/srv/irclogs.linaro.org/2011/08/24/#linaro.txt

* mwhudson smirk's at michaelh1's "don't use a random mix of -f optimization flags, just use -O2 or -O3" mail00:53
* mru usually throws in a few -fno-foo as well00:54
mwhudsonzyga: still here?01:33
=== asac_ is now known as asac
michaelh1mwhudson: it's more 'please qualify your use of flags as they seem strange/random.'01:50
mwhudsonmichaelh1: it just reminded me of talking to andrew at the connect about codesourcery getting contracts to determine the optimal set of compiler flags01:51
mwhudsonand him saying that they would go away and crunch some data and then 90% of the time say "-O3"01:51
michaelh1mwhudson: yeah, there's even a tool called acova that will run a genetic algorithm to figure it out for you...01:51
mwhudsonand 90% of the rest of the time it would be -O2 or -Os01:51
michaelh1mwhudson: you get some strange things like a small increase in code size causes a hot loop to no longer be cache aligned and run much slower...01:52
mwhudsonyeah01:52
mwhudson-Os was faster than -O3 for me, at some point on some code on some machine i no longer remember any details of01:52
mwhudsonbut still, randomly chucking -f flags around just makes you look like a gentoo user01:53
michaelh1heh01:55
rsalvetitgall_foo: I updated the form during morning 02:08
rsalvetibut unfortunately you got the broken one02:09
rsalvetibut still, it's supposed to work02:09
rsalvetiI'll test it here and see if I can reproduce it02:09
=== qiyao is now known as yao
Dr_Whoevening02:28
=== ScottK2 is now known as ScottK
Taiten987anyone suggest a lp bug number for the chromium-browser ftbfs on arm?02:45
* michaelh1 looks02:46
michaelh1Taiten987: I can't find it I'm afraid.02:50
Taiten987michaelh1, thanks anyway, I recall I saw it once but can't find it anymore02:51
michaelh1Taiten987: yeah, it was toolchain related and a month or two ago02:51
Taiten987michaelh1, sounds like not so easy to fix in few days :(02:53
rsalvetiAmaranth: ouch, seems unity is not even starting with latest compiz =\04:23
rsalveti*** glibc detected *** compiz: free(): invalid pointer: 0x0005b88c ***04:23
rsalvetiwhat is weird, I was able to install run it when I replaced the old one by hand04:24
rsalvetitime to debug04:24
rsalvetiAmaranth: new compiz-plugins-main seems to be the culprit 04:28
rsalvetiinstalled the older one and it worked fine04:28
rsalvetiAmaranth: seems to be related with the changes at the workarounds plugin05:14
ndeckiko: yep. this SMP bug is a mess. well, at least it's good that we could collect so much info... now it would be better if we could actually fix it... we didn't find it so far (in real products) because we don't use USB host but OTB... but that's clearly a problem that needs a fix.. 06:23
janimondec, doesn't the bug appear even without USB involved?06:57
ndecjanimo: what do you mean? isn't the bug about USB perf going down ?06:58
janimondec, last I looked (two weeks ago) it seemed SD cards were affected too IIRC06:58
janimoand USB speed going up even if pinged via wifi, not only via eth which is on USB as well06:59
ndeci don't think that's true. that was supposed at some point, but 'someone' reported that it was not correct06:59
janimoand that it was not even panda related but a generic scheduling bug. I have not kept up with development since though06:59
janimook06:59
ndecwhat i mean is that for our product mass storage is through OTB, not HOST. and we have never seen such issues. it's only on panda which exposes host.07:00
=== michaelh1 is now known as michaelh1|away
=== michaelh1|away is now known as michaelh1
=== agreen is now known as agreen-away
ssvbronynandy: hi, are you using origenboard?07:17
jcrigbyrsalveti, imx51evk u-boot seems to be broken07:24
jcrigbydisabling dcache fixes the problem07:24
jcrigbyand then the next problem is that mmc still appears to be broken, (no partitions) but I see the driver is loaded in dmsg07:25
zygare07:29
zygagood morning07:29
=== jk-- is now known as jk-
=== agreen-away is now known as agreen
fabolag: could you triage bug 823313 also?08:25
ubot2Launchpad bug 823313 in linaro-android "Android LEB fails to mount system and user partition interminttently " [High,New] https://launchpad.net/bugs/82331308:25
lagfabo: plang is going to work on it when he gets back from vacation08:27
lagfabo: It requires a JTAG08:27
fabook08:28
fabomabac: thanks for the review!08:34
mabacfabo, np, not done yet though :)08:34
asaczyga: how do we get a new bundle stream=?09:26
zygaasac: you can use lava-dashboard-tool to make one09:26
zygaasac: with the "make-stream" command09:27
asaczyga: also on validation.linaro.org ?09:27
zygaasac: lava-dashboard-tool make-stream /some/pathname/09:27
zygaasac: yes09:27
zygaasac: on any installation09:27
asaczyga: do i need authentication?09:27
=== michaelh1 is now known as michaelh1|away
zygaasac: only for non-anonymous streams09:27
asacinteresting09:27
asaczyga: is authetntication sorted out?09:27
zygaasac: anyone can create anonymous streams09:27
zygaasac: yes09:27
zygaasac: if you authenticate (the same way with scheduler, using add-auth) then you can create personal and team streams, both public and private09:28
zygaasac: all the data is inferred from the pathanem so there are no extra arguments to specify09:28
* zyga pushed initial image model merge proposal09:29
zygaasac: we are getting image-aware LAVA soon :-)09:29
asacawesome09:30
asaczyga: image == artifact?09:30
zygaasac: more than that, the details are here: https://code.launchpad.net/~zkrynicki/lava-server/lava-data-store/+merge/7268909:30
asaczyga: whats the ppa:... line to add and what package to install?09:42
mwhudson__ppa:linaro-validation/ppa probably & lava-dashboard-tool09:49
zygaasac: ?09:53
zygaasac: ah right09:53
zygamwhudson__: hi :-)09:53
zygamwhudson__: I started reviewing your changes09:53
mwhudson__zyga: cool09:53
* mwhudson__ is not really here09:53
zygamwhudson__: ok :)09:53
mwhudson__zyga: for the project model, do you envision android and ubuntu based images being different projects?09:54
zygamwhudson__: perhaps09:55
zygamwhudson__: or having Linaro prject with ubuntu, android subs09:55
zygamwhudson__: I need to give it some more thought09:55
ndecagreen: hi09:55
asackk09:55
ndecsorry about yesterday. 09:56
zygamwhudson__: one quick question, do you want to use the query string or more than that in https://code.launchpad.net/~mwhudson/lava-dashboard/jsonp-bundle-data/+merge/7266109:56
mwhudson__zyga: "perhaps" seems a good answer :)09:56
agreenndec: hi; no worries09:56
mwhudson__zyga: i want to use the trailing part yes, is that what you mean?09:56
ndecwhen is a good time now to talk about this kernel? 09:56
asacmwhudson__: zyga: the package lava-dashboard-tool doesnt exist ;)09:56
zygamwhudson__: no, I want to understand if you want JUST the query string09:56
zygaasac: it does, did you add the right ppa?09:57
asacsorry09:57
asaci didnt run update ;)09:57
zygaasac: ah09:57
asacadd-apt-repository should really do that ;)09:57
mwhudson__android build goes to /permalink/bundle/$sha1/json?callback=.....09:57
zygamwhudson__: right now the regular expressions you added catch more than the query 09:57
zygaah09:57
asacok here it coes09:57
asaccomes09:57
zygamwhudson__: hmm09:57
agreenndec: let me ping jassi09:57
mwhudson__zyga: query string is not matched by urlconf09:57
mwhudson__aiui09:57
zygamwhudson__: oh... right09:57
zygamwhudson__: hmmm... I don't know how I feel about that, it looks somewhat ugly, is the point to get to JSONP for a bundle?09:58
mwhudson__zyga: yes09:58
mwhudson__zyga: where is the ugly?09:58
zygamwhudson__: well, previously it was a permalink 09:58
mwhudson__it's like launchpad.net/people/me/+sshkeys doing something useful09:58
zygamwhudson__: now it is something I don't know how to name09:58
mwhudson__zyga: the first part is still a permalink of course09:59
zygamwhudson__: yeah but the permalink was based on a redirect...09:59
zygamwhudson__: ah09:59
zygamwhudson__: actually that makes sense09:59
zyga+109:59
zyganeat09:59
zygaI see it how09:59
mwhudson__:)09:59
mwhudson__btw09:59
asacmwhudson__: you are jenkins hero right? is it possible to inject info about the build to be displayed similar to "recent changes" here? https://ci.linaro.org/job/linux-linaro-3.0_omap2plus/09:59
mwhudson__the necessity for JSONP makes me cry09:59
zygamwhudson__: there is cross-domain safe request but it requires server magic and compliant clients 10:00
asacor even better an inline text banner instead of a folded lnk10:00
zygamwhudson__: you should hide now ;-)10:00
mwhudson__asac: no idea, tbh10:01
mwhudson__asac: i'm sure a plugin could do it10:01
mwhudson__asac: i can read the source to find out for sure, but not at 22:00, send me an email if you want me to poke more10:01
