/srv/irclogs.ubuntu.com/2010/05/26/#ubuntu-arm.txt

=== JaMa|Sports is now known as JaMa|Off
raihi all08:11
=== hrw|gone is now known as hrw
hrwmorning08:23
zygahrw, czesc08:23
hrwcześć zyga08:23
zygahrw, wiesz jaki tag PROGRESS, INPROGRESS, etc trzeba ustawic w work itemach gdy sie nad nimi pracuje?08:23
zygapatrzylem na strone pittiego ale tam nikt nie uzywa takich oznaczen08:24
hrwmoment08:24
hrwhttps://wiki.ubuntu.com/WorkItemsHowto08:24
amitkHi Poland08:24
zygaamitk, hi ;)08:24
hrwzyga: and do not use any other then listed there08:24
zygahrw, thanks08:24
zygaha, TODO is an alias of INPROGRESS08:25
zygathanks hrw :-)08:25
hrwzyga: I got that page from Luic yesterday after using REVIEW tag ;D08:25
=== XorA|gone is now known as XorA
=== JaMa|Off is now known as JaMa|Wrk
* NCommander waves09:13
ograNCommander, hey09:15
ograNCommander, can you make sure https://blueprints.edge.launchpad.net/ubuntu/+spec/mobile-maverick-arm-improved-subarch-detection and https://blueprints.edge.launchpad.net/ubuntu/+spec/mobile-maverick-softboot-loader have their series goal set to maverick (david needs to approve after you proposed, i could do the proposing  for only one of them)09:16
ograNCommander, oh, and for the subarch one david needs to become the approver09:16
ograelse i cant make them show up on http://people.canonical.com/~pitti/workitems/maverick/canonical-mobile.html09:17
ogracooloney, yay, you managed to upload omap4 \o/09:41
ogra(though it failed to build)09:41
hrwheh... why gcc test suite takes so much time...09:41
ograNCommander, did you see my request above ?09:42
cooloneyogra: oh, yeah, too bad. i uploaded it from our server09:47
ogracooloney, see #ubuntu-kernel :)09:47
NCommanderogra: no I didn't09:48
ograseems there is an issue with the version detection in the buiold scripts09:48
NCommandermy internet connection has been *blink* *blink* *blink*09:48
ogra<ogra> NCommander, can you make sure https://blueprints.edge.launchpad.net/ubuntu/+spec/mobile-maverick-arm-improved-subarch-detection and https://blueprints.edge.launchpad.net/ubuntu/+spec/mobile-maverick-softboot-loader have their series goal set to maverick (david needs to approve after you proposed, i could do the proposing  for only one of them)09:48
ogra<ogra> NCommander, oh, and for the subarch one david needs to become the approver09:48
ogra<ogra> else i cant make them show up on http://people.canonical.com/~pitti/workitems/maverick/canonical-mobile.html09:48
NCommanderogra: thanks09:48
NCommanderlet me do that09:48
ogragreat09:48
NCommanderogra: and fixed09:50
ograNCommander, great, please hunt down david once he's awake so he sets the right approvals09:51
NCommanderogra: will do09:51
ogragood :)09:51
* NCommander is just fixing his network setup so he can boot one of his ARM boxen09:51
ograNCommander, lol09:54
ograyou cheated !09:54
NCommanderogra: I did?09:55
* ogra never got the idea to make himself the approver to approve specs and then change them ot the real approver09:55
hrwNCommander: that reminded me that I have to remove one 8port ethernet switch and cables of it.09:55
hrwit is my FastEthernet switch which was used only for developer boards09:55
NCommanderogra: I cna't approve for series goals even as the approval. Just happens my brain did the wrong thing09:56
NCommander*approver09:56
ograNCommander, ah, you didnt, you just flipped approver and asignee ! .... do you really expect david to implement the subarch detection ?09:56
NCommanderI did?09:56
ogra*g*09:56
NCommander*sighs*09:56
* NCommander fixes it09:56
* NCommander also reachs for the coffe cup09:56
ograwell, you could leave it09:56
ograwould be funny if you approve davids work now :)09:57
NCommanderogra: I'm on the fence w.r.t. to going from genext2fs to loopmounting09:57
ograNCommander, please dont09:57
NCommanderogra: please dont what?09:57
ograplease stick with the specced way09:57
ogradid you try it on your babbage now ?09:58
ograwith a proper size value09:58
NCommanderogra: it works kinda now once I stuck a swapfile on my system09:58
ograright, make sure to have at least 2G of swap09:58
