=== omarrmz is now known as omarrmz_
=== michaelh1|away is now known as michaelh1
=== michaelh1 is now known as michaelh1|away
=== michaelh1|away is now known as michaelh1
=== michaelh1 is now known as michaelh1|away
=== michaelh1|away is now known as michaelh1
rsalvetirobclark: pm215 shold probably know better00:43
robclarkpm215, know anything about qemu-static + chroot on armhf rootfs?00:44
dokomichaelh1, online?00:47
michaelh1doko: yip00:47
dokomichaelh1, do you still know the name of the mexican restaurant in mountain view?00:47
michaelh1No, but my credit card does... standby00:47
michaelh1doko: Vive Sol00:48
dokomichaelh1, thanks!00:48
michaelh1(I really liked it)00:49
robclarkpm215, nm, it was operator error00:49
mwhudsontrying to use control seems particularly frustrating atm01:45
prpplagueho ho ho, merry tuesday02:11
mwhudsonwhy are there apache2 processes with 2.2 gigs RSS on control02:13
mwhudsonplars: ayt?02:13
plarsmwhudson: of course :)02:14
plarsmwhudson: hmm02:14
mwhudsonplars: we have dispatchers using 12, 8.7 and 5.5 % of ram02:14
mwhudsonand apache2's using 7.6 and 5.5%02:15
plarsmwhudson: how long have they been running??02:15
mwhudsonnot sure, not amazingly long02:15
=== michaelh1 is now known as michaelh1|away
mwhudsonoh heh well02:15
plarsmwhudson: ouch, it looks like we have a few that have been running amazingly long02:15
mwhudsonthe top dispatch has been running for 4 days02:15
* mwhudson kills with fire02:15
mwhudsonmaybe i'll work on https://bugs.launchpad.net/lava-scheduler/+bug/937443 today02:16
ubot2`Launchpad bug 937443 in lava-scheduler "should limit run duration and log file size" [Undecided,New]02:16
plarsmwhudson: one has been running longer02:16
plarsmwhudson: march 6th02:16
mwhudsonplars: nice02:16
plarsI'd like to understand where they're getting stuck too though02:16
plarsmwhudson: in the case of this one: http://validation.linaro.org/lava-server/scheduler/job/14761/log_file it seems it hasn't logged anything in the past few days at least02:17
plarsso it's not growing or anything02:17
mwhudsonit always seems to be origen jobs that go crazy02:17
mwhudsonmaybe the large apache2 processes are trying to analyze the log files from these crazy jobs...02:19
plarsmwhudson: ahhh02:19
mwhudson(but that would be uwsgi though, so...)02:19
plarsmwhudson: lava-android-test commands don't have a timeout :)02:19
mwhudsonplars: ah02:19
plarsmwhudson: because they don't use pexpect02:19
mwhudsona job with description "lt-origen-oneiric-20120311-0+linaro-o-alip-20120311-0" probably isn't running android command though?02:20
mwhudsoni agree that's quite likely though02:20
plarsmwhudson: nope, not that one02:20
plarsmwhudson: I was looking at the one that was stuck since March 602:20
mwhudsonah the imx one02:20
plarsmwhudson: haha, this is funny actually02:21
plarsmwhudson: it even allows a timeout parameter!02:21
mwhudsondo we just not use it?02:21
plarsmwhudson: yep02:22
mwhudsonproduction ready!02:22
mwhudsonthe job running on origen02 seems to be in the test install phase02:22
mwhudsonlast few meaningful lines are02:26
mwhudsonLAVA: (stdout) Unpacking gir1.2-atspi-2.0 (from .../gir1.2-atspi-2.0_2.2.2-0ubuntu1.2_armel.deb) ...02:26
mwhudsonmmc0: Timeout waiting for hardware interrupt.^M02:26
mwhudsonmmcblk0: error -110 transferring data, sector 12152824, nr 440, cmd response 0x900, card status 0x0^M02:26
mwhudsonmmc0: ADMA error^M02:26
mwhudsonmmc0: Got data interrupt 0x00100000 even though no data operation was in progress.^M02:26
mwhudsonmmc0: Got data interrupt 0x00100000 even though no data operation was in progress.^M02:26
mwhudsonoh ffs02:27
mwhudsonthe timeout for cmd_lava_test_run defaults to -102:27
mwhudsonit can be overridden by the test02:27
mwhudsonbut i bet approximately 0% of tests do taht02:27
plarsmwhudson: -1 for pexpect means just use the default, which was defined when the process was spawned02:28
plarsmwhudson: iirc, that's 1200 sec for us02:29
plarsmwhudson: hmm, it's using os.system too.02:29
mwhudsonplars: how much do you want to bet on that?02:29
plarsmwhudson: know of any os.system equiv with a builtin timeout?02:29
plarsmwhudson: on what?02:30
mwhudsonplars: no02:30
mwhudsonthat pexpect behaviour02:30
mwhudsoni agree it's what the docstring says though...02:30
mwhudsonplars: os.system with a timeout?  no not really02:30
plarsmwhudson: I'm fairly certain of that, however, there's a bug that I know of in the timeout exception02:31
mwhudsonplars: you can put 'timeout 10s ' in front of what you pass to system02:31
mwhudsonaaaaa using control is just impossible like this02:32
plarsmwhudson: if you c=pexepct.spawn('foo', timeout=300), then do c.expect('bar', timeout=1200), and the expect times out, the exception dump will tell you that the timeout was 300 (even though it waited 1200)02:32
plarsmwhudson: did you restart apache?02:33
mwhudsonplars: no02:33
plarsmwhudson: oh, we are getting overrun with lbsappd processes again02:35
* plars wonders if rsalveti or anyone discovered where that was happening02:35
mwhudsonwhat do they do?02:35
plarsmwhudson: it's running under qemu02:36
plarsmwhudson: I think it's some gps thing for snowball02:36
plarsmwhudson: seems to happen when we create images with lmc for snowball02:36
plarsmwhudson: my guess is something that happens as a side effect of the package install for that piece in the hwpack02:36
mwhudsonoh right02:37
mwhudsonthere is something magic you can do to kill all processes in a chroot i think?02:37
plarsmwhudson: it causes some general slowness sometime02:37
mwhudsonone of the tests we run seems to install the entire world02:38
mwhudsonah it installs ubuntu-desktop02:39
mwhudsonchroot /mnt/testrootfs lava-test install leb-basic-graphics02:39
mwhudsonplars: ok, i don't understand at all how these jobs are getting stuck then03:12
mwhudsonasdhaslkjf hdasljkfhlkasdjh vlkjasdhf vlkjsdfh lkjdsrhf vlkjsdf vh03:18
mwhudsonuh no03:19
mwhudsonthe job file does specify a timeout for this command: http://validation.linaro.org/lava-server/scheduler/job/15314/definition03:19
mwhudsonchithu fixed a bug where the timeout parameter wasn't being passed on by PrefixCommandRunner03:23
mwhudsonwhich lava_test_install uses03:23
mwhudsonso maaaybe we won't have this problem with new runs03:23
mwhudsonah no, that was far too long ago to be the difference03:24
=== Guodong-Xu is now known as guodong
=== guodong is now known as Guest33218
=== Guest33218 is now known as Guodong-Xu
mwhudsonreplacing conmux with cu might not be _totally_ straightforward04:22
mwhudsonbecause it takes a little while for cu to let go of the usb device04:22
plarsmwhudson: I don't think cu is a great thing directly05:54
plarsmwhudson: well, not for this purpose05:54
plarsmwhudson: I use cu over minicom for "I need a serial console" any day05:54
plarsmwhudson: but for test infrastructure, I think we want to either use terminal servers, or something that takes a usb port and puts it on a network port like a terminal server would05:55
plarsmwhudson: ttywatch for example05:56
plarsmwhudson: conserver is another one05:57
plarsmwhudson: of course we should be careful here that we aren't just replacing conmux with something that will have the same woes with a different name05:57
fabodeepti: morning, do you know who wrote lava_hwpack_submit_success/failure jobs?06:16
deeptifabo: good morning06:16
deeptifabo I did, why ?06:16
fabodeepti: cool :) I want to submit the armhf hwpack/images created on ci to lava06:18
deeptifabo ah! ok 06:18
fabodeepti: I'm wondering if you script does that or need to be forked06:18
deeptifabo It shld be the same. Let me give you a sample json to see if all the required information needed is present06:19
fabodeepti: lava_hwpack_submit_success/failure is a black box to me, I'll need the jobs input (hwpack_info?)06:21
deeptifabo yes, the hwpack_info is used to get the json to be populated with values06:21
deeptifabo how are u building your things on ci* as of now, do u use the lci-build-tools or your own scripts06:22
deeptifabo http://pastebin.ubuntu.com/882861/06:24
deeptifabo thats the json for testing on lava, we do send build results to lava to publish build results06:25
fabodeepti: my own script06:25
deeptifabo ok,  should  not be very difficult as long as you supply all the required values and then use the  lci-build-tools/get_latest_ci_hwpack 06:27
fabodeepti: the json is generated by your job? how do I change the values in there?06:27
fabodeepti: something like: export value1=foo; export value2= bar; lci-build-tools/get_latest_ci_hwpack;06:28
deeptifabo  the json is generated freshly by every job, refer to lci-build-tools/jenkins_common_lib line 1:2706:29
deeptifabo that is how you need to populate hwpack_info (file) and trigger a downstream job with file parameter06:29
deeptifabo those variables will then be available to the downstream job as env variables and can be picked up to create/submit job to json06:30
deeptifabo if u plan to submit the job from the build job itself then you can export as above you mentioned 06:31
fabodeepti: yes, that's the plan -> submit the job from the build job itself06:31
deeptifabo but then there might be false positives meaning even if there are failures bcs of the job submission then the entire job will be marked as failure 06:32
fabothat's fine06:32
deeptifabo  Almost all the jobs on ci* use https://ci.linaro.org/jenkins/view/All%20CI/job/lava_hwpack_submit_failure/, https://ci.linaro.org/jenkins/view/All%20CI/job/lava_hwpack_submit_success/06:33
deeptifabo if it suits then I would suggest you just present the values required by these jobs and leverage them 06:34
deeptifabo upto you 06:34
fabodeepti: it doesn't seem to fit exactly, the script needs to be more generic imo06:34
faboah no, seems code06:35
fabocode good :)06:35
deeptifabo ok if it does not suit then come up with a script and I will merge it in lci-build-tools06:35
fabonothing is hardcoded to kernel ci06:35
deeptifabo no nothing at all.. its all dynamic 06:36
fabook, let's experiment :) thanks06:36
deeptifabo so according to me it should help you, in any case do let me know06:36
deeptifabo cool06:36
wowonHi ... is there any mailinglist/forum to post a real beginner question ?07:48
fabowowon: ask.linaro.org07:49
fabowowon: you can also ask here. you'll get your answer or ... not :)07:51