asacmwhudson__: thats fine. just eager me asking. sleep well ;)10:02
asaci have zyga still ;)10:02
zygaheh10:03
zyga:)10:03
mwhudson__woo results link on http://validation.linaro.org/lava-server/scheduler/job/67810:05
zygamwhudson__: cool!10:06
zygaasac: did you manage to create a stream?10:07
mwhudson__not on some of the others though.. wonder why10:07
mwhudson__lack of trailing newline, maybe?10:07
AmaranthSo I've been trying and failing to run linaro-media-create for a while now, getting a fun error10:07
Amaranthapt-get update is failing with FATAL -> Could not set non-blocking flag Bad file descriptor10:08
mwhudson__oh... probably a race10:10
Amaranthmaybe I just can't do this from an ubuntu livecd10:10
asaczyga: thats my nice 20  job ;) ... let me try10:10
Amaranthgoing to reboot and do this as a proper install10:11
* zyga goes to retest android10:15
agreenndec: Jassi's not around, how does 09:00 UTC tomorrow sound, or we can do it without Jassi now  -- assuming it's the conf code I sent before10:20
ndecagreen: i can't. i can earlier. anything between 06:00UTC and 08:00UTC10:22
agreenndec: fine, I have large amounts of 'earlier' now ^^ 06:00 UTC is fine10:23
ndecagreen: ok. i will send an invite with the call details10:23
agreenndec: thanks10:23
ndecagreen: obviously it will be without rob, unless he has problems to sleep...10:24
agreenndec: ok, can do it later too... what is Rob's timezone?10:25
ndecdallas. but let's have this call. and we can sync with rob irc later10:26
agreenndec: Jassi is here, would you rather do it now?10:27
ndecagreen: ok. let's do something now. we will see if we need more tomorrow.10:28
agreenok10:28
ndecagreen: i am in, same bridge as usual10:32
asaclool: do you know if we run flash-kernel during lmc ?11:28
salgadoasac, no11:29
mabachey salgado, good to see you11:30
salgadohi mabac! how's it going?11:30
mabacsalgado, fine thanks! you're just back from Europe? Did you have a good trip?11:30
salgadomabac, it was great. got back yesterday morning :)11:32
mabacnice!11:33
asacsalgado: do we have any requirements on how the vmlinuz file is named to find the right one for uImage? 11:33
salgadoasac, iirc the config of each board specify a name pattern for that11:34
mabacasac, it has to match 'vmlinuz-*-%(kernel_flavor)s' and the kernel_flavor is set per board in l-m-c.11:35
mabacasac, when I get to merging the hwpack v2 code, it is specified in the hwpack instead11:35
asacmabac: aha11:36
loolasac: we don't11:36
asacmabac: ok thats good11:37
asaci see omap411:37
c10udhello there, do you know if there's a way to buy Pandaboards?11:59
c10udi tried some months ago through Digikey but no luck11:59
c10udthis might be kinda offtopic, but i plan running your software on it so..12:00
hrwc10ud: if digikey do not have them then no one has iirc12:02
hrwc10ud: TI is working on hw update for panda - new ones will have omap446012:02
c10udbut how <somepeople> got them then? that's what i wonder :D12:03
hrwdefine somepeople12:03
c10udi don't know, people that hack on it, and write posts about it 12:04
hrwmost of them bought them when they were available12:04
hrwsome got them from other developers or companies12:05
mrujust place an order12:05
hrwand iirc there are few smaller distributors too - checked pandaboard.org?12:05
SteveMcIntyrec10ud: it's quite common for the small dev boards to run out quickly, they're normally only made in small quantities12:05
mruit will arrive eventually12:05
c10udmru, tried twice, no luck12:05
mruwhere did you order from?12:05
mruand how did it not work?12:05
c10udmru, digikey, order expired -- twice12:05
c10udafter, like months12:05
mruwhen was this?12:06
c10udhrw, yea, but everyone has 0 in stock12:06
mrumost places allow to order nonetheless12:06
c10udmru, a few monhts ago, then i got busy with exams and got pretty frustrated so i didn't try further12:06
mruand will ship as soon as stock arrives12:06
c10udSteveMcIntyre, then is there some similar board that can be bought? 12:07
mruask in #pandaboard, someone there should know if there are production issues currently12:07
mruthe pandaboard should be in production12:07
SteveMcIntyrec10ud: I just bought several imx53 quickstart boards12:07
mruunless there's been an earthquake at the factory or something12:07
SteveMcIntyrebut they're often difficult to get hold of too... :-/12:07
c10udheh, that's my concern: i get one..then i need then and what? i wait..years'12:08
c10ud?12:08
mrulead time on pandaboards should be a few weeks at most12:08
c10udwas thinking about panda because it seemed the one with nicer hardware and user friendlyness12:08
mruunless demand has spiked12:08
SteveMcIntyrec10ud: depending on what you want, yes12:08
mrugo ask in #pandaboard, there are people there who know12:08
SteveMcIntyrepanda is nice for some things, quickstart is good for me with native sata12:09
c10udnow i'll try to ask in their channel, i tried here because i saw a blogpost on a new linaro rel and wondered: how do they get hardware? 12:09
SteveMcIntyrec10ud: we bought lots of them, starting ages back :-)12:09
ChiThuhi12:09
c10udSteveMcIntyre, oh, that's a nice info -- i didn't think about SATA!12:09
ChiThuzyga, hi12:09
mruSteveMcIntyre: which was why nobody else could buy them12:10
SteveMcIntyremru: yeah, maybe :-/12:12
c10udSteveMcIntyre, sorry for bothering: where are you ordering this imx53 board from? it could be interesting for me (!)12:14
rsalvetijcrigby: hm, then let's try to get some help from eric or the freescale lt12:16
rsalvetifabo: why did you remove the milestone from bug 784176?12:34
ubot2Launchpad bug 784176 in linux-ti-omap4 "CPU_FREQ support is broken for OMAP 3 and OMAP 4" [Medium,Confirmed] https://launchpad.net/bugs/78417612:34
rsalvetifabo: I just use the milestone so we can track later that this is not yet fixed, when producing the release notes12:34
rsalvetiand it's fine to keep postponing it, like what is done with ubuntu12:35
faborsalveti: how it could be targeted to 11.08 when it isn't fixed upstream?12:36
rsalvetifabo: at the distro we always can fix it in a different way when it's needed, but even if not yet upstream, it's good to track what bugs are affecting a milestone12:37
faborsalveti: could back to you in a moment, I'm on a meeting ;)12:38
rsalvetifabo: sure, np12:38
lagJesseBarker: Hey dude, you about?12:40
JesseBarkerlag, yup.  working group weekly call at the moment.12:41
lagJesseBarker: Ah, okay - can you give me a shout when you're done please?12:41
lagkrtaylor: Are you around atm?12:42
=== amitk_ is now known as amitk
jcrigbyericm|ubuntu, could you take a look at this config and see if anything is obviously missing for making the sd driver work: https://pastebin.linaro.org/208/12:52
jcrigbyericm|ubuntu, on mx51evk that is12:53
ericm|ubuntujcrigby, 'k12:53
jcrigbythanks12:53
ericm|ubuntujcrigby, and sd doesn't work w/ this config?12:53
ericm|ubuntujcrigby, what's the tree you are building against?12:53
jcrigbyI get a long pause at the end of booting then it drops into busybox saying the uuid does not exist so I am assuming it is an mmc problem.  Also catting /proc/partitions shows nothing12:54
jcrigbyericm|ubuntu, here is the git tree:12:54
jcrigbyhttp://git.linaro.org/gitweb?p=ubuntu/linux-linaro-oneiric.git;a=summary12:55
jcrigbyspecitically this tag:12:55
jcrigbyhttp://git.linaro.org/gitweb?p=ubuntu/linux-linaro-oneiric.git;a=commit;h=31b948b25768e2195d4c818c23567dc3b72bf36412:55
jcrigbyno12:55
jcrigbythis tag12:55
jcrigbyhttp://git.linaro.org/gitweb?p=ubuntu/linux-linaro-oneiric.git;a=commit;h=8f8db46fc623ade1e54b9ebef27542c5aa4f82fc12:55
* jcrigby should have stayed in bed12:56
jcrigbyrsalveti, on the u-boot issue I think I am missing a patch from aneeshv that basically does _NOT_ enable dcache for all plats, looking for that now12:56
hrwbug 823708 should interest toolchain team12:59
ubot2Launchpad bug 823708 in traverso "traverso version 0.49.2-4 failed to build on armel" [Undecided,New] https://launchpad.net/bugs/82370812:59
tgall_foomorning13:01
=== agreen is now known as agreen-away
ramanahrw: need a pre-processed file. 13:06
ramanahrw: ah take that back. :)13:06
hrw;)13:07
fabodmart: Bug 824545, could you comment which ones are failing?13:08
ubot2Launchpad bug 824545 in linux-linaro "Multiple hwpacks do not boot correctly with ext4 rootfs" [Medium,Confirmed] https://launchpad.net/bugs/82454513:08
hrwramana: installing gcc-snapshot now to check does it work with it13:09
janimohrw, ramana that is a very old bug :)13:13
janimoIIRC13:13
hrwjanimo: maybe 13:14
plarszyga: Hi13:16
hrwramana: gcc version 4.7.0 20110813 (experimental) [trunk revision 177733] (Ubuntu/Linaro 20110813-1ubuntu1) 13:17
hrwramana: this also fails13:17