NCommanderogra: but I still have consistancy issues with getting tune2fs to work if I don't pad a little spot of the image09:58
ograhmm09:58
ograi dont have that here and i dont see OE or angstrom adding extra space for it09:59
NCommanderogra: I still think 10MiB of scratch space is a good thing just so jasper has some room to write stuff before expanding the partition table09:59
ograso i wonder why you need it09:59
NCommandermaybe I just fail, but I got tune2fs: No space on device when I tried it09:59
ograthe partition will not be mounted before it is expanded09:59
NCommanderoh09:59
NCommanderhrm09:59
ograso there is no reason to leave any space09:59
NCommanderogra: the other issue with genext2fs is we can only use ext309:59
hrwogra: OE expands ext images10:00
NCommanderThat may not be a big issue, but I think we might want to consider using ext4 images so at least we're consistant w.r.t. normal installs10:00
ograthe first thing jasper will do is expand the partition, zero out the UUID and run fsck to create a new one10:00
ograhrw, but OE doesnt add extra space when using genext2fs10:00
hrwOE has default size set and if rootfs fits in it then no expanding.10:00
ograbefore running tunex2fs10:00
ograhrw, exactly10:00
hrwogra: and if rootfs does not fit then ext size is expanded by defined amount10:01
ograindeed10:01
hrwogra: I wrote expansion code10:01
ograin our case we determine the size of rootfs before running genext2fs10:01
hrwsame as in OE10:01
ograwhich means the image will have exactly the right size for it to fit10:01
hrwOE adds few MB10:01
ograhrw, hmm, where ? i didnt see that in the code10:02
hrwmoment10:02
hrwogra: 134         ROOTFS_SIZE=`du -ks ${IMAGE_ROOTFS}|awk '{size = ${IMAGE_EXTRA_SPACE} + $1; print (size > ${IMAGE_ROOTFS_SIZE} ? size : ${IMAGE_ROOTFS_SIZE}) }'`  in classes/image.bbclass10:03
ograhrw, what is the value of ${IMAGE_EXTRA_SPACE} ?10:04
hrwogra: 10240 by default10:05
ograok10:05
hrwso extra 10MB10:05
ograhttp://docs.openembedded.org/usermanual/html/image_types.html doesnt say that10:05
hrwOE manual is known to be not up-to-date ;(10:05
ograaha10:05
ograNCommander, so your code is fine then, i followed outdated docs apparently :P10:06
NCommanderogra: woo, I've been vidicated!10:06
ograthough i still didnt have any issues using tune2fs10:06
* ogra wonders why10:06
NCommanderogra: statistical improbability?10:07
NCommanderhrw: any reason you don't use loopmounting?10:07
hrwNCommander: in OE?10:07
ograNCommander, needs root10:07
hrwNCommander: OE refuses to work as root10:07
ograNCommander, you *could* use fuseext2 though10:07
ograbut i'm not sure about the permissions you end up with in the image afterwards10:08
ogragenext2fs seems a lot safer in that regard10:08
hrwNCommander: with OE you run one command and collect packages/images/sdks after some time. all is done as one process tree. running it as root would be insane (think "make install" in glibc/arm on !arm host). and with some build configurations one build can run for few weeks even10:09
hrwNCommander: my longest build took 2 weeks on dualcore amd64. but it was for ~30 targets on 6 architectures10:10
ograours only runs 30-40min for a 1.5G image10:10
NCommanderhrw: makes sense to me,10:11
ograand we'll create at most two image types10:11
hrwogra: with binary packages...10:11
ograyes10:11
NCommanderhrw: although we're generating the images on a machine running as root10:11
NCommanderso ...10:11
hrwogra: OE builds packages10:11
ograoh, indeed, OE builds everything10:11
hrwOe can be told to fetch packages anyway. I had such setups10:11
ograright, that would be similar to our setup10:11
ograNCommander, btw, being able to optionally build an ubuntu-minimal image (probably with openssh-server added) would be a nice to have, would be good if you could take that into account while coding ;)10:13
ograi exepct there are some use cases for having a preinstalled developer image10:13
NCommanderogra: we already have a base image type10:17
ograright, but not built in a preinstalled mannaer10:18
NCommanderogra: although without a seed, that won't be germinatable and d-cd will explode10:18
ogra*manner10:18
ograd-cd ?10:18
hrwdebian-cd?10:18
