/srv/irclogs.ubuntu.com/2011/11/30/#ubuntu-arm.txt

NCommanderGrueMaster: any luck with that image/01:29
GrueMasterHaven't rechecked.  Just automating core image tests for all 3 arches.  Running it now through jenkins.01:30
GrueMasterOk, it was waiting for me to nudge it past restricted.01:30
GrueMasterrunning now.01:31
GrueMasterfail.01:31
GrueMasterJan  1 02:25:00 base-installer: info: could not determine kernel flavour01:32
=== mayday_jay-|work is now known as mayday_jay
GrueMasterThis is telling:  Jan  1 01:58:17 base-installer: warning: Unknown armel subarchitecture 'omap4'.01:35
=== nemik_ is now known as nemik
=== twb` is now known as twb
=== Jack87|Away is now known as Jack87
=== ericm|ubuntu is now known as ericm-afk
=== doko_ is now known as doko
=== Quintasan_ is now known as Quintasan
WaltherF1Um, the latest I checked at raspberrypi.org, the final version is using Broadcom BCM2835, ARM11 compatible - isn't that one of the "newer" versions, that should be supported by Ubuntu?10:54
infinityWaltherF1: Unless something's changed, the Pi is armv6, and we build for v7.10:56
WaltherF1infinity: They published a new graphic for the final version, http://www.raspberrypi.org/wp-content/uploads/2011/11/Raspi-Model-AB-Mono-1.png11:00
WaltherF1infinity: and it exclusively states ARM1111:00
infinityWaltherF1: ARM11 = ARMv611:00
WaltherF1ah, okay, sorry for the confusion11:01
WaltherF1...and btw, imo that's not a neat way of arranging your versioning -.-11:01
WaltherF1(not your fault!)11:01
infinityhttp://en.wikipedia.org/wiki/List_of_ARM_microprocessor_cores11:02
infinityAnd yeah, it's confusing.11:02
ogra_send your thanks to ARM for that kind of crappy versioning :)11:02
infinityThe Cortexes continue the confusion with going from A8 to A9 to A7 and A15...11:02
ogra_yeah11:02
ogra_at least A7 will then be v711:03
infinityHah.11:03
lilstevieheh11:04
WaltherF1duh, at least ubuntu is sensible with yy.mm11:06
ogra_janimo, oh, btw, having the ac100 kernel built for el as well as hf would be nice11:06
infinityogra_: Kernels will come.11:06
ogra_WaltherF1, well, but we use funny names too :)11:06
infinityI'll bug people more insistently when the archive is actually installable.11:06
ogra_infinity, that kernel will need jani or me touching it ... its not maintained by the kernel team11:06
lilstevieogra_: el and hf?11:07
infinityogra_: I know.  Or me. :P11:07
ogra_indeed :)11:07
infinitylilstevie: armel and armhf.11:07
lilsteviekernels shouldn't have that issue11:07
lilstevieinfinity: yeah I am aware11:07
infinitylilstevie: The kernel needs to exist in the hf archive.11:07
ogra_lilstevie, but you want an armhf .deb ;)11:07
infinityWhich means, sadly, building it twice.11:07
lilsteviemer guys assured me that one kernel should be universal11:07
infinity(Or building it as arch:all, but we really don't need ARM kernels on powerpc...)11:07
ogra_its all about dpkg ... not actually the binary11:07
lilstevieah11:08
infinitylilstevie: Identical build.  Just needs to be in both _armel.deb and _armhf.deb form. :P11:08
lilsteviekinda crappy system11:08
ogra_we could as well just cp them on ports.u.c :P11:08
infinityWell, dpkg could handle it fine, to be fair.11:08
infinityWe could do multiarch kernels.11:08
infinityAnd should, eventually.11:08
infinityBut we'd need some installer mangling to make that work.11:08
lilstevieah11:09
infinityThe path of least resistance for now is just two builds.  It's not world-ending.11:09
infinityJust a waste of 4 hours of a panda's time.11:09
lilstevieheh11:10
ogra_and given that one of the arches will go away anyway ...11:10
lilstevieyeah11:10