plarszyga: so it looks like switching to uuid4() helped for a while... but the collisions are now back :(13:17
plarszyga: 6 collsions so far since yesterday13:18
plarszyga: so I was wondering... do we really even need the uuid? I'm not sure the purpose for requiring it is really all that important13:18
plarsChiThu: Hi, you around?13:31
ChiThuplars, hi13:31
ChiThuplars, how was the release ?13:32
JesseBarkerlag, ping?13:33
plarsChiThu: good, it's running live now and packages are done. with the exception of one that built last night while I was sleeping and I need to copy over13:33
lagJesseBarker: Hey buddy13:34
JesseBarkerlag, what can I do for you?13:34
lagJesseBarker: The person I wanted to introduce to you has left for the day (he's in India)13:34
ChiThuplars, great !13:34
JesseBarkerlag, new guy on the landing team?13:34
lagJesseBarker: After speaking with him, I think Kurt might be a better contact for him though13:34
lagJesseBarker: Which LT?13:35
JesseBarkerlag, thought you were saying you had a new guy on your team. nevermind.13:35
JesseBarkerlag, not sure when kurt starts his day (he's 2 hours ahead of me, but it's still only 08:30 where he lives).13:36
JesseBarkerlag, you still coming in for plumbers (if so, how much overlap will you have)?13:37
lagJesseBarker: I have the weekend after13:38
lagJesseBarker: Tuesday to Sunday I believe 13:38
lagJesseBarker: I guess in reality I only have Saturday 13:38
dmartfabo: I've added some more info to the bug description.  Basically, the config for CONFIG_LBDAF is fragmented rather than being applied globally -- it's been fixed in one or two places but not everywhere, and appears to have regressed for vexpress when moving to linux-linaro-oneiric13:40
JesseBarkerlag, maybe we can squeeze in a round (golf)....13:40
lagJesseBarker: Then a few rounds (not-golf) :)13:41
fabodmart: ok.13:41
lagJesseBarker: My flight doesn't leave until 9pm on Sunday either so ...13:41
mrua few rounds of hollow-point?13:41
ogra_arent sqeezed rounds actually oval ?13:41
faborsalveti: jcrigby: ^^^ bug 82454513:41
ubot2Launchpad bug 824545 in linux-linaro "Multiple hwpacks do not boot correctly with ext4 rootfs" [Medium,Confirmed] https://launchpad.net/bugs/82454513:41
JesseBarkerlag, ok, that gives us more possibilities (there's a course right next to the airport if need be :-)13:42
lagmru: What's hollow-point?13:42
mrulag: a type of bullet13:42
mrurather illegal too13:42
dmartfabo: any thoughts on the best policy for fixing it?  I don't know how useful CONFIG_LBDAF actually is for our boards, but it's required in ext4 filesystems by default (that requirement can be avoided at filesystem creation time)13:42
lagmru: Ah, I see. Similar effect to a dum-dum13:42
mrusame thing more or less13:42
lagmru: I guess we'll only be requiring those if we choose to play in "the hood"13:43
=== nytowl_away is now known as nytowl
fabodmart: it has been suggested previously. I guess it has been fixed by ericm|ubuntu for freescalle LT kernel.13:43
fabodmart: for the kernel configs changes, jcrigby is the man13:43
dmartWe can bypass this problem if we change l-m-c to generate cimpliant ext4 filesystems, but people may still hit it if they create filesystems by hand13:44
dmarts/cimpliant/compliant/13:44
faboconfirmed, eric has enabled CONFIG_LBDAF for linux-linaro-lt-mx513:46
jamestunnicliffeHi, has anyone got a recent (release candidate or todays) build of Ubuntu working on a PandaBoard?13:55
jamestunnicliffeThey are always boot hanging for me.13:55
jamestunnicliffe(ALERT!  /dev/disk/by-uuid/ce1b0748-95fe-4f65-b14a-09b6a415a6a3 does not exist.  Dropping to a shell!)13:55
pv_rmHi everyone, is there any of you who knows how to get a Exynos powered Origen Board? Is there any of you currently working on such board? thx13:56
mabacjamestunnicliffe, I have booted the latest nano build today with a few different hwpacks for Panda13:57
mabacjamestunnicliffe, that error usually means canning the sdcard and get a new one for me. or sometimes just mounting/unmouning the card in a pc works13:58
jamestunnicliffemabac: I just tried nano + panda hwpack from today and it failed.13:58
jamestunnicliffemabac: OK, I tried it on 5 cards with different hardware packs, so I am hoping they haven't all just died!13:58
mabacjamestunnicliffe, probably not :)13:59
mabacjamestunnicliffe, I can try with the current hwpack and see what happens13:59
jamestunnicliffemabac: Just did.13:59
krtaylorlag, pong?14:00
rsalvetiAmaranth: when, I just reverted the change at that plugin14:00
lagkrtaylor: Did you manage to speak with Sudip?14:01
hrwpv_rm: go to origenboard.org and order?14:02
krtaylorlag, no, for system metrics?14:02
lagkrtaylor: No, he's having trouble with gstreamer on Snowball14:04
pv_rmhrw, I ask here because both origenboard.org and/or insignal support hasn't helped so far, so I wanted to know if there is someone who has purchased for real one of the boards, and if there are active projects on it14:04
dmartfabo: do you know whether anyone has tried l-i-t 2011.08-0ubuntu1~linaro2 on maverick?14:04
lagkrtaylor: He's concerned that mplayer plays perfectly, but his smartt-player doesn't14:04
fabodmart: dunno. I guess most of us are using natty or oneiric. the image builders are using lucid.14:07
dmartfabo: for me, on maverick l-m-c always dumps a backtrace and prints "Invalid flag provided", even for --help:14:07
dmarthttp://pastebin.ubuntu.com/673849/14:07
asaczyga: will a stream be automatically created if i submit a job to a scheduler? do you know?14:07
jamestunnicliffemabac: I ran linaro-media-create --hwpack-force-yes --rootfs ext3 --mmc /dev/sdc --dev panda --binary /home/dooferlad/.cache/linaro/image-tools/fetch_image/snapshots.linaro.org/11.05-daily/linaro-nano/20110824/0/images/tar/nano-n-tar-20110824-0.tar.gz --hwpack /home/dooferlad/.cache/linaro/image-tools/fetch_image/snapshots.linaro.org/11.05-daily/linaro-hwpacks/panda/20110824/0/images/hwpack/hwpack_linaro-panda_20110824-0_armel_supported.14:08
jamestunnicliffetar.gz to get a nano image. Can you try the same?14:08
dmartThe issue is, customers are almost certainly not oneiric and probably not using natty yet14:08
pm215I still think the right answer to this is "customers shouldn't use l-m-c and we should provide prebuilt images"14:08
krtaylorlag, not sure how much help I'd be w/o snowball, has he tried another board?14:08
jamestunnicliffemabac: Or linaro-fetch-image --hardware panda --platform snapshot --image nano --build latest --hwpack panda --mmc /dev/sdc --rootfs ext3 if you don't want to type as much :-)14:08
krtaylorlag, I can contact him if you'd prefer 14:08
mabacjamestunnicliffe, sure 14:08
jamestunnicliffemabac: thanks14:08
dmartpm215: yes ... assuming we have the prebuilt images.  In my case, we don't14:08
asaczyga: ouch dashboard tool triggers kwallet for me14:09
asacthats wrong14:09
lagkrtaylor: He's gone for the day now14:10
fabodmart: we have some prebuilt images and we started to produce them14:10
lagkrtaylor: He's been working on it, but asked if I knew anyone who would be able to assist14:10
faboit isn't automated yet though14:10
dmartfabo: do we have anything for vexpress ... where should I be looking?14:10
zygaasac: that's lava-dev-tool and it uses the system-specific keyring to store your token14:10
lagkrtaylor: I said I'd speak to you (this was before I knew you'd worked together previously)14:10
fabodmart: http://releases.linaro.org/images/linaro-n/ubuntu-desktop/11.07/14:10
zygaasac: the stream will not be created automatically14:10
fabodmart: 11.08 are coming in the next days14:10
sjhill_something is up with the repo14:11
sjhill_platform/sdk14:11
zygaasac: why do you think storing the token in the system keyring is wrong?14:11
sjhill_Fetching projects:  95% (160/168)  remote: fatal: failed to read object c0f30688df7152202b032f5809b6b52d842ae09f: Invalid argument14:11
sjhill_remote: aborting due to possible repository corruption on the remote side.14:11
sjhill_fatal: protocol error: bad pack header14:11
sjhill_can't get past this14:11
sjhill_error: Cannot fetch platform/sdk14:11
pm215dmart: to the extent that non-linaro-developers need images which aren't prebuilt we should be increasing the set of images we prebuild and/or providing a web-ui that autobuilds them on demand14:11
asaczyga: i dont say that. i am saying that i am asked for a password through kwallet is wrong14:11
sjhill_linaro-android-11.08-release leb-panda14:11
asaczyga: ;)14:11
zygaasac: I never noticed how it integrates with kwallet, it depends on the host platform14:12
zygaasac: on bare systems it uses simple unsecure storage14:12
zygaasac: on gnome it uses gnome-keyring14:12
zygaasac: and on kde - kwallet (windows and mac have one framework each)14:12
asaci am on unity14:12
zygaasac: wow, that's strange, are you on oneiric or natty?14:13