NCommanderogra: it tries to germinate to populate pool. I'll fix it so it won't do that on pre-installed images10:18
ograthe image comes from livecd-rootfs10:18
NCommanderbut I haven't gotten quite that far yet10:18
ograwhat has germinate to do there ?10:18
raican anybody tell me the exact flow of rootstock script10:19
raiwhy it uses two kernerl10:19
NCommanderogra: its just being called currently during image creation, I need to fix that10:19
NCommander(on the d-cd side, not in livecd-rootfs of course)10:20
=== dmart is now known as dmart|armel
ograrai, debootstrap (into an armel chroot), roll an image out of that, run VM, install the tasks and configure, create a tarball from the image, clean up the build environment10:20
ograrai, it only uses one kernel, the kernel for the VM10:20
NCommanderogra: anyway, I think the base images are used for something special w.r.t. to sanity on antimony, we probably want an ubuntu-minimal image thats command line only + openssh + jasper10:20
ograNCommander, exactly10:20
ograand that from livecd-rootfs :)10:21
ograas ext3 image10:21
raiogra but it first download deb from external site and second for virtual machine10:22
ograrai, not by default10:22
hrwrai: first it fetch debs for minimal system and then runs that minimal and fetch next sets of packages10:22
ograrai, there is an unsupported --kernel-image option the beagle people use for test kernels in their images10:23
ograyou should only use that if you really want to run such test kernels10:23
ograthe default run just uses the VM kernel to run the VM10:24
raiso it is not necessary to pass any kernel image10:24
bercohello! I have a question on the debian packaging, specifically the debian/control file. This file contains a "Sections" field and have problems finding out what value I should chose for my case. Does anyone have familiarity with this topic?10:24
hrwberco: debian packaging policy is what you need I think10:25
bercohrw: according to this link http://www.debian.org/doc/debian-policy/ch-archive.html#s-subsections I need to chose from the list10:26
dmart|armelor install the ubuntu-policy package10:26
bercothanks, since I'm a bit new the packaging I'll first look at this ubuntu-policy package10:27
dmart|armelIt's just a load of docs, but I think the info you're after is in there10:28
raiogra: exactly what debootstrap creates10:29
* NCommander feels exhausted :-/10:29
NCommanderogra: if we want a minimal image that's pre-installed, we need to do it correctly with seeds and such10:32
ogra_cmpcNCommander, a new seed ? why not just use the one we have ?10:43
ogra_cmpc(we already have minimal and standard)10:44
raiogra_cmpc, what exactly debootstrap does10:44
bercodmart|armel: the documentation package contains what I had online in my link above. What I was looking for is what would be the appropriate section from this list: admin, cli-mono, comm, database, devel, debug, doc, editors, electronics, embedded, fonts, games, gnome, graphics, gnu-r, gnustep, hamradio, haskell, httpd, interpreters, java, kde, kernel, libs, libdevel, lisp, localization, mail, math, metapackages, misc, net, news, ocaml, oldlib10:44
bercoI believe if my packages contain drivers I could chose "embedded" but I'm not sure10:45
ogra_cmpcrai, man debootstrap should tell you :)10:45
raiogra_cmpc, i go through but i didnt get exactly please tell me in rootstock script what exactly it does10:46
dmart|armelberco: I guess it depends on what your drivers are for10:46
ogra_cmpcrai, it bootstraps a basic ubuntu system10:46
bercodmart|armel: omap platforms10:47
raiit creates basic ubuntu file system for arm10:48
rai????10:48
ogra_cmpcdebootstrap cretes an unconfigured munumal filesystem10:49
ogra_cmpc*minimal10:49
dmart|armelberco: I notice that kernel packages tend to be in "admin"10:49
ogra_cmpcin the case of rootstock its for arm, yes10:49
bercodmart|armel: thx. strange they don't fall into "kernel". It's not all crystal clear to me yet10:50
ogra_cmpcberco, i dont think we make any active use of the section filed in ubuntu, probably packages.ubuntu.com does though for the web indicies10:50
* ogra_cmpc waits for persia to correct him :)10:51