infinityogra_: I kinda want to rebootstrap armel as v4T as a community project (and maintain bi-arch/multi-arch with it)... Is that wrong? :P11:10
infinityI'd have to do it out of work hours, but I'm quite tempted, to be honest.11:11
ogra_well, the pi people would love you forever for that :)11:11
lilstevieinfinity: not really, that would solve the problems with all these *plug devices wanting ubuntu11:11
ogra_i would help :=11:11
ogra_:)11:11
WaltherF1now this sounds good, I'd be eager to help with all my knowledge as well11:12
lilstevieI would donate CPU time on my trimslice as a builder if that would help11:12
ogra_is theere actually much work to do apart from changing compiler flags and do a complete rebuild ?11:12
ogra_(and getting approval for the resource costs indeed)11:13
infinityogra_: And back out any braindead patches we have that assume thumb/vfp/etc.11:13
ogra_well, just re-roll debian then11:13
infinityogra_: Basically, revert to pure Debian compat for armel.  Which shouldn't be too hard.11:13
ogra_yeah11:13
davidminfinity, but why bother duplicating Debian, it's already there?11:14
infinityThe upshot is that we can do it on v7 hardware and just cheat.11:14
infinitydavidm: Because we get a TON of people who are grumpy when we say "use Debian for you < v7 device".11:14
ogra_davidm, there is no unity :)11:14
infinitydavidm: And when our official v7 solution becomes armhf, it might be a nice community service to offer a baselive v4T again.11:14
WaltherF1Gentlemen (and ladies?), I appreciate this. Thank you.11:15
ogra_indeed totally unsupported etc11:15
infinityWaltherF1: Nothing to appreciate yet. :P11:15
infinityWaltherF1: I need to get a lot of people to agree to my insanity before it could ever happen.11:16
lilstevieinfinity: heh11:16
infinityWaltherF1: (And the official answer for now on the Pi is, indeed, "use Debian")11:16
davidmand I suspect with the current lack of HW it's not going to happen11:16
WaltherF1infinity: But hey, I got you convinced to my insanity already, isn't that a start :P11:16
ogra_WaltherF1, even if it would happen that would be in a year from now or so11:16
infinitydavidm: Get me one 4U Calxeda box and we can build 4 arm* ports and keep up. :P11:17
WaltherF1yeah yeah, i'm not going to spoil this anywhere... But I can guarantee, (if)/when you'll announce the support officially, internet will shit bricks11:17
infinityWaltherF1: It would never be "official".11:17
infinityWaltherF1: It would be as unofficial as our powerpc port (which people keep trying to drop...)11:17
infinitySo...11:17
infinityI wouldn't hold your breath, is the take-home message. ;)11:18
infinityStill, I'd like to do it.11:18
ogra_we should also add nslu2 images back :P11:18
ogra_once we have such a prot11:18
ogra_*port11:18
xranbysneaky.. so when fedora moves to armhf we grab the legacy armv5 armel turf?11:20
ogra_xranby, its just a weird idea by a sleep lacking developer :P11:21
infinityxranby: Fedo--who?  Doesn't evern v5 user run Debian? :)11:22
infinitys/evern/every/11:22
xranbyat least some old fedora verson was preinstalled on the openrd boards11:24
xranbyalso some netgear NAS  like the Stora are based on a locked down fedora11:27
xranbyinfinity: but.. yeah.. i agree  I do run debian on my openrd machines and on my Stora :)11:33
=== Amaranth_ is now known as Amaranth
=== plars is now known as plars-vacation
=== Ursinha is now known as Ursinha-lunch
=== prpplague^2 is now known as prpplague
robbiewNCommander: ping16:13
robbiewinfinity: ping...any idea when we can expect to have an armhf image for folks to test out?16:15
infinityrobbiew: Desktop won't be for quite a while (hard to estimate while build-deps are being unsnagged), server might be doable (with some bits missing) in a week or so?16:16
