jbailey | mjg59: So it seems the patch killed any system without a swap partition, joy. | 12:37 |
---|---|---|
jbailey | mjg59: I thnk I have it fixed. | 12:37 |
jbailey | =) | 12:37 |
mjg59 | jbailey: Hrm? | 12:37 |
jbailey | The other suspend shell script had a guard to make sure that if the RESUME variable wasn't defined, that it didn't attempt to figure everything out. | 12:38 |
jbailey | This one didn't have the guard. | 12:38 |
jbailey | It turns out that [ -e ${RESUME} ] returns true when RESUME isn't defined. | 12:38 |
mjg59 | Ah | 12:39 |
mjg59 | Right | 12:39 |
jbailey | [ -e "${RESUME}" ] , however, returns false. | 12:39 |
mjg59 | Hm | 12:39 |
mjg59 | But it's still failing to resume, for some reason | 12:39 |
=== mkrufky [n=mkrufky@user-12lcl1s.cable.mindspring.com] has joined #ubuntu-kernel | ||
=== jorgp2 [n=jorgp@bnet-dial-32.bartnet.net] has joined #ubuntu-kernel | ||
jbailey | mjg59: Cerainly on my laptop resume doesn't seem to be defined. | 01:20 |
jbailey | But hopefully, I can kill all the critical bug reports at least. | 01:20 |
jbailey | *sigh* | 01:20 |
mjg59 | jbailey: "defined"? | 01:30 |
jbailey | //sys/power/resume is 0:0 | 01:37 |
mjg59 | Hm. When is this? After boot? | 01:38 |
mjg59 | jbailey: Ok. If I panic immediately before the resume call, I have no disk devices | 01:45 |
mjg59 | jbailey: Doing a udevstart gives me devices | 01:46 |
mjg59 | jbailey: Argh. Yes, udevstart is run after load_modules has finished | 01:46 |
jbailey | Ahahahaha | 01:47 |
jbailey | Hm | 01:47 |
jbailey | But why didn't the second call catch it? | 01:47 |
mjg59 | Dunno | 01:47 |
jbailey | It should at least still be set, since the suspend script is still there? | 01:47 |
jbailey | I think I want to spend another day thinking about how to lay this all out. | 01:48 |
mjg59 | I'd prefer that we had something in the archive that worked, to be honest | 01:48 |
jbailey | Most people will have swap on lvm, I think. | 01:48 |
mjg59 | Really? | 01:48 |
jbailey | IIRC, the installer does LVM by default now. | 01:48 |
mjg59 | Uhm. | 01:48 |
mjg59 | Not in my experience. | 01:48 |
mjg59 | Unless you tell it to use the whole disk? | 01:48 |
jbailey | Dunno. | 01:49 |
mjg59 | Hm. | 01:49 |
jbailey | All of my machines are whole-disk installs. | 01:49 |
jbailey | But I thought that it was doing LVM in general now. | 01:49 |
mjg59 | Right. Hrm. | 01:49 |
mjg59 | Oh, I'm told that it's easy to check if all the disks in an LVM are present | 01:50 |
mjg59 | One of the vg commands will tell you | 01:50 |
mjg59 | jbailey: Ok, yeah. A udevstart sorts it. | 01:54 |
jbailey | So in what order should this happen then? Attempt to walk the PCI bus, find card. Find ide-disk, sd, and friends. | 01:54 |
jbailey | udevstart | 01:54 |
mjg59 | Yeah. | 01:54 |
jbailey | Mm, no, step before. | 01:54 |
jbailey | Start md, lvm, evms. | 01:54 |
jbailey | I'm wrong, those are after udevstart | 01:54 |
mjg59 | udevstart needs to be before md and lvm | 01:54 |
mjg59 | Yeah | 01:54 |
mjg59 | Except we only want to start them if we have all member disks | 01:54 |
jbailey | Right, but that's a good hack to put the in lvm script anyway | 01:54 |
mjg59 | Sure | 01:55 |
mjg59 | Hurrah. Working hibernate. | 01:55 |
jbailey | Then walk the bus a second time, looking for usb and network drivers. | 01:55 |
mjg59 | Yeah | 01:55 |
jbailey | start md, lvm, and evms. =) | 01:55 |
jbailey | Sorry udevstart. | 01:55 |
jbailey | Then start those. | 01:55 |
jbailey | Then attempt hibernate resume again in case the device was on one of those and it happens to work for people. | 01:56 |
mjg59 | Yeah | 01:56 |
jbailey | Sick. | 01:56 |
mjg59 | Heh | 01:56 |
jbailey | Mind you.. | 01:57 |
jbailey | I guess I can shortcut that second pass if I have swap and the root device. | 01:57 |
jbailey | All I need is USB at that point in case they drop to a shell. | 01:58 |
=== doko_ [n=doko@dsl-084-059-067-020.arcor-ip.net] has joined #ubuntu-kernel | ||
jbailey | I wonder how many people will get screwed if I suddenly don't load their ethernet cards in the initramfs? | 01:58 |
=== zul_ [n=chuck@CPE0006258ec6c2-CM000a73655d0e.cpe.net.cable.rogers.com] has joined #ubuntu-kernel | ||
zul_ | heylo | 01:59 |
jbailey | mjg59: Are you getting lots of reports of fails to resume? | 01:59 |
jbailey | I have 15775 | 01:59 |
mjg59 | jbailey: Well, it's broken with current initramfs-tools | 01:59 |
mjg59 | As in, there's no way it can work | 01:59 |
jbailey | Right. I'm wondering if it's urgent enough that I should stay up, or if I can consider my workday over and dive in tomorrow. | 01:59 |
jbailey | I'd sortof like dinner and to see my belle. | 01:59 |
mjg59 | Heh | 02:00 |
mjg59 | Would it be possible to just add the one-line udevstart to fix some cases? | 02:00 |
mjg59 | I'll take a look at the lvm stuff and let you know what I find | 02:00 |
=== jbailey squirms. | ||
mjg59 | It leaves things no worse than they are right now | 02:01 |
jbailey | Oh, I figured out why the suspend script isn't running. | 02:01 |
jbailey | It needs a chmod +x | 02:01 |
mjg59 | Ah. heh. | 02:01 |
jbailey | When I recovered it from bzr, it didn't preserve that. | 02:02 |
jbailey | I'd really prefer to leave it until I can actually do a test run with it. | 02:02 |