wowonfabo : thanks :)08:03
wowonOk ,, I just install linaro and run first linaro-media-create ... it's booted sucessfully .... and now i'm at root prompt. my question is : how to make it 'multiuser' so there is root and non-root user08:04
mertsasI'm working on a omap4460 board, and I have found one kernel from linaro with rpmsg built in, and one which I can use to compile the pvr drivers. Problem is, these two are not the same. Anyone know if there is a kernel which have both, or do I have to start hacking on the kernels?08:05
fabowowon: you can use regular linux commands, adduser08:06
fabowowon: by default, you should have 'linaro' user08:06
fabospringz: should we move FSL testsuite to this cycle?08:07
wowonfabo : Thanks allot ..... but since it default have 'linaro' user , why every time I type 'exit' it always auto loged-in as root ?08:07
faboit's in the backlog atm08:07
fabowowon: which rootfs image have you used?08:08
wowonfabo : linaro-natty-nano-tar-20110527-1.tar.gz08:10
springzfabo, it may finish this or next cycle, so I suggest keeping it backlog, if I finished it this cycle, we can set it to this cycle, if not then next cycle, no need to postponed if delayed08:10
fabospringz: sounds good. thanks.08:11
fabowowon: that's why. nano is the smallest image. you probably want a full desktop like ubuntu-desktop or alip image.08:12
wowonfabo : is there any description what is in alip ? basically I just want a multiuser ubuntu with openbox 08:14
fabowowon: alip is xfce based08:15
fabowowon: you can start from alip and just install/remove (openbox) the packages you want08:16
fabowowon: https://wiki.linaro.org/GlossaryOfTerms08:16
wowonfabo : ah .. got it ... i really appreciate your enlightment. 08:17
faboyou're welcome08:18
davepigottplars: ping08:44
wowonfabo : i plan to use http://tinyurl.com/binotvbox01 as my platform , is http://releases.linaro.org/12.02/ubuntu/oneiric-images/alip/hwpack_linaro-lt-vexpress-a9_20120221-1_armel_supported.tar.gz will good for that box ? i'm downloading it08:59
fabowowon: it doesn't say much except the cpu is a cortex-a9. most likely, it isn't vexpress.09:07
wowonfabo :so .. what hw pack that potentialy will suit for that box ?09:09
wowonfabo : that box is widely manufactured in china, so potentialy it also widely available in my area (indonesia). if i go with such like beagle ... i'll need to do import09:10
frigglehi folks. Any ideas for encoding HDMI output from an ARM board to send over the network?09:14
bhojsuihkulokki,  Hello 09:15
frigglethe idea being you have a bank of dev boards, and anyone can 'log in' remotely09:15
friggleobviously there's ssh and vnc, but that's not helpful for 3d output09:15
frigglethis would be similar technology to onlive/gaikai. It seems like something that should exist and someone (Linaro?) might be doing already. Any pointers much appreciated09:15
bhojsuihkulokki, how are you supporting fastmodels in l-i-t ? 09:16
bhojsuihkulokki, I guess the bootable image  would be outside the mmc image . 09:16
fabowowon: the 1st step is to get more information on the box, which Soc/gpu. then you can look into hwpacks.09:17
suihkulokkibhoj: I was a few days in vacation, before going away the idea was that we'll make l-m-c output two files, the mmc image, and the axf file09:17
wowonfabo : ok ok ... i'll do it .. thankyou09:18
suihkulokkibhoj: was there a big.little call ~6h ago?09:19
bhojsuihkulokki, okay . was thinking of similar thing . currently for android I am creating an mmc image from the tarballs and one has to manually extract the axf  file . Let me know once you have something for ubuntu .I'll try to reuse.09:19
bhojsuihkulokki, yes there was . But I missed it since I had attended the one on yesterday .09:19
bhojsuihkulokki, I have named the builds as vexpress-rtsm . Let me know what naming are you planning to have in ubuntu . will try to use the same on android.09:20
suihkulokkivexpress-rtsm sounds good to me09:21
danboidI've installed and updated 12.02 on my Panda now but was sad to see that like Canonical ALSA sequencer hasn't been included with the kernel, not even as a module. Could I file ALSA sequencer support as a request for the Linaro kernel or do I do it via Ubuntu?09:29
jk-danboid: CONFIG_SND_SEQUENCER=m, for 12.0409:31
jk-or do you mean something else?09:32
davepigottfabo: ping09:32
danboidjk-, That will likely be the option I mean - are you saying this is fixed already in 12.04?09:32
davepigottfabo: Did you see my mail? All live boards now have audio loopback.09:33
jk-danboid: I think it's always been configured that way, which would suggest it's specific to your kernel build09:33
davepigottfabo: Will be out for the next hour or so. Have ARM FastModels team visiting to look at and understand LAVA09:33
jk-(this is a standard Ubuntu kernel)09:33
jk-danboid: so no need to raise it with the Ubuntu kernel team09:36
ofogIm trying to use linaro-media-create on a oneiric host, i keep getting the error "Cannot stat  resolv.conf", I'm using the hwpack and binary from the 12.02 release for an igep 09:38
fabodavepigott: ok, thanks for the cables.09:38
jk-danboid: sorry, looks like 11.10 had SND_SEQUENCER=n on armel09:41
jk-which would explain why that's the case with your kernel09:42
zygagood morning09:43
amitkamitdk: available for a call?09:51
amitdkamitk, yes 09:58
amitkamitdk: skype?09:59
amitdkamitk, shall i call the conf number09:59
amitkamitdk: let's try skype09:59
amitdkamitk, For skype I have to call from my cubicle otherwise meeting room. anyway lets try skype10:00
amitdkamitk, you online in skype?10:03
amitkamitdk: yes, I don't see you. hold on..10:05
ChiThudavepigott, hi10:07
alf_fabo: Hi! Is the only way to get a precise armhf desktop rootfs from jenkins?10:15
faboalf_: I think rsalveti used ubuntu precise image before we get them on ci10:16
faboalf_: both options should work10:16
alf_fabo: Right, I meant the linaro image :)10:17
faboalf_: if possible, give a try to linaro ubuntu desktop. I'd like some feedback10:18
alf_fabo: I am downloading it now (https://ci.linaro.org/jenkins/view/Ubuntu%20Build%20Service/job/precise-armhf-ubuntu-desktop/30/ is the most recent one that succeeded)10:19
frigglehas anyone ever implemented a system for remote graphical access to development boards?10:19
hrwfriggle: x11vnc on board + vnc on desktop10:20
amitkhongbo: available for a call?10:21
frigglehrw: yeah, I was wondering if there was a hardware solution for 3d support etc. e.g. an hdmi->h264 encoder10:21
hongboamitk: 3 minutes later10:27
hrwah, never check such ones friggle 10:28
friggleI'd have thought I can't be the only one wanting to do this...but googling doesn't bring up much. If anyone has *any* examples of someone doing the same thing I'd really appreciate being pointed in the right direction10:29
danboidjk-, So do I file a request to get that changed with Ubuntu or Linaro? I'm more concerned with Linaro but of course its based on Ubuntu so.. ?10:35
danboidjk-, I'd like it at least to be available as a module like in x86 buntu10:35
=== deepti is now known as deepti_away
Espersson|2mabac: Hi10:43
=== Espersson|2 is now known as Espersson
mabacEspersson, hi10:43
Esperssonmabac: when can i find the latest version of LAMC? 10:46
mabacEspersson, in the latest released l-i-t. just a sec10:46
mabacEspersson, https://launchpad.net/linaro-image-tools/trunk/2012.02/+download/linaro-image-tools-2012.02.tar.gz or apt-get install the package10:47
mabacEspersson, or if you're interested in the very latest change you need to get it from trunk10:47
Esperssonmabac: We are having some problems for Snowball and the latest 3.3 IK that we think are related to a chnage for vmalloc to 300M10:47
mabacEspersson, I landed that in trunk this week10:48
danboidALSA sequencer should really be available as at least a module as Pandaboard is quite an attractive platform for Linux musicians due to it being totally silent10:48
Esperssonmabac: ok, how do i get the latest from the "trunk"?10:48
mabacEspersson, bzr branch lp:linaro-image-tools10:48
mabacEspersson, and you do that while on the Linaro wifi, not from the office network, naturally ;)10:49
Esperssonmabac: Hmm, I think I need some more help understanding this, i will have to explain to the guys actually doing this10:50
mabacEspersson, not a problem, I can show you.10:51
Esperssonmabac: Ok, thanks,can you come to India where I am :)10:52
mabacEspersson, oh I think I'm supposed to save on the travel expenses... ;)10:52
mabacEspersson, you need to execute the command "bzr branch lp:linaro-image-tools" to get a local copy of the current working tree of linaro-image-tools10:53
mabacEspersson, then you'll have everything including the changes since the last relase10:53
Esperssonmabac: Sunil_K just joined, he will do this in,perhaps you can explain how to him? 10:53
mabacSunil_K, ^^10:54
Esperssonmabac: and this cannot be done from the company network?10:54
mabacEspersson, Sunil_K then you just exeute l-a-m-c from that directory 10:54
mabacEspersson, bzr won't work throught the proxies and firewalls no matter what holes you try to drill10:54
mabacEspersson, but I can create a tarball for you to download10:55
Esperssonmabac: great!10:55
Sunil_Kmabac: Thats great. It will be great. as i know we cannot use bzr.10:58
Sunil_Kmabac: do provide link or details where you are going to share this.10:58
mabacEspersson, Sunil_K http://dl.dropbox.com/u/400334/linaro-image-tools-2012.02-devel.tar.gz10:59
mabacSunil_K, extract that and you'll get the current tip of l-i-t11:00
Sunil_Kmabac: thanks alot. I am downloading it now.11:03
mabacSunil_K, you're welcome. then run l-a-m-c as ./linaro-android-media-create of course11:03
hongboamitk, may I edit the original suspend_test script when integrating it into pm-qa, or just copy it into pm-qa and call it simply?11:07
hongboamitk, I think I should edit  it to get uniform out put style, but I don't know if this is permitted or not11:08
=== deepti_away is now known as deepti
amitkhongbo: ofcourse you should edit it11:17
hongboamitk: got it.11:20
danboidSo who decides the default kernel options for the Linaro Ubuntu kernel or does it use exactly whats used by upstream?11:35
=== aviksil_ is now known as aviksil
danilomabac, salgado, deepti, gesha, pfalcon, jamestunnicliffe: I'll be missing out on today's hangout as well, please go on without me11:47
