[12:03] jdub: oh god [12:03] ok, processed [12:03] we are going to need to publish a script to tell people how badly broken their systems are :P [12:03] is the live CD looking better? specifically, is ipw2200 working properly now? [12:03] Kamion: i'm considering writing a mail about it, both to them and to ubuntu-users [12:03] thully: that's the fix I speak of [12:04] thully: on its way ... [12:04] how hard will it be for those people to upgrade to hoary? [12:04] ajmitch: many of the versions in those repositories are greater than those in hoary, so apt will not upgrade them automatically [12:04] Kamion: that's what I was worried about.. [12:04] causing much pain & anguish [12:04] I'm not sure it's possible to tell, but they might have pretty much arbitrary breakage [12:04] when will the live CD be ready? [12:05] ready for what? [12:05] the one w/the hotplug fix [12:05] it'll be ready for a round of testing in ~15 minutes or so [12:05] I'm just waiting for casper 0.22 to appear on mirnyy [12:05] maybe you can instruct them to adjust apt preferences, and pin Ubuntu's warty above all else and apt-get dist-downgrade' [12:05] I think more like ~45 minutes [12:05] uploaded at 2236 UTC === ironwolf [~ironwolf@c-24-6-169-124.client.comcast.net] has joined #ubuntu-devel [12:05] elmo's processing probably didn't beat cron.daily [12:05] (build log dated 2236, anyway) [12:06] yes, irrelevant 'cos it's byhand [12:06] I meant the casper build log [12:06] oh [12:06] I would have expected it to appear at :03 just now [12:06] sorry [12:06] I can push cron.daily [12:07] well once this one finishes [12:07] that would be good, thanks [12:07] :-) [12:07] cron.continuously! [12:07] OK - guess I'll go fetch the daily iso now for install [12:07] azeem: if they can pin properly, they'd not need the warning ;) [12:08] Oh man [12:09] How did I never notice xgl before? [12:09] Hooray, my heinously awful archive-copier hack works === ironwolf [~ironwolf@c-24-6-169-124.client.comcast.net] has joined #ubuntu-devel [12:16] lifeless: pong [12:16] mdz: pong [12:16] lifeless: run xrestop to see what's leaking your memory -- probably a client app [12:16] daniels: Have you played with xglx/ [12:16] hi, are the popularity-contest results publicly available somewhere? popcon.ubuntulinux.org seems to be a bit outdated [12:16] ? [12:16] mdz/jkamion: should be on mirrors now [12:17] mjg59: nope [12:17] mjg59: not much good without mesa-solo, anyway [12:17] since you need to have an x server to run it on top of [12:18] Yeah, but presumably faster for composite development than trying to do it on i855? [12:18] probably, yeah [12:18] How's mesa-solo coming along? [12:18] really well, actually [12:18] Cool [12:18] and soon we'll be able to build mesa and x separately, by the looks [12:18] mesa-solo = X-less mesa? [12:18] So then we start concentrating on 3D support? [12:19] daniels: Oh, yeah, the crack all works now [12:19] mjg59: sweet [12:19] bob2: basically, yeah [12:19] mjg59: well, then our build system gets less crap, and we're all using a modular X \o/ [12:20] so, DRI -> mesa-solo -> xglx -> xlib ? [12:20] That way we only end up with one hardware layer to support [12:20] battstat-applet-2 [12:21] lifeless: Oh, yeah, that leaks badly in 2.9.4 [12:21] davyd put out a 2.9.4.1 gnome-applets tarball that fixes it [12:21] elmo: thanks, building [12:21] Shame about legacy hardware without 3D engines... === mjg59 looks at his old Thinkpad with a non-3D SMI Lynx [12:21] sladen: er [12:22] lifeless: prey tell me battstat doesn't flash annoyingly by default? [12:22] sladen: dunno what you mean [12:23] daniels: still using 8879 root 6 -10 869m 701m 5092 S 0.7 69.3 21:43.65 Xorg [12:23] after killing battstatapplet [12:23] lifeless: yes, the allocations don't die [12:23] it's a leak in the truest sense [12:24] oh fragr [12:24] elmo: thanks [12:25] Christ [12:25] Why does xserver's autogen.sh not work? [12:25] I hate