubuntu (xfce) or lubuntu for arm? [12:21] we only build images for a few arches, on what SoC are you trying to run it ? [12:21] S5PV210 (samsung) [12:22] i got fedora-arm console image running [12:22] and installed lxde [12:22] http://cdimage.ubuntu.com/lubuntu/daily-preinstalled/20121008/quantal-preinstalled-desktop-armhf+ac100.tar.gz has a lubuntu tarball ... [12:23] orga_ thank you.. i will try.. did you read my first post regarding mounted-proc? [12:23] after untarring run: touch /var/lib/oem-config/run [12:24] ericbutters, most likely you are missing config options,. the tarball is originally for the ac100, look in /boot/ for the configuartion file and compare it to your config options [12:25] (modulo the arch specific oprions indeed) [12:27] ericbutters, oh, and there is an #ubuntu-arm channel btw [12:27] ogra_ ui :) nice! === _salem is now known as salem_ [13:10] cjwatson: oh, btw, one thing I noticed yesterday. We're adding memtest86 to grub even on EFI systems, but it requires linux16/initrd16 which don't exist [13:12] stgraber: awesome [13:13] stgraber: I'll fix that === MacSlow is now known as MacSlow|errand [13:18] cjwatson: thanks [13:20] mvo: i'm sorry, i'm not sure if we finished our discussion from yesterday about python-apt [13:27] doko: For google-glog is there a reason you only changed the build-dep to add it for i386 and not arm* as well? [13:27] ScottK, still ftbfs with it [13:28] OK. Thanks. === amitk is now known as amitk-afk [13:58] mvo: Does bug 1056752 still need to be open in Ubuntu? [13:58] Launchpad bug 1056752 in Ubuntu Quantal "Package description translations export not working" [High,Triaged] https://launchpad.net/bugs/1056752 [13:59] cjwatson: closing it now, that is fixed and uploaded [13:59] Thanks === MacSlow|errand is now known as MacSlow [14:20] barry: I think I'm good with python-apt, thanks. I ran into another aptdaemon issue today and tracked it down to http://paste.ubuntu.com/1273298/ - py2 seems to be rather unhappy about unicode DBusExceptions() [14:20] barry: this seems to be the root behind bug #846044 [14:20] Launchpad bug 846044 in aptdaemon (Ubuntu Quantal) "software-center crashed with UnicodeEncodeError in get_dbus_message(): 'ascii' codec can't encode character u'\xfc' in position 65: ordinal not in range(128)" [Critical,Fix released] https://launchpad.net/bugs/846044 [14:22] mvo: dang. i cannot wait until we can kill py2 once and for all. let me know if you want some input or help with that problem. [14:22] barry: I'm knocking up a patch right now (I hope) [14:23] ;) [14:31] mvo: btw, you know that "to knock up" == "make pregnant" in US English, right? :-) [14:32] pitti: I had no idea! [14:32] mvo: better get a room, you and your patch! [14:33] * mvo will refrain from using that phrase again in the future [14:33] pitti: ;) [14:33] mvo: no worries, I fell into that trap as well [14:33] mvo: but after I learned about it I finally understood the title of that Futurama episode [14:33] pitti: not a happy story right now anyway me and my patch not sure that there is going to be a long term future [14:33] ("Kyf gets knocked up a notch") [14:33] pitti: haha [14:33] lol [14:38] You could say "Knock together a patch" with no unfortunate connotations. [14:41] or "whip up" [14:41] Whip up is good. [14:44] doko: I was tempted to file a removal request for kompozer, upstream is dead [14:46] OMG this package has per-function changelogs [14:46] verbose much [14:49] micahg: Please do. [14:53] ScottK: Bug #1065547 , have fun :) [14:53] Launchpad bug 1065547 in kompozer (Ubuntu) "Please remove kompozer from quantal and blacklist" [Wishlist,Confirmed] https://launchpad.net/bugs/1065547 [14:53] thanks. [14:53] FWIW, the Debian maintainer is subscribed to the bugs, so maybe he'll fix it now :) [14:54] I was thinking about not blacklisting since if it comes back, it'll surely be fixed ~somewhat. [14:55] RC bug filed 18 months ago with no NMU...doubtful [14:56] OK. [14:56] still, doesn't seem a particular need to blacklist [14:59] ok, fine, skip the blacklist [15:01] Yeah, nowadays blacklisting is for "never in Ubuntu again, no really". [15:01] e.g. for kfreebsd-* it's quite reasonable [15:01] But because we used to have to use the blacklist much more, people are in the habit of massively overusing the blacklist and it creates a