bercoogra_cmpc: I see. "Section" is not mandatory but as it was "recommended" I thought I would fill it10:52
raiogra_cmps, in that script  run_vm() function what does exactly10:52
ogra_cmpcrai, it runs a VM :)10:52
raiogra_cmps , yes i got it but what does after running VM it displays Configuriang....    Unpacking.....    what is doing exactly????10:53
ogra_cmpcit is installing the package task you selected inside the VM10:54
raithat means seeds pass to script10:54
ogra_cmpcwell, to apt, but yes10:55
raiok and qemu-system-arm what it does actully??10:55
ogra_cmpcit is the VM10:56
raibut what it does10:56
rai????10:56
cooloneyogra_cmpc: my second upload was rejected because of The source linux-ti-omap4 - 2.6.33-900.1 is already accepted10:57
ogra_cmpcit executes a virtual machine10:57
cooloneyogra_cmpc: is that possible to remove my first upload10:57
cooloneyand i upload my new version?10:57
NCommandercooloney: you have to bump the version number.10:57
ogra_cmpccooloney, hmm, ask in #launchpad, i think it requores an admin, not sure10:57
ogra_cmpcor that10:58
NCommandercooloney: there's a hardcoded sanity check to prevent you from uploading the same version number more than once10:58
cooloneyNCommander: i don't wanna to bump the ABI, since that's my first upload,10:58
raiogra_cmpc ,  is this VM uses ubuntu basic minimal file system created using previous process og debootstrap10:58
rai*og of10:59
cooloneyalthough the first upload build failed10:59
ogra_cmpci think there is a button to remove uploads, let me check10:59
ogra_cmpcrai, right10:59
raiogra_cmpc , is it right??11:01
NCommanderogra_cmpc: cooloney that will not let you upload with the same version :-/11:02
ogra_cmpccooloney, i deleted it but please see https://help.launchpad.net/Packaging/PPA/Deleting11:03
ogra_cmpccooloney, according to that it can take 30min or more11:04
cooloneyogra_cmpc: thanks a lot, i will wait to 6:35pm in my TZ11:05
ogra_cmpc(up to 7 days)11:05
ogra_cmpcif it doesnt work at all, you need to talk to the lp admins in #launchpad11:05
ogra_cmpcrai, yes, this is right11:06
raiogra_cmpc ,  ok after can i use that armel.tgz file kernel for beagleboard??11:07
ogra_cmpcrai, why not use one of the officil beagle images ?11:08
* ogra_cmpc points rai to https://wiki.ubuntu.com/ARM/Beagle11:08
raiok but for testing11:08
raiok thanks11:09
ogra_cmpcthere are also unofficial images (with the test kernels i mentioned above) at http://elinux.org/BeagleBoardUbuntu11:10
raiogra_cmpc , can we get repository or packages required during debootstrap and second stage processing from local machine insted of internet11:11
ogra_cmpcyes, if you have an archive mirror or a package proxy11:12
raibut how ??11:12
ogra_cmpcjust point to the mirror11:12
raii dont know the exact way???11:12
raino i did not get exactly wht u say11:13
hrwrai: install squid-deb-proxy11:13
hrwrai: and then "sudo http_proxy=localhost:8000 rootstock ...."11:13
hrwrai: first use will populate cache for next builds11:14
ogra_cmpcor use the --mirror arg11:14
* ogra_cmpc points once again to the manpage11:14
hrwogra_cmpc: squid-deb-proxy is easier to seetup then local mirror11:15
ogra_cmpchrw, well i massively prefer approx, but yes11:15
raiok but wht it does exactly??11:16
ogra_cmpchrw, oh, and localhost will definately not work11:16
hrwogra_cmpc: so 127.127.127.127:8000 :D11:16
ogra_cmpcyou always need to use the external IP else the VM will point to itself ;)11:16
hrwogra_cmpc: I used vm-builder not rootstack11:17
ogra_cmpcrootstock ... --mirror http://192.168.2.87:9999/ubuntu-ports is what i use with my approx instance for example11:17
hrwogra_cmpc: does it works for debs not in your mirror?11:18
ogra_cmpcsure11:18
ogra_cmpcapprox is just a package proxy11:18
hrwok11:18
ogra_cmpcit automatically pulls whats missing11:18
ogra_cmpcand *only* pulls the packages i have used before11:19
ogra_cmpcway way smaller than a local mirror11:19
raiogra_cmpc, what is vm-builder11:20
ogra_cmpca tool to build vm images11:20
ogra_cmpcit has no support for arm yet, hrw is just adding it11:21
hrwyeah, need to boot laptop and grab latest ver11:21