autotools [12:27] Maybe it's GTK's fault [12:27] mjg59: hm, how does it fail? [12:27] mdz: done [12:27] aclocal: configure.ac: 40: macro `AM_PROG_LIBTOOL' not found in library [12:28] cjwatson@little:~/cdimage/www/full/daily-live/current$ isoinfo -R -i hoary-live-i386.iso -x /install/initrd.list | grep ^hotplug-udeb [12:28] hotplug-udeb 0.0.20040329-16ubuntu10 [12:28] mjg59: errr [12:29] argh, why is base-config not starting gdm automatically? [12:29] somebody sucks and I don't know who [12:30] Ah. S/AM/AC/ on AM_PROG_LIBTOOL seems to help. [12:30] Kamion: gtk [12:31] No, then autoconf complains [12:31] re-run libtoolize --force --copy? [12:32] Nope [12:34] mjg59: it probbly needs certain versions of automake and fails to check for them [12:35] mjg59: try AUTOMAKE=automake-1.9 and see if that works [12:36] (this seems to be a common theme atm, god knows who started it, i'm sure it used to be normal for autogens.sh's to check versions :/) [12:38] (revert that last s// 1st though ;) ) [12:39] oh, and ACLOCAL=aclocal-1.9 =) [12:39] obviously .. [12:41] mjg59: any luck? [12:41] robtaylor_: configure.ac:41: error: possibly undefined macro: AC_PROG_LIBTOOL [12:41] (aclocal-1.9 && autoconf) [12:42] mjg59: did you `libtoolize -fc' ? [12:42] mjg59: aclocal-1.9 && automake-1.9 && autoconf [12:43] crimsun: Yes [12:44] daniels: Same error [12:44] mjg59: what source tree? i'll take a look locally [12:45] robtaylor_: xserver from cvs.freedesktop.org [12:45] mjg59: where is your libtool.m4? [12:45] mjg59: in . ? [12:45] crimsun: Yes [12:45] mjg59@tyrosine:/tmp/xserver$ grep AC_PROG_LIBTOOL *.m4 [12:45] libtool.m4:AU_DEFUN([AC_PROG_LIBTOOL] , [LT_INIT] ) [12:45] mjg59: aclocal-1.9 -I . && automake-1.9 && autoconf [12:46] crimsun: Ok, that works [12:46] woo! [12:46] [I run across that all the time] [12:47] daniels: http://www.minion.de/files/1.0-6629/ [12:48] daniels: no idea what this is for, but some dude asked about it [12:48] mjg59: hmm, maybe a patch for autogen.sh is in order.. [12:48] jdub: christian worked for Nvidia; he provides those patches. [12:48] check automake/aclocal version, and add -I . to the autoreconf line, i guess [12:48] jdub: he goes by 'zander' here on freenode. [12:48] Haha [12:49] Patches that we aren't actually legally allowed to apply at the moment [12:49] oh [12:49] (but that's being sorted) [12:49] i've heard about thoe [12:49] Nvidia fuckup on their licensing text [12:49] mjg59: ah === robtaylor_ spanks nvidia one more time === lexhider [~lexhider@dip-220-235-87-122.vic.westnet.com.au] has joined #ubuntu-devel [12:50] crimsun: oh, that's zander? [12:50] mjg59: well, actually, we can now [12:50] mjg59: before I try, do you reckon your fixed IRQ stuff avoids the need for pci=routeirq ? === thully [~knoppix@wuser44-league.umnet.umich.edu] has joined #ubuntu-devel [12:50] daniels: yep, zander. [12:51] sladen: It should === mjg59 goes to get a bath [12:51] Hi - sorry, I haven't tested the install CD yet (I was testing the timezone behavior) - Imay not be able to do so for a while [12:52] Does anyone know if live cd 0119.4 fixes the hotplug issue? [12:52] which 'hotplug issue'? [12:52] daniels: so what are these patches? [12:52] jdub: various patches to the kernel code for the nvidia driver [12:52] issue w/wi-fi not loading properly [12:52] I was discussing it w/mdz [12:53] jdub: looks like most of them are bugfixes for certain situations === lamont takes his chalk and goes to the chalk board [12:53] I WILL NOT TRY TO DIVERT CONFFILES [12:53] *50 === Mithrandir pats lamont on the back [12:53] thully: there is a new live CD build up which should fix that [12:53] thully: I'm downloading it now to test === robtaylor_ twangs an elastic band at lamont === robtaylor_ looks innocent [12:54] OK - I'm downloading it now too, 0119.4, correct? [12:54] heh [12:54] /current/ [12:54] should be the same [12:54] OK [12:54] mdz: and that's what was wrong with the livecd, duh [12:54] I have to go now - have some things to do [12:54] bye [12:54] is the frambuffer thingie with nvidia fixed ? [12:54] uh, how do I convert a random iso-8895-nn file to xx from the command line? [12:55] elmo: iconv or recode [12:55] lamont: diverting conffiles, eh? [12:55] iconv says "conversion from iso-8895-1 not supported" [12:55] mdz: any other locales? Pre-generate some locales (en_{US,GB,ZA}.UTF-8) [12:55] ogra: according to daniels, yes [12:55] mdz: /etc/init.d/dbus-1 would be a conffile... [12:55] elmo: it's spelled 8859 [12:55] I'll be testing it shortly [12:55] ogra: which framebuffer thingy? the livecd stuff? (if so, should be) [12:55] lamont: let's start with those [12:55] it should definitely work on i386 [12:55] lamont: how big are they? [12:55] not 95. :P [12:55] daniels: amd64 :) [12:55] ogra: (and amd64) [12:56] mdz: the livecd needs some games, you know... :--) [12:56] daniels: and i have a nice bug for you, i'll file on the weekend if i have time to switch to nv to track it [12:56] lamont: are you ready to implement the live seed? [12:56] ogra: awesome [12:56] Mithrandir: fair point, well made [12:57] mdz: comment on #5495? [12:57] ISO8895 -- Shaped insulating refractory products -- Determination of cold crushing strength. I kinda understand iconv won't convert that. ;) [12:57] it would be cool if it did [12:58] mdz: we could do that. [12:58] just need an archive to check it out of, and we're golden [12:58] elmo: btw, if its text you can open it in gedit, it offers to select the locale on save [12:58] Kamion: no higher than 'enhancement', ok? [12:58] daniels: mm? [12:58] you get to hit scott really hard for that one [12:58] er, ww [12:59] Kamion: the bug on iconv :) [12:59] oh :) [12:59] mdz: the nice part is that the live seed just feeds straight into apt-get... no dependency deals or anything.. [01:00] for that matter -it should just go into ubuntu-meta as ubuntu-livecd, which depends: ubuntu-desktop [01:00] lamont: let's have a round of ubuntu-devel about the live seed and what should go in it [01:00] the last time, I remember there was some dissent [01:00] (live should be the same as a desktop install, etc.) === kent [~kent@c83-249-61-131.bredband.comhem.se] has joined #ubuntu-devel [01:01] I think I need to tweak germinate a bit to make the live seed really work properly, too [01:01] hm, or indeed work at all [01:01] Kamion: it wouldn't need to be germinated [01:01] for seedname in ("base", "desktop", "ship", "installer", "supported"): [01:01] it should be, IMO [01:01] mdz: only silly pant stealers dissented with liveseed [01:02] Kamion: ok. [01:02] I don't think there should be such a thing as a seed that doesn't get fed to germinate [01:02] because if nothing else you want to make sure the stuff stays in main [01:02] hrm... yeah - liveseed might be broader than supported, eh? [01:02] Kamion: games in main??? the heresy [01:02] live < supported, definitely [01:02] but there might be stuff in the text of the live seed not mentioned in the text of the supported seed [01:03] right [01:03] I think we want base < desktop < ship < supported; base < desktop < live < supported; casper < (?) installer < supported [01:03] it ought to be germinated, so we get a consistent set of stuff on the CD [01:03] mdz: should be done staring at the changes and pushing a round of updates to all the buildds within the hour. If you want a dbus-1 happy livecd before that, holler now and I'll go fix it in place and run it. [01:03] maybe just casper < supported and installer < supportd [01:03] ed [01:04] mdz: bz email address changed (finally) [01:05] lamont: livecd was dbus-1 unhappy? [01:07] jdub: when you divert /etc/init.d/dbus-1 and then do a non-interactive install, you get /etc/init.d/dbus-1.dpkg-new, which is just not what you want later. [01:07] daniels: thanks [01:08] lamont: I'm about 1/4 of the way through downloading a full set [01:08] lamont: oh. why does it