jamestunnicliffedanilo: will do. Hope you feel better soon.11:48
xranbyandrey_konovalov: ping please take a look at https://bugs.launchpad.net/ubuntu-omap4-extras-graphics/+bug/84515812:28
ubot2`Launchpad bug 845158 in openjdk-6 "Frequent java task hang on ARM server" [Undecided,Confirmed]12:28
xranbyit would be nice to know if you can reproduce this bug using the testcase in #1312:29
* andrey_konovalov is looking12:29
salgadodanilo, if you want to have a look at the new workitems UI, it's available on qastaging already.  we are QAing now and already found a couple improvements we can do, but it's looking good12:31
danilosalgado, cool, let me glance at it12:32
salgadoan example BP with migrated work items: https://blueprints.qastaging.launchpad.net/linaro/+spec/tr-kernel-demo-complete-dt12:32
danilosalgado, like that help text icon that should have been on the same line as "Work items" header? :)12:33
salgadodanilo, that's been reported a while ago.  not easy to fix, unfortunately12:33
danilosalgado, I am looking at infrastructure BPs, they seem fine as well12:33
danilosalgado, as in 'putting style="float:right"'? :P (at least it'd be better, not perfect imo)12:34
salgadodanilo, I don't think that looks better12:37
danilosalgado, sure, tastes differ12:37
salgadodanilo, if it ended up next to the save/cancel buttons it'd be fine, but it ends up very far from them12:39
danilosalgado, sure, it can be worked about with proper div nesting, but not a big deal imho12:39
fabozyga: could a lava job fails to download if the website doesn't have a valid certificate?12:41
danilosalgado, yeah, there do seem to be few kinks with error handling as well, but it does work, which is good12:41
zygafabo: is the download over https?12:41
fabozyga: yes, http://validation.linaro.org/lava-server/scheduler/job/15518/log_file12:42
zygafabo: which tool was used to download the file? wget/curl or some python code12:42
danilosalgado, it would be good to get rid of the "Work items:" inside the box as well, but I guess that's in for compatibility reasons12:42
zygayes it is possible12:42
zygafabo, some tools require an option to ignore certificates12:42
salgadodanilo, we reported a few related to validation. they're linked to https://blueprints.launchpad.net/linaro-infrastructure-misc/+spec/work-items-as-first-class-objects but feel free to file others if you find any12:42
zygale me read the log12:42
salgadodanilo, exactly.  no format changes on this first pass12:42
salgadothe time to break the external tools will come soon :)12:43
zygafabo, error reporting is implemented incorrectly there12:43
zygafabo, but from what I see we just got 40412:43
zygaso this would not be related to ssl12:44
rsalvetialf_: were you able to test precise and compiz?12:44
rsalvetialf_: I finally got all packages to build again yesterday12:44
rsalvetialf_: will give them a try later today12:44
fabozyga: I can download with wget --no-check-certificate12:44
faborsalveti: unity-2d only atm12:45
salgadodanilo, btw, is it ok to track those bugs we just filed on that same BP?  I'd think so and they should be easy fixes so we can squeeze them in this cycle12:45
rsalvetifabo: yup, at https://launchpad.net/~linaro-graphics-wg/+archive/daily/+packages?field.name_filter=&field.status_filter=published&field.series_filter=precise you can find the current unity 3d related packages 12:45
zygafabo: perhaps it was not there at the time12:45
rsalvetiif you also want to give them a try12:45
zygafabo: I recall we had some issues with stuff not being pushed on time, before a job is triggered from jenkins12:46
danilosalgado, I'd rather we have a new one for "cleanups" if this thing is rolled out12:46
danilosalgado, it gives me satisfaction to call something done, even if there are improvements to be made (and that's always)12:47
fabozyga: it doesn't seems that. I triggered a lava job this morning using yesterday images12:47
salgadodanilo, what do you mean "if it's rolled out"? of course it will be! :)12:47
danilosalgado, well, if our current work is not blocked for roll out before we fix those12:47
zygafabo: I would suggest fixing error reporting there12:47
salgadodanilo, I only call something done when there are no more bugs to fix in it.  that's why I've never called any piece of software done in my life ;)12:47
zygafabo: and retrying that job a few times12:47
fabozyga: I guess bug against lava-dispatcher?12:48
salgadodanilo, I'll register a new BP for those improvements12:48
zygafabo: better patch the code, ignore the paperwork12:48
fabosubmitted as job id: 1553112:49
fabolet's see12:49
alf_rsalveti: fabo: Just finished trying unity3d and it works! One caveat: I think that unity_support_test is broken at the moment because running from lightdm doesn't work, but running unity from a recovery console works! \o/12:52
faboyay \o/12:52
alf_rsalveti: fabo: There are some artifacts now and then, I can't get transparency in gnome-terminal, no global menu but nothing too serious12:53
rsalvetialf_: hm, will check the test case12:53
rsalvetialf_: great!12:54
alf_rsalveti: unity_support_test complains about pcilib: Cannot open /proc/bus/pci, which makes sense on a board that doesn't have pci :)12:57
rsalvetialf_: I guess I always had that complain in the past, but could be that this is affecting the results now 12:57
alf_rsalveti: ahh, I don't know then12:58
rsalvetiI know imx6 has pci-express, so we'll soon have boards with pci in it :-)12:58
rsalvetialf_: I'm also having quite a few glitches on my x86 precise desktop13:07
rsalvetiwith the gl based compiz13:07
Stskeeps     13:17
mabacdanilo, salgado https://blueprints.launchpad.net/linaro-infrastructure-misc/+spec/lp-workitems-qa-tracking13:23
danilomabac, thanks13:25
Sunil_Kmpoirier: hello, i want to clarify something on kernel 3.3 and ICS boot.13:35
mpoirierSunil_K: good day.13:35
mpoirierSunil_K: sure, go ahead.13:35
Sunil_Kmpoirier: We have used new LAMC and also uboot.. and now we are not facing vmalloc issue13:36
mpoirierSunil_K: good good.13:36
Sunil_Kmpoirier: but we want to understand from you, if you are able to boot ICS fully with kernel 3.313:36
mpoirierSunil_K: I have booted it yes.13:37
mpoirierSunil_K: but first i have a question.13:37
mpoirierSunil_K: above you write " We have used new LAMC and also uboot"13:37
Sunil_Kmpoirier: yes pls.. :)13:37
mpoirierSunil_K: what do you mean by "uboot" ?13:37
Sunil_Kmpoirier: i mean uboot-env.bin. But i understand that for ICS, it reads from scr file.13:38
Sunil_Kmpoirier: so it doesnt matter.13:38
mpoirierSunil_K: well,13:38
Sunil_Kmpoirier: its more like using latest changes in LAMC. 13:38
mpoirierSunil_K: u-boot-env.bin is needed to save the environment to emmc.13:38
mpoirierSunil_K: ok, got it. thanks for the clarification,13:39
Sunil_Kmpoirier: yes. we have used that as well.13:39
mpoirierSunil_K: about ics and kernel 3.3.13:39
mpoirierwhat is the problem you are seeing ?13:39
Sunil_Kmpoirier: we understand that there is HDMI issue and also when we tried to boot ICS 3.3 kernel image. Kenel prompt comes up. but then it stops booting.13:39
Sunil_Kmpoirier: it hags in android animation.13:40
mpoirierSunil_K: is the board powered in front of you ?13:40
Sunil_Kmpoirier: yes. 13:41
mpoirierSunil_K: do ls /dev/graphics/13:41
mpoirierSunil_K: you probably have a fb device but I want to make sure.13:41
Sunil_Kmpoirier: just one min pls.13:42
Sunil_Kmpoirier: it says fb013:46
mpoirierSunil_K: ok, that is good.13:47
mpoirierSunil_K: did you compile this image of you got it from an official build ?13:47
Sunil_Kmpoirier: we compliled and build by ourselves.13:47
Sunil_Kmpoirier: not an official build image13:48
mpoirierSunil_K: then i highly suspect an environment problem.13:48
mpoirierSunil_K: is there a way you can post the boot, system and userdata images somewhere ?13:49
mpoirierSunil_K: I'll try them on my side.13:49
mpoirierSunil_K: I need them as they came out of the build.13:49
Sunil_Kmpoirier: do you know how usually images are shared like this?13:49
Sunil_Kmpoirier: between, even we want to try working ics image.13:50
mpoirierSunil_K: I'm sorry, pls rephrase your last statement.13:50
Sunil_Kmpoirier: i mean we want to flash an working ICS kernel 3.3 image and try. do you think we can take any of the tar balls sets which are released?13:51
Sunil_Kmpoirier: because LAVA report mentioning as boot failures for all 3.3 buils.13:52
mpoirierSunil_K: that is because the vmalloc hasn't been fixed properly.13:52
mpoirierSunil_K: and despite...13:52
mpoirierSunil_K: the tracking builds will be broken a lot.13:52
Sunil_Kmpoirier: okay, but we also have heard about HMDI issue on 3.3 kernel. 13:53
Sunil_Kmpoirier: thats not blocking? 13:53
mpoirierSunil_K: I haven't seen any hdmi issues on android with 3.3 kernel.13:53
mpoirierSunil_K: it is an issue with ubuntu yes.13:53
mpoirierSunil_K: but with android I haven't seen it.13:53
Sunil_Kmpoirier: then in that case, can you suggest any official build from which i can take 3 tar balls and create single image for our present test.13:54
Sunil_Kmpoirier: we first of all want to reproduce the issues. 13:54
Sunil_Kmpoirier: later we will try replacing only the kernel image in that.13:54
mpoirierSunil_K: of course... i wrote many emails on this...13:55
mpoirierSunil_K: https://android-build.linaro.org/builds/~linaro-android/snowball-ics-gcc46-igloo-stable-blob/13:55
mpoirierSunil_K: this is the stable build.13:55
mpoirierin there you have stable toolchain and kernel 3.213:55
mpoirierSunil_K: I usually start from those and replace my kernel image with 3.3 that I compile myself.13:56
mpoirierSunil_K: that is the trick..13:56
mpoirierSunil_K: does that make sense ?13:57
mpoirierSunil_K: you can also take the tracking build, repo the code on your machine and compile with the stable toolchain.13:58
mpoirierSunil_K: but I still need your boot, system and userdata images.13:58
mpoirierSunil_K: as Espersson to fix you an account on the igloocommunity server.13:58
mpoirierSunil_K: that way we can exchange images.13:58
mpoirierSunil_K: still with me ?13:59
Sunil_Kyes i am..14:00
Sunil_Kmporior: Having some issue with my IRC editor.14:00
Sunil_Kmpoirier: Sorry for that.14:00
mpoirierSunil_K: did you get my last 10 posts ?14:01
