=== ikepanhc1 is now known as ikepanhc === hzhang__ is now known as panda|phenom [07:27] apw: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/407824 , did you fix get reverted again? :) [07:27] Malone bug 407824 in linux "BOTH Network controller: Intel Corporation Wireless WiFi Link 5100 AND Ethernet controller: Marvell Technology Group Ltd. 88E8040T PCI-E Fast Ethernet Controller (rev 12) FAIL TO LOAD!" [Medium,Fix committed] [07:27] your* [09:41] I followed the instructions on https://wiki.ubuntu.com/KernelTeam/GitKernelBuild and ended up with a 310MB kernel image deb [09:41] is this normal? [09:44] I built v2.6.30 and used .config from the ubuntu mainline build of 2.6.30 [09:48] akheron, The effect sounds rather like using a config which has debugging enabled. But if you took the git sources (or apt-get source) and used debuild there would be no need to get the configs from the mainline builds [09:51] I cloned sources from git.kernel.org, git checkout v2.6.30, cp /boot/config-2.6.30-020630-generic, make oldconfig, make-kpkg [09:51] akheron, Our configs normally have the debugging enabled but the build system strips the code before packaging [09:51] *sigh* [09:51] ah, ok [09:51] I'm debugging a problem where my system doesn't boot [09:51] I wish the make-kpkg would get eradicated from peoples memory [09:51] this appears somewhere between 2.6.30 and 2.6.31 [09:52] (we should do that from the wiki) [09:52] so I'd like to bisect the kernel between these versions [09:52] how to do it properly? :) [09:53] If you have the git sources (which contains the configs in debian/configs) a simple "fakeroot debian/rules binary-generic" would do the generic kernel package [09:54] does the ubuntu kernel repo follow linus' tree so that I can bisect? [09:55] and another question: can I build the kernel without cleaning it first? The build takes like an hour and I expect that I'll need to build tens of images [09:56] akheron, Before releasing its a rebase tree, so bisecting gets painful. Its following Linus tree, but maybe a starting point would be to try some of the mainline builds to narrow the thing [09:56] from what repo are the mainline kernels built? [09:57] Those are Linus tree sources at -rc points [09:57] with the ubuntu config [09:57] yes, but where do I get debian/configs? [09:58] for them [09:59] the mainline builds are prebuild debs, so for them you would need no configs seperately [09:59] smb: +1 about make-kpkg. We should have a bot that tells people that make-kpkg isn't supported [10:00] I am helping another guy on #ubuntu-mobile with the same problem [10:00] For your own builds you might be able to speed up/shortcut by removing the debian/stamps/build (would need to check the correct name) file [10:01] and then call the build again [10:01] smb: I've tried mainline builds alread [10:01] or use the ccache package which helps to speed up rebuilds [10:01] ah ccache is a good idea [10:01] mainline 2.6.30 works, mainline 2.6.31 doesn't [10:01] so I'd like to build something in between [10:02] Hm, there should be 2.6.31-rc? ones as well, right? [10:02] yes [10:02] .31-rc1 doesn't work [10:02] .30.9 does work [10:03] so 2.6.31-rc2 is before 2.6.31 [10:03] yes [10:03] but as -rc1 doesn't work I expect that -rc2 doesn't work either :) [10:03] Correct :) [10:04] so there are no more mainline builds left to narrow it down [10:04] So you would need to check "only" whats in the first rc [10:04] And the stable kernels seem to be ok too [10:04] ?? [10:04] Right [10:04] .30 works, .31 and .32-rc* don't [10:05] so there's a breakage somewhere between .30 and .31-rc1 which haven't been fixed [10:05] Wait 2.6.32-rc? [10:05] I tested mainline 2.6.32 builds, too, to see if it's already fixed upstream [10:05] ah ok [10:06] with "somewhere between" I mean that there's (hopefully) one commit in linus' tree before which my system boots and after which it doesn't :) [10:06] Ok, yep. So now you really are down to bisect the changes between .30 and 31-rc1 [10:06] yes [10:08] and building a kernel with ubuntu configs is my problem [10:09] right, so I would suggest to take a upstream tree, then copy into that the debian/* and ubuntu/* dirs from the 2.6.31 ubuntu repo and try to bisect that [10:09] altough ubuntu configs may not be needed [10:10] ok, I'll try that [10:10] thanks [10:10] If that gives problems, come back and we try to figure out what goes wrong [11:10] <^arky^> Hi ogasawara can please look at bug 464442 , it caused screen reader users and voice chat users lot of pain [11:10] Malone bug 464442 in alsa-driver "alsa-util.c: snd_pcm_delay() returned a value that is exceptionally large: -4496252 bytes (-25488 ms)." [Undecided,New] https://launchpad.net/bugs/464442 [11:31] is the kernel version for lucid decided? if so what is it? [11:31] greetings === apw changed the topic of #ubuntu-kernel to: "Home: https://wiki.ubuntu.com/KernelTeam/ || Lucid Kernel Version: 2.6.32 || Ubuntu Kernel Team Meeting - Next Meeting after UDS - 17:00 UTC" [11:31] mpt_, hiya [11:32] tgpraveen1, i've updated the topic [11:32] apw: thanks [11:32] I'm trying to specify the details of a power menu that will include all chargeable items that the kernel knows about === apw changed the topic of #ubuntu-kernel to: Home: https://wiki.ubuntu.com/KernelTeam/ || Lucid Kernel Version: 2.6.32 || Ubuntu Kernel Team Meeting - Next Meeting after UDS - 17:00 UTC [11:32] https://wiki.ubuntu.com/SystemIndicators?action=AttachFile&do=get&target=power-menu.jpg [11:32] So what I'd like to know is: When a notebook has multiple battery slots, what kind of identifier does the kernel have for those slots? [11:32] What does the ID look like? [11:33] i thought that they were like BAT0 BAT1 [11:34] ah, I see [11:34] e.g. in http://lkml.indiana.edu/hypermail/linux/kernel/0401.0/1155.html [11:34] seems they also have serial numbers where available [11:35] apw@dm$ cat /sys/class/power_supply/BAT0/serial_number [11:35] 20993 [11:35] Shame it doesn't say something like "front slot" and "back slot" :-) -- then it might be easier for people wanting to know which one to replace [11:35] but I guess even the kernel doesn't know that [11:37] no you seem to even be able to work out its Li-ion etc [11:37] but not where it is [11:38] battery type: JUNGLE BEAST [11:38] ok, thanks apw [11:40] mpt_, i'll try and not guess what the machine is called if it has a jungle beast battery [11:41] i note that the mains connector is also mentioned in /sys/class/power_supply ... nice [11:43] apw: the power_supply class has existed since .31 I believe [11:43] but battery hotplugging doesn't yet work. damn. [11:44] amitk, interesting thanks ... it looks like a more sensible way of naming things [11:44] though what ADP1 means ... [11:44] I just tried to plugin by second battery to see if there is any way to differentiate the two, so anwer mpt_'s question [11:45] you presumably get BAT0 and BAT1 and thats it? [11:46] when I _boot_ with both batteries plugged in, yes. [11:46] but I usually unplug the second one when at home, and hotplugging doesnt work [11:46] oh poop thats useless === mpt_ is now known as mpt [12:50] smb: the build system seems to install modules in /debian/linux-image-2.6.31-16-07a8c6-generic/lib/modules/2.6.31-16-generic/ [12:50] err, debian/linux-image-2.6.31-16-07a8c6-generic/lib/modules/2.6.31-16-generic/ [12:51] altough they should go to debian/linux-image-2.6.31-16-07a8c6-generic/lib/modules/2.6.31-16-07a8c6-generic/ [12:51] my ../.karmic-env has: [12:51] LOCAL_ENV_CC="ccache gcc" [12:51] AUTOBUILD=1 [12:51] CONCURRENCY_LEVEL=3 [12:53] hmm, actually this might be because AUTOBUILD was not 1 from the beginning... [12:54] https://help.ubuntu.com/community/Kernel/Compile is outdated as AUTOBUILD et al are not accepted from command line [12:58] akheron, For the wrong location, there is one option that causes that. Let me check, I always forget the name. The rest of the doc, yes actually the whole thing should be looked over I guess [13:05] akheron, Hm, it usually is caused by having LOCALVERSION_AUTO being set, but that should not be with the ubuntu configs... [13:05] Lemme try to do what I suggested locally and see what happens [13:06] I guess it's because I set AUTOBUILD=1 on the command line, let it build, and saw some checks fail [13:07] then found out that AUTOBUILD=1 has to go to ../.karmic-env and just run debian/rules build-generic again [13:13] akheron, there should be no need to set anything for your need [13:13] oh? [13:13] concurrency normally is #cpus and ccache might be used by having /usr/lib/ccache in the path [13:14] at least it fails because