need to be diverted? [01:08] jdub: it had a bug [01:08] (past tense) [01:09] Kamion: is there an easy way to point cron.daily-live at an old version of the cloop images [01:09] oh [01:09] assuming lamont keeps them around? [01:09] jdub: and the bug was fixed [01:09] mdz: never been deleted :-) [01:09] although the location did change [01:09] that way, lamont can go ahead and build new images to test the new stuff [01:09] but if there is a casper problem, I can build new live CDs without downloading another 1.5 gigs [01:11] mdz: the other option is that if you need to build a new one, you can holler and I'll move the 'current' links back [01:11] of course, that may become obsolete soon with the rsyncable stuff [01:11] which would ROCK [01:11] mdz: yes, poke /srv/cdimage.no-name-yet.com/debian-cd/tools/add_live_filesystem [01:12] I'll let you do that though, I'm off to bed [01:12] night [01:13] amd64 live CD on a bandwidth-limited download overnight [01:18] has anyone seen this? [01:18] W: GPG error: http://localhost hoary Release: Unknown error executing gpgv [01:22] haggai: do you have gnupg installed? [01:22] haggai: you building in a chroot and running apt in the chroot? [01:22] or outside the chroot without gnupg installed. [01:23] haggai: that means that waitpid() returned something unexpected after trying to fork-and-exec gpgv [01:23] I have gnupg installed, and am running in the chroot [01:24] -rw-r--r-- 1 buildd root 549947608 Jan 19 23:27 livecd-20050119.5-ia64.cloop-1024:65536 [01:24] woot! [01:24] lamont: I've seen the same thing in 2 seperate chroots [01:25] lamont: and, I'm trying to get sbuild working. Have you had the situation where it fails to install any build deps? [01:27] haggai: what's the error message? [01:27] azeem: I don't get any error message, just that W: line, and another one suggesting an apt-get update [01:29] I think you need to run apt-get update in the chroot once to actually have the lists around, did you do that? [01:29] yes I did [01:29] same error === JanC [~JanC@dD5764FA0.access.telenet.be] has joined #ubuntu-devel === usual [~colin@alb-69-202-36-213.nycap.rr.com] has joined #ubuntu-devel [01:30] are you running sbuild from ~/build? do you have links to your chroot in there? [01:31] daniels: for some reason, when I reboot from GNOME on the live CD, I don't see the shutdown messages until the very end [01:32] daniels: it seems unlikely to be livecd-specific [01:32] I haven't tested on a standard install [01:32] I get vc1 for a moment, and then a blank screen [01:33] mdz: ack, I've seen that also; depends how I shut down [01:33] mdz: if I just click through, it's fine [01:33] mdz: it's likely to be due to our start-gdm-earlier stuff [01:34] mdz: which has already had weird interactions with XKeepsCrashing if it fires at the wrong time [01:34] daniels: yeah, I've seen that too [01:34] daniels: I end up at the XKeepsCrashing dialog with my keyboard in raw mode === daniels shrugs ... [01:34] I just pound keys until I get out by sheer luck === thully [~ubuntu@wuser16-league.umnet.umich.edu] has joined #ubuntu-devel [01:34] The hotplug issue is fixed [01:35] daniels: so how can an X client force the server to leak. Surely thats bad ? [01:35] lifeless: no, not at all [01:35] This actually should fix another issue I had - in the normal installer, my wi-fi wasn't available as a network interface to configure [01:35] lifeless: you ask X to allocate a pixmap for you, then you never free it [01:35] lifeless: the whole server/client thing means that pixmaps, et al, need to be stored server-side as well [01:35] hda: dma_intr: error=0x01 { AddrMarkNotFound }, LBAsect=135032251, sector=135032072 [01:35] that does not look good at all [01:35] daniels: but when the client dies, surely you can walk back to the clients resources and free them. [01:36] lifeless: not as such, because you can share them aiui [01:36] lifeless: as they're just identified by