/srv/irclogs.ubuntu.com/2005/04/05/#ubuntu-kernel.txt

=== zul [~chuck@CPE0006258ec6c2-CM000a73655d0e.cpe.net.cable.rogers.com] has joined #ubuntu-kernel
zulheylo12:21
fabbionemorning06:14
=== Seveas [~seveas@seveas.demon.nl] has joined #ubuntu-kernel
=== lamont [~lamont@mix.mmjgroup.com] has joined #ubuntu-kernel
=== zul [~chuck@CPE0006258ec6c2-CM000a73655d0e.cpe.net.cable.rogers.com] has joined #ubuntu-kernel
zulhey03:09
=== hakGER [~hak@ip191.13.1411N-CUD12K-02.ish.de] has joined #ubuntu-kernel
zulfabbione:  for #8189 have a look at http://marc.theaimsgroup.com/?l=linux-scsi&m=109638967308598&w=203:30
zulbut it might break abi03:31
=== hakGER [~hak@ip191.13.1411N-CUD12K-02.ish.de] has left #ubuntu-kernel ["Verlassend"]
lamontmoo04:34
lamontfabbione: are you using baz mv for the abi files, or add /delete?04:36
lamontlooks like baz mv, if I'm understanding the cryptic output from baz update04:38
zulfabbione: uh never mind its alrady applied06:04
fabbionehi guys06:15
fabbionebaz mv :-)06:15
zulhey fabbione how is it going?06:15
fabbioneit's going finr thanks and you?06:15
zulgood...just enjoying a day off06:15
fabbionesome bugs have bounced to kernel-team06:17
zulyeah noticed that06:18
zulso uh i take it we are going to be using 2.6.11 for breezy/bender06:21
fabbionenope06:23
fabbionei am planning 2.6.1206:24
zulok06:24
fabbionei don't think we should spend too much time on 2.6.11 for 2 reasons06:24
fabbione2.6.12 is already at RC stage06:24
fabbionethat means it will be out in final pretty soon06:24
zuland....06:25
fabbione2.6.11 has a lot of security holes phone... sec.06:25
zulk06:25
fabbionesorry i was saying..06:25
fabbione2.6.11 has a lot of security holes that are fixed in 2.6.12 and we have as patches for 2.6.1006:26
fabbionethat means a lot of extra work in portforwarding the patches06:26
zulsounds reasonable i was thinking about version numbering last night06:27
fabbionei started looking into the XEN patch06:27
fabbioneand how to build the debs for it06:27
fabbioneat least the kernel side06:27
zuli say let schweeb handle the user side06:28
fabbioneyes06:28
fabbioneuserland is very simple06:28
fabbionei could package it in half day probablu06:28
zulfor the kernel version stuff i was thinking something 2.6.12.abi.patchlevel06:28
fabbioneour plan is to drop the patch level completely06:29
fabbioneMEH06:29
fabbionethe abi06:29
zuland use what?06:29
fabbionenothing. s/abi//g06:30
fabbionedilinger did a proposal on debian-kernel that sounds interesting06:30
zulok...then how are you going to keep track if they are using for example different acpi snapshots06:30
zulill have to see06:30
fabbionewe still track the package version06:30
fabbionethat cannot be removed06:30
fabbionebut we can kill the ABI06:31
zulok06:31
fabbionewe need to see how complex would be to manage without06:31
zuldo have a url for dilinger's proposal06:33
fabbioneit should be somewhere here in channel logs06:34
fabbionefrom yesterday iirc06:34
zulk06:34
zulheh i wont be lazy and ill find it myself ;)06:34
zulso you would have a base kernel-image and you would have the user compile the modules themselves ?06:37
zulwith module-assitant?06:38
fabbionesomething like that, but using a new standard hook where to store the sources of the external modules06:38
fabbioneso that they can be compiled at postinst time06:38
fabbionegiven that there is an ABI change06:38
zulso what if they have a slow ass system?06:39
fabbionenasty, eh?06:39
zulyeah reminds me of gentoo kind of06:39
fabbionewell in all cases if they use thirdy part modules they will have to recompile them anyway06:39
fabbionewe can still do in such a way to ensure that the modules we provide are always in sync06:39
zultrue but what if the kernel-headers are mixed up, for example if someone who thinks they are know what they are doing but they dont messes up and files a bug report06:40
