/srv/irclogs.linaro.org/2012/03/25/#linaro.txt

hieuletrunghi, i want to build linaro-nano image, is there any guideline existed?09:16
=== koud_ is now known as koud
=== arun__ is now known as arun
zygagood evening19:49
mwhudsongood morning19:54
fsvendgood night19:57
fsvendanyone here recommend any affordable jtag for home usage. target is the samsung exynos 4210. i've used the bdi3000 at work for ppc targets, but it's out of reach for my "home budget".. i'm considering the amontec (+openocd), flyswatter2 or maybe a cheap ebay variant of the realview ice.20:05
rsalvetiflyswatter2 +openocd would probably be the best choice 20:07
mruI'm sure prpplague would agree too :)20:07
mruit's too bad lauterbach is so expensive20:08
zygamwhudson, hi, how are you doing?20:10
mwhudsonzyga: ok20:10
fsvendrsalveti and mru: ok. thanks for the input..guess i'll go for the flyswatter2 then.20:11
mwhudsonzyga: not too much email over the weekend, that's always good :-)20:11
zygamwhudson, I'm just catching up20:12
zygamwhudson, was on a trip in barcelona20:12
zygamwhudson, and bricked my efika20:12
mwhudsonoops20:12
zygamwhudson, well I'm on precise on efika20:13
zygamwhudson, with the old maverick kernel20:13
zygamwhudson, and udev loops at 100% till restarted (spamming syslog about unimplemented features)20:13
mwhudsonugh20:13
zygamwhudson, so I've decided to echo manual > /etc/init/udev.override20:13
zygaand that was silly20:13
zygaand now I know ;)20:14
zygafun fact: udev is pretty essential nowadays 20:14
steevzyga: which old maverick kernel?  you should be using linux-legacy from git://github.com/genesi/linux-legacy.git ; i added the syscalls udev needs long ago20:26
zygasteev, nice to hear that, I'd love to try that kernel20:26
zygasteev, I cannot recall now, I need to figure out which micro image to download to recover first, I strongly believe that I was running the current latest kernel available in the genesi archive20:27
zygazf20:27
steevzyga: not sure, i'm running precise just fine, though i started from ubuntu-core, not one of the linaro packs20:27
zygasteev, I'd like to try to build a LAVA master image for my efika but last month our releases were not really building at all, let alone boot20:27
zygasteev, I've started with the maveric image and never reinstalled, just upgraded20:28
zygasteev, could you point me at something that I could boot from SD to dig deeper?20:28
steevzyga: http://www.powerdeveloper.org/platforms/efikamx/linux the 20111010 image, when it prompts you to reinstall, hit N, it will drop you to console, just login as oem/oem20:29
zygasteev, thanks, let me download that now20:29
zygasteev, is there a better kernel I could upgrade to without building it myself?20:30
zyga(to resolve that udev issue)20:30
steevzyga: the one on that sdcard should20:30
steevso you could be lazy and copy pasta it over20:31
zygasteev, hmm, I think that's the kernel I'm running20:31
steev.26 definitely has the syscalls20:31
zygasteev, let me download and double check20:31
zygaI'll probably report an interesting bug then :)20:32
zygalet me fetch my little efika from the bag downstairs20:32
=== phantone is now known as phantoxe
mwhudsonpfefferz: are you around?22:01
=== michaelh1|away is now known as michaelh1
mwhudsonzyga: around still?22:28
zygayes22:29
zygathe nice movie I just watch has finished22:29
mwhudsonzyga: i'm thinking about how to transitioned to owned bundle streams22:29
zygaand now I'm copying efika image to try to boot my laptop again22:29
zygamwhudson, ah22:29
mwhudsonzyga: is there a wat to move bundles en masse between streams?22:29
zygamwhudson, worthy problem, do you have any ideas?22:29
mwhudson(other than sql, obviously)22:29
