Kanohi, is the linux-meta 7 package already done02:46
fairuzHI, morning07:40
apwloic yes, we should be using orig from now, i didn't bother for the last upload for expediency08:24
=== jjohansen is now known as jj-afk
fairuzHi, I do #include <plat/omap44xx.h> in kernel space and it recognizes it, but it did't recognizes it in user space. Is it normal ? 08:57
apwfairuz, yep, those are kernel headers, only a select few are exported to userspace in the linux-libc-dev package and they are primarily for libc use09:06
fairuzapw: ok thanks09:06
ppisatiericm|ubuntu: ping09:36
ericm|ubuntuppisati, pong09:45
ericm|ubuntuppisati, what's up dude?09:45
ppisatiericm|ubuntu: i can't get any audio out of mvl-dove, did you?09:46
ericm|ubuntuppisati, I used to - but not really sure if that's still the case09:46
ppisatiericm|ubuntu: uhmmm...09:46
ppisatiericm|ubuntu: i mean, a vanilla 10.04 installation was enopugh to get sound? or do i have to tweak anything?09:47
ericm|ubuntuppisati, I don't think so09:47
ericm|ubuntuppisati, wait09:47
ericm|ubuntuppisati, I seem to remember a alsamixer thing to get it work but I'm not really sure09:48
ericm|ubuntuppisati, did you check the gnome-volume-control?09:48
ppisatiericm|ubuntu: volemu control seems ok09:49
ericm|ubuntuppisati, hrm.... did you do a search on LP?09:53
ppisatiericm|ubuntu: i'm hoovering all the mvl-dove bugs09:53
ppisatiericm|ubuntu: there are 2 audio-related bugs for mvl-dove09:53
ppisatiericm|ubuntu: the first is about the headphone not recognized (while in a LSP update there was specific mention about it)09:53
ppisatiericm|ubuntu: the second is about an alsa config09:54
ppisatiericm|ubuntu: headphones: https://bugs.launchpad.net/ubuntu/+source/linux-mvl-dove/+bug/55124909:54
ubot2Launchpad bug 551249 in linux-mvl-dove "Audio driver fails to detect Headphone insertion/removal." [Wishlist,Confirmed]09:54
ppisatiericm|ubuntu: alsa config: https://bugs.launchpad.net/ubuntu/+source/linux-mvl-dove/+bug/65128109:55
ubot2Launchpad bug 651281 in linux-mvl-dove "Need alsa configuration files for Dove A0 SOC platform" [Undecided,New]09:55
ericm|ubuntuppisati, the 2nd one is suspicious09:55
ppisatithe second one, looks like you need a specific config file to get any audio out of this board09:55
ericm|ubuntuppisati, I didn't really ever check the audio difference between X0 and A009:55
ppisatibut since i _guess_ you (i mean the mvl-dove team) tested the audio09:55
ppisatiso you did all the stuff on X0?09:56
ppisatiand i got an A0, right?09:56
Kanohi, how to disable the modules check globally10:09
apwKano, add the ignore.modules file10:43
apwto the abi for the flavour, see the check-modules script10:44
Kanoand globally?10:58
Kanothe funny thing is, i only increased the version and changed the hz value10:59
Kanobut it failed with modules10:59
Kanoalso it still builds too much11:00
Kanoit builds tools and i dont want tools11:00
Kanoi already added a skip abi check11:01
Kanodid i forget something after increasing the version with dch -lxxx11:02
Kanoas there are no modules for 7.35 only 6.3411:03
Kanohow to increase version correctly?11:05
Kanohmm there is a skipmodule too11:07
apwwhen you start a new release you need to provide abi information for the previous version11:08
Kanohow to do that?11:08
apwthis is all documented in how to bump the abi information11:08
apwthe getabis script handles the body of the work11:09
Kanoyou call that manually?11:11
apwafter the builds complete yes, you run that and it sucks down the abi info from the archive11:11
apwthen that gets commited to allow checking of following builds11:12
apwbut you have been building kernels for years, so not sure why this is a new problem for you11:12
Kanobut that means you have to build it before you can put it into pbuilder11:12
Kanowell the last year i just used your mainline kernels11:12
Kanoi just began with .38rc6 again recompiling your kernels because i needed aufs11:13
apwno cause you are always using the previous builds stuff11:13
apwthe abi is never for the new kernel, its for the one before, which by definition is already built11:13
Kanoyour getabis script tries to dl from a server11:15
Kanoi think i just try skipmodule in arch.mk11:17