crimsunhmm.  How about always triggering a recompile of l-r-m whenever a new linux-source is uploaded?  True, more space, but more transparent to the end-user06:41
crimsunor is that precisely what we're trying to avoid?06:41
fabbionezul: i am not sure i understand...06:41
fabbionecrimsun: we are avoiding 2 things here:06:41
fabbione1) random recompilation of l-r-m and thirdy part modules06:42
fabbione2) having to deal with ABI changes, since they are embedded in the package name06:42
fabbioneboth of them are not easy06:42
zulwell lets say they have a vanilla kernel-headers installed and they have our kernel-headers installed and something messes up06:42
fabbionezul: our kernel headers lands alwys in a predictable place.06:43
fabbioneif they fuck up the system is kinda of not our problem06:43
zulok...but what if the user puts it in the same place...06:43
fabbioneplus we can add sanity checks before performing operations06:43
zulim just trying to play devio advocate here06:43
zultrue..06:43
fabbionezul: sure.. i appreciate that :-)06:43
fabbionethat's why we need really brainstorm a lot face2face06:44
zulno gentoo here ;)06:44
fabbioneit's really difficult to evaluate all the corner cases on IRC06:44
zulyeah too bad i wont be at udu :)06:44
fabbionewe can also make it an option06:44
crimsunfabbione: ok, I see regarding (1).  So that means we avoid automatically triggering a rebuild on the buildds whenever a new l-s is uploaded?06:44
=== svenl_ [~luther@AStrasbourg-251-1-52-128.w82-126.abo.wanadoo.fr] has joined #ubuntu-kernel
fabbionecrimsun: there are a lot of implication in hiding an ABI change. l-r-m is really the last of our problems06:45
fabbionesince we have control of it06:45
crimsunfabbione: right, gotcha06:45
fabbioneour problem is to find an efficent way to detect external modules (out of Ubuntu), and do in such a way that if there is an ABI change, print a big fat warning to the user in the first place06:46
fabbioneand provide an automatic system to manage recompilation of these modules on request06:46
fabbioneof course there would be the nice config file with: do_everything_for_me=[yes|no|ask] 06:47
zulwell why not have a list of base modules that we support like ipw2200, hostap, and then inform the users that they abi has changed we are going to break shit now06:47
zuler...list of 3rd party modules06:47
fabbionezul: i am not talking about 3dy part modules in our tree.06:48
fabbionefor example i compile the quickcam module myself06:48
zulduh..oops06:48
fabbionethat is not part of either linux-source or l-r-m06:48
fabbionehow can i make it so that if i change the ABI i can tell the user: hey dude.. this module you compiled yourself will not load....06:48
fabbioneand so on...06:48
fabbione+06:48
fabbionehey dude: if you put the source in this dir06:49
fabbioneit will be recompiled automatically for you06:49
zulsounds reasonable06:49
fabbionewe can't use a Xen kernel as our default kernel... yet07:06
fabbioneso that means that we to create 386-xen, 686-xen, k7-xen07:06
fabbioneSMP is not supported in 2.6.1007:06
fabbionebut it might in 2.6.11.. not sure yet07:06
zulin linux-sources? cool07:06
fabbionewell, that's the idea07:07
fabbioneso that would add 6 more linux-images on i38607:07
fabbionethat will differ of one option07:07
fabbione+ we lose ACPI and CPUFREQ07:07
fabbionestill not supported by xen07:08
fabbioneah and >4GB mem support.. but i don't know that many people using that much anyway07:08
fabbionemay i add a royal mess to build the package? ;)07:09
fabbionesince ARCH=xen :-)07:09
fabbionethat might attempt to create _xen.deb07:09
zul++07:09
zuli think im going to go watch the incredibles ill be back later07:11
fabbionehave fun07:12
zuli will07:12
fabbioneit's a nice movie07:12
fabbione:-)07:12
zulespecillay the extras for the dvd07:12
fabbioneheheh07:20
=== lamont_r [~lamont@phantom.acmeps.com] has joined #ubuntu-kernel
zulthat was a good movie11:18

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