mjg59 | I've just done a test run with that case | 02:02 |
jbailey | I'm basically going to go offline for the night in a few minutes and don't want to look for more "OMG can't boot" bugs. =) | 02:02 |
mjg59 | The one change I made was to add the udevstart | 02:03 |
zul_ | heh liar | 02:03 |
jbailey | mjg59: You have main upload privs, don't you? =) | 02:03 |
mjg59 | jbailey: Heh. Indeed. | 02:03 |
mjg59 | jbailey: Have you uploaded the one with the resume guard quotes? | 02:03 |
jbailey | mjg59: Feel free. That way they beat on your for touched-it-last priviledges. | 02:03 |
jbailey | mjg59: I have. If you're using bzr, I've updated that to 0.27 | 02:04 |
mjg59 | jbailey: Hm. vgscan ought to return an error, I'm told | 02:04 |
mjg59 | jbailey: Ok, where's your tree? | 02:04 |
jbailey | http://people.ubuntu.com/~jbailey/bzrtree/initramfs-tools | 02:04 |
mjg59 | Thanks | 02:04 |
jbailey | vgscan doesn't take a vg as a parameter, though. | 02:05 |
jbailey | The root vg is the only thing that needs to promise to be present at boot time. | 02:05 |
jbailey | Other arrays and stuff could require fancier setup that requires an otherwise running system. | 02:05 |
mjg59 | Hmm. Yes. | 02:06 |
jbailey | I wonder if vgdisplay has it. | 02:06 |
mjg59 | There must be a reasonable way to get it out of the metadata. | 02:06 |
jbailey | With lvm, it's not obvious what's reasonable or not. | 02:06 |
=== jbailey runs off for food. | ||
=== mkrufky is now known as mkrufky-away | ||
BenC | mjg59? | 03:42 |
mjg59 | BenC: Hi | 03:42 |
=== mkrufky-away is now known as mkrufky | ||
=== TheMuso [n=luke@dsl-202-173-132-131.nsw.westnet.com.au] has joined #ubuntu-kernel | ||
desrt | BenC; poke | 06:30 |
fabbione | morning guys | 06:40 |
=== TheMuso [n=luke@dsl-202-173-132-131.nsw.westnet.com.au] has joined #ubuntu-kernel | ||
=== Yagisan [n=jamie@220-244-253-135-nsw-pppoe.tpgi.com.au] has joined #ubuntu-kernel | ||
fabbione | BenC, jbailey: ping? | 07:43 |
=== jorgp2 [n=jorgp@bnet-dial-32.bartnet.net] has joined #ubuntu-kernel | ||
=== jorgp2 [n=jorgp@bnet-dial-32.bartnet.net] has joined #ubuntu-kernel | ||
=== JaneW [n=JaneW@wbs-146-149-243.telkomadsl.co.za] has joined #ubuntu-kernel | ||
=== jorgp2 [n=jorgp@bnet-dial-32.bartnet.net] has joined #ubuntu-kernel | ||
=== jbailey [n=jbailey@testhaus.cns.utoronto.ca] has joined #ubuntu-kernel | ||
=== jorgp2 [n=jorgp@bnet-dial-32.bartnet.net] has joined #ubuntu-kernel | ||
zul | morning | 02:43 |
fabbione | hey zul | 02:48 |
zul | hey fabbione how goes it? | 02:51 |
fabbione | zul: working as usual | 02:51 |
fabbione | you+ | 02:51 |
zul | ditto | 03:01 |
=== jorgp3 [n=jorgp@bnet-dial-178.bartnet.net] has joined #ubuntu-kernel | ||
Diziet | Ah, hello fabbione. I just wanted to have a quick word with you about that kernel-package dependency thing. | 03:20 |
Diziet | Do you have a moment ? | 03:20 |
fabbione | Diziet: can you give me 5/10 minutes? | 03:22 |
Diziet | Sure. | 03:23 |
fabbione | Diziet: ok... | 03:30 |
fabbione | i am back | 03:30 |
fabbione | Diziet: ?? | 03:33 |
Diziet | Hello. | 03:33 |
fabbione | so remind me where we left the discussion | 03:34 |
Diziet | So this is about 15488. Can you remember what I was saying on Friday ? | 03:34 |
fabbione | yes | 03:34 |
fabbione | till we agree to stop for the day :) | 03:34 |
Diziet | kernel-package has a hardcoded value `gcc', which it inserts into the Depends of the source .deb's that it causes to be created. | 03:34 |
Diziet | mdz and the submitter seem to think just changing that to gcc-3.4 would be correct. | 03:35 |
fabbione | Diziet: one second that i am digging code and stuff... | 03:35 |
Diziet | Since when you use kernel-package with one of our kernels they end up trying to build against gcc-3.4. | 03:35 |
fabbione | Recommends: libc-dev, gcc, make | 03:36 |
fabbione | ok | 03:36 |
fabbione | gcc is only Recommended | 03:36 |
fabbione | not a Depends: | 03:36 |
fabbione | now we have 2 situations we need to analize | 03:36 |
fabbione | and that's why i didn't agree with your changes | 03:36 |
fabbione | 1) building our linux-source | 03:37 |
Diziet | Yes, I see what you were saying about my change which I agree was wrong. | 03:37 |
fabbione | 2) building a generic vanilla kernel | 03:37 |
fabbione | ok | 03:37 |
fabbione | let's take it again point by point | 03:37 |
Diziet | Right. | 03:37 |
fabbione | so we are sure we don't overlook anything | 03:37 |
fabbione | for the sake of release in 3 weeks :) | 03:38 |
fabbione | our kernel must be built with gcc-3.4 | 03:38 |
fabbione | this is hardcoded | 03:38 |
fabbione | so | 03:38 |
fabbione | the Package: linux-source-2.6.12 must Depends on gcc-3.4 | 03:38 |
Diziet | Hardcoded in the Makefile you get when you unpack one of our kernel source trees (no matter how you got it). | 03:38 |
fabbione | or Recommends: gcc-3.4 | 03:38 |
fabbione | and I agree with BenC that should be a Recommentds | 03:39 |
Diziet | Right. | 03:39 |
fabbione | Diziet: exactly | 03:39 |
fabbione | it's hardcoded in the makefile | 03:39 |
Diziet | When you say `the Package: linux-source-2.6.12' you mean _our_ `linux-source-2.6.12'. | 03:39 |
Diziet | But of course someone could sensibly use make-kpkg to make their own linux-source-2.6.12. | 03:39 |