Kanothe last kernel with heavy mods i which i compiled using pbuilder was 3 years ago. using the custom flavour options11:21
apwKano, yep but it shows you how the abi stuff is meant to extract the info if you care.  otherwise you can skip it11:26
Kanois there a linux-meta 38-7?11:41
loolapw: Oy11:42
loolapw: Re: armhf11:42
loolapw: Essentially, this is the future for ARM support; this effort started perhaps about 6 months ago from a Genesi initiative using first Ubuntu armel and then introducing a real Debian armhf port, which is not an official Debian port yet but will eventually be one11:43
loolapw: On the Ubuntu / Linaro side, we want to eventually switch to this port instead of armel as it provides strictly better performance, but it has to be bootstrapped as it changes ABI11:44
loolapw: We had a good UDS session on this port but we were lacking buildds for very long11:44
apwlool, ok so when are we going to need this support, natty or O ?11:44
loolNow, I think we might have buildds, but it's clear that we wont have time to bootstrap armhf for Ubuntu before 11.04, still we're trying to merge as many support patches now rather than later11:44
apwoneiric (sp?)11:44
loolapw: I've just asked a couple of days ago for a FFE to merge sourceful fixes in the relevant packages11:45
loolapw: IOW, we're trying to have the packages be buildable with -aarmhf, but we don't expect Ubuntu to carry binaries this cycle11:45
apwlool, ok so i should wait for the FFE to be approved before merging them?  otherwise they ahve no operational effect in theory at least11:46
loolMost of the time, it's a matter of either fixing rules/control to check for both armel and armhf rather than just armhf, or to fix upstream configures which might not support the new triplet (arm-linux-gnueabihf instead of just arm-linux-gnueabi for armel)11:46
loolapw: You could opt to wait for the FFE, or decide that they are unintrusive  :-)  if you find there is a chance for breakage, then maybe wait for the FFE11:48
apwlool ok, will have a look see if they work ok in my cross env here11:48
loolapw: What's the best to rev one patch?  I'd like to update armhf.mk to also skipabi=true and skipmodules=true, but it feels heavy to resend the series just for that11:49
apwlool do you need to do that, can't you just include the ignore/ignore.modules in the abi directory11:50
apwwhen you inject the configs etc11:50
loolapw: I decided to not add any config or ABI data as to not pollute the kernel tree with armhf churn11:50
loolthat is, I figured you wouldn't want to see a lot of armhf in the diff, when you are not actually building these packages11:51
apwright, but you must be injecting the configs etc when you build, you could just inject those at the same time?11:51
apwyeah you are spot on there11:51
loolI'm not building any kernel ATM; I'm bootstrapping so I just need libc-dev11:51
apwok then just send the skip stuf as anohter patch, make make it a reply to the one it need squishing into11:51
apwand i'll squish it when applying11:51
loolapw: Ah one important thing I should also mention: the different is only userspace ABI; it's the same kernel ABI, so that armel kernels can run an armhf userspace and vice-versa11:51
apwright, as there is no fp in the kernel (in theory at least)11:52
Kanoapw: you tagged the linux-meta but the package is not created?11:54
apwKano, not created ?11:57
Kanoor not uploaded or whatever11:57
apwright, you cannot upload the meta until all the builds are completed and accepted by an AA11:58
apwelse the architecture becomes uninstallable, and cd builds fail, and kernels get de-installed from systems11:58
apwbut you run a debian based distro, you must have the same problem11:59
Kanosure,but i upload all the same time11:59
Kanoi upload prebuild packages11:59
Kanousually i fetch the packages from ubuntu which are not critical12:00
apwthat leave dist-upgraders at risk of partial installs, and most resoltions are fatal for meta packages12:01
Kanoi decide when i update the index12:01
apwyou have an advantage i don't have then, hense its 12:01
apwnot uploaded in ubuntu12:01
Kanowell you only need to upload source12:02
apwi can't upload the source, as that will build and upload the binaries, and they are not held for AA review12:02