an integer [01:36] azeem: I'm running from ~build and there is a link to the chroot in there, and it finds it too [01:36] mmm, you could still know that the client that allocated it has gone, and I presume there are protocol to enable sharing (else it would be a security hole), so you can count as they get shared. [01:37] lifeless: how would it be a security hole? [01:37] lifeless: if you have access to the display, you own it [01:37] lifeless: (just hide all the other windows and pop up one of your own that looks like Firefox doing internet banking) [01:37] (or, more subtly, just grab all input) [01:37] client A has pixmap with the map for foo on it, client B is rogue and copies that pixmap. [01:38] if client B is rogue, you have already lost in so many ways it's irrelevant [01:38] daniels: eek [01:38] client B just pops up a window that's a direct screen-scrape of client A's [01:38] or looks like a password dialog or something [01:39] again: if an untrusted client has access, you have already lost [01:40] azeem: hmm, not sure what it was that I changed but it's giving me a proper error now [01:40] thully: thanks for testing [01:40] there was a pretty half-arsed XSECURITY extension to try and control inter-client exchanges, but it was badly thought out, badly implemented, and even sun has long considered it a very poor idea [01:40] the selinux guys are doing interesting things on that front, though [01:40] thully: that was fast; I thought you were on a modem :-) [01:40] daniels: so if you have client that allocates pixmaps (i.e. everything nowadays, no?) and it crashes regularly, that causes an irrecoverable memleak in the server? [01:40] haggai: heh, can you make something out of it? [01:41] robtaylor_: quite possibly [01:41] mdz - I downloaded it at another PC (not at home) where there's high speed net [01:41] robtaylor_: iirc gtk's segfault handler will clean up quite well, for one [01:41] I got the hoary install iso too [01:41] will test that next [01:41] so, evolution borks regularly for me... ah gtk ok. [01:41] daniels: ah thats good to know ;0 [01:41] azeem: yeah, it's a build-dep not in main [01:42] mdz: This should take care of an issue with the installer not finding my wi-fi (in install mode) as well... [01:42] daniels: surely theres a way to refcoun them, as long as they're not MIT-SHM? [01:42] hell you can even refcount then, but that gets icky [01:42] thully: the install iso will not have the wireless firmware fix yet [01:42] azeem: I wonder if apt-proxy and authenticated apt have problems; I switched sources.list [01:42] robtaylor_: probably, but you could also just pass the reference over to someone else out-of-band and they could use it later :) [01:42] the next daily build will [01:42] we did a new live CD build manually [01:43] mdz: is there a config file option to disable the authentication stuff, or is it only a command line switch? [01:43] haggai: both [01:43] mdz: apt, I mean [01:43] daniels: gosh, didnt think that was possible... urrgle [01:43] mdz: ah, an undocumented one then :) [01:43] haggai: man apt-get [01:43] haggai: bob2 had problems with authenticated apt and sbuild as well [01:44] I have to go now... I'll check out install iso later [01:44] haggai: by design, all apt command-line flags have a corresponding configuration item [01:44] haggai: which version of sbuild are you using? [01:44] very weird ones [01:44] mdz: ah, I see, the config option is with the command switch [01:44] azeem: from unstable [01:44] robtaylor_: aiui, this is the case, but client-side x library stuff isn't my forte [01:44] azeem: I patched in hoary as a distro [01:44] mdz: thx [01:44] daniels: is it anyones? ;) [01:44] (rhetorical..) [01:45] robtaylor_: heh ... maybe owen taylor's [01:45] haggai: if you tell me how to disable authenticated apt, I'll add that to the options sbuild passes to apt [01:45] daniels: theres some scary s**t in there, i remeber that much =) [01:46] azeem: --allow-unauthenticated [01:46] azeem: I don't know if it's necessary, though, assuming it all works [01:46] I guess that will break on apt from unstable [01:47] azeem: I think it might just be a problem with my local setup and apt-proxy. Still, you do have to manually install gnupg in the sbuild chroot before it works [01:47] azeem: true [01:47] azeem: yah, is there a need for it to limit the valid distros? [01:47] robtaylor_: heh, yah [01:47] robtaylor_: 'swhy I steer clear [01:47] bob2: no, and I proposed to just check whether there is ~/build/chroot-$dist [01:48] hrmpf. somebody upload something, k? [01:48] daniels: heh, sounds a sane plan to me ;) [01:48] azeem: oh, that would be awesome [01:48] daniels: keithp probably once understood it ;) [01:49] bob2: I'm gonna send of a 'ok to commit mail?' to nudge frankie [01:49] off, even [01:49] woo [01:50] of course, now I have to patch that check out locally here again, or finally figure out how to use sbuild with a chroot on GNU/Hurd [01:51] amd64 live CD looks good [01:51] daniels: fb fix confirmed [01:51] mdz: ok to build a new livecd fs? [01:52] lamont: yes, thanks [01:52] mdz: phat [01:52] daniels: whatever that shutdown bug is, it's definitely worse now [01:52] I never saw it before today [01:53] our sbuild doesn't limit the valid distros [01:54] azeem: bob2: ^^^ [01:54] mdz: shutting downd the hd and repowering it before complete shutdown? [01:54] lamont: the one you're running on buildds or the one in hoary? [01:54] I also got a segfault there once in a while.. [01:54] sivang: blank screen instead of shutdown messages [01:54] actually, that reminds me... I need to send neruo my patches for separating the distros in to their own directories [01:54] buidlds [01:54] the one in hoary is the same nobody-uses-it version as debians [01:54] mdz: eh, havn't tested reboot with latest update then. [01:54] mdz: it's always been present for me [01:55] mdz: except it's OK when I shut down from within GNOME [01:57] lamont: what exactly do you mean, seperating the distros in to their own directories? [01:58] build-$dist/chroot-$dist is the build directory [01:58] lets me build different $dists in the same buildd, with different chroots [01:58] or (what we'll first use it for): rebuild everything from scratch, stealing idle time from the buildd, to look for ftbfs's [01:58] ah, ok [01:59] lamont: very nice [01:59] the non-trivial part is that some things need to stay in build/, while other stuff moves to build-$dist. [01:59] yeah [01:59] and I still have a few loose ends [01:59] buildd-watcher and I are not particularly on good terms right now. [02:01] $conf::separate_distributions [02:02] happy happy joy joy [02:04] daniels: I'm talking about within GNOME [02:04] daniels: that used to work fine [02:04] daniels: and now it doesn't [02:04] lamont: ran and stimpy? [02:04] mdz: unreproducible. i've gotta run for an hour (new phone!), so i can take a look after that [02:09] sivang: that could be where I got that from [02:09] it's been a while [02:09] mdz: the build logs already have all of the entry points linked from DeveloperResources [02:10] lamont: ok, thanks [02:10] daniels: do you know about this crazy capslock per-top-level window foo? [02:11] lamont: how did you do the ccaching on the buildds? [02:11] /home/buildd/ccache is bind mounted as /home/buildd/.ccache in each of the chroots on the machine [02:11] then we shim gcc into 'ccache gcc' everywhere [02:12] how do you do that? changing the default path to include the trampolines? [02:13] lamont: in /usr/lib/ccache? [02:13] gcc-3.3 is dpkg-diverted === lamont customizes with a crowbar, you see.. [02:13] heh [02:13] did you not know about /usr/lib/ccache, or did you find it did not always work? [02:14] haha [02:14] ccache actually came second.. tweaking gcc options came first [02:14] ah :) [02:15] which reminds me - I need to hijack g{cc,++}-3.4 as well. [02:15]