of missing modules without AUTOBUILD=1 [13:15] hm, strange. give me a few mins to see what happens [13:16] you must have quite a farm of distcc hosts to compile it in "a few mins" :) [13:47] read 2664 modules : new(0) missing(115) [13:49] akheron, Oh so you need skipabi, not autobuild [13:50] skipabi=true ? [13:50] or was that skipmodules...? too many things [13:50] :) [13:52] skipmodule [13:52] probably both "skipabi=true skipmodule=true" before the binary-generic [13:52] so "fakeroot debian/rules skipabi=true skipmodule=true binary-generic" [13:53] AUTOBUILD=1 sets those and additionally skipdbg=true and abi_suffix= > [13:53] ah [13:53] dpkg-gencontrol: error: package linux-image-2.6.31-16-07a8c6-generic not in control info [13:54] a bad property of abi_suffix :F [13:54] akheron, Doh, right so thats where the suffix comes from [13:55] * smb wonders who used AUTOBUILD last [13:56] akheron, Sounds like you will need to do a clean at least once to get the control files right "fakeroot debian/rules clean" not the "make clean" [13:56] I did that [13:59] akheron, Have you gotten rid of the AUTOBUILD=1 in the env file? [13:59] That will cause the strange suffix to be attached to the version number and the build to fail [13:59] yes [14:00] err, no I haven't [14:00] but yes, I agree that I probably should delete it [14:00] the only problem is that then all the debs I create have the same name [14:01] the suffix is derived from git commit id and this is a good thing [14:01] akheron, for the test you can edit the first line in debian.master/changelog and append something like +bisect123 [14:01] but that would be manually [14:01] that's true [14:01] It doesn't matter [14:02] I'll try that [14:02] Just to be sure, the same would need to go to debian/changelog (that would normally updated with each clean). So if you do cleans between builds that is not needed [14:07] is there a way to do it without cleans between builds? [14:08] akheron, With somewhat more pain. Like said you need to update both changelogs in step and then remove the stamp file [14:09] debian/stamps/stamp-build-generic for the generic build [14:10] ok [14:10] and then hope that all the kernel isn't built from scratch [14:11] Right, but that depends on how much changes are between the steps. If an important include file was changed you still might end up doing a lot [14:14] yes [14:14] I'll see that after I manage to complete the first build [14:21] smb: you said earlier you wish make-kpkg would get eradicated from peoples memory, but as far as I can see, if someone does not get the source from your git there is no other way to build kernel packages. Or is there another package I'm missing that would give us the ubuntu build system? [14:22] mozmck, The ubuntu build system for the kernel is just a chroot of the series (which is described in some related wiki docs) and then either clone the git repo and check out a version tag or do an "apt-get source linux-image-xxx-generic" === BenC1 is now known as BenC [14:24] mozmck: I'm currently building from linus' kernel tree with the ubuntu build system [14:24] so an apt-get source will get it? I tried that the other day but something was wrong with my setup or something. Thanks for the info [14:24] I only use the build system from ubuntu tree [14:24] mozmck: you could also use the source packages for the mainline builds [14:25] mozmck: that will give your the ubuntu build system and kernel configuration w/o any of the ubuntu patches to linus' tree [14:25] I got the linux-source package from the repository and built from that using make-kpkg [14:26] amitk, I would be a bit careful about that. I have seen the kernel-source packages sometimes a bit useless as the miss the configs and some stuff. It could be apw fixed that for the mainline builds but I am not sure [14:26] * amitk sobs about make-kpkg [14:26] I don't think the souce from the repository had the debian directory at all. [14:26] smb: that is just a bug that needs to be fixed if there is enough demand for it. And apw has fixed it AFAIK [14:26] The problem with make-kpkg is that nobody from us uses it and if it works its rather coincidental than expected [14:27] make-kpkg is a horrid bit of overkill [14:27] that is the only way we can tell people to compile a mainline tree w/o using