fabbione | Diziet: we have a source called linux-source-2.6.12 that B-D on gcc-3.4 | 03:39 |
Diziet | That's one of the things it's for. | 03:39 |
Diziet | Right. | 03:39 |
fabbione | and a package that ships our source + patches that Recommends: gcc | 03:39 |
fabbione | the latter is wrong | 03:40 |
Diziet | Indeed so. | 03:40 |
=== mkrufky [n=mk@68.160.103.77] has joined #ubuntu-kernel | ||
fabbione | so we all agree that P: l-s-2.6.12 must either Depends: gcc-3.4 or Recommends: gcc-3.4 | 03:40 |
fabbione | because i might download the source only for reading | 03:40 |
Diziet | _Our_ l-s-2.6.12, yes. | 03:40 |
fabbione | yes | 03:40 |
Diziet | Right. | 03:40 |
fabbione | our | 03:40 |
fabbione | i didn't talk about pint 2 at all | 03:40 |
Diziet | OK. | 03:41 |
fabbione | that's why i isolated the cases | 03:41 |
fabbione | i am still on point 1) | 03:41 |
Diziet | Carry on ... | 03:41 |
fabbione | so for 1) that's the only change required | 03:41 |
fabbione | now | 03:41 |
fabbione | 2) is more complex | 03:41 |
fabbione | it is way more complex | 03:41 |
fabbione | why: | 03:41 |
fabbione | a) we don't know what kernel the user is building | 03:42 |
fabbione | b) and we don't know why | 03:42 |
fabbione | so we can't force a gcc to him | 03:42 |
fabbione | he might be as well compiling a kernel with gcc-4.0 for debuggin | 03:42 |
Diziet | Currently, make-kpkg does do that. That is, if the user uses make-kpkg to make a linux-source.deb, it will say Recommends: gcc. | 03:42 |
fabbione | we have no idea | 03:42 |
Diziet | Right. | 03:42 |
fabbione | so i still believe that Recommends: gcc is the right solution | 03:43 |
fabbione | i see no winner in forcing a gcc | 03:43 |
Diziet | In this second case ? Certainly there shouldn't be a Depends. | 03:43 |
fabbione | ok.. | 03:43 |
Diziet | But yes, it seemed to me that either Recommends: gcc or nothing at all would be best. | 03:43 |
fabbione | and that's how the situation is | 03:43 |
Diziet | Nothing at all is a reasonable option - after all, the user knows what they're doing and we shouldn't trip them up. | 03:44 |
fabbione | also.. who compile its own kernel, needs to know what it is doing | 03:44 |
Diziet | Exactly. | 03:44 |
Diziet | Now, AIUI both of these cases are generated by the same code in make-kpkg. | 03:44 |
fabbione | so for me.. personally.. 15488 is NOTABUG :) | 03:44 |
=== Seveas [n=seveas@seveas.demon.nl] has joined #ubuntu-kernel | ||
Diziet | Um, you're saying that _our_ linux-source packages are _not_ made with make-kpkg ? | 03:45 |
fabbione | Diziet: yes it is made with make-kpkg | 03:45 |
Diziet | So surely atm the dependencies are wrong ? | 03:45 |
fabbione | Diziet: yes. | 03:45 |
=== Diziet goes to check. | ||
fabbione | but we need to fix it in debian/control in linux-source-2.6.12 | 03:46 |
fabbione | and not in kernel-package :) | 03:46 |
fabbione | our l-s-2.6.12 debian/rules is insane | 03:46 |
Diziet | But linux-source-2.6.12, if it is make with make-kpkg, has an autogenerated debian/control made by make-kpkg. | 03:46 |
Diziet | s/is make/is made | 03:46 |
fabbione | iirc it is overwritten | 03:47 |
fabbione | there are scripts that copies parts of debian/control to differnt packages | 03:47 |
Diziet | Scripts where ? | 03:47 |
fabbione | inside debian/ | 03:47 |
Diziet | In linux-source-x.y.z's source package ? | 03:47 |
fabbione | it's all contained in one dir (at least!) | 03:47 |
fabbione | yes | 03:47 |
Diziet | But make-kpkg makes up a source package out of whole cloth. | 03:48 |
fabbione | Diziet: be aware that our source doesn't call make-kpkg on itself | 03:48 |
fabbione | the source is copied/mangled/trashed/reconfigured several times | 03:49 |
fabbione | and that happens in debian/build | 03:49 |
fabbione | building our kernel is not only executing make-kpkg | 03:49 |
fabbione | there are plenty of more problems we need to address | 03:49 |
=== Diziet reads debian/rules build in linux-source 2.6.12's source. Blimey. | ||
fabbione | Diziet: you also want to check the scripts that it calls | 03:50 |
Diziet | Yes, so I see. | 03:50 |
Diziet | kernel-wedge gen-control > debian/control | 03:50 |
fabbione | yes.. kernel-wedge is for udebs | 03:51 |
fabbione | control file is generated via control.stub | 03:51 |
fabbione | nothing too fancy | 03:51 |
fabbione | Diziet: line 162 | 03:52 |
fabbione | sorry 163 | 03:53 |
fabbione | cp debian/control $(srcdir)/debian | 03:53 |
fabbione | that one copies our control file into debian/build/linux-source-2.6.12/debian/ | 03:53 |
Diziet | Right. | 03:53 |
fabbione | the dir that will generate the source afterwards | 03:53 |
Diziet | But even debian/control.stub looks like it should be autogenerated. | 03:53 |
Diziet | It's full of version numbers. | 03:54 |
fabbione | Diziet: ALT | 03:54 |
fabbione | control.stub and control are only for udebs | 03:54 |
Diziet | ALT ? | 03:54 |
fabbione | ALT = STOP :) | 03:54 |
fabbione | sorry... | 03:54 |
fabbione | itaglish | 03:54 |
fabbione | you will learn that my itaglish > pure english :P | 03:54 |
fabbione | don't get confused by control.stub and control | 03:55 |
fabbione | that's not relevant for we need here | 03:55 |
fabbione | just think at them like debian/control | 03:55 |