ogra_cmpchrw, ogra@osiris:~$ du -hcs /var/cache/approx/11:21
ogra_cmpc1,8G/var/cache/approx/11:21
ogra_cmpcthat has all pacckages i used from main since jaunty11:21
ogra_cmpccompare that to a real mirror for three releases :)11:22
ogra_cmpc(and i'm building netbook and desktop for each release at least once)11:23
hrwnice11:23
raigrw, i trying to install squid-deb-proxy but it gives error E: Couldn't find package squid-deb-proxy11:23
raihrw, i trying to install squid-deb-proxy but it gives error E: Couldn't find package squid-deb-proxy11:23
hrw12:23 hrw@home:~$ apt-cache search squid deb proxy11:23
hrwsquid-deb-proxy - Squid proxy configuration optimized for deb packages11:23
ogra_cmpcits in univers though11:24
raiyes i found it..11:25
raibut it gives error package not found11:27
NCommanderogra_cmpc: http://paste.ubuntu.com/439822/ - progress is being made :-)11:29
ogra_cmpcNCommander, sweet !!!11:30
NCommanderogra_cmpc: it gets pretty far into d-cd before exploding into a pile of exceptions, but I think I got the hard part done.11:30
ogra_cmpcwhy is it so big though ?11:30
ogra_cmpcmy armel netbook is only 1.4G11:31
NCommanderogra_cmpc: guess its bigger on amd6411:31
ogra_cmpcoh, you build amd6411:31
ogra_cmpcyeah11:31
NCommanderYeah, I'm just doing that so I can spin faster11:31
ogra_cmpcOO.o and evolution11:31
NCommanderyup11:31
ogra_cmpcNCommander, btw, you chould make sure the ext3 image is depleted too, your code only deletes the squashfs one11:32
aaron_liujIllegal instruction11:32
aaron_liujwahy11:32
aaron_liujwhy Illegal instruction11:32
NCommanderogra_cmpc: yeah, I saw that and fixed it already ;-)11:32
ogra_cmpcNCommander, and make sure the .size file contains the size you actually use11:33
ogra_cmpc(not sure the file is actually used but it should have a correct value, so overwrite it if you compute your value)11:33
aaron_liujwhy Illegal instruction11:33
aaron_liujIllegal instruction11:33
ogra_cmpcaaron_liuj, context ?11:33
NCommanderogra_cmpc: size is used by livecd-rootfs, I don't think its used by d-cd although I'm not 100%. The diff for d-cd is going to be fugly11:33
aaron_liujno11:34
ogra_cmpci think its used by the web indicies etc11:34
ogra_cmpcand might be udes by ubiquity/oem-config, not sure11:34
ogra_cmpc*used11:34
aaron_liuji compile a applicaion ,but it used other lib with another compiler11:35
aaron_liuji compile a applicaion ,but it used other lib compiled by another version compiler,when i runs the application ,errors occur Illegal instruction11:36
ogra_cmpcso use the correct lib for your compiler or the right compiler for the lib11:37
ogra_cmpcwhat are you building and how do you build it ?11:37
aaron_liujbut i have not the  lib source and11:38
ogra_cmpcthats tricky then11:38
aaron_liujcompiler verion even don't known11:39
ogra_cmpcyou really need to give more context, what lib, what app, how do you try to build it etc etc11:39
aaron_liujjust readelf -a and find the compile different from my used11:39
ogra_cmpcwell, if it was built using a different toolchain and compiole options its unlikely you can make it work11:40
aaron_liujthe app is our company11:40
aaron_liujthe app is our company ,but the lib is another company11:41
aaron_liujthe app is comer from our  company ,but the lib is come from another company11:41
NCommanderogra_cmpc: I'm thinking to simplify d-cd code, it might be better to have the preinstalled image generated to always be livecd.project.preinstalled instead of livecd.project.filesystem unless we have a usecase for wanting multiple filesystems per subarch/target11:55
ogra_cmpcplease discuss that with cjwatson, i'm not sure we dont rely on the naming scheme somewhere11:56
ndecogra: hi! looks like the problem we used to have on OMAP3 with the boot partition and ROM code not able to find MLO is gone for OMAP411:56
ogra_cmpcelse, no objection11:56
ogra_cmpcndec, wohooo !!!11:56
NCommanderogra_cmpc: this is completely new code, so if we just add it as .preinstalled instead of .*, it makes life a lot cleaner11:56
ogra_cmpcNCommander, sure, just make sure it doesnt break existing code11:57