maintenance problem for the future. [15:01] stgraber: will it matter for us at all if we take Dwight's patch to switch 0.8.0-rcX with 0.8.0.rcX ? [15:04] micahg: Done. Please fix the recommends in ezgo-network. [15:05] ScottK: is a missing recommends a problem? we have no diff ATM === kengyu_ is now known as lexical [15:09] pitti: thanks I commented on the bug too [15:11] hdf-eos5 is officially awful [15:11] Insanely repetitive code, and it seems to use memmove if and only if it doesn't need to [15:19] cjwatson, can you confirm that i should not need to specify anything special to have dpkg unsafe-io used during installs [15:21] smoser: confirmed [15:22] why does installaiton take so long. [15:23] i know, it is completely not helpful an di have no data. but it justseems like install takes longer than it should. [15:27] cjwatson: In the last hour, i've just had two netinstalls fail to boot. Do you think it is SB related? [15:27] seemingly stuck at bios [15:27] It seems unlikely but I suppose I cannot rule it out [15:28] smoser: Why doesn't the server team profile it then [15:28] It's two differ net [15:28] I can't do everything for you! [15:28] It's two different machines, which were reliably working earlier today.. has anything changed [15:28] cjwatson, i cant be expected to give actual data! [15:29] i want to make arbitrary unsubstantiated claims! [15:29] Daviey: SB is the obvious thing, I suppose; they might object to the extra certificate data at the end of the kernel [15:29] Oh [15:29] I did test in BIOS kvm but who knows [15:29] Daviey: We keep multiple versions of d-i images for this kind of reason [15:29] Daviey: Compare old and new kernels to see if that's all it is [15:29] smoser: Do you want to try stripping some noise from the kernel command line? [15:30] noise? [15:30] or that.. [15:30] smoser: Don't we have some essential things that can be removed [15:30] * Daviey needs to dash for 20-30 mins [15:30] The kernel command line hasn't changed recently [15:30] cjwatson: Sorry, this is a MAAS based install. :) [15:31] Sorry, not my fault that you can't get your overly complex infrastructure to do a simple test [15:31] hie hie [15:31] Surely it can manage a different installer version [15:31] surely it cannot. [15:32] cjwatson, where would i get an older installer to start poking [15:32] http://archive.ubuntu.com/ubuntu/dists/quantal/main/installer-amd64/ [15:32] 181 is unsigned kernel, 183 is signed kernel [15:33] I assume you know which bits under that you're using [15:33] probably something under images/netboot/ [15:34] cjwatson, thanks. i can figure that out. [15:34] ok. if that turns out to be it then I can either do a straight revert or else (preferably if I have the time) come up with something to ship unsigned for BIOS and signed for UEFI [15:35] smoser: fwiw if we didn't have to support minimalvm then the base squashfs could be much more extensive and I expect it would be rather quicker as a result. [15:35] it was that way during quantal for a while until we realised that it broke minimalvm. [15:36] (alternatively some kind of multi-squashfs scheme, but that's not been a road down which very much success has ever lain ...) [15:37] oh shoot. [15:37] ugh [15:37] hmm [15:37] (never mind, misread) [15:37] we *should* hvae been installing precise [15:38] in relation to slowness, or to boot failures? [15:38] actually both. [15:39] jus ignore me for the moment. [15:42] cjwatson, but for slowness, precise should still have unsafe-io, right? === dendrobates is now known as dendro-afk [15:42] smoser: Yeah [15:43] debootstrap rather than squashfs-base, though, which makes some difference [15:43] some. [15:43] but just watching syslog messages of install [15:43] just seems slower [15:50] cjwatson: is a missing recommends for kompozer a problem (it shows up on the NBS list), we're in sync on the package now (ezgo) [15:51] micahg: Technically, yes, but if there's no diff, meh. [15:51] We can break missing Recommends if we need to [15:51] Better not to if possible, but yeah, what ScottK said [15:51] right [15:51] I'll NBS that now [15:52] thanks === amitk-afk is now known as amitk === dendro-afk is now known as dendrobates [16:00] slangasek, hi [16:03] Hi. What's the preferred method for submitting a bug if apport hangs while collecting the information? Just