fabbione | the mangling required for udebs is another story | 03:55 |
fabbione | (brb.. mother nature is calling ;)) | 03:56 |
Diziet | So when you update to a new upstream kernel, how to the version numbers get furtled in these control files ? | 03:56 |
Diziet | OK. | 03:56 |
fabbione | (= i need to take a piss ) | 03:56 |
Diziet | Yes, that's not itaglish :-). | 03:56 |
fabbione | Diziet: sed -i -e 's/2.6.12/2.6.whatever/g' debian/control.stup | 04:01 |
fabbione | and for a little bunch of more files | 04:01 |
fabbione | it's suboptimal | 04:02 |
fabbione | but I never had the time to rewrite that mess | 04:02 |
fabbione | so we can live with it... | 04:02 |
fabbione | and we should merge our build system with the Debian one | 04:02 |
Diziet | Ah, right. So the right answer is just to change control and control.stub and leave kernel-package alone. | 04:02 |
fabbione | exactly | 04:02 |
fabbione | control.stub is enough :) | 04:02 |
fabbione | control is regenerated | 04:03 |
Diziet | OK :-). | 04:03 |
Diziet | Do you want me to attempt to do that ? Is this stuff in arch or something or just normal uploads ? | 04:03 |
fabbione | Diziet: see /topic :) | 04:03 |
fabbione | i can do it.. it's one line change | 04:03 |
Diziet | Hey, I looked at the topic _yesterday_ :-). | 04:04 |
fabbione | yeah... | 04:04 |
fabbione | debian/ is in baz | 04:04 |
Diziet | Right. Well, I'll leave you to do it I think; probably quicker than me faffing, unless I'm going to turn into an actual kernel bod. | 04:05 |
fabbione | Diziet: that's ok.. | 04:05 |
fabbione | Diziet: it is important that we coordinate changes to the kernel build system all together | 04:06 |
fabbione | there are a lot of bits that need to fall down in the right place | 04:06 |
fabbione | otherwise we can wave kthxbye to it | 04:06 |
=== doko [n=doko@dsl-084-059-067-020.arcor-ip.net] has joined #ubuntu-kernel | ||
Yagisan | how much space is needed when compiling linux-source ? | 04:14 |
Yagisan | I've tried to add a driver to my local copy, but it dies in my pbuilder complaining it is out of space. | 04:14 |
fabbione | Yagisan: it depends.. it can be quite a lot | 04:18 |
fabbione | up to 550MB for a build | 04:18 |
fabbione | sorry per flavour | 04:19 |
Yagisan | ah | 04:19 |
Yagisan | so my 11GB tmpfs wasn't enough | 04:19 |
fabbione | Yagisan: they should be plenty... | 04:20 |
fabbione | but well... if it complains about space, you must have done something wrong | 04:20 |
Yagisan | well, all I did this morning was "sudo pbuilder build linux-source-2.6.12_2.6.12-8.13.dsc" | 04:22 |
fabbione | Yagisan: it depends how you did configure pbuilder | 04:22 |
fabbione | by default it uses /var/cache | 04:22 |
fabbione | so if your /var/cache/pbuilder/build is a tmpfs, than yes.. you were building in RAM | 04:23 |
fabbione | otherwise you were writing on fs | 04:23 |
Yagisan | I set BUILDPLACE=/tmp/ in pbuilder and /tmp is tmpfs | 04:24 |
Yagisan | so I think pbuilder is set up ok | 04:25 |
Yagisan | I've never build anything this big in pbuilder though | 04:25 |
fabbione | Yagisan: just be sure it is really using /tmp than | 04:25 |
Yagisan | ok. I've just freshly unpacked linux-source-2.6.12_2.6.12-8.13.dsc and confirmed it is building in /tmp, now to wait and see if it crashes again - this time with logging on. | 04:46 |
mkrufky | fabbione: about v4l/dvb kernel updates (for breezy): what kernel should I patch against? | 04:51 |
fabbione | mkrufky: if you really want to spend your time on it, against our tree | 04:52 |
fabbione | but we are really too close to release to update an entire subsystem | 04:52 |
fabbione | i think it's more worth to start working from the dapper kernel | 04:52 |
mkrufky | ah, then i wont bother... no big deal... | 04:52 |
mkrufky | i guess that kernel will be ready in about a month, right? | 04:53 |
=== Diziet [n=ian@xenophobe.extern.relativity.greenend.org.uk] has left #ubuntu-kernel [] | ||
fabbione | mkrufky: i think in 10 days we can start working on dapper kernel | 04:54 |
fabbione | we won't be able to upload it | 04:54 |
fabbione | but given that breezy is uber frozen | 04:54 |
fabbione | we can start playing around in a devel branch | 04:54 |
mkrufky | hmmm..... so will i get access to this branch? | 04:55 |
fabbione | mkrufky: if you use baz yes of course | 04:57 |
fabbione | you won't be able to commit directly to it | 04:57 |
fabbione | but you can branch off | 04:57 |
fabbione | and publish your branch | 04:57 |
fabbione | and we can merge back | 04:57 |
fabbione | no big deal | 04:57 |
fabbione | zul did this for all the hoary release... | 04:58 |
fabbione | it did work pretty well | 04:58 |
mkrufky | thats fine... i guess i have a new tool to learn then.... (never heard of baz, but there's always a good time to learn) | 04:58 |
fabbione | ehhe | 04:58 |
fabbione | apt-get install bazaar | 04:58 |
fabbione | there are plenty of howto's | 04:58 |
mkrufky | got it | 04:58 |
fabbione | for basic stuff i can even help you :) | 04:58 |
mkrufky | i'm downloading it now... will play around a bit | 04:59 |
fabbione | sure | 04:59 |
=== jorgp3 [n=jorgp@bnet-dial-178.bartnet.net] has joined #ubuntu-kernel | ||
zul | i did what? | 05:25 |
zul | oh...heh | 05:25 |