ndecogra: as such, we could have a single image with all flavors of MLO, uboot, ... on the boot partition, e.g. MLO-board1, MLO-board2 and have a script to switch from 1 board to the other11:57
NCommanderogra_cmpc: not sure how, considering I just changing the name in squash_ext2 :-)11:57
NCommanderbut will test heavily regardless11:57
ogra_cmpcndec, yeah, to sad vfat doesnt support symlinks though :)11:58
hrwuf.. gcc-4.5 build goes to packaging12:17
NCommanderogra_cmpc: for the image name itself, is $(CODENAME)-preinstalled-$(FULLARCH) acceptable, or do we want something like $(CODENAME)-preinstalled_desktop-$(FULLARCH) (I'm kinda afraid to make the if in the d-cd Makefile any bigger though; its already scary)12:49
NCommander(or well, preinstalled netbook)12:49
hrwNCommander: so maverick-preinstalled-armel?12:50
ograhrw, nope12:51
ograneeds to have the flavour12:51
ograand the subarch12:51
ograNCommander, can you shorten preinstalled to preinst or some such12:51
hrwmaverick-preinstalled-netbook-omap3 etc atleast12:52
* ogra would love to use oem but that gets to confusing12:52
ograhrw, maverick-netbook-preinst-armel+omap.img.bz212:52
ograand maverick-netbook-preinst-armel+omap4.img.bz212:52
hrwnice12:53
ograwe always have $release-$flav-$arch+$subarch.$type12:53
NCommanderogra: ugh, that's going to make that Makefile if statement even worse :-/12:54
NCommanderSuppose it can't really be helped12:54
ograyou just add preinst and compression to the naming scheme12:54
ograso instead of $release-$flav-$arch+$subarch.$type you have $release-$flav-preinst-$arch+$subarch.$type.$comp12:54
NCommanderogra: that's not hte problem :-)12:54
NCommanderthe problem is:12:54
ograelse we get massive probs with the make-web-indicies scripts12:55
NCommander     ifeq ($(PROJECT),ubuntu-server)12:55
NCommander CDBASE = $(CODENAME)-live-$(FULLARCH)12:55
NCommander     else12:55
NCommanderAnd a hell of a lot mor elines like that12:55
* NCommander thinks he just needs to rewrite the in a less-than-evil sorta way12:56
NCommanders/in/if12:56
NCommanderogra: I think we need to do compression in the publish step, not part of the actual image creation12:56
NCommanderbut we can solve that problem as we come to it12:57
* hrw → lunch12:57
ograNCommander, i'd like to do it as early as possible to save disk space12:57
* NCommander likes hrw's use of unicode12:57
ogradont forget we're low on space on antimony12:57
NCommanderogra: how low?12:57
ograwe couldnt publish the omap images until space was freed up for lucid12:58
NCommanderThe current design mimics the way normal live images store their squashfs until cleaned automatically in scratch12:58
NCommanderI'm guessing that won't work so well by having 2-3 2GiB images sitting around :-/12:58
ograi would have proposed to do the compression in livecd-rootfs ... but the imx51 CPU will take ages to compress 1.5G12:59
NCommanderogra: can't compress until after we run the post-boot scripts12:59
ograwhy ?12:59
NCommanderWe won't have a valid filesystem on the SD card12:59
ogranothing will ever touch the image12:59
ogra??13:00
NCommanderif we compress on the live image builders, we'll have a compressed ext2 image13:00
ograyou use the imx51 scripts to build it, right ?13:00
NCommanderright13:00
ograwell, ext3 but yes13:00
NCommanderthose expect an uncompressed image coming in13:00
ograoh, crap, indeed13:00
NCommanderIt really doesn't buy us much to compress it just to uncompress then recompress it13:00
NCommander:-)13:00
ograwe compress at the very end when we have both partitions in place13:01
NCommanderright13:01
ograhmm13:01
ograwell, i guess we have to live with that then13:01
NCommanderBut we'll have to dump the folder with the preinstalled images downloaded from the live builder13:01
NCommander:-/13:01
NCommanderSo we will probably need upwards of 4GiB to build an image until we delete the incoming files13:02
NCommanderUgh13:02
NCommanderthat's going to be freaking tight13:02
ograthe cdimage dir has 74G free atm (from 1.6TB)13:02
ograits not different to a DVD after all13:02
NCommanderThat's fine then. Once we compress the image, we can probably get it to the size of a normal image in www/13:03