untested tools [14:27] hiya BenC [14:27] hey amir [14:27] *amit [14:28] The kernel I built works fine on some computers but won't boot on one that has intel 82845G graphics. [14:28] BenC, Hi2 [14:28] I think debian talked about retiring make-kpkg at plumbers [14:28] hey smb [14:28] weird thing is that it booted fine from the livecd I made with that same kernel, but after installing it won't boot. [14:28] isn't there a "make deb" in the kernel tree? [14:29] BenC: it was broken and someone on the debian kernel team pledged to fix it IIRC [14:30] Have never tried that. And even if, the is too often confusion with the kernel-source package (which is not very usefull imo) and the sources you get with git or apt-get source [14:30] I used the source and config from 2.6.31-15.50-generic and added an RTAI patch and changed a few config settings. The generic kernel will boot if I tell it "nomodesetting" [14:33] I'll try building again with the ubuntu build. Most of the docs I see say to use make-kpkg and to use the kernel-source packages. [14:35] mozmck, have you read https://wiki.ubuntu.com/KernelTeam/KernelMaintenance ? [14:36] Well yes, there are still too many docs referring to that. [14:36] BenC: dude, good to see you online. [14:36] rtg: hey, thanks [14:37] hmm, I don't know... no I don't think so. Thanks! [14:38] rtg, It feels we should spent a bit of our plenty time to update some of the other building bits [14:39] smb, I'm not really interested in supporting non-Ubuntu build methods. [14:39] rtg, I don't want to support then, more saying they are not [14:40] smb, I'm not fully comprehending your intent. [14:41] rtg, My intend would be to find pages claiming that make-kpkg works and clearly state it does not [14:42] smb, in our wiki? or in someone elses? [14:42] At least when it seems to come from our own wiki pages [14:42] ubuntu wiki and one link previously told seems to be even in the KernelTeam space [14:42] sounds like a damn good idea to me. when will you have it done? [14:44] rtg, You know how good we are at finding documents... ;-) [14:45] smb, maybe we can get pgraner to do it in his copious spare time. [14:46] * BenC wonders why kernel-package is even in main [14:46] rtg, Heh, he likely just waits to be asked. [14:46] BenC: I think it should have been dropped after dapper. [14:47] smb, I only find 19 instances of 'make-kpkg' in wiki.ubuntu.com [14:47] drbd8-source recommends it. [14:47] only one in the kernel team heirarchy [14:48] rtg: go to google and try: site:wiki.ubuntu.com make-kpkg [14:48] rtg: google is far better at w.u.c than the wiki search engine [14:48] <_maks> mozmck: make deb-pkg [14:48] pgraner, again, only 19 [14:48] rtg: wow impressive they match [14:49] <_maks> you need >= 2.6.31 to have most of its bugs fixed mozmck, but it is not yet feature complet (saying it yet misses linux-libc-dev, source and headers packages :) [14:49] pgraner, but there are 2 references in _our_ pages. gotta fix that. [14:57] rtg: will be up on Wed, he called and asked if I could wait, the installer had a family issue come up :-/ [14:57] rtg: I have to run into town today and get a USB serial cable so I can get to the BIOS to redirect the serial console back to the main on the build box [14:58] pgraner, -EWRONGCHAN [14:58] rtg: All my cables are in the travel box enroute to here via FedEx Ground [15:02] rtg: could you snag my ssh pub key from launchpad and have an account created on zinc, please? [15:03] BenC: it may still be there. lemme check. [15:03] rtg: it's not, because mine was an ldap synced canonical account [15:03] BenC: right, I thought of taht as soon as I said it. [15:04] BenC: rtg@zinc:/home/bcollins. this'll take a bit. [15:04] cloning git over http should be a sign of the apocalypse [15:05] BenC: use git://kernel.ubuntu.com/ubuntu [15:05] BenC: you have a firewall? :-p [15:05] ah, forgot git:// was setup [15:05] amitk: no, it's just plain evil :) [15:05] BenC: still want a zinc account? [15:06] rtg: yes, please [15:50] for some reason, I get empty LINUX_VERSION_CODE [15:50] in include/linux/version.h [15:50] smb: ideas? [15:52] no immediate ones. and as I got a bit sidetracked I have not started to try the same in parallel. Try to do now [15:54] can someone explain the debian.master thing in karmic kernel source? [15:54] I