zul | heh i wasnt special enough to get direct access ;) | 05:26 |
mkrufky | dont be jealous... im not special enough either :-P | 05:27 |
BenC | mjg59: ping | 05:27 |
mjg59 | BenC: Hi | 05:28 |
mjg59 | Hm. Are we allowed to break the ABI now? | 05:28 |
BenC | hey, can I close 14790? | 05:28 |
BenC | no :) | 05:28 |
mjg59 | Shit. Right. | 05:28 |
mjg59 | BenC: Yeah, looks like that can be closed | 05:28 |
mjg59 | Argh. | 05:30 |
=== mjg59 tries to figure out a way of doing this without breaking ABI | ||
mjg59 | I fear it may be impossible | 05:30 |
mjg59 | Right. It breaks ABI in the pcmcia socket driver | 05:31 |
mjg59 | If we want to support recent TI PCMCIA chipsets properly, they need an extra prod to power up properly by the looks of it | 05:32 |
zul | mjg59: i cleaned up the patch a bit last night so it applies will be doing a test build tonight and will commit it to my archive tonight for BenC | 05:32 |
mjg59 | Maybe I can just drop that hunk. | 05:32 |
mjg59 | zul: Which patch? | 05:32 |
fabbione | meh breaking the abi is still possible, but mdz is not going to like it | 05:34 |
mjg59 | I'll see if I can do this without it | 05:34 |
fabbione | BenC: i am working on fixing 15488, 14736 and 13506 | 05:36 |
fabbione | if you already did any of them, please let me know | 05:36 |
zul | mjg59: sk98lin | 05:38 |
zul | BenC: i have a patch for the buslogic stuff, i need to find a way to test it though | 05:39 |
mjg59 | zul: Cool | 05:39 |
BenC | buslogic == vmare fix? | 05:41 |
zul | yep..i added a struct for hotplug it seems the logical thing to do | 05:42 |
zul | based on previous examples | 05:43 |
BenC | fabbione: cool, thanks | 05:44 |
BenC | my vmware install is v5, so I can't test it | 05:44 |
Yagisan | I have vmware 5.5beta | 05:46 |
Yagisan | want a copy ? | 05:46 |
mjg59 | Blah. | 05:48 |
zul | sure...just send me a key or something | 05:48 |
mjg59 | Suck. | 05:48 |
mjg59 | It looks like we need to break ABI if we want these controllers to work | 05:49 |
BenC | Yagisan: bug shows in vmware v4 | 05:49 |
mjg59 | (Which includes bsaically everything made by HP at the moment) | 05:49 |
mjg59 | Sorry, my fault. I should have noticed this earlier. | 05:49 |
zul | meh...i have a key for vmware4 | 05:49 |
BenC | mjg59: well, if it's that bug a deal, we will break it | 05:49 |
mjg59 | They work if you boot with the card in, but there are no events for insertion/deletion | 05:49 |
BenC | s/bug/big/ | 05:49 |
Yagisan | BenC: I have 4.5.2 somewhere too | 05:49 |
BenC | I think supporting cardbus on a large range of laptops qualifies as worthy of abi breakage | 05:50 |
Yagisan | vmware 5 keys work in 5.5beta - that's part of the deal of being a beta tester | 05:50 |
BenC | Yagisan: A copy of that would be nice, if it's not too much trouble | 05:50 |
zul | Yagisan: same here | 05:53 |
zul | lunch time | 05:53 |
mjg59 | BenC: Mailed | 05:56 |
BenC | mjg59: the ABI breaker? | 05:56 |
mjg59 | Yeah | 05:56 |
mjg59 | The break is in cs.c | 05:56 |
BenC | ok, thanks | 05:57 |
BenC | btw, did you resend a second sk98lin patch? | 05:57 |
BenC | I only got one | 05:57 |
BenC | the first one | 05:57 |
mjg59 | Hrm. I did. | 05:58 |
mjg59 | The only difference was that -Ia/drivers was replaced with -Idrivers in the EXTRA_CFLAGS statement | 05:58 |
=== mjg59 wonders what's up with his mail | ||
fabbione | BenC: we need to add an extra package to upload when we break the ABI | 06:03 |
fabbione | BenC: klibc B-D | 06:03 |
Yagisan | heh - I did it again. linux-source-2.6.12_2.6.12-8.13 again failed to build in pbuilder - cites "lack of space" | 06:03 |
fabbione | Yagisan: i blame pbuilder | 06:04 |
Yagisan | fabbione: at least I have a log now | 06:05 |
mjg59 | BenC: Did you get the patch I just mailed you? | 06:05 |
Yagisan | I'll point pbuilder at a 600GB RAID array, start it up, and go to bed. If it isn't done by the time I get up - I'll file a bug on pbuilder. | 06:08 |
BenC | mjg59: not yet, I did get the pcmcia patch though | 06:08 |
mjg59 | BenC: Yes, that's the one I sent | 06:09 |
mjg59 | Should I resend the sk98lin one? | 06:09 |
BenC | nah, I can just add your change manually | 06:10 |
mjg59 | Ok | 06:11 |
BenC | I get spurious a's in my stuff too, I can deduce that you use vi :) | 06:13 |
BenC | mjg59: are you sure the change to cs.c breaks the ABI? | 06:25 |
BenC | the change is to socket_setup(), which is a static function | 06:26 |
mjg59 | BenC: Sorry, I should have been clearer. It adds a new callback to struct pcmcia_socket (see ss.h) | 06:26 |
mjg59 | So that changes the version | 06:26 |
BenC | ah, ok | 06:27 |
BenC | zul: ping | 06:39 |
mkrufky | jbailey: fedora kernel people dont like my tree-merge solution, RE: missing dvb headers.... have you ever heard from the original bug poster? | 06:43 |
mkrufky | jbailey: im hoping to get those moved into /include/linux/dvb .... hopefully for 2.6.15 | 06:43 |
BenC | mjg59: does the pcmcia fix apply to bug 15734, or is that bug a dup of the one this fixes? | 06:44 |
zul | BenC: yo | 06:44 |
mjg59 | Should be unrelated | 06:45 |
BenC | zul: you have sk98lin in your tree, right? So I don't need to merge this directly? | 06:45 |
BenC | apply directly I mean | 06:45 |
BenC | mjg59: ok | 06:45 |
BenC | what bugs is it | 06:45 |