apwso they will swan past the binaries, and leave us in a heap12:02
apwwhich is why we don't do it12:03
apwwe don't do our work in this order to make it hard for you, we do it cause it is right for our users12:05
Kanooh no, i got past the modules error but now: http://paste.debian.net/11085612:05
Kanodpkg-gencontrol: error: install new files list file: No such file or directory12:06
Kanowhat is that12:06
apwno idea, never seen that phase forget in all the builds i've done, and its a lot12:06
Kanoi use kernel-wedge from ubuntu but rest is debian12:07
Kanobut it has got an error, maybe it is related12:07
apwthe error is a rename failure, renaming a file it made from foo.new to foo 12:09
apwnot something i'd expect we'd trigger really12:09
Kanodo you get this error too?12:09
apwKano, those are normal, they are related to empty lines IIRC12:09
apwand yes we get them12:09
apwi'd check you arn't out of inodes or something12:09
apwrename("$fileslistfile.new", $fileslistfile) || syserr(_g("install new files list file"));12:10
apwis what barfed on you12:10
Kano17 gb free just for generic+generic-pae12:12
Kanois there something major missing?12:18
Kanothose are changes against your package12:18
Kanofirst i use the other script to disable parts of the build12:18
fairuzI do request_irq in kernel space, but is it possible to have the handler code in user space?13:15
mjg59fairuz: No13:16
fairuzmjg59: ok13:16
mjg59fairuz: I mean, you could generate a netlink event or something on interrupt and have userspace do something as a result13:16
mjg59But that's going to end up racy13:17
fairuzmjg59: yup, I will stick on kernel space then13:17
=== ogra_ is now known as ogra
=== tm_ is now known as tm
fairuzwhats writel or readl equivalent in user space?14:13
apwlool, this nls_cp437 patch seems to have an incorrect bug number in it14:13
loolapw: indeed, should be 73204614:13
loolI'll blame it on Chromium's paste buffer handling  :-P14:14
loolI wonder where I picked the other bug id from, it's private14:14
loolapw: Do you want me to resend?14:15
=== yofel_ is now known as yofel
sorenlool: * (drop after 2.6.38) ahci: AHCI mode SATA patch for Intel Patsburg SATA RAID controller - LP: #73524014:16
sorenlool: From an older changelog entry.14:17
loolAh, right, I probably forgot to update it after pasting then, thanks14:17
smoserlool, thank you for pushing on bug 73524014:20
ubot2smoser: Error: Could not parse data returned by Launchpad: list index out of range (https://launchpad.net/bugs/735240)14:20
loolsmoser: np14:20
apwlool, nope i can correct it now i know, thanks14:21
smoseri think it would be quite generally useful if kernel build process created a -extra-modules with the non-whitelisted files whenever the inclusion-list was used14:21
apwbug #73204614:21
ubot2Launchpad bug 732046 in linux "Missing filesystem modules in -virtual package" [Undecided,Won't fix] https://launchpad.net/bugs/73204614:21
loolsmoser: I agree with Stefan that the packaging changes aren't trivial, but this would also be my preference14:21
apwlool and how did you set linux back to new, without natty going back to new ... you've unconjoined the pair somehow14:22
loolI think you can do that as soon as the development task is closed14:23
smoserand generally, that list culd use some attention (i am probably at least partially responsible for doing so)... it has net/wireless/* in it.14:24
loolFor some value of the bug field14:24
apwlool well you learn something new and odd about launchpad every day14:26
* ogasawara back in 20min14:27
* apw giggles at ogasawara 14:27
apwsmoser, so it does ... hrm ... probabally something for O now14:34
smoseryeah, i dont really care, but the mac80211 is also there... porobably not that important :)14:35
smoseri suppose we could do a session on it in uds-o or at least you and smb and i and jj-afk could discuss14:35
fairuzHi, all variables in a kernel module will stay until rmmod is made?14:36
apwfairuz, a module can rely on its memory existing until the end of its destructor14:36
fairuzapw: destructor == module_exit?14:37
apwfairuz, believe so yes14:37
fairuzapw: thanks14:37
smbsmoser, apw sounds reasonable... hopefully we will have again any sensible layout for sessions14:38