Sunil_Kmpoirier: yes.14:01
mpoirierSunil_K: do you have any questions ?14:02
mpoirierSunil_K: earth is calling Sunil14:04
Sunil_Kmpoirier: shall we send you the trace? will that help?14:04
mpoirierSunil_K: no.14:04
Sunil_Kmpoirier: i am back okay now.14:04
Sunil_Kmpoirier: will b able to respond immediately.14:05
mpoirierSunil_K: I need the images you're working with.14:05
Sunil_Kmpoirier: so u need image.14:05
mpoirierSunil_K: ya, much easier.14:05
Sunil_Kmpoirier: okay. but whats the procedure to uplaod image. we need your help as we are doing for first time.14:06
mpoirierSunil_K: once you have an account on igloo you can sftp the images in your home.14:07
mpoirierSunil_K: I'll pick them up from there.14:07
Sunil_Kmpoirier: okay..14:07
mpoirierSunil_K: does that sound like a plan to you ?14:07
jk-danboid: luckily, Ubuntu's updates already fix the problem14:08
Sunil_Kmpoirier: yes, it does, Just that we need to cross check about access.. if its already provided.14:08
jk-danboid: so you could just wait for your kernel to be re-synced with the Ubuntu troo14:08
jk-danboid: or, contact the Linaro folks to make the change if you need it done before the next re-sync14:08
mpoirierSunil_K: yes, make sure you get an account on there.14:09
mpoirierSunil_K: in the mean time you can get the stable images and boot them on your boards.14:10
Sunil_Kmpoirier: yes, sure.14:10
mpoirierSunil_K: that will prove your environment is working.14:10
Sunil_Kmpoirier: https://android-build.linaro.org/builds/~linaro-android/snowball-ics-gcc46-igloo-stable-blob/ is using landing-snowball.xml and is 3.2 kernel right?14:12
mpoirierSunil_K: exactly.14:13
mpoirierSunil_K: install the image and boot your system.14:13
Sunil_Kmpoirier: that i have done before. I have used 176 stable release and its working14:13
mpoirierSunil_K: then clone the 3.3 android integration kernel, compile and subsitute.14:13
Sunil_Kmpoirier: we now want to try kernel 3.3 based ICS tar balls14:14
mpoirierSunil_K: they are the same in AOSP.  just the kernel is changing.14:14
Sunil_Kmpoirier: i also understand that tracking-snowball points to kernel 3.314:14
mpoirierSunil_K: correct.14:14
Sunil_Kmpoirier: can we use any working image from trcking-snowball?14:15
mpoirierSunil_K: sure you can.14:15
mpoirierSunil_K: but they may be buggy.14:15
mpoirierSunil_K: again, it's tip kernel and tip toolchian.14:16
plarszyga: Hi14:16
ChiThumpoirier, hi14:16
zygaplars, hi14:16
mpoirierChiThu: good day.14:16
ChiThumpoirier, I have latest ubuntu desktop installed in my snowball V11.14:17
mpoirierChiThu: ok.14:17
ChiThumpoirier, I also have a mouse connected to it with correct USB adapter. The mouse works14:17
mpoirierChiThu: ok.14:18
plarsdavepigott: you were looking for me earlier?14:18
ChiThumpoirier, now I want to connect a USB keyboard as well together with the mouse. How should I do _14:18
plarsChiThu: usb hub?14:18
davepigottplars: Yeah, just to let you know that all the audio cables were fitted.14:18
plarsdavepigott: awesome, thanks a lot for doing that!14:18
davepigottplars: I've also added the audio-loopback tag to all the appropriate boards14:18
mpoirierChiThu: do you have a usb hub ?14:19
plarsdavepigott: \o/14:19
ChiThumpoirier, any usb hub or other requirements such as powered usb hub ?14:19
plarsdavepigott: any ideas on why the load average of the server is running so high?14:19
mpoirierChiThu: it needs to be a powered hub.14:19
ChiThumpoirier, no but I plan to get one or buy one.14:19
plarsdavepigott: it's been between 8 and 11 since I got up this morning14:19
davepigottplars: No. I was wondering that too. Was really slow earlier. I had the ARM FastModels team in most of the morning so haven't had much chance14:19
mpoirierChiThu: ok, it has to be a powered hub.14:20
plarsthere are 16 jobs running at the moment14:20
davepigottplars: to look at anything14:20
ChiThumpoirier, any prefer branch or model ? which is the one you are using ?14:20
plarsdavepigott: right14:20
mpoirierChiThu: mean and cheap works best.14:20
plars...and definitely powered :)14:20
mpoirierChiThu: I have a super cheap 'Dynex'.14:20
ChiThudavepigott, do you have an other snowball to replace the snowbal02 ?14:20
davepigottplars: We now have a direct support line into the fast-models team too. I'll send out a mail with the details for those who need to know14:20
davepigottChiThu: No14:21
danboidjk-, Thats good to hear! Do you know when this next sync will occur?14:21
plarsdavepigott: there should be 7 on order right?14:21
davepigottplars: Last time I checked, yes14:21
ChiThudavepigott, I thought the LAVA lab has ordered more than 10 units ?14:21
plarsdavepigott: yeah, I've retried multiple times on snowball02 and it is still failing with the same problem pretty regularly, I think it's just a bad board14:21
davepigottChiThu: ^^14:21
zygaplars, snowball02 is bad?14:22
plarszyga: looks like14:22
plarszyga: keeps losing eth0 randomly14:22
Sunil_Kmpoirier: thanks for all inputs. we are trying here. if any input needed, i will ping you back. thanks again14:22
zygaplars, I'd like to get it if possible14:22
plarszyga: on health jobs.  same job works very reliably on the other boards14:22
mpoirierSunil_K: really push to get an account on igloo.14:22
zygaplars, right14:22
mpoirierSunil_K: that way we can start exchanging images.14:22
zygaplars, did we try re-imaging it?14:22
Sunil_Kmpoirier: yes. definitely.14:23
plarszyga: having a few issues right now... I bumped the priority of this serial truncation issue on panda14:23
plarszyga: I'm seeing it happen quite a lot14:23
zygaplars, I see14:23
zygaplars, what is the bug number?14:23
plarszyga: https://bugs.launchpad.net/lava-dispatcher/+bug/95346014:23
ubot2`Launchpad bug 953460 in lava-dispatcher "wget command truncated" [High,New]14:23
zygaplars, frankly it is hard to work on such bugs as our serial setup is unique14:23
zygawget ?14:23
zygawget is truncated or serial output is truncated14:23
plarszyga: it always seems to fail on that command14:23
* zyga looks14:23
plarszyga: the command itself14:23
plarszyga: not the output14:23
plarszyga: you'll see, read the bug14:24
zygaplars, so wget fails to properly download the image14:24
plarszyga: no14:24
plarszyga: it's not that14:24
plarszyga: look at the bug14:24
zygaI see it now14:24
ChiThuNo problem.14:26
zygaplars, this one: ?14:26
zygaroot@master:~#  [rc=0]: wget -qO- |tar --numeric-owner -C /mnt/lava -xjf -14:26
zyga[  165.196075] fmdrv: Timeout(5 sec),didn't get regcompletion signal from RX tasklet14:26
zygawget -qO- m.tar.bz2 |tar --numeric-14:26
zygatar: You must specify one of the `-Acdtrux' or `--test-label'  options14:26
zygaTry `tar --help' or `tar --usage' for more information.14:26
ChiThuwrong keyboard. Sorry.14:26
plarszyga: yes14:26
zygaplars, I see14:26
zygaplars, may I suggest something?14:26
zygaplars, I take that back...14:26
zygaplars, I need to think about it for a moment14:27
plarszyga: we are also getting some decently high load right now, and mwhudson noticed that there were some apache processes eating memory last night14:27
zygais alt-tab broken for anyone too?14:31
=== ofog_ is now known as ofog
alf_zyga: pfff, real programmers use wmctrl :D14:41
zygaalf_, :-)14:42
hrwfabo: did you ever got something like this with live-build? /usr/share/live/build/scripts/build/lb_chroot_archives: line 250: chroot/etc/apt/sources.list.d/linaro-overlay-ppa.list: Nie ma takiego pliku ani katalogu14:45
fabohrw: in english?14:46
hrwfabo: no file or directory14:46
hrwfabo: config/archives/linaro-overlay-ppa.list.chroot exists14:46
fabohrw: it doesn't trigger anything.14:47
fabohrw: have you been able to build an image or did you always hit this issue?14:48
hrwfabo: I was able to build nano-dev image - this one is alip-dev (only s/nano/alip/ in configure and package-list)14:50
tgall_foohmmm mobile device says meeting in 10 minutes ? no so mobile device says meeting in 1 hour 10 minutes ? which one to believe ??14:50
tgall_foohrw, that error message is usually because something is wrong with live-build14:51
hrwtgall_foo: the one with proper dst?14:51
tgall_fooleast if I'm reading that message right14:52
hrwrestarted build14:53
fabohrw: could you upload your nano-de and alip-dev on people so I can reproduce14:53
tgall_fooI should give your -dev versions a try hrw make sure this a45 stuff doesn't break it14:54
tgall_foorsalveti, ping14:54
hrwfabo: sure14:55
hrwfabo: wget -r -np http://people.linaro.org/~hrw/live-build/14:57
tmax2Hello i followed this guide to install Ubuntu on PandaBoard http://releases.linaro.org/12.02/ubuntu/leb-panda/. Is there anyway to create a permanent image instead of doing all the process every time i have to put Ubuntu on an SD Card ?14:59
manssonrsalveti, WG sync is one hour from now? (calendar confusion)15:03
fabotmax2: yes, you can use --image-file option to create a file instead of pushing to mmc15:04
tmax2and then how i put the img on my SD Card ?15:04
fabotmax2: with dd15:06
hrwsudo dd if=image of=/dev/sdcard bs=64k15:07
tmax2hooo ok thanks you :) i am investigating about that dd i dont know what is that :)15:09
danboidThee Data Duplicator15:13
tmax2mmm what is the function of bs ? It writes blocks of that size ? it is just for security ? I mean i can put 4M instead of 64k ?15:15
=== zyga is now known as zyga-afk
doanactmax2: you can do 4m15:16
dzindanilo: hey15:16
tmax2and which will be the difference ? just faster ?15:17
hrwhave a nice rest of day people15:19
hrwtmax2: all depends on card15:19
plarszyga-afk: load avg is back down to reasonable levels, was something crazy running earlier?15:20
pm215tmax2: the default block size is way too small and will take forever to write to an sd card. 4 meg or whatever should be fine though15:20
tmax2hoooo ok so a large value can corrupt writing also ?15:21
pm215tmax2: I don't think there should be a problem with a large block size, no15:31
tmax2ok thanks you for your explanation :)15:31
tmax2it is ok to create an image file called ubuntu ? sudo linaro-media-create --image-file ubuntu --dev panda --hwpack hwpack_linaro-lt-panda-x11-base_20120221-1_armel_supported.tar.gz --binary linaro-o-ubuntu-desktop-tar-20120223-1.tar.gz15:33
davepigotttmax2: of course. You can call it anything you like15:34
tmax2ok i have to put ubuntu.img or ubuntu ?15:34
davepigotttmax2: It's just a file. You can do either15:34