mjg59 | BenC: Don't have the number to hand, but it bites me here | 06:45 |
zul | BenC: it will be going in tonight when i get home, should i remove the other yukon driver patch? | 06:46 |
BenC | zul: yeah, I'm dropping sky2 completely | 06:46 |
mjg59 | 15196 looks likely, 15083 certainly is | 06:46 |
BenC | if you pull it from your tree, even better, so I can just merge and take care of both | 06:46 |
zul | BenC: ah so ill add it tonight, im just doing a build now to make sure everything is hunky dory | 06:47 |
BenC | zul: thanks | 06:47 |
BenC | I plan on making tomorrow prep day for upload on Thurs | 06:48 |
BenC | just FYI | 06:48 |
zul | ok..ill commit it tonight then ;) | 06:48 |
mjg59 | BenC: One more easy patch for you, possibly a couple of more awkward ones | 06:48 |
mjg59 | (won't break ABI) | 06:48 |
BenC | mjg59: ok, I'm going to take those bugs, mark 15196 a dupe of 15083, and mark 15083 PEND | 06:50 |
mjg59 | Ok, cool | 06:52 |
=== jorgp3 [n=jorgp@bnet-dial-178.bartnet.net] has joined #ubuntu-kernel | ||
zul | whoops | 07:24 |
mjg59 | Hurrah! | 07:26 |
mjg59 | I may have found the solution. | 07:26 |
BenC | pcmcia? | 07:27 |
mjg59 | PCMCIA breaking hibernate | 07:29 |
mjg59 | I'm afraid you're stuck with the ABI breakage :) | 07:29 |
BenC | damnit :) | 07:29 |
jbailey | mkrufky: Ah, okay. I haven't taken a look at the script. | 07:32 |
jbailey | mkrufky: Basically what's in the bug report is all I've got. | 07:32 |
jbailey | mkrufky: Given that it's new files going in, is there no way to get them in for 2.6.14? | 07:33 |
mjg59 | I win | 07:34 |
=== fabbione commits fix for 15488 | ||
fabbione | guys .. one suggestion | 07:37 |
fabbione | i am reading here... kill sky2.. add sk98lin.. blabla | 07:38 |
fabbione | be aware that this might be the last kernel upload | 07:38 |
fabbione | and that somebody will have to support for 18 months | 07:38 |
fabbione | is the archive fucked up??? | 07:41 |
fabbione | oh fuck | 07:42 |
fabbione | Mithrandir: you blocked the kernel archive | 07:42 |
fabbione | Mithrandir: you need to set a proper umask on people | 07:43 |
fabbione | Mithrandir: /home/lamont/public_html/Archives/kernel-team@ubuntu.com--2005/kernel-debian--preX,14--2.6.12/ | 07:43 |
fabbione | check for patch-5 and fix permissions | 07:43 |
fabbione | i guess i will commit later | 07:43 |
mjg59 | fabbione: While it's bad to ship a kernel that's hard to support, it's worse to ship a kernel that doesn't actually work on useful hardware | 07:47 |
mjg59 | BenC: I've just sent you pcmcia_hibernate.dpatch | 07:47 |
mjg59 | BenC: This fixes hibernation when PCMCIA cards are physically present in the machine at hibernate time | 07:47 |
=== dmk [n=dmk@81.156.25.127] has joined #ubuntu-kernel | ||
zul | fuck fuck.. | 08:12 |
BenC | mjg59: cool, thanks | 08:12 |
=== jorgp3 [n=jorgp@bnet-dial-178.bartnet.net] has joined #ubuntu-kernel | ||
mjg59 | BenC: Another patch. Fixes reboot on HP laptops without affecting anyone else (I had one report of a desktop that didn't reboot with the reboot-thru-bios patch - this one should only hit affected machines) | 08:15 |
Mithrandir | : tfheen@rookery ~ > umask | 08:16 |
Mithrandir | 002 | 08:16 |
Mithrandir | fabbione: it looks sane to me | 08:18 |
=== doko [n=doko@dsl-084-059-067-020.arcor-ip.net] has joined #ubuntu-kernel | ||
=== jorgp3 [n=jorgp@bnet-dial-178.bartnet.net] has joined #ubuntu-kernel | ||
jbailey | BenC: Do you think it's safe to merge 10307 and 9115? | 08:31 |
mkrufky | jbailey: looks like it's safe to move the headers... i just worked it out with the dvb maintainer | 08:46 |
mkrufky | we should probably have them (officially) moved by 2.6.15 | 08:46 |
jbailey | mkrufky: Cool. | 08:46 |
jbailey | mkrufky: If you have a list of what headers are moving where, I can take care of it in Breezy. | 08:46 |
fabbione | Mithrandir: well the commit you did locked the archive | 08:47 |
jbailey | mkrufky: No chance of sneaking it in for 2.6.14? It's less invasive than the other crap that went in to -rc1 | 08:47 |
mkrufky | those 5 .h files in /drivers/media/dvb/dvb-core | 08:47 |
jbailey | -rc2, rather | 08:47 |
fabbione | Mithrandir: remember tha scp/sftp needs umask very eary | 08:47 |
fabbione | early even | 08:47 |
mkrufky | jbailey: I dont wanna push Linus' buttons | 08:47 |
mkrufky | and... there are some out-of-tree dvb drivers that still depend on them to be in the current location | 08:48 |
jbailey | mkrufky: 'kay. Where will they go? | 08:48 |
mkrufky | include/media/dvb | 08:48 |
jbailey | include/media? | 08:49 |
jbailey | A new subdir? | 08:49 |
mkrufky | checking..... | 08:49 |
jbailey | Maybe include/linux/media or somethin | 08:49 |
jbailey | mkrufky: Is there another channel I should join? | 08:49 |
mkrufky | nah, i took care of this over the linux-dvb mailing list | 08:50 |
jbailey | mkrufky: Right now there's stuff in /usr/include/linux/dvb | 08:50 |
mkrufky | but the applicable channel is #linuxtv | 08:50 |
jbailey | Well, I'm not in a rush to add windows 76 on my irssi. =) | 08:50 |
mkrufky | heh | 08:50 |
mkrufky | looks like ur're right... that directory doesnt exist | 08:51 |
mkrufky | i guess johannes wants to create the new directory | 08:51 |