* apw overlaps with tgardner :)14:44
* smb tries to picture that...14:44
tgardnerapw, uh, sorry. I was just grinding through the list14:44
apwtgardner, nope no problem, you carry on14:45
apwtgardner, i am only seeing the fat change, not the others, yet email says its pushed?14:47
tgardnerapw, which changes are missing?14:48
apwtgardner, oh ignore that, they moved, and i confused two threads14:48
apwtgardner, i'll handle updating ubuntu-debian, if thats ok, i want to resync them anyhow14:49
tgardnerapw, I figured you would.14:49
apwtgardner, i think lool wants this stuff in natty so they can do early testing, much as we have already elsewhere14:55
loolapw: Should I have sent a series against ubuntu-debian instead?15:01
loolI vaguely remembered the existence of this repo, but wasn't quite sure which one to send against15:01
apwlool, nope, its half and half for your use case15:09
apwi'm sorting it out now15:09
=== iulian_ is now known as iulian
tgardnerppisati, I was just chatting with ikepanhc about OEM mvl-dove kernels. He's still consuming CVEs and stable updates from the distro kernel. could add 'rebase to master' to your todo list for mvl-dove and send me a pull request?15:21
ikepanhctgardner: IIRC the mvl-dove lucid branch is maintained for 18mo?15:24
ppisatitgardner: isn't the -proposed kernel already rebased against the lastest-ish?15:24
tgardnerppisati, hang on a sec. otp.15:25
tgardnerppisati, I can't remember, but I think its a ways behind. I'm sure you can figure it out.15:32
tgardnerikepanhc, the mvl-dove branch only gets 18mo worth of love15:33
ppisatitgardner: anyway, ok, i'll do that15:33
ikepanhctgardner: ok, I need to inform PM this15:33
* apw_ is off the sir, internet in a heap15:50
tgardnerapw: did you mean "off the air" ?15:52
apw_indeed, touch screns 15:52
apw_make lousy keyboards15:53
bjfogasawara, you got a sec ? 15:57
ogasawarabjf: yep15:57
bjfogasawara, i just uploaded a linux-ubuntu-modules-2.6.24 to the canonical-kernel-team ppa15:58
bjfogasawara, the builds for it all failed: Missing build dependencies: linux-headers-2.6.24-29-generic15:59
bjfogasawara, however -29 is in proposed15:59
=== jj-afk is now known as jjohansen
bjfogasawara, i'm don't get it16:01
* ogasawara tries to remember if it'll pull build deps from -proposed16:01
apw_no it won't it is security only16:03
* smb thinks only updates/security and ppa itself16:03
apw_one reason to not delete kernrl till all drps built16:03
apw_remember that is why wr have the ppa, to build against security only16:04
* tgardner is enjoying apw's gibbers typing16:05
apw_bjf you may be able to copy kernel back from proposed16:06
bjfapw_, ogasawara ok, so i'm getting feedback that the hardy kernel in -proposed can't be installed due to ubmet dependencies. 'linux-ubuntu-modules-2.6.24-29'16:06
tgardnerapw, bjf: why don't I just upload LUM to the archive?16:06
bjfapw_, ogasawara so i uploaded a new LUM to fix this issue16:06
bjftgardner, you want my src pkg ?16:06
apw_but if kernel not still in ppa, it wont build16:07
tgardnerbjf, I'll just build from the repo. I assume its what you want?16:07
apw_assume someone cleaned itvout16:07
apw_we needcit built in ppa16:08
tgardnerapw:  2.6.24-29.87 is in -proposed. LUM should build against that OK.16:08
bjftgardner, will that work ? I want it for both -security and -updates. that's why we use the ppa16:08
ogasawarabjf: I'd assume you'd need to also upload the headers package to the ppa to get it to build ok16:08
apw_sobir can be copied to security  too16:08
ogasawarabjf: once it builds successfully then you could just delete the headers package from the ppa?16:09
apw_we should notbe deltingbhr16:09
bjfwe should not be deleting the ppa packages16:09
smbI would agree with apw (though he sounds like he ran against a wall with his nose)16:09
tgardneruntil all dependencies are built16:09
apw_the headers come from the kernel source16:09
smbEverything should be first in the ppa and then copied16:10
bjfsmb, thats all well and good except when you forget something, it sucks being human16:10
=== herton is now known as herton_lunch
apw_i suspect to fix youll have to do a no changecupload of the kernel16:10
apw_tovthe ppa, then lum16:11