dmartpm215: sure, we should aim for this ... but we need to make sure customers know.  Maybe we should stop releasing l-i-t for old ubuntu versions14:13
asaczyga: dont know14:13
zygaasac: do you have kwallet installed?14:13
asaczyga: whats the alternative14:13
asaczyga: obviously i have that installed14:13
asac;)14:13
zygaasac: it could be a bug in the library we use (python-keyring)14:13
asaclet me see if i can run make-stream on slave in jenkins14:13
zygaasac: if you feel like it please report a bug against lava-tool14:13
asaczyga: whats --dashboard-url for our validation.linaro.org?14:15
zygaasac: http://validation.linaro.org/lava-server/RPC2/14:16
asackk14:16
zygaasac: we ought to get "bookmarks" for those14:16
zygaasac: it would simplify life a lot14:16
zygaasac: note that for auth you should use ssl (so https://) and designate the user (https//user@validation...)14:16
sjhill_no one online who manages repo/git ?14:20
asacsjhill_: whats your prob?14:23
asacsjhill_: #linaro-android might have better attention to this 14:23
sjhill_ah14:24
sjhill_asac: my apologies14:24
sjhill_thx14:24
* zyga takes a break14:28
=== zyga is now known as zyga-afk
janimoppisati, thanks for the stap guidance :)14:30
=== mayday_jay is now known as mayday_jay-away
amitkhttps://launchpad.net/linaro-power/+milestone/2011.0914:37
asaczyga-afk: 14:38
asac+ lava-dashboard-tool make-stream --dashboard-url http://validation.linaro.org/lava-server/RPC2/ /anonymous/ci.linaro.org-linux-linaro-3.0_panda-omap2plus14:38
asacXML-RPC error 403: Junk after pathname: '/anonymous/ci.linaro.org-linux-linaro-3.0_panda-omap2plus'14:38
rsalvetijcrigby: do you know if that patch went upstream already?14:40
jcrigbyrsalveti, u-boot dcache bug patch?14:41
rsalvetiwonder why he would enable dcache for all plats14:41
rsalvetiyeah14:41
asaczyga-afk: can i disable this wallet thing?14:41
jcrigbyit is in patchworks but not applied upstream, I have applied it and it fixes mx5114:41
asaci want to get rid of it14:41
jcrigbyrsalveti, testing on panda now14:41
rsalvetijcrigby: great14:41
rsalvetijcrigby: and another bug for you: bug 82454514:43
ubot2Launchpad bug 824545 in linux-linaro "Multiple hwpacks do not boot correctly with ext4 rootfs" [Medium,Confirmed] https://launchpad.net/bugs/82454514:43
rsalvetirelated with config changes14:43
ppisatijanimo: i had a failure with one script14:43
ppisatijanimo: but everything else worked so, try it out :)14:44
jcrigbyrsalveti, is that what dmart was mentioning earlier? 14:44
rsalvetijcrigby: yes14:44
rsalvetijames_w: fabo: bug 83298414:45
ubot2Launchpad bug 832984 in linaro-image-tools "recent linaro-image-tools fails on maverick" [Undecided,New] https://launchpad.net/bugs/83298414:45
hrwramana: https://bugs.launchpad.net/gcc-linaro/+bug/809761 14:46
ubot2Ubuntu bug 809761 in oss4 "oss4 version 4.2-build2004-1ubuntu1 failed to build on armel" [Undecided,Confirmed]14:46
tgall_foowookey, ping14:47
=== rsalveti changed the topic of #linaro to: Linaro ARM Porting Jam Now! Grab your bug from http://bit.ly/mtV9cZ | Linaro 11.07: http://bit.ly/qgt8Qd | Logs at: http://irclogs.linaro.org/ | Ask Linaro: http://ask.linaro.org/ | ARM Porting bugs: http://bit.ly/mtV9cZ
=== mayday_jay-away is now known as mayday_jay
=== doko_ is now known as doko
* zyga-afk sprinkled his ankle15:12
=== zyga-afk is now known as zyga-ouch
zyga-ouch:-(15:12
zyga-ouchplars: I'll be away, about to go to the local hospital15:13
dmartjcrigby: The linaro-image-tools failure on maverick was a false alarm: it's a python-parted bug which is fixed in maverick-updates.  I've marked that bug as invalid.15:13
plarszyga-ouch: what? what's up?15:13
plarsoh15:14
plarsouch15:14
plarszyga-ouch: ok, talk to you later15:14
rsalvetidmart: great :-)15:14
rsalvetifabo: ^15:14
zyga-ouchplars: trying to figure out what to do15:15
zyga-ouchplars: I fell down the stairs15:15
zyga-ouchplars: my leg hurts badly, proablby it's just the ankle but I should check15:16
dmartrsalveti: for testing purposes, I ignore -updates and -security by default, because these change over time and aren't reproducible15:16
zyga-ouchasac: did you have a space between RPC2/ and /anonymous/... ?15:16
zyga-ouchasac: currently you cannot but we could easily force it to the insecure storage mode15:17
rsalvetizyga-ouch: ouch, good luck on that15:18
plarszyga-ouch: glad it was just an ankle15:18
plarszyga-ouch: I got him sorted15:18
rsalvetidmart: yeah, makes sense15:18
zyga-ouchit's kind of hard to go to the hospital without anyone to look after the kids though15:18
zyga-ouchwell, that's life15:19
hrw~curse bugs.launchpad.net for slowness15:19
asaczyga-ouch: i didnt have a / at the end of the bundle stream name15:20
asaczyga-ouch: at least plars said i need that. ... so trying again15:20
zyga-ouchasac: do file usability bugs please15:20
zyga-ouchasac: I suspect there are a lot of thorns there15:20
wookeytgall_foo: pong. You wanting a multistrap answer? I'll look in 10 mins15:41
tgall_foowookey, well yes some advise would be helpful tho I think I know more today then I did yesterday ...  related were you aware that the live-build community is looking to building something completely new to replace multistrap15:43
tgall_foowookey, no rfc even put together yet.   I was approaching them about my multistrap support patches for live-build, suffice to say that kind of went nowhere15:45
amitkmturquette: sent15:46
hrwhave a nice rest of day15:56
=== salgado is now known as salgado-lunch
wookeytgall_foo: hmm, OK. Did they say why they didn;t like the multistrap design? I hope it's not just because it's not python...16:41
voidspaceok, so how do I create a boot image for an i.mx53 from the 11.08 download?16:42
voidspaceMy i.mx53 has just arrived, it boots into lucid from their supplied 4gb sdhc card - but doesn't recognise the mouse or keyboard16:44
voidspaceso I'd like to try an updated version16:44
voidspaceI'm downloading 11.08 for i.mx53 now16:44
* rsalveti lunch17:08
voidspaceso, there's a google doc for installing 11.08 on i.mx5317:28
voidspacehttps://docs.google.com/spreadsheet/viewform?formkey=dHZNMHVkVnRqY2xTM0ctVEtIZWFrbEE6MA&ndplr=117:28
voidspacethe final command is:17:28
voidspacesudo linaro-media-create --dev mx53loco --mmc /dev/mmcblk0 --hwpack hwpack_linaro-lt-mx5_20110823-0_armel_supported.tar.gz --binary linaro-n-ubuntu-desktop-tar-20110823-3.tar.gz17:28
voidspaceI assume this creates the media card 17:29
voidspaceso I replace /dev/mmcblk0 with my actual media card 17:29
fabovoidspace: yes, you can check dmes to get the mmc device17:30
fabodmesg17:30
voidspacefabo:  apparently /dev/sde17:31
voidspaceit's sd rather than mmc - I assume that makes no difference17:32
fabovoidspace: linaro-media-create will ask you if it's the right device and you'll get the partitions table17:32
voidspacefabo: thanks17:32
voidspacestill downloading...17:33
fabovoidspace: most likely if you use ubuntu, it's mmcblk0. if you use debian, it might be sdX17:33
voidspaceubuntu server :-)17:34
fabo;)17:35
voidspacedbus.exceptions.DBusException: org.freedesktop.DBus.Error.ServiceUnknown: The name org.freedesktop.UDisks was not provided by any .service files17:37
=== zyga-ouch is now known as zyga
zygare17:41
plarszyga: hey17:42
ZeZuwb17:42
plarszyga: how's the ankle17:42
ZeZuugh, now i _really_ remember why i usually build everything by hand, my terrible luck with all buildscripts, even the ones 50 people have tested are doomed to fail locally17:44
zygaplars: not broken that's good17:44
ZeZu#define HAVE_MALLOC 017:44
plarszyga: sprains some sometimes be worse17:44
ZeZu#define malloc rpl_malloc17:45
zygaplars: our kids went crazy on the way to the hospital, my wife accidentally closed the car door while our son had his finger there17:45
zygaplars: both of us were waiting in the emergency room ;)17:45
zygaplars: fortunately it was nothing serious either17:45
ZeZuundefined reference to `rpl_malloc'   << FML17:45
zygaplars: as for me, ice + holding leg up for a while should do it17:45
zygaplars: I'll get back to work soon today17:46
zygaplars: did you see my initial MP for images/17:46
zygaplars: on the up side everyone in the waiting room saw my grey+orange ubuntu t-shirt :)17:49
=== nytowl is now known as nytowl_away
plarszyga: I saw the email, but I haven't even had a chance to look yet... been a busy day18:02
zygaplars: ok18:09
zygaplars: do you think we are ready to move the front page to the lava-server front pae?18:09
plarszyga: almost, we can change the layout and add things with the template override right?18:10
zygayes18:10
zygawhat would you like to change?18:10
zygabtw: I noticed the dispatcher package, great work18:10
mkedwardsZeZu: I apologize for not responding to your emails promptly18:11