fabbione | Mithrandir: i do set the umask from .bash_profile, because i think sftp doesn't load bashrc | 08:51 |
mkrufky | jbailey: apparantly this has already been the plan, i just kicked it with a little jumpstart, thanks to you | 08:51 |
mkrufky | hang on, i will show you the email thread | 08:51 |
jbailey | mkrufky: Tx. | 08:52 |
jbailey | If they're going under /usr/include/linux/dvb, I can do this no prob. | 08:52 |
jbailey | If he wants to create a new subdir, I need to wait for the flamewar, sadly. | 08:52 |
mkrufky | hmm | 08:53 |
mkrufky | would it cause any harm if u put them in include/linux/dvb ? | 08:53 |
mkrufky | i doubt it. | 08:53 |
jbailey | mkrufky: None at all. =) | 08:53 |
jbailey | mkrufky: 'cept that if nothing is going to look for them there... | 08:53 |
mkrufky | heh true | 08:53 |
jbailey | That's why I'd rather drop them into wherever they're going to wind up eventually. | 08:53 |
mkrufky | well, i wouldnt start up a flame war just yet | 08:54 |
mkrufky | http://linuxtv.org/pipermail/linux-dvb/2005-September/004861.html | 08:54 |
jbailey | But I imagine that there'll probably be some resistance to adding another directory to /usr/include from the kernel folks. | 08:54 |
jbailey | hmm | 08:54 |
mkrufky | ya maybe it makes sense to wait on it for now | 08:54 |
jbailey | Dude, I see the words "are considered half-private i.e. not for userspace" | 08:55 |
mkrufky | and tell your users to use the tree-merging scripts instead, as a temporary workaround | 08:55 |
mkrufky | yes..... thats the same as any other kernel headers | 08:55 |
mkrufky | but u see, those headers are needed for v4l compile , which is kernel space | 08:55 |
jbailey | Is it a kernel module? | 08:55 |
mkrufky | v4l is a set of kernel modules | 08:55 |
jbailey | Hmm. | 08:55 |
mkrufky | and so is dvb | 08:55 |
mkrufky | ivtv is NOT a kernel module | 08:56 |
jbailey | Are those going to just be integrated eventually? | 08:56 |
mkrufky | they are already integrated | 08:56 |
mkrufky | the only difference is: | 08:56 |
mkrufky | users find themselves updating their v4l / dvb drivers to take advantage of support for new hardware | 08:56 |
jbailey | Oh, hmm. | 08:57 |
mkrufky | you see, i just wrote all these drivers for new hardware that came out last year | 08:57 |
mkrufky | unfortunately, due to kernel release cycle, these drivers wont be available in a mainline kernel until 2.6.15 | 08:57 |
fabbione | BenC: ping? | 08:57 |
=== jbailey reaches for his 'no sympathy' stick... | ||
jbailey | =P | 08:57 |
mkrufky | so many users want to play | 08:57 |
mkrufky | so they get cvs | 08:57 |
mkrufky | now do u see ? | 08:58 |
jbailey | I do. It sounds very much like problem we can't solve. | 08:58 |
mkrufky | no | 08:58 |
mkrufky | it is easily solved | 08:58 |
mkrufky | the user gets newer drivers from cvs | 08:59 |
mkrufky | simple as that | 08:59 |
jbailey | benc, fabbione: What do we currently do for folks who want to compile their own out-of-tree kernel drivers? | 08:59 |
mkrufky | about the missing headers, until we standardize the final location, the tree-merging scripts will suppice as a temporary workaround | 08:59 |
mkrufky | s/suppice/suffice | 08:59 |
jbailey | Right. But whatever mechanism we provide for other kernel modules to be built out-of-tree should also be available for these things. | 09:00 |
jbailey | That might provide some hints for an esaier solution than the tree-merging scripts. | 09:00 |
jbailey | The userspace headers that I provide in linux-kernel-headers have all of the #ifdef KERNEL bits ripped out of them. | 09:00 |
jbailey | They can't be used for compiling kernel modules at all. | 09:01 |
mjg59 | BenC: Fix for 13382 just emailed to you | 09:01 |
mjg59 | Uh, 13882 | 09:01 |
fabbione | jbailey: we usually hand them a nice piece of RTFM :) | 09:01 |
mkrufky | heh | 09:02 |
fabbione | mjg59: re 13506... | 09:02 |
fabbione | ahci is a scsi module... | 09:02 |
fabbione | why should it land in sata? | 09:02 |
mkrufky | the tree-merging scripts do more that just fix this header problem | 09:02 |
mkrufky | there are other (non-critical) dependencies that v4l has in dvb-kernel that is necessary fro hybrid analog/digital capture cards | 09:02 |
mkrufky | oh, wait... it all just clicked... | 09:03 |
mjg59 | fabbione: It's not a SCSI module. It's a sata one. | 09:03 |
mkrufky | jbailey: did you at first think that v4l was userspace?!? | 09:03 |
jbailey | mkrufky: Yes. If I had known it was a kernel module, I would've asked Fabio the question earlier and not gone any further with it. =) | 09:03 |
mkrufky | aaah | 09:03 |
fabbione | mjg59: ok added.. but i have no way to ensure it's loaded before ata_piix | 09:05 |
mjg59 | fabbione: Well, we'll have to work that out later | 09:05 |
mjg59 | That can be handled in userspace | 09:05 |
fabbione | mjg59: yup | 09:05 |
fabbione | but there is no much time to do it | 09:05 |
fabbione | so i rather suggest we talk with Kamion | 09:06 |
jbailey | Do I smell more initramfs magic? | 09:06 |
mjg59 | jbailey: Sadly so | 09:06 |
jbailey | Greeat. | 09:06 |
mjg59 | jbailey: Actually, probably module-init-tools magic. But it'll need to go in initramfs. | 09:06 |
mjg59 | jbailey: We just need a modprobe.d that attempts to load ahci before ata_piix | 09:07 |