tmax2ok :)15:35
davepigotttmax2: This isn't a Window's system where the extension implies something about the file. :)15:35
* davepigott goes to wash his mouth out after saying the "W" word. :D15:36
tmax2Hoooo in Linux it is just a file, just bytes whatever extension is ? 15:36
davepigotttmax2: Linux doesn't care. You can make it take certain actions on file extensions, but you don't have to15:37
tmax2hooo ok thanks you :)15:37
doanactmax2: you'll probably need to include the --image_size flag. For a desktop image you are doing, you'll need it to be 3.5G at a minimum I think15:38
tmax2sudo linaro-media-create --image-file ubuntu --dev panda --hwpack hwpack_linaro-lt-panda-x11-base_20120221-1_armel_supported.tar.gz --binary linaro-o-ubuntu-desktop-tar-20120223-1.tar.gz15:39
doanactmax2: sorry - just checked. you can do 3G - but thats the smallest15:39
tmax2i have to put the img size ?15:39
doanactmax2: I think so. Otherwise it defaults to 2G which isn't big enough. You might want to do even bigger to use more of your SD card though.15:39
tmax2so where i should put the size ? I used it: sudo linaro-media-create --image-file ubuntu --dev panda --hwpack hwpack_linaro-lt-panda-x11-base_20120221-1_armel_supported.tar.gz --binary linaro-o-ubuntu-desktop-tar-20120223-1.tar.gz15:41
fabodefault size has been increased a long time ago. you don't need --image-size15:41
faboit's 3G now15:42
tmax2hooo ok thanks you :)15:43
heeendoes anyone have information about x11-less GLES on PowerVR hardware / omap415:51
rsalvetindec: ^15:59
rsalvetitgall_foo: pong15:59
rsalvetimansson: I believe it is 15:59
tgall_foohey rsalveti,  question for precise, are you of the opinion to basically just have armhf hwpacks and skip the armel ones or ?16:00
rsalvetitgall_foo: just armhf hwpacks16:00
rsalvetias it's the official abi ubuntu will be supporting for arm16:00
tgall_foorsalveti, ok 16:02
dzinamitk: is the WG-Platform meeting still in your calendar?16:02
dzinrsalveti: mansson: ping16:03
rsalvetidzin: hangout?16:03
rsalvetinobody is there it seems16:03
dzinI had same problem, close and try again16:05
=== omarrmz_ is now known as omarrmz
=== zyga-afk is now known as zyga
rsalvetialf_: I'll not join the unity meeting, with the timezone mess this is now conflicting with the plat-wgs meeting16:34
alf_rsalveti: np16:35
robtaylordoes anyone know who could I talk to about using hi-speed usb in host mode on the Origen?16:35
ndecheeen: rsalveti: well, we don't support x11-less quite well now... 16:48
ndecthough we wish we support it better ...16:49
ndecrobclark: any news or hope on ^^16:49
Stskeepswhy was the wsegls removed, just out of curiousiy?16:49
=== salgado is now known as salgado-lunch
ndecit was not removed per-say, it's just that the most recent changes we did in DDK broke it..16:52
kikorobtaylor, most people that would know are in bangalore and gone for the day17:00
kikorobtaylor, use linaro-dev@lists.linaro.org17:00
robtaylorkiko: ah, i shall, thanks :)17:01
kikopeople that would know are chander, tushar, probably thomas abraham17:01
zygadamn typo17:19
* zyga got something working just now17:19
plarszyga: ?17:22
zygaplars, almost there17:22
plarszyga: cool :)17:22
plarszyga: /me is eager to see it :)17:22
=== e-ndy|afk is now known as e-ndy
robclarkndec, wayland support adds back half of a nullws..    basically you can use kms+gbm plus surfaceless_gles2 extension in the same way that weston does..17:26
robclarkand with small bit extra we could make eglSwapBuffers() work17:26
rsalvetirobclark: yeah, that's true17:33
Espersson|2lag: great progress on the DT work! 17:37
robclarkfwiw, there is some work to add gbm_surface, which will make the whole egl connection a bit cleaner.. from egl perspective, gbm will be your winsys on server side (or if you want to use gles directly), and wayland-egl is your native winsys on the client side (if you use wayland apps)17:38
=== salgado-lunch is now known as salgado
lagEspersson|2: Thanks :)17:39
=== ofog_ is now known as ofog
abnerHi guys, I'm planning to create my own hwpack to be used by Lava. Is there any tutorial or specification for building hwpacks? I saw that there's a tool called linaro-hwpack-create.18:33
zygadanilo, ^^18:34
doanacabner: this may help: https://wiki.linaro.org/Resources/HowTo/AddingNewHardwareSupport18:36
abnerdoanac, I will take a look. Thanks!18:38
=== davidrusling_ is now known as davidrusling
plarsChiThu: hi, didn't see you around earlier, still have time for a sync in just a bit?18:49
alf_ogra_: Check the email I sent. You can start experimenting with creating/using the distro patch whenever you feel like it :)18:50
ogra_ah, sweet !18:50
abnerdoanac, my intention is to create a hwpack for a x86 that will boot only on qemu for now. Do you know if there's an existing config file that I could base on? It would be better than writing one from scratch..18:50
ogra_thanks so much, awesome job !18:51
ChiThuplars, sorry I missed. shall we synch now ?18:52
plarsChiThu: sure, but I need a few min first, ok?18:52
doanacabner: I'm confused. You want to build x86 target? or some type of ARM target running under an x86 host?18:54
abnerdoanac, no, real x86 target :)18:55
doanacabner: I'm not sure if you'd really need a hardware pack for this. hardware packs were a Linaro invention to faciliate deployment of ARM images. x86 is a bit different18:56
heeenndec: can you be more specific? what does work, what doesn't?18:58
ChiThuplars, I am on Mumble.19:03
plarsChiThu: I need like 5 or 10 min. still, ok19:05
ChiThuplars, no problem19:05
plarsChiThu: thanks :)19:05
mwhudsongood morning19:11
ndecheeen: null WS does not work at all. iirc (it was long time ago i looked into that) it doesn't even compile...19:15
zygamwhudson, morning19:35
zygamwhudson, have you seen: https://www.djangoproject.com/weblog/2012/mar/13/py3k/19:46
abnerdoanac, sorry, didn't see your msg. Yes, it's not really needed. But we would like to have this x86 image going through the same process as the arms ones. Which means the job specifying a hwpack + rootfs. We are aiming to run it using the qemu client support in lava and it's currently using lmc IIRC.19:47
mwhudsonzyga: i hadn't19:51
mwhudsonzyga: seems fairly sensible19:51
zyga1.6 will likely be the time we'll see lots of switchers19:51
kaparezyga, hi what at linaro depends on django? all web pages/framework?19:51
zygakapare, linaro has many different web parts19:52
zygakapare, lava uses django19:52
zygakapare, as for other services, I don't know19:52
kaparezyga, aaa thx ;) didn't try lava yet19:52
zygakapare, jenkins is based on java so that's ruling django out 19:52
zygakapare, if you have a specific question that might be easier19:53
kaparezyga, you are using django version?19:53
mwhudsonthe android-build frontend to jenkins is django though :-)19:53
kaparezyga, no specific question was just curious ;) thx19:53
mwhudson(well, it's mostly javascript)19:53
=== e-ndy is now known as e-ndy|afk
abnerit seems that it will be harder than I thought.. =\19:57
doanacabner: I would think it would be easier to update the LAVA dispatcher to handle some new command like "deploy iso image" that qemu could use.19:59
doanacBut I'm not familiar with how qemu currently works in LAVA19:59
zygadoanac, iso images are not suitable for testing19:59
doanaczyga: he wants to run an x86 vm not arm19:59
zygadoanac, lava really wants to run on a installed system19:59
zygadoanac, I know19:59
doanaczyga: so you think updating l-m-c to create x86 hwpacks makes more sense?20:00
zygadoanac, we could have more commands in lava to run from something else but getting a x86 hwpack/rootfs would be IMHO better20:00
abnereven lmc only support a specific set of targets.. so no i386 afterall20:00
zygadoanac, coupled with ubuntu core? yes20:00
zygadoanac, I think ubuntu core will be very important20:01
zygadoanac, and a tool to remix that in a preinstalled image would be a great asset20:01
zygadoanac, that does not have to be lmc 20:01
doanacseems odd to me, but I'm sure you know best on this topic :)20:03
zygadoanac, well a typical ubuntu iso is very hard to use for testing20:05
zygadoanac, it has an installer and requires deploymet to happen first20:06
zygadoanac, while it would be possible to support that I think it makes more sense to start testing on a pre-installed system20:06
zygadoanac, then there is ubuntu core which seems to be come a hot topic20:06
zygamwhudson, fun fact: celery has two APIs20:08
doanaczyga: good point.20:08
zygamwhudson, first one, originally for django, has global application/configuration20:08
zygamwhudson, and the second one, where you explicitly instantiate Celery object20:08
zygamwhudson, and configure it20:09
plarszyga, doanac: should maybe look at what ubuntu folks are doing here, they do iso installs with preseeding20:14
=== michaelh1|away is now known as michaelh1
zygaplars, yeah but they are interested in ubiquity testing20:14
zygaplars, testing the installer is nice but installing ubuntu to run stream sounds bad20:14
michaelh1lool, ibiris: just noticed the meeting.  I have a regular 1-on-1 at that time sorry.20:15
plarszyga, doanac: I think it could be made to work, of course if it has to be combined with the concept of a master image, things get nasty because it wants to explode itself all over your available disks.  But I don't think that's a concern here"20:15
plarszyga: it's more than that, they use it to preseed with the things they want to test, so that dependencies are already installed once the system comes up20:15
zygaplars, preseeding is just telling the installer what to get, right?20:16
zygaplars, and it still involves really installing the system under test20:16
plarszyga: basically, yes - setting all the install options so that you don't have to answer questions, but you can also specify extra packages to install20:16
ibirismichaelh1: right, your calendar shows free time at that slot, what time would suit you best? 20:16
zygaplars, so that may be fine for some cases but if you are interested in application testing you really want fast feedback: installing everything all the time is wasting time20:17
plarszyga: from their perspective, the iso is the thing they release, not the installed system.  So that's where they like to start from20:17