infinityrobbiew: Very hand-wavy at this point, hard to commit until I can get a good overview of what I still need to manually bootstrap.16:16
robbiewack16:16
infinityrobbiew: Plus, I'll need to force some server bits to the front of the queue to make that estimate. ;)16:16
robbiewinfinity: cool, thx16:16
robbiewheh16:16
robbiewinfinity: looks like the ARM Supercomputer folks are willing to testdrive our bits16:17
robbiewso the sooner the better ;)16:17
infinityrobbiew: I'd like to testdrive theirs too!16:17
infinity(Man, that sounds bad)16:17
robbiewI'm working on that16:17
infinityrobbiew: I'll see what I can do about reordering and jamming some more stage-2 bootstrap bits in for server dep loops.16:18
infinityrobbiew: Of course, we won't have installable images for those folks (as I assume they use kernels we don't ship), but I also assume they're smart enough to figure out how to install minimal chroots and go from there.16:18
robbiewinfinity: yeah...given they've been using Ubuntu already...I think so :P16:19
infinityrobbiew: Feel free to keep giving me gentle prods over the next week or two.  I'll see what I can do about delivering most of server-ship.16:20
infinity(Hey, you already have MySQL, we're done, right?)16:20
dokoinfinity, robbiew: a debootstrappable buildd chroot would be appreciated as the first step ;-P16:20
infinity(Right?)16:20
infinitydoko: Yeah, yeah.  Working on it.16:20
dokocouldn't resist ...16:21
robbiewinfinity:cool...who should I talk to about producing an ARM HPC kernel for them...is that you all or #ubuntu-kernel folks16:21
infinitydoko: You can cheat and debootstrap --minbase from http://people.canonical.com/~ubuntu-archive/armhf/ for now, add ports.u.c to the sources.list, and install build-essential.16:21
robbiewThey need kernel scheduler and parameters tuned for single-threaded execution, as there is no user interaction or time-sharing in an HPC platform16:21
infinityrobbiew: Well, I assume it would be targetted specifically at their hardware as well.  You could talk to davidm about it.  I'm fairly sure it won't be his team that ends up doing the work, but he'll know who should. :P16:22
infinitydoko: --variant=minbase, that is. But you know what I mean.16:23
infinitydoko: But I'll get --variant=buildd working from ports fairly soon.  Need a kernel and a few other bits.16:23
robbiewinfinity: ack, thx16:23
dokoinfinity, minbose is supposed to work?16:24
robbiewinfinity: yeah...could be something done as part of a Canonical/BSC relationship16:24
infinitydoko: minbase works from people.16:24
dokoahh16:24
infinitydoko: Probably not from ports yet.16:24
infinitydoko: buildd would theoretically work from people too, except my stage-2 archive doesn't take overrides into account, so sections/priorities are wrong, and debootstrap has a bit of a hissy fit about the whole deal.16:25
infinitydoko: But yeah, if you want to play, minbase from people, add ports at the top of sources.list (and keep people), install build-essential, have fun.16:26
ogra_well, shouldnt -core be buildable already ?16:29
ogra_we could probably make A1 with it ;)16:29
ogra_(for hf that is)16:29
infinityogra_: Not yet.16:30
* infinity tests this statement.16:32
ogra_if minbase and apt work i would have guessed it should be buildable16:32
dokoinfinity, did you ever get a reply from apw about linux-libc-dev built from separate source?16:33
infinityNote that I said to debootstrap minbase from my stage-2.16:33
infinityNot from the archive. ;)16:33
ogra_(which i would expect to be reqs for the buildd setup)16:33
ogra_ah16:33