mkedwardsI'm afraid I've been *crazy* busy18:12
plarszyga: oh, I didn't look yet, but do we have attribute support for the whole bundle? or just things under it?18:12
plarszyga: or for adding a bundle name, are we going to have to extend the schema?18:12
zygaplars: for just things under it, I imagined that the bundle would be largely an uninteresting container18:14
zygaplars: no, bundles already have a name but it is not stored in the json document18:14
plarszyga: ...and now it seems exactly the opposite :)18:14
zygaplars: it is specified when you put() it to the dashboard18:15
zygaplars: yeah, I was just thinking about how that changes things and how we should proceed18:15
plarszyga: ahhh, could we use that though? any restrictions on it?18:15
zygaplars: no, no restrictions18:15
zygaplars: any text will do18:15
zygaplars: I just noticed that lava-test was not released/packaged18:16
plarsI think asac wants to make a bundle called "!!# ZOMG KernEl RE$ul+s... and flying IGUANA's!!!@!"18:16
plarsis that ok?18:16
zygaplars: yes18:16
plarscool18:16
zygaplars: (this is one of the few things that are really free form while being designed to store typical filenames)18:17
zygaplars: ZOMG lava is a good name :)18:17
asaclol18:17
* zyga copies packages around18:18
rsalvetifabo: need to respin ubuntu-desktop and lt-panda-x11-base18:18
rsalvetialready tested the new rootfs and hwpack18:18
zygaplars: I'll work on the image a few more minutes and would like to look at packaging things 18:18
asacso first i want to be able to give a job a  name and a description18:18
plarszyga: lp still fighting you? I need to check dispatcher, I think it failed for most of them I tried to copy this morning18:18
rsalvetifabo: rootfs because of broken compiz-plugins-main, that blocks unity-3d on panda18:18
asacthat info gets - together with job info - inherited by the bundle18:18
zygaplars: actually it just copied 10 packages so it's really nice18:18
rsalvetifabo: and lt-panda-hwpack to disable broken sound support and generate the tools package18:18
rsalvetiso people can use perf18:18
asacso you always know which job a bundle came out of and you can give it a good name in our default bundlestream view18:19
plarszyga: oh, I thought all the packages were done, but maybe not propagated to all series18:19
rsalvetifabo: should I just update the links at the form?18:19
zygaasac: not all bundles are coming from jobs though18:19
asac2nd. on bundlestream view level - like http://validation.linaro.org/lava-server/dashboard/streams/anonymous/ci-linux-linaro-3_0/bundles/4af2fdb5368aa2cbf938fd78a3f1241f730b3da0/18:19
plarszyga: that reminds me... I need to mark the dispatcher packaging bp done18:19
zygaasac: I see what you want but we should either say that some things are mandatory and use that or be able to handle missing data18:19
asaci want to give the tests also a name ... like the lava test should read "Boot and Basic Health"18:19
asaciunstead of the UUID18:19
asaczyga: default bundle_name is uuid18:20
zygaasac: bundle have no name, uuid is object identity18:20
zygaasac: I want to have nice names too18:20
asacyou can set it optionally in a bundle; on top the job dispatcher will set that on the bundle18:20
asaczyga: thats understood18:20
asacbut i think it will work18:20
zygaasac: yeah18:20
zygaasac: we can make it pretty quickly18:21
zygaasac: I need to talk to mwhudson about one thing18:21
zygaasac, plars currently a lot of steam is spent trying to make sure the distributed nature of lava works 18:21
asaczyga: ok. i would like to move my CI job to already put in the nice parts in the job18:21
zygaasac: like scheduler and dashboard being two separate systems on (possibly) two separate servers18:21
asaczyga: but i guess you need time to think about schema changes?18:21
zygaasac: schema changes are easy we just need to coordinate with mwhudson 18:22
zygaasac: I'll grab him this evening and brainstorm how to do it18:22
zygaplars: we should try to do edge.validation.linaro.org soon, it would help us18:22
plarszyga: indeed... trying to get through the current hardware order first though18:22
zygayeah18:22
asacso18:22
asachttp://validation.linaro.org/lava-server/dashboard/tests/lava/18:22
asacthat view is kind of useless18:23
zygalots of work to do :)18:23
asachow about we display a job list of jobs that include that test?18:23
plarszyga: we could maybe do it in a VM with more ram, but I think we really want a second server eventually... if for no other reason than a hot standby18:23
zygaasac: that view will grow "subscribe to failure"18:23
plarszyga: but there have been several times where it would have been helpful to have a different system in there other than the one server18:23
=== salgado-lunch is now known as salgado
zygaasac: dashboard knows nothing about jobs18:23
zygaasac: this is something we have to solve eventually18:23
zygaasac: if we do hybrid views like that then we need to say "that's what we are doing" and merge dashboard with scheduer18:24
zygaasac: or build an API for that but it would be rather painful18:24
zygaasac: on the other hand, the scheduler should know all that18:24
zygaasac: and could permalink to ... some dashboard to see test data18:25
zygaasac: but I agree that the view could be more useful18:25
zygaasac: actually, listing test runs there would be interesting in itself, jobs are not that special18:25
asacjobs should submit to dashboard18:28
asaca bundle is a job for me atm18:28
asacmaybe not all bundles are jobs18:28
asacbut all jobs seem to relate to a bundle18:28
zygaasac: well that's high level view 18:29
asacexactly18:29
zygaasac: dashboard only stores results atm, jobs are plans of what to do - not results18:29
asacand i say you can add a parameter to submit_results in the job description18:29
asacthat sets the name ;)18:29
zygaasac: the key is to link jobs to results sensibly18:29
asacand the submit_results also adds job id and details to meta18:30
zygaasac: storing jobs id in the dashboard is an excellent idea and we should expand on that18:30
zygaasac: then you only need to solve the distributed nature of LAVA to get what you want18:30
zygaasac: and that is solvable18:30
zygaasac: although we need to draw the line and decide on how to actually look at pieces of LAVA18:31
asaczyga: which ways would you propose?18:33
zygaasac: I'd love to say "distributed lava is a good idea and this $MAGIC will make working in a distributed environment just as easy as in a monolithic system"18:33
asacyou mean if you have multiple job schedulers?18:34
zygaasac: we should brainstorm with mwhudson and just decide18:34
zygaasac: or multiple dashboards currently there is no connection between scheduler and the dashboard and all scheduler does is call and XML-RPC method on an URL specified in the job file18:35
=== mayday_jay is now known as mayday_jay-away
zygaasac: although that is changing the QA tracker replacement is tightly integrated (at least initially as it was much easier)18:35
dokomabac, still online?18:35
asacso i found the linux-arm-soc git and built the for-next tree ... that is still 2.6.39 :/18:36
* zyga curses his aching leg18:36
asaczyga: whats up with your leg?18:36
* zyga was going to experiment with diving with weight belt today :-(18:36
zygaasac: I sprinkled my ankle a few hours ago18:37
zygaasac: it hurts like crazy when I move my leg even a little18:37
asacouch18:37
asaczyga: so to be clear on the bundle thing:18:38
asachttp://validation.linaro.org/lava-server/scheduler/job/71518:38
zygaI fell down the stairs wearing slippery flip flops18:38
zygaasac: yeah?18:38
asacmy feeling would be that i would love to give the bundle a name and description in that command:18:38
asac      "command": "submit_results", 18:38
asacby adding a "bundle_name" parameter ;)18:38
zygaasac: that's doable and we don't even need schema changes18:39
asacrigt18:39
zygaasac: I'll talk to mwhudson and we'll try to do it tonight18:39
asacso and second thing:18:39
zygaasac: bundles _have_ a name, we just never used it before18:39
zygayes?18:39
asaci want the "lava" test to be always named "Boot and Basic Health"18:39
asacand i want to be able to set a name and description for       "command": "lava_test_run", 18:40
zygaasac: lava test is special18:40
asacsure18:40
zygaasac: what you want is 'all daily tests should be called like that'18:40
asacthats why i say it should be "Boot and Basic Health" fixed18:40
asaczyga: right. Boot and Basic Health for lava and the others can be set in lava_test_run command18:40
zygaasac: no, because _all_ jobs sent to the scheduler will have a lava test18:40
zygaasac: you want to have "boot and basic health" for the tests plars runs daily18:41
zygaasac: not on all tests submitted, ever18:41
asacok18:41
asaclets use Basic LAVA Health18:41