zygaplars, a tool to make preinstalled images allows you to turn on kvm/qemu/virtualbox and really execute only your test code20:17
zygaplars, right20:17
zygaplars, that's why I said ubuntu might be more interested in testing the installer20:18
michaelh1ibiris: it shouldn't.  The event is marked as busy.  I wonder why...20:18
zygaplars, if you look at application developers they want to test their code in a pristine environment of their specification 20:18
zygaplars, this is where ubuntu core and preinstalled images come into play20:18
ibirismichaelh1: maybe a bug of google. Regardless what is the timeslot we have this chat?20:20
michaelh1Rest of today is fine, but it's 9:30 pm for lool already.  I'm free from 0800 UTC to 0900 UTC20:21
michaelh1bero: happy 10th birthday!20:22
ibirismichaelh1: ok thanks I updated the invitation20:23
ibirislool: ^^^20:23
beromichaelh1: Huh? How so? 10 years of an account in gcc's bug tracker or something?20:24
plarszyga: hmm, so I guess I will really have to take a look at python 3 someday... I've only dabbled with it a long time ago, not looked recently or taken a hard look at all the changes20:26
zygaplars, it's nothing special20:26
plarszyga: I didn't get the impression it was *that* big a deal20:26
plarszyga: it wasn't like the debacle that is perl6 for example20:26
zygaplars, io is new, strings are new, old stuff is gone20:26
zygaplars, no, py3k was a flop because of the need for flag day and the continuous development of python220:27
alf_rsalveti: unity_support_test is not broken after all... it turns out I hadn't updated the nux-tools package and was still using the GL one20:27
michaelh1bero: oops.  My mistake.  /Arch/ Linux is 10 years old, not Ark.  20:27
rsalvetialf_: oh, good then :-)20:27
beroThanks anyway ;) Ark is about 3 months older20:28
loolmichaelh1: I'm around20:28
loolibiris: poke20:28
ibirislool: hi20:28
beroSo not really late ;)20:28
lool8 UTC is fine20:28
loolchecking the invite20:28
loolyep, looks god20:28
ibirisgod is good (usually) too ;-)20:29
alf_lool: poke = ircbook ? ;)20:29
loolpoke: http://1.bp.blogspot.com/_Pqpmm2T6j4g/TBTFgZwFJGI/AAAAAAAAIR4/dDDKbxr7SAw/s400/poke.jpg20:29
loolnot sure what ircbook is20:30
michaelh1Ack, that time is fine.20:31
alf_lool: irc + facebook mix :D20:32
loolgosh, what an awful idea!  :-)20:32
loolarf, I was wondering why I had a folder named merkel in my home dir, I was thinking WTF is Germany's chancellor doing there, but it's a backup of a Debian machine that I accessed some months ago "ha ha" </stupid me>20:36
svuorelalool: why is debian having a server named after germany's chancellor ?20:37
mwhudsonzyga: did you see my mail about sentry?20:38
zygamwhudson, aww, yes20:38
zygamwhudson, sorry for not responding, I had a hectic day20:38
zygamwhudson, in short:20:38
zygamwhudson, kill that stuff20:38
zygamwhudson, setup another screen20:38
zygamwhudson, change to sentry user, make sure to use sudo -H to get $HOME right20:39
zygamwhudson, run sentry start20:39
zygamwhudson, ^ac to get another screen20:39
zygamwhudson, do the same sudo dance to get sentry user20:39
zygamwhudson, run sentry celeryd 20:39
zygamwhudson, sorry, I should have done that myself20:39
mwhudsonzyga: there's a missing "sentry celeryd " in there right?20:41
mwhudsonno not that20:41
mwhudson. virtualenv/bin/activate20:41
zygamwhudson, that plus if you ran it as another user without fixing HOME you got the wrong config20:41
mwhudsoni ran sudo su -l sentry20:41
mwhudsoni'm pretty sure that gets $HOME and all that jazz right20:41
zygamwhudson, hmm, I'm not sure what --login does, is $HOME right?20:41
mwhudsonmy scheduler changes prevent one from submitting to an anonymous stream i think20:43
mwhudsonhm no20:44
zygamwhudson, have you found the bug yet?20:46
mwhudsonzyga: well sentry is working now20:46
mwhudsonbut i don't know why submit_job is always failing on staging20:46
mwhudsonwhich is why i wanted sentry :-)20:47
zygamwhudson, did you create the project for staging?20:48
mwhudsonzyga: yes20:48
mwhudsonand updated the dsn20:48
nekohayomwhudson, o/20:48
mwhudsoni was only running start though, not celeryd20:49
zygamwhudson, right, that was my second question20:49
nekohayoI'm looking at the lava qa tracker code a bit20:49
nekohayowas wondering if you had some suggestions of what to do with that hardcoded /home/mwhudson/tmp/test-cases path :)20:49
* zyga likes it when more people are hacking lava20:50
mwhudsonnekohayo: heh, i don't really know, no20:50
mwhudsonnekohayo: i guess you could put it in settings20:50
nekohayoI can't simply do a os.join(os.environ["HOME"], "lava-test-cases")20:50
nekohayosince the server is not run as the user :)20:50
mwhudsonin general something i want all the time in django is a way to have settings-like things in the db20:50
zyganekohayo, what do you want to achieve with that?20:50
mwhudsonso you can change them without restarting the process20:51
nekohayozyga, hopefully not require people to hack the code when installing it20:51
zygamwhudson, django-appconfig20:51
zygamwhudson, from django core dev20:51
nekohayomwhudson, I don't see any settings file in there...?20:51
zyganekohayo, ah 20:52
zyganekohayo, no that needs redesign20:52
zyganekohayo, it's not meant to work like that at all20:52
mwhudsonzyga: ah20:52
mwhudsonnekohayo: hardcoding the branch like that was just a temporary hack20:52
zygamwhudson, I gave you the wrong project name20:53
nekohayomwhudson, well yeah... pretty much the whole thing was a proof of concept, as I understand it20:56
nekohayonow it depends what your plans are, if you're going to rewrite major parts of it then it makes little sense for me to try to fix code nitpicks at the time being?20:57
nekohayoactually the thing seemed to "do the job"20:57
zyganekohayo, I think the best thing that could happen is that you'd adopt that part and help finish it20:59
mwhudsonzyga: you did?20:59
zyganekohayo, according to your own needs20:59
zygamwhudson, ?20:59
mwhudson<zyga> mwhudson, I gave you the wrong project name20:59
zygamwhudson, django-appconf is not what I meant20:59
mwhudsonzyga: you didn't mean https://github.com/jabapyth/django-appsettings#readme21:00
zygait's something different21:00
zygaI did look at it long time ago21:00
nekohayozyga, the question is, what path should be there instead? the instance's path? $INSTANCE/share/tests   ?21:02
nekohayoand if so I'm not sure how to get the instance's path from inside lava qa tracker's views.py21:02
zyganekohayo, you need to take a step back21:02
zyganekohayo, look at it from two angles: first - you should want to solve a problem, that drives design to working features21:02
zyganekohayo, from other angle, you should make it usable for other by designing a sensible library21:03
zyganekohayo, then having said that21:03
zyganekohayo, do you want to use the feature that loads test scenarios from bzr trees/21:03
zyganekohayo, do you want to switch to django database models for that instead21:03
zyganekohayo, or do something completely different21:03
zygamwhudson, sigh21:08
zygamwhudson, why oh why cannot django-and-friends adopt something other than "import that for settings"21:09
zygamwhudson, celery has the same disease21:09
zygamwhudson, it's really great if you use it but somehow I just don't like it21:09
mwhudsonzyga: it's terrible21:09
zygamwhudson, celery has settings big enough to fill a titanic of command line options21:10
mwhudsonzyga: more than uwsgi!!!?!21:10
zygamwhudson, start with all the backends first21:10
zygamwhudson, oh yeah21:10
zygamwhudson, each one presents a few options, some many more than others21:10
zygamwhudson, then results21:10
mwhudsonzyga: have you looked at django-configglue?21:10
zygamwhudson, yeah I did21:10
mwhudsoni believe james_w et al use that21:10
zygamwhudson, it did seem sane and some folks used it at canonical21:11
=== davidrusling_ is now known as davidrusling
james_wwe do21:11
mwhudsonzyga: i mostly like our extensions + django-debian thing we have now21:11
zygamwhudson, but at the time I already wrote django-debian and that kind of gave a second settings layer21:11
mwhudsonzyga: although json is not the friendliest thing to edit21:11
zygamwhudson, plus AFAIR django-configglue had something I disliked, but I may be wrong21:11
zygamwhudson, it may have been in configglue itself21:11
zygamwhudson, I do agree21:11
zygamwhudson, I was leaning towards yaml21:12
zygamwhudson, yaml is nice, except when you make a pointer loop by accident21:12
zygamwhudson, I wish someone made small-yaml or something, without all the "this-punctuation-is-valid-exotic-option" gotcha moments21:12
mwhudsonzyga: and the fact that you can even make a pointer loop is a pretty black mark for a frigging config file format21:12
mwhudsonbut i agree, everything is horrible21:13
mwhudson.ini files are pretty bad21:13
james_w+ the fact that there is no readable definition of the format anywhere I can find :-)21:13
zygamwhudson, and that no parser probably parses everything the spec defines21:13
james_wI gave up looking last time I wanted to edit some yaml21:13
zygajames_w, yeah, so I share your view here :)21:13
zygayaml is nice but don't tell people it's yaml21:13
zygaand stick to : - and *21:14
zygamwhudson, json is nice for strictness, sometimes to a point where I hate it21:14
zygamwhudson, ini may be a good-enough compromise21:14
zygamwhudson, especially with that odd syntax from ConfigParser or something, where you can nest sections21:14
mwhudsonzyga: one of the problems with ini is the configparser is just horrrrrrrrrrrrrrrrrrible21:14
zygamwhudson, I wrote dpkg-splitpackage a while ago; after initial flame I extended the config system to support ini alongside .json without any changes to application code really21:15
zygamwhudson, haha21:15
zygamwhudson, how horrible? I did not use it much TBH21:15
zygamwhudson, (I did like that I could still run json validation on the stuff it gave me)21:15
mwhudsonzyga: it just doesn't present an api i find useful21:15
mwhudsonit does some things nicely, but i just hate using it21:16
mwhudsonyou have to have sections, for one thing21:16
zygamwhudson, well the code can be thrown away, it's really the format that matters21:16
zygamwhudson, I think that's INI requirement (informal)21:16
james_wbecause everyone loves [__main__] :-)21:17