ogramy compressed netbook image was 411MB13:03
raiogra , squid is decreases server load and fast content delivary so how it help in rootstock script13:03
NCommanderWe probably just want to do some early cleanup to prevent space from being a huge issue13:03
ograso with the second partition (and given i didnt install oem-config in the image) we'll likely end up with a 500 to 550MB image13:04
ograNCommander, as soon as we have compressed we should clan up scratch13:04
ogra*clean13:04
raihrw, squid is decreases server load and fast content delivary so how it help in rootstock script13:04
ograeven though i already hear cjwatson screaming, that seems to be the best13:05
ograrai, you wont have to download over and over again for subsequent builds13:05
raiok.. but if i wnt to use my local repository to build basic minimal rootfs how can i create ??13:06
NCommanderogra: http://paste.ubuntu.com/439858/ - I win13:09
NCommander(I think)13:10
ograi dont see it finishing any build13:11
ograi pinged cjwatson in -release btw to see if we can wipe scratch/ right after build13:11
NCommanderogra: ?, it does13:11
ograwhere ?13:12
NCommanderI get the raw (the publishing scripts don't know to publish preinstalled images yet, so I haven't written that code yet)13:12
ograah13:13
NCommanders/so//g13:13
ograwell, i only see it attempting an alternate build13:13
NCommanderogra: it looks like an alternate build because some of the scripts to refresh APT are in build_all.sh, and I don't want if statements all over the place; they're harmless for what we see now13:33
NCommander(the same messages are there on live builds as well)13:33
ograok13:34
=== freeflyi1g is now known as freeflying
NCommanderogra: you don't put MLO/u-boot.bin on omap images right? (you made it so that we require NAND bootloader on first boot, right?)13:40
* NCommander has a script that makes the VFAT partition correct on the first go13:40
ograNCommander, for lucid ... for maverick we want MLO and u-boot.bin in the vfat13:41
ograas well as for 10.0713:41
NCommanderogra: oh good, I was just concerned when I ended up with an empty vfat13:41
NCommander:-)13:41
ograthat will be the most tricky part btw13:41
NCommanderogra: not really, there's a shell script to do it right already13:41
ograMLO dees to live in block 1 of the vfat for omap313:41
ogra*needs13:42
NCommanderhttp://www.xora.org.uk/2009/08/14/omap3-sd-booting/13:42
NCommandersame with Blaze13:42
ograelse it will not boot13:42
NCommanderand that works fine with Blaze13:42
ografor omap4 it doesnt matter where MLO lives13:42
ograyou can just copy it in13:42
NCommandermy Blaze won't boot if MLO wasn't first block13:42
NCommanderregardless, its a solved problem, we can fix it as we go13:43
NCommanderjust wanted to make sure that an empty vfat was sane13:43
ograNCommander, well, according to ndec above MLO can live where it likes on the vfat13:43
NCommanderogra: hrm13:43
* NCommander shrugs13:44
ograthe point is that we want MLO to be replaceable easily13:44
NCommanderI admit I haven't toyed with MLO a lot after I got my Blaze to boot ;-)13:44
ograsince the image will be used on different HW that uses different bootloaders but the same kernel13:44
NCommanderogra: ah13:44
ograsee the spec :P13:44
ogra[ogra] Create documentation and/or scripts to replace the bootloader in the boot partition for using the images with all HW the kernel supports: TODO13:45
ograNCommander, so we're fine to wipe the raw ext3 images after d-cd finished14:10
ograso add some code to clean up scratch/ for that image14:10
=== rgreening_ is now known as rgreening
NCommanderogra: where'd you talk to Colin?14:17
ogra#ubuntu-release14:17
NCommanderogra: hrm, that fell off my AJOIN :-/14:17
ograNCommander, btw, how do you determine the number of cyls in the vfat script you pointed to ?14:22
ogra(in an image i mean)14:23
hrwlool: have a minute?14:24
NCommanderogra: I'd assumed it was sfdisk options, but this script gave me a VFAT that worked for both omap3 and 4, even if MLO wasn't copied in first14:34
ograNCommander, in an image ?14:34
ograor in SD ?14:34
loolhrw: I do14:40
hrwcool14:40