smbThen the kernel package need to be uploaded and the cleaning mechanism should only delete packages that are superseded in the ppa itself16:11
apw_and get ourchecklist sorted so it includes lum etc16:11
tgardnerapw: in this particular case, Is ee no reason why we can't just upload LUM to -proposed.16:12
smbapw_, Could be that the version needs to be tweaked a bit because the real one had been in the ppa already16:12
apw_ivf the compiler is the same likely itll work16:13
tgardnerapw: I'm not too worried about that. gcc for Hardy hasn'16:14
bjfapw, tgardner, smb, i'm tempted to just ignore the one that is currently in -proposed and just upload the kernel for the new cycle, then the matching lum16:14
tgardnerhasn't changes in a bit16:14
* smb wonders how long it would take if we only uploaded to our ppa and requested more disk-space until lp learned to auto-clean ppas...16:14
tgardnerbjf, have you had any complaints about folks not being able to upgrade their LUM packages? I assume the meta package in -proposed is pointing at a LUM that does not yet exist.16:18
=== cmagina is now known as cmagina-lunch
bjftgardner, bug 725138 comments 2 & 316:20
ubot2Launchpad bug 725138 in linux "linux: 2.6.24-29.87 -proposed tracker" [Undecided,Fix committed] https://launchpad.net/bugs/72513816:20
bjftgardner, don't know why QA hasn't run into this16:21
tgardnerbjf, 'cause LUM is an elective install16:21
bjftgardner, that just occurred to me16:21
bjftgardner, it's only an issue if you have installed it and then are trying to update16:22
smbtgardner, I am not so sure about the electiveness of LUM16:22
smb(as it has all the sound of hardy)16:22
tgardnersmb, , oh it LBM and LRM that were elective?16:23
smbI think so. 16:23
tgardnerright, so I think all those packages need to get uploaded to -proposed.16:23
* smb is surprised to see his lucid ec2 branch building after all he has done to it...16:26
tgardnerbjf, so I assume Hardy LBM/LRM/LUM are now all on your dependencies checklist for ABI bumps? Hardy is a significant pain in the ass wrt ABI bumps. LBM/LRM are still currently lagging the ABI update.16:27
bjftgardner, yes, i'll work on straightening this out today16:27
apwdamnable broadband is still in a heap, at least i have a keyboard now though16:38
smbapw, Did you buy one for your phone ?16:39
apwheh, no but i also have a data stick for emergencies, which i've just activated16:39
smbapw, Ah :) Actually I guess you could have used the phone for that16:40
apwi can but it sucks up my allowance pretty quick, this is 'unlimited'16:40
smbThey did not promise unlimited speed... Though its hard to get unlimited data with a speed near to zero16:41
=== sforshee is now known as sforshee-lunch
=== bjf is now known as bjf[afk]
* bjf[afk] heading out to deal with a passport issue16:56
=== herton_lunch is now known as herton
ogasawaraso how are CVE's being handled now?  Does the security team still ping us to process a batch of public CVE's or do we just pick them up when we have time or are there dedicated days to work on CVE's?17:06
* ogasawara just noticed the list of CVE's is growing17:07
tgardnerogasawara, its been kind of jit and miss. most of us try to get to at least one CVE per week. you can coordinate on https://spreadsheets.google.com/a/canonical.com/ccc?key=0AgFTUDTDyXredE92dFdsVGkxN1FMMWJabS0wZENLRWc&hl=en&ndplr=1#gid=017:08
=== cmagina-lunch is now known as cmagina-afk
=== cmagina-afk is now known as cmagina
* tgardner --> lunch17:48
=== bjf[afk] is now known as bjf
bjftgardner, ogasawara https://wiki.ubuntu.com/Kernel/Dev/FixingCVEs18:20
=== sforshee-lunch is now known as sforshee
kristian-aalborghi ppl18:29
kristian-aalborgwhich is the smallest/fastest ubuntu installation that will let me build a kernel.deb?18:30
apwkristian-aalborg, i don't think any of them will let you do that without installing additional stuff18:54
kristian-aalborgyup, I'm aware18:54
apwapt-get builddep linux or similar will sort you on any18:54
kristian-aalborgbut would Server Edition do?18:54
apwyeah as long as you have the builddeps you are good, there is not impediment18:55
kristian-aalborgwould this be true for debian as well?19:00
kristian-aalborgI have an ultra modest machine that I plan to build kernels on - would love as little "fluff" as possible and there's only 2 gigs to spare19:01