jbailey | fabbione: Which which FM shall I tell them to read when I smack this bug report with it? | 09:07 |
=== jorgp2 [n=jorgp@bnet-dial2-243.bartnet.net] has joined #ubuntu-kernel | ||
jbailey | mjg59: 'attempts to'? | 09:07 |
mkrufky | jbailey: so, i take it, that now that you know the true story... that ubuntu will frown upon such cvs module updates? | 09:07 |
mjg59 | jbailey: Well, it may fail | 09:07 |
jbailey | mjg59: So anyone with ata_piix will have a stale ahci lying around? | 09:07 |
mjg59 | jbailey: It can always attempt to rmmod it afterwards :) | 09:07 |
jbailey | mkrufky: No idea. It means that I'm not worried about it for the userspace headers. =) | 09:07 |
mjg59 | Actually, I dunno if that would work | 09:08 |
mjg59 | jbailey: It's better to have a stale ahci module than non-working systems, I think | 09:08 |
jbailey | mkrufky: That's why I'm asking Fabio which manual we send them to when they want to do their own modules? Between that and the tree merging scripts, hopefully they'll sort it out. | 09:08 |
fabbione | jbailey: it's called README + Makefile + Kbuild | 09:08 |
fabbione | jbailey: seriously... modules that are done properly use kbuild | 09:08 |
fabbione | if they use kbuild users needs only the headers for their running kernel and the correct gcc | 09:09 |
mjg59 | jbailey: There are 4 Intel bridges that will work with *one* of ata_piix or ahci | 09:09 |
mjg59 | And have identical PCI IDs | 09:09 |
jbailey | fabbione: Which package is kbuild in? | 09:09 |
mjg59 | Actually, that's probably not true. ata_piix will probably drive them most of the time | 09:09 |
jbailey | mjg59: 'kay. Hmm. Right now I'm just copying modprobe into the initramfs, and not the config files. | 09:09 |
mjg59 | jbailey: Yeah. We'll work something out. | 09:10 |
jbailey | mjg59: Do y ou think I should also take all of /etc/modprobe.d ? | 09:10 |
mjg59 | jbailey: Not sure. Scripts there may call stuff that you don't have. | 09:10 |
fabbione | jbailey: linux-headers.. it's kernel build system :) | 09:10 |
fabbione | jbailey: but once you install the headers for your specific kernel | 09:11 |
fabbione | you also automatically get the generic and kbuild | 09:11 |
mjg59 | BenC: Have we got the Apple touchpad driver? | 09:11 |
mkrufky | jbailey: ah, gotcha | 09:12 |
mkrufky | jbailey: this is exactly why i will backport the newer v4l/dvb code for the dapper kernel... hopefully users wont need to mess with cvs if all the latest support is already included in ubuntu's kernel | 09:13 |
Mithrandir | fabbione: anyway, is it correct now? | 09:14 |
fabbione | Mithrandir: yes.. thanks | 09:15 |
fabbione | at least i could commit | 09:15 |
jbailey | mkrufky: Ah, cool. I expect that it's quite late for getting an update into breezy with less than a month to ship. | 09:15 |
jbailey | mkrufky: I'll let you sort that out with BenC though. | 09:16 |
jbailey | mkrufky: I see that the kernel-headers-2.6.12-8 package does have an include/media directory | 09:17 |
jbailey | So this all makes sense now. =) | 09:17 |
jbailey | Getting the headers into there for Breezy is probably quite doable to keep everyone happy. (Again, it needs to go to BenC) | 09:18 |
mkrufky | ah, cool | 09:19 |
mkrufky | it DOES have include/media | 09:19 |
mkrufky | but not include/media/dvb | 09:20 |
mkrufky | (that is where they will end up going) | 09:20 |
BenC | mjg59: don't think so | 09:20 |
mkrufky | fabbione told me that he welcomes updates like i've mentioned, but yes, it probably IS too late for breezy | 09:20 |
mkrufky | although if you guys are in fact interested, I can probably have a patch ready by tomorrow (or the next day) | 09:21 |
fabbione | BenC: i did commit fixes for 14736 and 13506 | 09:21 |
fabbione | BenC: the former needs a full test build on ppc | 09:21 |
fabbione | the latter on all arches | 09:21 |
mjg59 | BenC: Want me to find the diff for you? | 09:21 |
fabbione | or at least ppc, i386, amd64 and ia64 | 09:21 |
BenC | fabbione: thanks | 09:23 |
BenC | mjg59: yes, please | 09:23 |
fabbione | BenC: no problem | 09:23 |
BenC | dpatch is really slow for taking in a lot of small patches | 09:24 |
jbailey | mjg59: Hey, did you ever make it possible for usplash to handle input text? =) | 10:27 |
mjg59 | I didn't | 10:27 |
jbailey | I have a wishlist bug with a patch attached for encrypted root. =) | 10:27 |
mjg59 | Haha. | 10:27 |
mjg59 | Prefix it with usplash_write "QUIT" | 10:27 |
jbailey | Sweet, that'll do nicely. =) | 10:28 |
jbailey | usplash_write "FOAD" | 10:28 |
=== Seveas [n=seveas@seveas.demon.nl] has joined #ubuntu-kernel | ||
=== jorgp3 [n=jorgp@bnet-dial-37.bartnet.net] has joined #ubuntu-kernel | ||
mjg59 | BenC: Just sent you a patch to fix http://bugzilla.kernel.org/show_bug.cgi?id=3927 | 11:40 |
mjg59 | (Well, "fix") | 11:40 |
=== mkrufky [n=mk@68.160.103.77] has left #ubuntu-kernel [] | ||
mjg59 | BenC: I've sent you a patch for #7904 (apple trackpad driver) | 11:47 |
=== chmj [n=chmj@wbs-146-160-201.telkomadsl.co.za] has joined #ubuntu-kernel | ||
=== apokryphos [n=apokryph@host-87-74-2-161.bulldogdsl.com] has joined #ubuntu-kernel | ||
=== apokryphos [n=apokryph@host-87-74-2-161.bulldogdsl.com] has left #ubuntu-kernel ["So] |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!