ogra_that one i missed indeed :)16:33
infinitydoko: linux-libc-dev is buildable out-of-band with DEB_STAGE=stage116:33
infinitydoko: Works well enough for bootstrapping.16:33
infinityogra_: Yeah, minbase is still missing a few bits in the archive.  Will sort it.16:34
infinityOh, or not just yet.16:35
infinityNeed ruby1.9 to build libselinux.16:35
infinityLa la la.16:35
infinitySorting slowly. ;)16:35
ogra_no hurry, i wasnt actually serious about A1 anyway16:35
dokolooks like the bootstrap is stalled now, universe packages start building16:36
=== Ursinha-lunch is now known as Ursinha
NCommanderrobbiew: pong16:40
infinitydoko: I'm revisiting some snags.16:40
robbiewNCommander: nevermind...all sorted ;)16:41
NCommanderrobbiew: k :-)16:42
robbiewNCommander: got another question for ya...any reason why hdf5 isn't tagged for armhf builds? ->    https://launchpad.net/ubuntu/precise/+source/hdf516:44
robbiewhdf4 is, but our supercomputer friends in barcelona would like hdf5 :)16:44
NCommanderrobbiew: its probably in P-a-s16:44
NCommanderlooking16:45
ogra_it probably simply missesw armhf in debian/control16:45
ogra_*misses16:45
infinityWhat ogra_ said.16:45
NCommanderogra_: any all16:45
infinityIt's not P-a-s.16:46
NCommanderhrm, no its got a armhf build record16:46
NCommanderrobbiew: https://launchpad.net/ubuntu/+source/hdf5/1.8.4-patch1-3ubuntu1/+build/296328216:46
infinityOh, then it just needs to build. :P16:46
robbiewah16:46
robbiewcool...nevermind :)16:46
NCommanderrobbiew: always use the source overview page to look at what builds are doing what where: https://launchpad.net/ubuntu/+source/hdf516:47
ogra_robbiew, http://qa.ubuntuwire.org/ftbfs/primary-precise-armhf.html might also be intresting for you in case you look for failed builds (not actually instresting at this stage of the bootstrap but will be soon)16:48
infinitydoko: We have a fair few packages that need -lm fixes.  You could queue those up for uploads right after the freeze is over, if you're feeling helpful. ;)16:48
robbiewNCommander: thanks...I **thought** I did :/16:48
NCommanderrobbiew: you were looking at the percise version page. LP's URLs for finding stuff are kinda irritating16:48
infinityrobbiew: having the release name in the URL is a drastically different page.  Annoys me to no end.16:49
robbiewindeed16:49
NCommander+1 infinity16:49
infinity(Try having the release name in a bug URL and see how that screws with your head)16:49
infinityLike, being unable to target things, etc.16:49
dokoinfinity, hmm, new -as-needed ftbfs?16:58
infinitydoko: No, it was the "something used to needlessly link against libm and doesn't anymore" thing.17:00
infinitydoko: Remember that landed late in the oneiric cycle and then seb128 rapidly reverted it?17:01
dokoahh, the seb128^Bgnome mess17:01
infinitydoko: Well, it's landed for good now.17:01
infinitydoko: https://launchpad.net/ubuntu/+source/gnome-desktop/1:2.32.1-0ubuntu6/+build/2962501 <-- For example.17:02
infinityI'm working on the "stuff that's FTBFS due to changes in if_packet.h" failures right now.17:02
dokoahh, like klibc?17:15
infinitydoko: klibc, iproute... I'm sure I've seen a couple more.  Will comb.17:16
infinityOf course, I also need a nap at some point.17:17
infinitypython distracted me last night.17:17
dokoctypes?17:17
infinityctypes is hilariously broken.17:17
infinityYeah.17:17
dokolool sent a patch17:18
infinityYeah, I'm building with it.17:18
infinityBut that's just papering over awful code.17:18
=== zyga is now known as zyga-afk
=== WaltherF1 is now known as WaltherFI
=== zyga-afk is now known as zyga
=== Jack87 is now known as Jack87|Away

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