zygaasac: that is sensible18:41
asaci dont care ;)18:41
zygaplars: ^^18:41
asacif i want a boot and basic test we will alwasy add a simple "zero" test on top18:41
asacand for android might be something different18:41
zygaplars: what do you think?18:42
zygaasac: "zero test"?18:42
asacdont bother18:42
zygaasac: booting is a tests in itself, is that what you mean?18:42
asacNOP test_run ;)18:42
asacyes18:42
asacif i want to have a "boot and basic" test_run 18:42
zygaright 18:42
zygaI get it18:42
asaci have to do this on my own ;)18:42
asacwhile LAVA Health is always run18:42
asacand is kind of similar ... but doesnt go that far18:43
asacas it tries to boot18:43
asac(afaik)18:43
zygahmmm18:43
plarszyga: sorry, was looking at your branch, let me catch up18:43
zygaplars: let's separate what asac said into topics, I'm interested in your feedback on the dispacher topic 18:43
asacwhast the dispatcher topic?18:44
asac:-P18:44
plarsok...18:45
zygaasac: technically this is handled by the dispatcher :)18:45
zygaasac: basically the last thing you spoke about18:45
plars1. I don't think we need to change the submit_results command to give it a bundle_name  parmaeter18:45
plarsthere's already a job_name in the job that you can specify18:45
asacbut i want to display the bundle_name in the stream bundle view18:46
plarsit should use that, or just set it to something generic "lava test run" or "unknown" or "please name me, I feel so ashamed that I have no name"18:46
asacnot sure if you always want to display the job_name18:46
asacif thats the case then i am fine i guess18:47
plarsasac: the job name is what would display instead of the SHA1 in this view:18:47
plarshttp://validation.linaro.org/lava-server/dashboard/streams/anonymous/lava-daily/bundles/18:48
asacyes thats bundle stream view for me (if i parse that correctly)18:48
plars2. going a step further to http://validation.linaro.org/lava-server/dashboard/streams/anonymous/lava-daily/bundles/f43ca432eed1f443d639f3c7fefa2cb247338bcb/ we want to display something other than the UUID for the test run18:48
plars(on a side note, zyga did you see my earlier comments about "is uuid really neede?"18:48
asacright18:49
plarsThat's where we should have something like "stream results" "boot and basic health", whatever18:49
zygaplars: no18:49
plarswe could just use the test name18:49
plarsno?18:49
asacand i was saying we should select a default for the "lava" test18:49
zygaplars: where did you post them, here?18:49
asacrun18:49
zygaplars: (no I did not see your eralier comments)18:49
asacand the other should be configurable or settable18:49
plarszyga: yes, but that was before you decided to go stair diving18:50
zygahehe18:50
plarszyga: will talk to you about it later18:50
* zyga looks back18:50
rsalvetijcrigby: just to let you know that the latest tilt kernel update is working fine18:50
plarswe already have a link for the test name18:50
plarsexample: http://validation.linaro.org/lava-server/dashboard/tests/pybench/18:50
plarsbut I don't think it's all that useful at the moment18:50
zygaadding ability to browse all stuff is easy but remember about one thing18:51
zygawe have bundle streams for a reason18:51
zygaIMHO those should be features of bundle stream views18:51
plarsit could maybe be a link for historical results under the details for that test?18:51
zygaplars: yup18:51
plarszyga: can you explain further what you mean?18:52
zygaplars: how about moving this (test-centrict) view as a "mode" of a bundle stream view?18:52
zygaplars: I would like to add different way you can look at a particular bundle stream18:52
plarszyga: like a tab?18:52
zygaplars: and not on the whole global data-set18:52
zygaplars: yes18:52
jcrigbyrsalveti, good to know, thanks18:52
zygaplars: because that defeats the purpose of streams, to separate data18:52
plarszyga: oh, yes for sure18:53
zygaplars: similarly with our new "image health" view18:53
plarsI'm not suggesting that we look at things across streams18:53
zygaplars: that is just another way to look at a bundle stream18:53
zygaplars: bundled with unfortunate instance of "dynamic filtering" in one stream18:53
zygaplars: right, then I'm all for doing that (adding test-centric mode to bundle stream view, etc)18:54
zygaplars: as for looking at more than one stream I think we should eventually have "dynamic streams" that contain search data, we could then allow users to select, say "all bundles from stream A, and all instances of test 'foo' from stream B"18:55
zygaand see that using the beefy stream view18:55
plarsmmmm beefy18:55
zygathat is what we currently do for: http://validation.linaro.org/lava-server/dashboard/image_status/developer+omap3/18:55
=== mayday_jay-away is now known as mayday_jay
tgall_foorsalveti, for sound and panda, is it realistic that would be fixed for 11.09 ?19:11
tgall_fooI hope I'm wrong but I thought UCM was a bit further off into the future19:11
zygaplars: read your comments, thanks19:12
plarszyga: I think it's a great start, just want to understand some things better19:12
zygaplars: I need to think about how to specify that on the other end19:15
zygaplars: I started with a mental model of "what is really out there" 19:15
plarsyeah19:16
ZeZuanyone here know offhand any major instruction behaviour differences between ARMv4 and ARMv7 (differences in instructions that do exist in both ISA, but differ. ie:  unaligned access addr bits rotating ,  store of pc offset +8/+12)19:28
mruZeZu: hoping for a different answer here?19:41
=== michaelh1|away is now known as michaelh1
ZeZumru: it was before anyone answered, and most of the people in there really can't answer many of my questions,  save for yourself and a few select others :)19:49
michaelh1lool: hey, do you still need someone to cover the techleads call?19:50
mruZeZu: ah yes, I didn't check the timestamps19:51
loolmichaelh1: I actualled asked Ilias and he said yes, but there are two, so perhaps you could offer to pick up one, or you two could drive them; I'm sure you'll be helpful in the actual contents in any case  :-)19:53
loolmichaelh1: thanks a lot for getting back to me though, appreciated19:53
michaelh1ZeZu: the ARM ARM (or A9 TRM?)  has sections on that near the end...20:20
rsalvetitgall_foo: well, guess it's not that complicated, will see20:21
rsalvetineed to sync with mwg and tilt20:22
rsalvetiwe need to fix it for ubuntu either way20:22
rsalvetiotherwise we'll have another release without sound20:22
tgall_foorsalveti, good to hear,  def need that fixed20:23
voidspacehey guys, trying to create a linaro boot disk on sd20:35
voidspacerunning the linaro-media-create command (on ubuntu server) I get the following error:20:35
voidspacedbus.exceptions.DBusException: org.freedesktop.DBus.Error.ServiceUnknown: The name org.freedesktop.UDisks was not provided by any .service files20:35
voidspacethe error is being thrown in "confirm_device_selection_and_ensure_it_is_ready"20:36
voidspacefull traceback if anyone is interested20:37
voidspacehttp://paste.pocoo.org/show/463858/20:37
=== michaelh1 is now known as michaelh1|away
=== jconnoll1 is now known as jconnolly
=== michaelh1|away is now known as michaelh1
faborsalveti: do you have time to run a test on origen?21:12
=== mayday_jay is now known as mayday_jay-away
=== mwhudson__ is now known as mwhudon
=== mwhudon is now known as mwhudson
mwhudsonplars: good morning21:17
plarsmwhudson: hi21:18
mwhudsonplars: call in a few minutes?21:19
zygamwhudson: hi21:19
mwhudsonzyga: hi21:19
zygamwhudson: could we brainstorm?21:19
mwhudsonzyga: about images and so on?21:20
plarsmwhudson: I can't today, need to take my son to soccer in just a bit21:20
zygamwhudson: I'm not certain I'm up to it now but we need to consider how to handle a few topics21:20
zygamwhudson: more than that21:20
mwhudsonzyga: gimme a few minutes to get settled, but sure21:20
mwhudsonplars: ok21:20
zygamwhudson: no mike today, I cannot walk upstairs21:20
mwhudsonplars: tomorrow then?21:20
plarsmwhudson: possibly tonight depending on when I get home, or tomorrow, sure21:20
zygamwhudson: (small accident earlier today)21:20
mwhudsonzyga: eek, hope you get better soon21:20
zygamwhudson: well, yes but it still hurts, akle sprinkled bad21:21
zygamwhudson: too bad you don't have scrollback21:21
zygamwhudson: I'll try to find the log21:21
mwhudsonzyga: it's logged, right?21:21
mwhudsonit says so in the /topic anyway21:21
zygamwhudson: up, looking at the public log now21:21
zygamwhudson: http://irclogs.linaro.org/2011/08/24/%23linaro.html#t18:3421:23
zygamwhudson: read back starting from that point once you have a moment21:24
mwhudsonzyga: ok21:24
zygamwhudson: the bottom line: tighter coupling between server side components, we need to think how to provide that without fusing everything (or should we?)21:24