jjohansenkristian-aalborg: uh you are aware that kernel builds suck up lots of space19:04
kristian-aalborgyes, painfully aware... I'm thinking of doing the building itself on an usb stick19:05
kristian-aalborgthen I can just throw the Eee 2g surf (!!) in a corner and check it every few hours ;)19:05
jjohansenkristian-aalborg: so a minimal debian would work as long as all the proper build deps are installed, but the build deps are actually pretty weighty19:07
jjohansenseveral hundred megs, /me doesn't remeber the exact figures19:07
apwyou do need about 5GB plus 1GB per flavour you build19:08
kristian-aalborgI have two eight gb sticks19:08
jjohansenkristian-aalborg: while you can do builds on a stick you will want them to be fast sticks and you can burn them out pretty quick19:10
kristian-aalborgno need to burn, I'll just move the stick over19:11
kristian-aalborgthe perfect crime ;)19:11
* jjohansen -> lunch19:14
kristian-aalborgI see that ubuntu NetInst with the debs is 1,2 gig19:20
* kristian-aalborg has settled for Server Install and is about to give it a try19:59
* bjf -> lunch20:16
=== bjf is now known as bjf[afk]
JFostepping away for a bit. call my cell if you should need me  20:20
JFo(not that I expect you will need me)20:20
* vanhoof calls JFo for fun20:34
kristian-aalborgby the way, is there somewhere you can get a pre-built kernel with APM support (i.e for old hardware)?20:38
tgardnerkristian-aalborg, every release since Dapper has CONFIG_APM=m20:39
kristian-aalborgtgardner: I was thinking of something app-gettable?20:40
kristian-aalborga project somewhere for freaks like me running old machines ;)20:40
tgardnerkristian-aalborg, I'm not really sure what you're asking then. Every _kernel_ we've released supports APM to some extent. Have you tried 'sudo modprobe apm' ?20:44
* kristian-aalborg needs to boot up another machine to do it live (like O'Reilly puts it)20:45
kristian-aalborgFATAL: error inserting apm, no such device20:50
kristian-aalborghowever, the file is very much there20:51
mjg59kristian-aalborg: apm will fail if acpi is running20:53
kristian-aalborgso, the vanilla ubuntu kernel supports apm?20:55
* kristian-aalborg has to shop for cigs, brb20:56
njinhello to all, in a system with a pae kernel installed, can i suggest to install the ppa kernel?20:59
apwnjin, depends on the PPA ... but normally there are generic and generic-pae flavours in a ppa21:01
kristian-aalborgmjg59: which flags in grub? apci=off and apm=on ?21:02
mjg59kristian-aalborg: That would probably be a start21:02
njinapw: i know only http://kernel.ubuntu.com/~kernel-ppa/mainline/, there's another ppa?21:03
kristian-aalborgI'd swear I tried any and possible flags...21:03
apwthe mainline builds, we don't build just -generic ones21:04
apwi mean we do just build -generic there21:04
njinand where i can find the pae then?21:05
apwof mainline kernels, we don't make those i am afraid21:05
apwthey are not designed for running regularly, they are for testing only21:05
njinah, i understantand, so better not suggest to try the mainline in a machine with a pae kernel?21:06
kristian-aalborghurm, same error with those flags21:06
kristian-aalborgto clarify, this is the kernel that came with the install21:07
mjg59You're sure that the machine has an APM BIOS?21:09
kristian-aalborgpretty much21:09
mjg59Yeah, ought to work21:10
mjg59What does dmesg say after modprobe apm ?21:11
=== bjf[afk] is now known as bjf
njinKernel driver in use: i915 and intel ips 0000:00:1f.6: CPU power or thermal limit exceeded is a linux issue ?21:15
njinbug 72857921:16
ubot2Launchpad bug 728579 in linux "Imagen in external VGA monitor trembles" [Undecided,New] https://launchpad.net/bugs/72857921:16
mjg59kristian-aalborg: Ok, I blame the code that's meant to set up the APM data21:17
mjg59Should probably just file a bug21:18
kristian-aalborgas I understood it, apm has to be compiled into the kernel to work now?21:19
kristian-aalborgok, that's good21:21
=== MTeck is now known as MTecknology
loolDoes someone have a recipe for doing a PPA upload of an Ubuntu kernel?22:56