managed to get v2.6.30 to build, but nothing after that [15:54] it always fails in a file that tries to use LINUX_VERSION_CODE [15:55] ## [15:55] ## Ubuntu Kernel Team Meeting - Today @ 17:00 UTC - #ubuntu-meeting [15:55] ## agenda: https://wiki.ubuntu.com/KernelTeam/Meeting [15:55] ## [15:55] BenC, debian.master is the common things that go to debian. We used that to make multiple topic trees that rebase simpler [15:55] BenC, More ore less its debian.master you look at. When you do a clean, files get generated in debian [15:56] BenC, So this gets rid of the generated files in git as well. But you always need a clean before build now [15:56] so debian/ is completely generated off of debian.master? [15:57] BenC, Mostly. A small debian/rules was needed, but most of the rest, yes [15:57] gnarl: always need a clean? [15:58] akheron, One time after getting the sources. Oh darn, did I tell you to copy debian.master (completely) as well? [15:58] don't know if it was you but someone did [15:59] akheron, The one time clean generates some files (as debian/changelog) and abi files (when needed) [15:59] akheron, me or smb which is mostly the same :) [15:59] heh [16:00] still can't get a value for LINUX_VERSION_CODE [16:01] I'm basically using debian/, debian.master/ and ubuntu/ from ubuntu-karmic.git [16:01] on top of linus' tree and have checked out an arbitrary commit [16:03] I propose 50€ to fix a bug with suspend/resume on my macbook (issue #400413) [16:03] http://www.cofundos.org/project.php?id=178 [16:07] akheron, Just started of the same. Just with v2.6.31 instead of a random point. Then changed the version number in debian.master/changelog and did "fakeroot debian/rules clean" followed by "fakeroot debian/rules binary-generic", got one config qustion asked and now it builds [16:08] I did this first for v2.6.30, and it built ok [16:08] then after checking out v2.6.31-rc2 it didn't build anymore [16:09] nor did an arbitrary commit [16:10] akheron, as bad in time it might be, have you tried a "fakeroot debian/rules clean" in between? [16:10] yes [16:10] now I even deleted debian/, debian.master/ and ubuntu/ and copied them all over again [16:10] but still I get an empty LINUX_VERSION_CODE in include/linux/version.h [16:11] $ cat debian/build/build-generic/include/linux/version.h [16:11] #define LINUX_VERSION_CODE [16:11] #define KERNEL_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c)) [16:11] maybe give me the sha of the commit in between you tried. When my build here finished I'll try that [16:11] v2.6.31-rc2 [16:11] and now I'm at d27eadc7612404b06f99888c02726ab7d5036e0f [16:11] but this is a build system issue, as the #define has been there forever [16:12] or is generated there [16:12] ok, give me a bit. Some file was, but that define looks like one around for quite a while [16:14] Hm, but seems to get generated at some point [16:16] could the changelog mangling be a reason? [16:16] So it is rather an issue of how doing the build. The whole of debian/build/build-generic is generated as an out of tree build source [16:16] if I don't touch debian/changelog it works [16:17] how do you change it (paste before and after of the first line) [16:17] hmm no [16:18] and debian.master/changelog is the master file (when trying to avoid clean you would need to change that and copy it to debian/changelog) [16:18] now it generates ok [16:18] I'm confused [16:18] got no time to debug this anymore today, I'll be back tomorrow [16:52] is there a reason that kernel.shmmax changed from 256MiB to 32MiB between 2.6.31-14 and -15 ? [16:53] I can't find it in any changelogs [17:19] SUB_INST checking: sound/drivers/pcsp/snd-pcsp.ko [17:19] SUB_INST installing: /lib/modules/2.6.32-5-generic-pae/kernel/sound/drivers/pcsp/snd-pcsp.ko [17:19] SUB_INST installing: /lib/modules/2.6.32-5-generic-pae//lib/modules/2.6.32-5-generic-pae/kernel/sound/core/snd-pcm.ko [17:19] install: cannot stat `debian/linux-image-2.6.32-5-generic-pae//lib/modules/2.6.32-5-generic-pae//lib/modules/2.6.32-5-generic-pae/kernel/sound/core/snd-pcm.ko': No such file or directory [17:19] why is that doubling the path? [17:19] it worked with generic [17:55] ## [17:55] ## Ubuntu Kernel Team Meeting - in 5 minutes - #ubuntu-meeting [17:55] ## === bjf__ is now known as bjf === bjf is now known as bjf-afk