zygamwhudson: I'll limp back to the living room, try to get back here in 30 min21:24
zygaaround even hour21:24
mwhudsonzyga: ok, don't push yourself too hard21:25
plarszyga: around?21:33
plarszyga: cjb is trying to install from the ppa and it seems to want to be in development mode by default or something21:33
plars<cjb> installed lava-dashboard, but lava-server createsuperuser raises "django.core.exceptions.ImproperlyConfigured: Development mode REQUIRES VIRTUAL_ENV to be set"21:33
mwhudsonah yes21:34
mwhudsonwe should fix that21:34
mwhudsonyou need to run /usr/lib/lava-server/manage.py createsuperuser21:34
plarsmwhudson: no, I thought that was fixed... I used it in the last release I think21:34
mwhudsoncan someone file a but on lava-server21:34
plarsmwhudson: lava-server that is21:35
rsalvetifabo: yup, testing it in a few minutes21:35
faborsalveti: \o/21:35
mwhudsonplars: doesn't look fixed to me21:36
faborsalveti: we want to check if add-origen-android-support introduce a regression with Origen board21:37
faborsalveti: https://code.launchpad.net/~linaro-landing-team-samsung/linaro-image-tools/add-origen-android-support/+merge/7246521:37
rsalvetifabo: ok, will make sure I'm flashing it using this branch21:38
rsalvetifabo: any other specific test with origen?21:38
faborsalveti: we would like to do the same with android but we don't have the images yet21:39
faborsalveti: let's stick to the combo patch/ubuntu/origen21:39
rsalvetifabo: oh, ok, let me know if you also want me to test android on it21:39
rsalvetilater today21:39
fabook21:39
fabothanks21:39
rsalvetisure, np21:39
mwhudsonzyga: ok, caught up on the backlog21:56
mwhudsonzyga: it swaps between the high and low level in the usual disorienting way :-)22:04
zygamwhudson: re22:06
zygamwhudson: yeah22:06
zygamwhudson: asac wants results, err, effect, we need to map that ;)22:06
zygamwhudson: so what options do we have?22:07
zygamwhudson: 1) fusing lava-server parts together to lava-blob22:07
zygamwhudson: 2) doing fully distributed asynchronous $BUZZWORD system22:07
zygamwhudson: 3) finding the sweet spot between 1) and 2) where just enough data is passed/shared22:07
mwhudsonzyga: i have to say my opinion on this area in general has changed a bit22:08
zygamwhudson: good22:08
mwhudsonzyga: at first it seemed insane to have the separations we have22:08
zygamwhudson: so speak your mind22:08
mwhudsonzyga: but in practice, it seems to work quite well22:08
mwhudsonlike this thing with the scheduler storing a url to link to the results rather than a fkey22:09
mwhudsonzyga: another peanut gallery comment is that if we are going for more tight integration, then the scheduler should be inserting the submit_results action to the job22:09
zygamwhudson: wow, nice idea22:10
zygamwhudson: yeah22:10
mwhudsonjust assuming that the json specifies the same dashboard as the one we assume would be utterly mental22:10
mwhudsonwell, maybe that's over strong22:10
mwhudsonbut it wouldn't be a good idea :)22:10
zygawell to some extend it might22:10
zygamwhudson: but to others yeah, it's not ... I'm not really sure22:10
zygaa though22:11
zygawe should also look at keeping dependency (as in A needs B to be installed _in the same database_) from cooperation (A talks to B)22:12
zygaas you did in the qa tracker rewrite22:12
zygadependency provides direct internal APIs and full DB access22:12
zygacooperation requires up front planning and any changes in the scope need to be coded and deployed22:12
zygaalso security (privacy) might be a problem if we use cooperation more as services will need to impersonate users across lava boundaries22:13
zygaso I'm still not sure which model is better to be honest22:13
zygaI hoped that we could get clean separation22:13
zygadashboard stores results22:14
zygascheduler worries with jobs22:14
zygabut it seems, as we add more pieces, that people want tight integration on the UI front22:14
zygamwhudson: let's take a step back and think about the problem at hand22:15
mwhudsonzyga: yes, i was going to say22:15
mwhudsonzyga: give me use cases or give me death22:15
mwhudsonor something22:15
zygamwhudson: so we want to have cross knowledge22:15
zygamwhudson: knowing job details on one end (to reuse, say, job "title/name/description"22:15
zygamwhudson: and on the other end knowing where the results are (I dare to say we have yet to see a request to show some failures inline in the scheduler)22:16
=== mayday_jay-away is now known as mayday_jay
mwhudsonzyga: in this specific area, it seems the dashboard doesn't want information from the _scheduler_ really22:16
zygamwhudson: no, but from the original submitter (job name)22:16
mwhudsonit wants information from the submitter22:16
mwhudsonand we already have a mechanism for this -- the job metadata22:16
zygamwhudson: yes and no22:17
zygamwhudson: it works on test run level22:17
mwhudsonah right22:17
zygamwhudson: and it seems that bundles are becoming promoted to first-class objects 22:17
zygamwhudson: which is problematic22:17
zygamwhudson: early on the bundle did nothing really22:17
zygamwhudson: and I always thought of it as uninteresting component22:17
zygamwhudson: aka the transport layer22:18
mwhudsonso for all it's non-idealness, we do have sufficient data -- testrun -> bundle -> testrun named lava -> metadata 22:18
zygamwhudson: so try to solve this use case:22:18
zygamwhudson: you see the bundle browser, there's a bundle with contains fragments of two jobs22:18
zygamwhudson: I can always find a bad scenario for "Bundle as a composable 'bag' for test runs"22:19
zygamwhudson: it's not that "not all bundles have job association22:19
zygamwhudson: it's that "bundles can have arbitrary associations"22:19
zygamwhudson: one way we could do it is make the scheduler/dispatcher reuse bundle NAME for this22:20
zygamwhudson: so bundles would get nice name that is sensible in the bundle stream / bundle view22:20
zygamwhudson: and then we could use the chain you just described in the test run view22:20
zygamwhudson: but then we need to either promote the test run "name" to a first-class field or keep hacking around it with attribute names22:20
* mwhudson scrubs his brain and reads again22:21
mwhudsonzyga: i'm all for not hacking around with attribute names (unless we move to a nosql stor... no, let's not go there today)22:22
zygaone problem at a time22:22
zyga(offtopic: i had a _great_ idea about nosql stuff)22:22
mwhudsonzyga: the more i think, the more i don't want the dashboard to depend on the scheduler (in the sense you described above)22:23
zygaagreed22:23
zygaI share the same view22:23
mwhudsonzyga: use case for this: there's only one board available and it's sitting on a developers desk and she's running the dispatcher by hand22:24
mwhudsonand uploading to the member-internal dashboard22:24
zygahmm, is that a question?22:25
mwhudsonno22:25
zygaI'm not looking at use cases, just at model coreherence22:25
zygaah22:25
mwhudson:)22:25
zygaokay22:25
zygaah I get it sorry22:25
mwhudson<zyga> mwhudson: I can always find a bad scenario for "Bundle as a composable 'bag' for test runs"22:25
mwhudsonwhat did you mean here?22:25
zygamwhudson: that the fact that bundles are exposed in the UI caused people to somewhat abuse them, they were the transport layer, people now want to use them as something that comes "at once" from the dispatcher and I think this is causing various internal conflicts (like having a malicious bundle that has conflicting attributes on various test runs etc)22:27
mwhudsonzyga: ah ok, so the two approaches would be to hide bundles completely, or give people what they claim to want and make bundles more of a 'thing' ?22:28
zygayup22:28
zygaand if you do the former then you still need to make people happy somehow22:28
zyga(by introducing another entity)22:28
mwhudsonright, some kind of grouping of test runs?22:29
zygasomething like that22:29
zygaa test_run_group uuid?22:29
zyga:D22:29
zygapaul will kill me for _another_ uuid :D22:29
zygaone that dispatcher could generate22:29
zygathis would make bundles hidden/internal detail22:29
mwhudsonthe question is, i guess, will punning the lump the test runs arrive in (the bundle) and this grouping we're talking about now cause problems22:29
zygaand allow us to look at that instead22:29
zygahmm22:30
zygapunning the lump?22:30
mwhudsonzyga: is there a use case for not wanting this identification though?  like a situation where one bundle would contain two logical lumps, or one logical lump would be spread across two bundles?22:30
mwhudsonzyga: sorry, probably confusing language, by punning here i guess i mean "identification", "regarding as the same concept"22:31
zygamwhudson: ah22:31
mwhudsoni guess it's this kind of usage: http://en.wikipedia.org/wiki/Type_punning22:32