hrwlool: https://bugs.launchpad.net/ubuntu/+source/gcc-4.4/+bug/585439 - doko gave few suggestions14:40
loolhrw: How may I help?14:40
ubot2Launchpad bug 585439 in gcc-4.4 (Ubuntu) "migrate to debhelper7 (affects: 1) (heat: 10)" [Undecided,Incomplete]14:40
hrwlool: I can generate version without --autodest for gcc-4.4 I think but one thing is differ in how dh_install works and doko base on it14:41
hrwlool: dh_install only copies files so after all packaging was done you cant easily check which files were not packaged.14:42
loolhrw: So the two remarks in the bug are about --autodest, I didn't know whether it was needed or not myself either14:42
loolhrw: if it's not needed, don't add the flag14:42
loolhrw: Now concerning the other remarks of checking whether files are all installed, there are two ways to achieve this, one is --list-missing14:43
hrwit was first version of patch thats why autodest was used14:43
lool(and --fail-missing)14:43
loolHowever I believe this only works if you have a single dh_install call14:43
loolThe other way to achieve this is by hand14:43
hrwdo dh_movefiles -> dh_install + rm?14:43
loolBasically doing find on debian/tmp (or wherever things are installed) and on the packages dirs to see whether all files are there14:43
hrwgenerate list of installed and packaged and then show diff? that kind of?14:44
loolhrw: I would personally find dh_install + rm ugly, but there's an argument for it in terms of needed space (if you move, you use less space than if you copy)14:44
hrwspace is cheap14:44
loolthe reason rm is ugly is because if you fail in the binary-indep/binary-arch targets, you need to rerun the make install target14:45
hrwand if buildd has a problem with space then let someone connect 1.5TB drive to it14:45
hrwlool: will check with comparing installed/packaged list of files14:46
loolhrw: You might want to ask about which way doko would prefer it in the bug too14:46
hrwthats what I plan now14:47
loolhrw: You can ping him on #ubuntu-devel if you have questions14:47
hrwwill do this too14:47
NCommanderogra: on an SD card directly14:58
ograright14:58
ograi suspect for images that will become problematic14:58
ograbut i trust you to find a solution ;)14:59
ograNCommander, did you ask davidm already to change your specs ?15:00
ograi still dont see them on the tracker15:00
=== lag is now known as lag-afk
NCommanderogra: looking at the OMAP scripts, you create the FAT partition, but you don't seem to stick an entry in the MBR for it properly, am I missing something?15:22
ograNCommander, the omap scripts are a bare copy of the dove scripts :) in lucid we dont need MBR or anything since we dont ship a bootloader there15:23
NCommanderogra: ah, that's why I'm loosing my mind trying to find a non-existant bug :-). I ran the preinstalled image generation through the imx51 generation scripts, and presto, I got two partitions15:25
NCommanderNumber  Start   End     Size    Type     File system  Flags 1      0.51kB  33.6MB  33.6MB  primary                     2      33.6MB  496MB   463MB   primary  ext3         lba15:25
NCommanderso I think we can call this a working prototype15:25
NCommanderI need to look at the code and remove some of the crack I added15:25
NCommanderbut I say we're moving along nicely15:25
NCommanderwb ogra15:28
ogramumble15:29
* ogra cuses that norwegian guy15:29
comradekinguogra: ooooodin?15:42
ogracomradekingu, nope ...15:42
ograpeer15:42
ograhe resets my connection !15:42
ogra(Connection reset by peer).15:42
comradekingu:)15:42
comradekinguIt would be sweet to have odin error messages though15:43
ograheh15:44
=== lag-afk is now known as lag
comradekinguLoki was found in your network configuration, by draupnir, odin shall make amends, meanwhile at your place, no wifi15:51
ogralol15:52
=== robbiew1 is now known as robbiew
=== XorA is now known as XorA|gone
hrwbye all17:00
=== hrw is now known as hrw|gone
=== sbambrough is now known as sbambrough-lunch
=== JaMa|Wrk is now known as JaMa
=== sbambrough-lunch is now known as sbambrough
prpplaguedavidm: there a trick to getting ubuntu-arm working with a nfs root?23:29

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