zygamwhudson, well, bundle.stream = ... ;-)22:29
mwhudsonyeah ok22:29
zygamwhudson, maybe the problem is different22:30
zygamwhudson, maybe there's a stream you want to chown?22:30
mwhudsonzyga: so my not-very-thought through plan was this:22:30
mwhudsonscratch that22:30
zygamwhudson, also recall we have denormalized data22:30
mwhudsonzyga: the problem seems to me to be coordination between moving data around the dashboard and with the scripts that submit jobs22:31
zygamwhudson, true22:31
mwhudsonzyga: BundleStream.pathname?  or something else?22:31
mwhudsonone approach would be:22:31
zygamwhudson, hmm? I'm not sure what that last question is about22:31
zygamwhudson, I think you need to fix the scripts first, right?22:31
mwhudsonzyga: the denormed data22:31
mwhudsonfor each active anon stream22:31
mwhudson(a) create a owned stream22:32
zygamwhudson, I think there's a copy of the bundle stream somewhere, but just the primary key22:32
mwhudson(b) move all bundles from anon to owned stream22:32
mwhudson(c) update the script22:32
mwhudson(d) when the updated script has been deployed, move all bundles from anon to owned stream again22:32
zyga(d) fix denorm data22:32
mwhudsonor variants on that22:32
zygahmm22:32
zygaI think solving c is essential and actually a22:33
zygamoving data around later is really easy in comparison, you can do that in 10 minutes over python shell22:33
mwhudsoni guess22:33
zygaand a is a coordinate/political problem22:33
zyga(read: convince others to follow your agenda)22:33
zygayou cannot do that in repl22:33
zygathat's always harder22:33
zygado we know who submits jobs today?22:34
mwhudsoni think mps in the right projects will be reasonably quickly accepted22:34
mwhudsonthat's an excellent question22:34
zygamps22:34
zyga?22:34
mwhudsoni know of android, ci.linaro.org and the image tests22:34
mwhudsonmerge proposals22:34
zygaah22:34
zygamwhudson, I agree22:34
zygamwhudson, although I fear those are on git.* and are pretty obscure in location22:35
zyga(last time I've checked an essential infrastructure script lives in +junk)22:35
zygamwhudson, anyway22:35
zygamwhudson, I'd run a select over all jobs and group + count by submitter22:35
zygaI suspect we may still have a long tail22:35
mwhudsoni know where the android * kernel ones are22:35
zygado you want to get rid of all anonymous streams?22:35
mwhudsonyeah, that's a good idea (the count)22:35
zygamwhudson, that's good, I think we need a plan to coordinate this, (at least the nontechnical parts)22:36
zygamwhudson, it would also help22:36
zygamwhudson, if we did this for everyone else by fixing how our remote api works (read: it's bloody hard to use right)22:36
mwhudsoni think aiming for 100% is probably fruitless, i only really care about the automated stuff for now22:36
mwhudsonzyga: yes, i'm trying to develop a plan :-)22:36
zygamwhudson, that's a good goal I hink22:36
mwhudsonif you're rather talk about this when it's not sunday for you i could totally understand that :)22:37
zygamwhudson, in very high level steps: 1) make it technically possible (done) 2) move over the bulk of automated submissions 3) fix the usability issues in client tools 4) talk about it to try to convince the rest 22:37
zygamwhudson, no that's fine22:37
zygait's Monday22:38
zygasummer time22:38
mwhudsonheh22:38
mwhudsoni guess22:38
zygageez, I picked as slow card22:38
zyga2GBs copies forever22:38
zygaI have to find a sheet of small stickers on amazon22:38
zygaand label my cards "crap" "good"22:38
mwhudsonheh22:41
mwhudsoni'm sure this is very naive, but i don't completely understand how flash memory is the basis for both the most awesome (ssds) and most terrible (sd cards) storage22:41