zygapip install mordor && python -c 'import mordor' <- twitter fun21:18
zygamwhudson, sentry keeps talking to me21:22
zygamwhudson, I like the frequency graph21:22
asac_michaelh1: i dont see the log with the version string attached to your mail :)21:22
michaelh1asac_: it's a dialog on my phone and no obvious way to save :(21:23
mwhudsonzyga: yes, that's me21:23
asac_michaelh1: what does our version string look like :)?21:23
zygamwhudson, wow, niceness everywhere21:23
zygamwhudson, we're running 3.x sentry now21:23
mwhudsonzyga: it's bizarre, the same code works on my machine21:23
* asac_ tries to find a binary build21:23
* michaelh1 types...21:24
mwhudsoni guess i should check i'm running _exactly_ the same code...21:24
michaelh1asac_: 'gcc version 4.6.3 20120105 (prerelease) (crosstool-NG linaro-1.13.1-2012.01-20120125 - Linaro GCC 2012.01)'21:25
zygamwhudson, we need to change sentry sender address to something filter friendly21:34
mwhudsonzyga: it's sending email!?21:35
zygamwhudson, to me, yes21:35
mwhudsonoh ok21:35
zygamwhudson, that's why I said it's started talking21:35
zygamwhudson, I like the new 500 error page21:36
zygamwhudson, I mean I like the OOPS id so to speak, we really need artists to make it pretty :)21:36
mwhudsonah right21:36
mwhudsonzyga: it's sending me mail too21:36
mwhudsonzyga: how do i stop that?21:37
* mwhudson pokes around21:37
zygamwhudson, look at ~sentry/.sentry/sentry.conf.py21:39
zygamwhudson, I've enabled that, it's off by default21:39
araujohello ... the omap3 hwpack should work fine for beagle/beagle-xm boards right?21:40
zygamwhudson, eh21:40
zygaaraujo, correct21:40
zygamwhudson, can I just surrender and give in to import_me_for_settings21:40
araujothanks zyga++21:40
mwhudsonzyga: for what?  celery?21:40
zygamwhudson, yes21:40
loolsvuorela: actually Debian names machines after composers -- but I bet you knew that already  :-) -- http://en.wikipedia.org/wiki/Gustav_Merkel21:40
zygamwhudson, btw: if you have 3 minutes _do_ look at kombu/__init__.py21:41
zygamwhudson, it's pretty interesting, I always wanted something like that so that I can have pretty imports and safe __init__21:41
loollistz, gluck, handel etc.21:41
* svuorela smiles to lool 21:41
mwhudsonzyga: would you object violently if i unchecked "Send emails to all members of this project." on http://sentry.validation.linaro.org/projects/2/plugins/mail/21:41
zygamwhudson, not at all21:41
mwhudsonzyga: done21:42
zygamwhudson, but we really need _someone_ to look at that21:42
mwhudson(also for staging)21:42
zygaI'd like digest mode 21:42
zygadaily stats21:42
zygacould be cool21:42
zygamwhudson, python aint no c++, that's one for sure :-)21:43
mwhudsonzyga: what's this __init__ doing?21:43
mwhudsoni can see it's replacing the module with a custom class21:43
zygamwhudson, creating a module that lazily imports stuff21:43
mwhudsonso you can say kombu.submodule without having to say import kombu.submodule?21:44
zygano, you can always do that in python21:44
zygait means you can say import kombu21:44
zygaand not start importing stuff 21:44
mwhudsonno you can't :-)21:44
zygayet when you do kombu.foo you do21:44
zygamwhudson, os.path?21:44
zygaor better yet21:44
zygaI'll check21:44
mwhudson>>> import lava_dispatcher21:44
mwhudson>>> lava_dispatcher.job21:44
mwhudsonTraceback (most recent call last):21:44
mwhudson  File "<stdin>", line 1, in <module>21:44
mwhudsonAttributeError: 'module' object has no attribute 'job'21:44
mwhudsonos.path is different21:45
mwhudsonin os.py there is code that says "import posixpath; path=posixpath"21:45
mwhudson(which predates packages in python at all)21:45
zygamwhudson, or should I write a simple module class that loads json settings again?21:46
zygamwhudson, (back to celery config sucks)21:46
zygamwhudson, I mean, celery is better21:46
zygamwhudson, it has explicit apps you can config21:46
zygamwhudson, but then if you spawn celeryd you're back to square 121:47
* zyga keeps poking21:47
mwhudsonzyga: how would you spawn celeryd?21:47
zygaceleryd --help?21:47
zygaceleryd is celery.bin.celeryd21:47
mwhudsoncan you not do it via lava-server manage somehow?21:47
mwhudson(and so get all our settings love)21:47
zygathat just imports apps.worker and gives it a ton of config options21:47
zygamwhudson, ah, we dig deeper, there is another problem with that21:48
zygamwhudson, lava-tool is standalone21:48
zygamwhudson, perfect solution will be seamless in lava-tool / lava-server mode21:48
zygamwhudson, I was thinking about this 21:48
zygamwhudson, if lava server could be a lava-tool command (running like sentry now does)21:48
zygamwhudson, then perhaps it _would_ make sense to have ~/.config/lava/21:49
zygamwhudson, and a system where anything in lava.* asks for configuration21:49
zygamwhudson, and then that comes from "the right place"21:49
zygamwhudson, be it per-user, per-instance or some global config21:49
zygamwhudson, imagine this: 21:49
zygamwhudson, pip install lava-stack21:49
zygamwhudson, lava init 21:49
mwhudsonzyga: i think i need a coffee to grok this :)21:50
zygamwhudson, lava server start21:50
zygamwhudson, I guess the question I'm asking is: should settings vary based on the presence or lack of lava-server 21:50
mwhudsoni was assuming that we'd have the same code on all nodes21:54
mwhudsoni'm sure it's possible to not run like that21:54
mwhudsonbut sanity seems much more likely if we do21:55
mwhudsonzyga: or do you mean something else?21:55
zygamwhudson, sure but this is for the dispatcher21:57
zygamwhudson, I'm cloudifying it21:57
zygamwhudson, and dispatcher is still happily agnostic about anything lava-server21:58
mwhudsonah right21:58
zygamwhudson, I had a plan to turn it into lava-tool command21:58
zygamwhudson, and cloudify lava-tool21:58
mwhudsonwill it still have a boring old non-cloudy version?21:58
zygamwhudson, and then I started doing settings21:58
zygamwhudson, yes21:58
zygamwhudson, all commands could actually have transparent cloud mode21:58
zygamwhudson, and even the dispatcher, as we start taking it apart21:58
mwhudsonthen i kind of thing it would be ok to have the cloudy version depend on lava-server21:59
mwhudsonbut i haven't thought about this much21:59
zygamwhudson, could run eagerly inside the current process without any cloud 21:59
mwhudsonhere's a different question: what answer would let you get this into production the quickest?21:59
* mwhudson afk for 5 again21:59
zygaI need  to take a break and think about it21:59
heeenndec: what exactly is null WS, no windowing system?22:06
ndeci thought this is what you were looking for, right?22:07
ndecwhat's your use case?22:07
ChiThumwhudson, plars : panda09 truncated the serial port again. http://paste.ubuntu.com/884009/22:18
plarsChiThu: yep, seeing a few of those today22:18
plarsChiThu: I don't think it's new, but it makes it easier for us to see it now22:18
plarsChiThu: I have at least one idea on how to reproduce it22:19
plarsChiThu: will try to get around to it later tonight22:19
ChiThuplars, is all the panda board connect to some kind of serial box server ?22:20
plarsChiThu: terminal server, cyclades22:21
plarsChiThu: just like the other boards22:21
ChiThuplars, except snowball boards.22:22
ChiThuplars, I have no clue myself.22:22
ChiThuplars, snowball01 has a dead sd-card. It boot the image on eMMC. http://paste.ubuntu.com/884020/22:26
plarsChiThu: did you open a bug and assign it to dpigott?22:26
plarsChiThu: I can't do much about it from 5000 miles away :)22:26
ChiThuplars, I known. dpigott getting busy with the lab.... I will open a bug now22:27
plarsChiThu: he's *always* busy with the lab, we need a clone for him :)22:27
mwhudsonthis might be a silly question22:30
mwhudsonbut does talking to a board involve flow control at all?22:30
mwhudson(over a serial line)22:30
ChiThumwhudson, good point about the flow control of serial line to the board.22:32
mrumost boards don't seem to use flow control at all22:33
zygamwhudson, I don't think it does22:34
mwhudsonso if you send data too fast, it just drops the extra?22:34
zyga-> tcp/i22:38
zygaserial for maybe some part 22:38
michaelh1mwhudson: I've done evil things like use a inter-character delay before.  Hardware flow control should work but you need to check it all the way through.22:38
ChiThumwhudson, will it work with lower baudrate, like 9600 ?22:39
zygaChiThu, no, it will be worse22:39
zygaChiThu, the point is that if something writes too fast it _may_ get lost22:40
plarsmichaelh1: we've thought about it, yeah but it could get pretty painful even adding 1/10th of a sec between each char22:40
akgranerwookey, ping you around?  22:40
michaelh1You can also pick the FIFO length, send a burst of ~16 bytes, then pause.22:40
ChiThuzyga, why lower baudrate not better ? 22:41
zygaChiThu, the problem is caused by lack of any flow control22:41
mwhudsonplars: i don't think it would be that bad22:41
mwhudsonplars: we never send very large strings over serial22:41
zygaChiThu, if you have less bandwidth and still push the same volume of data more bits are lost22:41
mwhudson(large ones come the other way of course)22:41
zygamwhudson, do you think it depends on size?22:42
ChiThuzyga, the volume of data is minimunm. just a couple of lines.22:42
zygaChiThu, that's not the point22:42
ChiThuzyga, it is how fast it deliver to the board.22:42
zygaChiThu, well, no22:43
mwhudsonzyga: the size matters if we delay between each character22:43
zygaChiThu, etherent does not loose frames because you write them too fast, there are things along the stack that prevent you from doing this22:43
zygamwhudson, ah22:43
zygamwhudson, that's insane :)22:43
zygaChiThu, serial does because there is no synchronization on the reader22:43
zygamwhudson, hmm22:45
zygamwhudson, idea22:45
zygamwhudson, what if we don't really need to send anything at all over serial22:45
zygamwhudson, (well apart from the boot process)22:45
zygamwhudson, nor over ethernet22:46
mwhudsonthat said, it seems i can spam large amounts of data down the serial line to my panda without losing any22:46
zygamwhudson, what if we convert the scenario/job file to a script (or even install an interpreter)22:46
zygamwhudson, perhaps we should write a test22:46
zygamwhudson, and then run that on boot and simply wait for the test to finish22:46
zygamwhudson, we'd loose real-time output22:47