submit the bug manually? [16:06] tkamppeter: hi there [16:28] slangasek, does cups crash often for you? In which situation? [16:30] Laney, https://launchpadlibrarian.net/119497991/buildlog_ubuntu-quantal-armhf.cabal-debian_1.25-1_FAILEDTOBUILD.txt.gz [16:30] tkamppeter: it crashes once a day for me [16:30] runhaskell not available on arm? [16:31] doko: indeed [16:31] ok, thanks [16:31] you can use ghc manually [16:33] tkamppeter: I had speculated that it might have been related to cron.daily and logrotate-based force-reload, but that doesn't seem to happen at the right time of the morning === dendrobates is now known as dendro-afk [16:36] slangasek, does it happen once a day? At random time points? Always the same time? [16:42] tkamppeter: I always get the crash notice in the morning; I don't know if it's always at exactly the same time; today's crash file has a timestamp of 8:04, the main cups process started running at 07:53; the access log appears to have been rotated at 07:48 [16:43] tkamppeter: testing to see if I can reproduce the crash by running invoke-rc.d force-reload [16:45] slangasek, you can find the exact time point of the crash by the segfault message in /var/log/syslog. [16:47] slangasek, if I run "sudo invoke-rc.d cups force-reload" I do not get a crash, but I also do not have a crash report every morning. [16:47] tkamppeter: I haven't gotten a crash /yet/ after doing a force-reload; but that does not seem surprising, given the gap in time between the process starting and the crash file written [16:48] tkamppeter: I don't have any mention of this crash in syslog (though, I'm not using the stock syslog config) [16:49] tkamppeter: ok, I found it in /var/log/apport.log, apport was invoked at 7:53 [16:49] so something changed the timestamp on the crash file afterwards [16:49] (I blame whoopsie) [16:50] tkamppeter: so that corresponds to the time when the current cups process started; but there's still a gap of 5 minutes between when the log file was rotated and the crash happened, which doesn't make much sense [16:51] slangasek: could you fiddle with the ulimits in the cups startup to let it dump core, so you could at least get a stack trace out of it? [16:51] (that might be tomorrow, of course, but a core in the morning is a good way to start the day...) [16:51] slangasek, perhaps crash notices emitted by the kernel can also appear in kernel.log, dmesg, ... would be an alternative place for uses who suppress these messages in syslog. [16:51] tkamppeter: also, the apport log shows that it did crash again in response to my most recent reload - apport just didn't re-report it [16:52] sarnold: it *is* dumping core, to apport [16:53] slangasek: ah, good :) [16:54] slangasek, I have 3 bug reports now with similar stacktraces: bug 1034045, bug 1061457, and bug 1051799. Probably therefore Launchpad does not take more of these. [16:54] Launchpad bug 1034045 in cups (Ubuntu) "cupsd assert failure: *** glibc detected *** /usr/sbin/cupsd: free(): corrupted unsorted chunks: 0x00007f3dc478c0f0 ***" [High,Triaged] https://launchpad.net/bugs/1034045 [16:54] Launchpad bug 1061457 in cups (Ubuntu) "cupsd crashed with SIGSEGV in find_subtree_recurse()" [Medium,Incomplete] https://launchpad.net/bugs/1061457 [16:54] Launchpad bug 1051799 in cups (Ubuntu) "cupsd crashed with SIGSEGV in dbus_connection_dispatch()" [Undecided,Incomplete] https://launchpad.net/bugs/1051799 === dendro-afk is now known as dendrobates [16:55] slangasek, sarnold, the crash happens in D-Bus code called by Avahi code called by CUPS code. Within the D-Bus code the crash happens on different places but the path in the Avahi and CUPS code seems to always be the same. === mcclurmc is now known as mcclurmc_away [16:56] slangasek, sarnold, all other known CUPS crashes seem to be fixed with only this one remaining. [16:57] slangasek, it seems probable that the problem is somewhere in Avahi or D-Bus. [17:03] slangasek, sarnold, according to the stacktraces the problem happens for sure on the shutdown of CUPS (can also be the shutdown part of a reload). [17:03] * slangasek nods [17:06] anyone know what would be adding a foreign arch for me? [17:06] unit222.pas(30067,17) Fatal: Procedure too complex, it requires too many registers [17:06] nice compiler ... [17:08] smoser: is the foreign arch i386? [17:08]