zygamwhudson, it's like booze, they sell all kinds 22:42
mwhudson:)22:42
zygamwhudson, we just get the discount aisle22:42
mwhudsoni want the moet & chandon of sd cards then22:42
steevzyga: sorry my connection dropped, any luck?22:43
zygasteev, hi, I'm glad you're back22:44
zygasteev, well that depends on my SD card, still on the way, I'll let you know very soon22:44
zygamwhudson, what's the block size you use to copy the image to a card?22:44
zygamwhudson, I used to say=whateverM 22:44
zygamwhudson, then I standardized on 10M22:45
zygamwhudson, but now I'm worried that 10M is wrong and 4MB is less wrong22:45
mwhudsonzyga: i don't remember tbh22:45
zygamwhudson, but none of that is scientific22:45
mwhudsoni haven't actually dded a card for ages22:45
zygamwhudson, really?22:45
mwhudsoni think anything over about 4k probably doesn't make much difference?22:45
zygamwhudson, I guess we don't test the dispatcher too often ;-)22:45
mwhudsonyeah, i have a master image that works22:45
zygamwhudson, oh, it does22:45
zygamwhudson, <1M is much slower from what I see22:45
mwhudsonand i don't want to touch it until i have to :)22:46
zygahmm22:46
zygaI wonder what's the size we use to put the test image on the card?22:46
mwhudsonhuh?22:51
mwhudsontar x doesn't take a blocksize22:51
mwhudsonor do you mean something else22:51
mwhudsonzyga: http://paste.ubuntu.com/899765/22:51
zygamwhudson, ah, now I want for the last 3 months22:51
mwhudsonyes, on that22:52
zygamwhudson, deepti is probably automatic as well22:52
mwhudsonyeah, i think the counts for plars and deepti are old automated submissions22:53
* mwhudson edits sql over a high latency link22:53
mwhudsonplease excuse any screaming22:53
mwhudsonzyga: http://paste.ubuntu.com/899771/ <- in 201222:53
zygamwhudson, I need to show you something one dy22:54
zygamwhudson, right22:54
zygamwhudson, so fixing auto submissions is 80%+ of the problem, good22:54
zygamwhudson, chithu is probably also automatic22:54
mwhudsoni'm not sure22:54
mwhudsonif we had a per-user job view...22:54
zygamwhudson, that raises another interesting concern22:56
zygamwhudson, since the vast majority of our jobs are done by robts22:56
zygamwhudson, we should support expression that better22:56
mwhudsonyeah22:57
zygamwhudson, I'm not sure how exactly but it feels wrong to have a fake user 22:57
zygaand no real trace22:57
mwhudsoni was thinking of an 'automated' flag on testjob or something22:57
mwhudsonbut that was just a fleeting though22:57
mwhudsont22:57
mwhudsonblargh blargh too many problems :)22:58
zygait'd be a start to some extent but then who can use it? is it everyone?22:59
zygaI think there's a model lurking22:59
zygathat model could tell us something about the automated bit22:59
zygamaybe even fix it if it's broken but that's probably too good to be true22:59
zyga\o/22:59
zygacopying done22:59
zygawhooping 1MB per second22:59
zygain 4M blocks22:59
zygaI sure hope reading's faster23:00
zygaI think I'll look at basic SD perf test for candidate cards for LAVA 23:00
zygasteev, what's the user/password on the 1010 recovery image23:02
steevzyga: oem/oem23:02
zygathanks23:02
zygaso this is running 2.3.61.14.26-efikamx23:02
zygalet me mount my rootfs and compare23:02
zygaand fix udev override23:03
steevokay.  i'd be curious since i have a smartbook here that's running precise; what kernel version are you using on the rootfs?23:05
zygasteev, I'm not sure if it's my fault but I always get /dev/sda2 is going to be checked on next reboot message and no actual checking, do you know if the script that does that from initrd is somehow customized and I could have wiped the original with precise upgrade?23:05
zyga2.6.31.14.20-efikamx23:05