zygamwhudson: IMO we either say that you cannot do this kind of abuse (random test runs in one bundle) and sanction that or make it impossible to express and allow people to upload arbitrary crazy bundle configurations 22:32
zygaI kind of like the latter solution, by making it imposible to do wrong you don't need to handle the "error" cases :)22:33
mwhudsonzyga: by abuse here, you mean things like two test runs named 'lava' with conflicting metadata?22:33
zygayes22:34
mwhudsonzyga: maybe this is too blue sky, but perhaps illustrative: what about one logic test 'situation' that is run across multiple boards?22:34
mwhudsoni say 'situation' because i'm running out of nouns :)22:34
zygainteresting challenge22:34
zygayes I though of that (although after the fact)22:34
zygaas calxeda guys want to do stuff like that22:35
zygahere it's a matter of perspective 22:35
zygayou can say that test "Foo" ran on 2 boards22:35
zygaas long as you don't care about test result -> board association22:35
zygayou are fine with the confines of our current model22:35
zygajust the hardware/software context is hard to untangle22:36
zygaand would (should) need changes22:36
mwhudsonright i probably wouldn't have thought of this without the calexda branches...22:36
zygaI did briefly think about this but in a different settings22:36
zygaabout expressing the network "neighbourhood" of a board, like the server it was talking to22:37
zygathe tests would be confined to the board but you want to know about the server/network configuration22:37
zygastill22:37
zygaokay22:37
zygawhat was the noun, situation22:37
zygawe could have a test_situation object22:37
zygaone that would describe an arbitrary collection of test runs22:37
mwhudsonright22:38
mwhudsonthe question is whether or not this would be a win22:38
zygathat would definitely need new UI, we'd promote old test runs to test runs in situation of the bundle22:38
zyga(I'm not sure, just trying to follow the idea)22:38
mwhudsongiven that our users seem to understand the bundle concept afaict22:38
zygaoh yes, that too -- any new noun will be painful to push22:39
zygait is equivalent to promoting bundle to first class object22:39
mwhudsonanother approach would that for the multi machine case, something or other combines the bundles in some way22:39
mwhudsonand submits a single bundle22:39
zygahmm22:39
zygalike a meta-dispatcher?22:40
zygathat could work22:40
zygado we care about the tree structure or just the "collection of test runs"? (another topic)22:40
zygaso22:40
zygaone thing that could help with putting this into bunds is that ideally bundle would be just the on-disk format that you can send or view with a simple application22:41
zygaIMHO if we do care about high-level grouping we should go with a new noun, as you described22:41
zygaah ah22:41
zygadrifting away from the task at hand22:41
mwhudson:)22:41
zygacan we -- for the momemnt -- look at naming bundles sensibly?22:41
mwhudsonah sure22:41
zygaand then think about how to name test runs sensibly?22:41
zygaoh, what do you think about that "Basic health" thing asac spoke about?22:42
zygalava-the-test-id is super-special22:43
mwhudsonzyga: this is a bit about test_run.name vs test_run.test.test_id right?22:44
zygayup22:44
zygaI'm not really sure what to say there22:44
zygatest_id is already good22:44
zygamaybe we should have test_run.test.name?22:44
mwhudsoni think test_run.name is already there in the db but usually empty or something?22:45
mwhudsonanyway22:45
mwhudsonit is sort of clear that the test_run lava is special22:45
mwhudsonit has the serial log attached, for example, which has the log of running _all_ the tests22:46
zygayeah22:46
zygabut we should avoid calling it "basic health xyz", it is just the ... hey22:47
zygathe situation :D22:47
zyga! 22:47
mwhudsoni guess a question is whether it lives on as a test run object, or becomes part of the bundle22:48
zygayeah22:48
zygahmmmmm22:48
mwhudsonor the test-run-group-object-that-might-be-bundle-or-not22:48
zygaI kind of worry the scope of this change22:48
zygaespecially because lava-test-run is still a test 22:48
zygaer, test run22:48
zygawith all the stuff that it carries22:48
mwhudsontrue22:53
mwhudsonzyga: to me, elevating bundle to more of a concept and adding metadata and the serial.log and maybe a few other things to it (and changing the bundle schema i guess)22:55
mwhudsonfeels sort of reasonable22:55
mwhudsonzyga: we can back fill by hoisting data from the 'lava' test runs to these new fields22:55
mwhudsonzyga: i guess a relevant question _here_ is how many bundles one expects to be submitted that didn't come from the dispatcher22:56
mwhudson(toolchain stuff?)22:57
zygare23:05
* zyga reads23:06
zygahmm23:07
zygayeah23:07
zyga+ random stuff we have yet to see23:07
zygawill that be bad though? we could make the new part optional23:07
mwhudsondunno, that's why i asked23:08
mwhudsonit wasn't meant to be a leading question at all23:08
zygaI think we can do two things:23:09
zyga1) not care about other bundles _yet_23:09
zyga2) for other bundles (not via dispatcher) use the traditional model unless we are forced to re-evaluate this question23:09
mwhudsonzyga: +1223:10
mwhudsoner +123:10
zygaok23:10
mwhudsonnot _that_ enthusiastic :-)23:10
zygaI'm about to fall asleep23:11
zygaso conclusion time23:11
zyga1) new top-level container23:11
zyga2) new UI23:11
mwhudsonzyga: you'll have to tell me about nosql some other day :-)23:12
zyga3) + rethinking what else to store there to make cooperation easy23:12
zyga0) cooperation trumpts dependency23:12
mwhudsonzyga: did we decide if "new top-level container" and "bundle" need to be different things?23:12
zygamwhudson: I'd love to implement a nosql myself ;)23:12
rsalvetilool: james_w: any specific reason why we're not using neither u-boot nor x-loader for igep?23:12
james_wrsalveti, no idea, sorry23:12
zygamwhudson: no, I think we should evaluate that with plars and 'marketing' as bundle became recognizable name over the time23:13
rsalvetiboth should be supported with our binaries already23:13
mwhudsonzyga: ok23:13
zygamwhudson: for the moment I'd rather have a new top level23:13
rsalvetijames_w: ok, thanks, lool should know about it23:13
zygamwhudson: as appending is better than renaming for people's memory23:13
mwhudsonzyga: very nosql!23:14
zygamwhudson: as for nosql ... heh, I think synchronization (or more likely - synchronizability, if that's a word at all) is a valuable property of lava 23:14
rsalvetijcrigby: igep is another board we could use to test pxe 23:14
rsalvetismc911x-023:15
rsalvetisince u-boot works fine with it already23:15
rsalvetinot sure with ours/upstream, but with the stock one already available at nand it works fine23:15
zygamwhudson: and being to able to map all our stuff onto some underlaying storage part that can share data/do queries would be a good next step23:15
zyga(next as in next leap)23:15
zygaI was thinking about how to do indexing23:16
zygaincluding non-precise indexing that speeds up traversing the data set in search of something23:16
jcrigbyrsalveti, good idea who has igep hw?23:17
rsalvetijcrigby: I got one23:19
zygamwhudson: I'll tell you about it tomorrow :)23:19
mwhudsonzyga: good idea23:19
rsalvetijcrigby: first thing is making sure we're using our u-boot :-)23:20
zygamwhudson: I'll reply to my merge request (to plars' comment)23:20
rsalvetilinaro-media-create is not using neither x-loader nor u-boot23:20
jcrigbyoops23:20
michaelh1rsalveti: I gave PXE a try on my Panda the other day and it worked fine...23:21
michaelh1rsalveti: I like it.23:21
rsalvetithanks to jcrigby :-)23:21
rsalvetilet me try our x-loader/u-boot with it23:21
rsalvetiseems I'm not able to have input over uart23:22
rsalvetihttp://labs.igep.es/index.php/How_to_setup_the_IDC10_cable this should be why23:26
mwhudsonzyga: oh, in other news, it seems that there are no really weird quoting issues with redirecting query strings23:26
mwhudsoni think probably META['QUERY_STRING'] is not parsed much23:27
zygagood23:28
voidspacelinaro lucid, as distributed with i.mx53 "out of the box" can't see my keyboard or mouse23:51
voidspaceI'm making an 11.08 mmc - just wondered if anyone else had seen the problem or if I may have dodgy hardware23:51
voidspacelinaro-media-create wouldn't run until I had installed udisks - so it should probably be marked as a dependency23:52
fabovoidspace: please report a bug https://bugs.launchpad.net/linaro-image-tools/+filebug23:56
voidspacefabo: done23:58
voidspacefabo: does this look like a healthy completion?23:58
voidspacehttp://paste.pocoo.org/show/463951/23:58
voidspacefabo: several hundred preceding lines of output trimmed23:59

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