loolbasically I'd like to apt-get source linux, do some changes, push to my PPA, but I keep running into issues; I did rename the ABI dir last time around, but now I'm getting module hashes mismatches22:56
loolsorry symbol hashes22:58
* lool is going to try with a custom ABI name23:00
kamallool: hi, I publish kernel PPA's quite regularly -- maybe I can help.  took me several iterations to finally whip the ABI and module check issues into submission also :-)23:02
loolkamal: Happy to hear your tricks  :)23:03
loolI could obviously set skipabi and skipmodules on all arches, or touch the ignore files, but if I could keep the changes to a minimum and keep the safety nets, that'd be nice23:03
kamallool: the secret is to disable the ABI check *and* the module check23:04
kamallool: ah, so you know about the ignore files already -- that's the big "trick" :-)   always works for me, 100%23:04
loolkamal: Ever tried uploading your own ABI?23:05
kamalI used to worry about the "safety net" also, but have found that none of my PPA "customers" ever have any trouble with my published kernels -- I don't worry about it any more, I just shut those ABI and module checks right off and don't think twice about it.23:05
loolkamal: Also, what version number do you use, and do you update both debian.master/changelog and debian/changelog?23:05
kamallool: no, I've never tried uploading my own ABI23:06
jjohansenlool: generally debian.master23:06
jjohansendebian/changelog should be copied from debian.mast/changelog23:06
loolBut when you prepare the source, tools usually poke debian/changelog e.g. dpkg-source -b23:06
kamallool, jjohansen: yes, I always edit *both* as jjohansen describes23:07
loolso I end up editing both -- if I edit only debian/changelog it gets replaced during build, and that doesn't end happily23:07
jjohansenlool: fdr clean, setups up debian/changelog23:07
kamallool: yes, correct -- you must keep them in sync for any sort of sanity to occur.23:07
kamaljjohansen: but beware!...   fdr clean only gets run *after* you've produced the source package!...23:07
kamaljjohansen: so if you don't *manually* copy debian.master/changelog to debian/changelog before you produce that source package, it will get produced with a bogus version number.23:08
kamal(doesn't matter for git-push-and-build, but does matter very much for PPA uploads).23:08
jjohansenkamal: ah, yeah that is a problem, /me has always done fdr clean before making the source package23:09
kamaljjohansen: ah, I don't do that -- I *just* "cp debian.master/changelog changelog"23:09
jjohansenkamal: well I wasn't sure what other bits might be needed, so I figured it was safest23:09
kamaljjohansen: that's the only one needed.   I've thought about trying to fix up our packaging system to resolve that issue also (since I built an awful lot of mis-versioned packages before I figured out the reason).23:10
jjohansenlool: as for the abi stuff its an absolute pain, I suggest you just follow kamal's suggestion of disabling23:10
jjohansenkamal: its the abi stuff that killed me originally, I was preparing the package from a tree where I had done test builds, but the test builds had generated the abi files I needed incrementally, so they were available locally but didn't become part of the upload, and hence the ppa builds would fail23:12
kamaljjohansen, lool: yes, I used to tear my hair out over build failures due to ABI ... but having never experienced any problem with just disabling the check.23:15
loolThanks all23:16
kamallool: you also asked about version numbers...  I've settled on a "+kamal~suchandsuch" scheme so if I'm publishing a build of "the foobar fixes" on top of say 2.6.38-7.35, then I'll use a version number of "2.6.38-7.35+kamal~foobar~fixes".  ...23:18
kamalIMHO, its good to have my name there for easy blame^Widentification by uname -a.   The "+" sorts correctly and conveys some useful meaning and is a legal character.  The "~" is also a legal character (which does have another special meaning, not relevant here).   Note that neither "-" nor "_" are acceptable characters for you to add on to the string.23:20
kristian-aalborgmjg59: might you be interested in helping me file a bug report later?23:33
kristian-aalborgI'd want to do it properly, if I should do it23:34

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!