zygaah23:05
steevmmm, no, we modify maybe 4 packages23:05
zygastrange23:05
zygalet's see23:05
steevzyga: ah, yeah 2.6.31.14.24 should be the first release of the kernel with sys_accept4 support.  https://github.com/genesi/linux-legacy/commit/a84fac75f38de592e530a2f9f982d7aafec6c28823:06
zygasteev, is that packaged anywhere?23:06
zygasteev, a quick apt-cache search tells me it's not in the ubuntu archive or the genesi repo I've got23:07
steevmmm, doesn't look like it's in packages.efikamx.info.  i don't have any real space i can uploaded23:07
steevupload it*23:07
zygasteev, can you get it up there?23:08
steevzyga: not for a bit, the server is being moved tomorrow23:08
zygaI know there's a 3.x coming but I'd rather have a system that works without this annoying udev restart (I don't even know why restarting udev fixes the CPU loop)23:08
steevzyga: not sure either.  are you using precise hardfloat?23:09
steevif so i can possibly upload the 2.6.31.14.27 i'm using on my smartbook23:09
zygasteev, I may be ignorant but arm-linux-gnueabi does not sound hardfloatish, right?23:10
zygacan I somehow magically (multiarch) upgrade to hf image?23:11
steevzyga: i honestly have nfi23:11
zyga(I could reinstall, I only care about ~ anyway23:11
zygaI can see there's a preview hardfloat image on powerdevelopers.org23:11
steevthat's debian23:11
steevi'm working on a precise one, but i have so little free time these days23:12
zygasure, thanks for doing it23:12
zygamarkos_, is there a chance for any 2.6.31 release before 3.x is out?23:13
steevzyga: of precise?23:14
steevor you mean of the kernel? because yes there is, we're working on an hdmi bug we've run into23:14
zygasteev, well, of just the kernel that would make udev happy23:14
zygaahh23:14
zygaexcellent than23:14
zygathen23:14
* mwhudson wonders how token management works for ci.linaro.org these days23:14
zygamwhudson, I think I know23:14
steevinterestingly enough, the bug exists in 3.2 as well 23:14
zygamwhudson, you don't want to know ;-)23:14
zygamwhudson, we need to fix our token story, it's so hard to use and so easy to screw up 23:15
mwhudsoni _think_ the job that submits stuff to lava runs on the master as an account that has a token installed23:15
zygamwhudson, and keyring sucks in practice23:15
mwhudsonzyga: yeah23:15
mwhudsonyes yes yes23:15
mwhudsonthat part was disappointing :(23:15
zygamwhudson, I think jobs are going from some cloud things and the token is in ~/keyring.cfg23:16
zygaor something equally ugly23:16
mwhudsonyeah23:16
zygamwhudson, I wrote a silly blueprint-in-a-bug about this a few weeks ago23:16
zygamwhudson, I wish I had the time to implement that23:16
mwhudsoni don't know, i think the approach ci.l.o takes is ok23:16
mwhudsonat least it keeps the token away from the code being built23:17
zygamwhudson, yeah, my bug was about the loops people have to hoop through to use this23:17
mwhudson(modulo general jenkins scariness)23:17
zygamwhudson, (read: how hard it is to just get a token and use it properly today)23:17
mwhudsonyeah23:17
mwhudsoni agree about that23:17
zygawith all the URL crap, keyring woes and got-to-page-and-copy stuff23:17
mwhudsonbut it's not what i'm worrying about just now23:17
zygaI think they are fine for the moment (jenkins)23:18
zygaand nothing we change will make that better 23:18
zyga(unless we break the current code, that'd be bad)23:18
mwhudsonzyga: afk for a bit23:34
zygamwhudson, I guess I'm off 23:34
zygamwhudson, talk to you tomorrow 23:35
mwhudsonzyga: good night23:35
zygamwhudson, I'll work on the release leftovers and upgrade23:35
zyganight everyone!23:35
cjohnstong'nite23:36

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