zygamwhudson, but the process would be more robust22:47
ChiThuzyga, how do you get the script down to the board ?22:48
zygaChiThu, same way we install everything else22:48
zygaChiThu, right now that's ethernet22:49
zygaChiThu, we could just copy it over22:49
mwhudsonhm, maybe there is a limit of how much date can be sent at once22:50
mwhudsonseems like it's about 4k though22:50
michaelh1mwhudson: I've seen problems with u-boot on the OMAP3 when setting environment.22:51
michaelh1If you dump ~3 lines of setenvs into a terminal it falls behind.  Keep it in < FIFO length chunks and it's fine22:51
zygamichaelh1, uboot serial could be implemented differently from kernel serial22:53
zygakernel may be doing a better job 22:53
zygaat trying to ensure the buffer is sanely purged and further writes blocked22:53
mwhudsonmichaelh1: what is FIFO?22:53
michaelh1zyga: it is different, which was my point.  u-boot polls the UART which means if it's off doing something then the FIFO (buffer) fills up and you lose characters.22:53
michaelh1The kernel uses DMA which will survive much longer.22:53
michaelh1mwhudson: FIFO is the first in first out receive buffer built into the UART.  They're normally 16 or 64 bytes long.22:54
mwhudsonmichaelh1: ok22:54
mwhudsonmichaelh1: i can definitely spam way more than that over the serial line in one go22:55
zygamichaelh1, I was hoping that after a few decades modern serial controllers would get interrupts and local buffers so that writing a few K would not loose data like that but perhaps I misunderstand how serial works at that level22:55
mwhudson(to my panda connected @ 112500 baud via usb serial)22:55
michaelh1Where is the data being lost?  What's running on the Panda at the time?22:55
mwhudsonmichaelh1: bash22:55
mwhudsonmichaelh1: basically we run a command like "wget -qO- $url | tar -xfz --numeric-owner $location"22:56
mwhudsonmichaelh1: and occasionally it seems bash only runs up as far as "wget -qO- $url | tar -xfz --nu" or something22:56
mwhudsonmichaelh1: example jobs linked in https://bugs.launchpad.net/lava-dispatcher/+bug/95346022:58
ubot2`Launchpad bug 953460 in lava-dispatcher "Serial input truncated on panda" [High,New]22:58
michaelh1I wonder if it's something else?  My boards won't even respond to pings when I'm untaring as it hammers the USB too hard22:58
mwhudsoncould be, but this is before the tarring happens22:59
michaelh1BTW, why are you using serial instead of SSH?  Gives you flow control, checksuming, and retransmission.22:59
mwhudsonand in http://validation.linaro.org/lava-server/scheduler/job/15200/log_file at least its the first transfer that fails22:59
zygamichaelh1, ssh is problematic 22:59
mwhudsonmichaelh1: dunno really, we need to do serial for boot control22:59
zygamichaelh1, keys, etc22:59
zygamichaelh1, I'd use normal TCP/IP instead of ssh in the lab23:00
zygamichaelh1, plus ssh is slowish on slow cpus23:00
zygamichaelh1, we could probably switch to driving execution over tcp/ip but we don't have the code for it yet23:00
mwhudsonof course the wget|tar approach sucks terribly for other reasons23:01
michaelh1zyga: no, it's straight forward and the CPU usage is tiny for what you're doing.  A common passwordless key and StrictHostKeyChecking=no gives you telnet like loging without yet another protocol23:01
michaelh1Anyhow, not sure about the serial problem.  You can always tee it to a laptop as well to see if the fault is in the Panda or rest of the system.23:02
zygamichaelh1, for shell, maybe, I think that using a dedicated agent is better as the same agent API could run on android with crippled userspace 23:02
michaelh1zyga: don't invent new tools when ssh and bash will do the job.23:03
zygamichaelh1, once the agent starts up we just talk to it, like ssh but focused on lava, running everything as shell _is_ fragile23:03
zygamichaelh1, bash scripting gave us /etc/init.d/, no thanks23:03
zygamichaelh1, I debugged a CGI script written in bash, running on a "consumer" device, it was driving the web UI23:04
michaelh1zyga: a random, specialised tool should only be created when other options fail23:04
zygamichaelh1, iPod proves otherwise23:04
michaelh1Go on?23:04
zygamichaelh1, integrated solutions work best if done properly23:04
zygamichaelh1, writing a new service that runs on our boards might seem like an overkill23:05
michaelh1Yes, and NIH is the best way to blow out a schedule and introduce bugs23:05
zygamichaelh1, that's also true23:05
zygamichaelh1, but if you look at the volume if issues we have now, at least some of which are easily traced back serial + shell being flaky, how much time did we loose on that?23:06
zygamichaelh1, mind you, virtually all of the team is focused on fixing that23:06
mwhudsonwe have three total mystery bugs currently23:07
mwhudsonone is this truncation thing, one is why the origen jobs don't get timed out, and the other is why wget | tar sometimes fails23:08
mwhudson(although there is a theory for the last one)23:08
michaelh1I disagree that shell is flakey.  Serial should only be used where it has to be unless you run a layer like PP on top of it.23:08
zygamichaelh1, did you see how the dispatcher is implemented?23:08
zygamichaelh1, the kind of stuff it resorts to to work23:09
michaelh1No, I haven't looked into LAVA.23:09
zygamichaelh1, that would all be removed and replaced by board.run_test('stream')23:09
zygamichaelh1, sure, shell might be fine for stuff23:09
zygamichaelh1, but shell is our RPC layer23:09
zygamichaelh1, with PS1 edits23:09
zygamichaelh1, and grep-friendly output so that we spot "aha, this is done"23:10
zygamichaelh1, all driven by pexpect23:10
michaelh1Sorry, are you saying shell when you mean console?23:10
zygamichaelh1, ?23:10
zygamichaelh1, I mean we really run sh there and type commands from lava23:10
zygamichaelh1, what is console that you refer to?23:10
michaelh1A console is where you run sh and type commands23:11
zygamichaelh1, ah, no I mean that we use 'sh' 23:11
michaelh1Well, normally the even worse version where it's also over serial23:11
michaelh1Anyhow, it's your architecture.23:12
zygamichaelh1, I agree23:13
michaelh1GDb used expect as at the time it only had a command line interface.  It's fragile and people  brought the same infrastructure to GCC which is better driven other ways.23:13
michaelh1Right, lunch.23:13
mwhudsonzyga: how do you think an agent should work?23:15
mwhudsonwe could do something as basic as something that listens on a socket and passes what it gets to os.system23:15
zygamwhudson, it should have a link layer (could be serial, tcp or anything else that people come up with) and action layer where we get to do stuff, I hate to say it but it looks like json-rpc 2.0 to me (with asynchronous notifications) or something equivalent23:16
mwhudsonzyga: i guess the question is how fat the client side should be23:16
zygamwhudson, it's FAT ALREADY :-)23:16
zygamwhudson, we install bzr 23:16
zygamwhudson, what are we talking about here23:16
mwhudsonwould it take actions like "deploy the rootfs from this url"23:17
zygamwhudson, one thing to think about23:17
mwhudsonor something more like what we do today?>23:17
mwhudsoni mean conceptual fat :)23:17
zygamwhudson, initially I'd add a few primitive commands (run shell)23:17
mwhudsonwe'd need it in both master and test image of course23:17
zygamwhudson, but yeah, later on it'd translate directly into the stuff from the dispatcher23:17
zygamwhudson, yes23:17
zygamwhudson, an update/install mode might be nice23:17
mwhudson... run a celery node on the board? :-)23:17
* mwhudson runs away23:17
zygamwhudson, I'd like the agent to be frozen after a few weeks of development23:18
zygamwhudson, we ... could do that23:18
zygamwhudson, I'd like to push code to it so that master could be updated without updating the real image23:18
zygamwhudson, and so that test images don't need to get it 23:18
zygamwhudson, (we could put lava-agent in nano if it is small enough)23:18
zygamwhudson, android is interesting23:19
mwhudsonhm yeah android is interesting23:19
zygamwhudson, we may build java agent23:19
mwhudsoni mean, if it was a super thin agent you could implement as a statically linked c binary maybe23:19
zygamwhudson, maybe that's pointless, I don't know much about what android is capable of with adb23:19
mwhudsonbut that also seems insane23:19
zygamwhudson, u-boot agent?23:19
zygamwhudson, that's equally interesting23:20
mwhudsoni didn't mean that23:20
zygamwhudson, I know23:20
zygamwhudson, uboot probably deserves pyexpect driven tests more than everything else23:20
zygamwhudson, for now my main concern is android and ubuntu23:20
zygamwhudson, is android immune to our three key bugs?23:21
mwhudsonyes, sure23:21
mwhudsonzyga: well, two apply during deployment23:21
zygamwhudson, I see23:21
zygamwhudson, wait23:21
mwhudsoni don't know about the third, i've only seen that during lava-test install23:21
zygamwhudson, but if ubuntu master ran an agent23:21
zygamwhudson, we'd probably not encounter them right?23:21
mwhudsonthat would certainly be a great improvement23:22
zygamwhudson, the bugs are: serial drops, ... (looks at scrollback)23:22
zygamwhudson, quick thought: a python one liner that spawns a socket and runs eval on that ;-)23:22
mwhudsonzyga: well yeah23:23
mwhudsonthere are some issues23:23
mwhudsonlike how do you find out the ip address to connect to?23:23
zygamwhudson, a few options here: hardcoded address is good (1) discovery (deps suck) (2) explicit kernel config option (3) preseeded config file from lmc hook or post-process step done on deployment23:24
zygamwhudson, discovery has this extra magic feeling to it23:24
zygamwhudson, imagine plugging a specially-crafter panda to your network and seeing it pop up on lava listing 23:25
zygamwhudson, (well discovery is also known to fail)23:25
mwhudsoni guess we can hard code an address in /etc of the master image23:25
mwhudsonnot sure what dave would think of that though23:25
zygamwhudson, I think 2 is best if we can (no image shuffling) and 3 might work anywhere23:25
zygamwhudson, if we put the agent in the image somehow23:25
mwhudson(having all master images be the same has some advantages for him, i imagine)23:25
zygamwhudson, we can feed the config at that time23:25
zygamwhudson, ah23:26
zygamwhudson, I was playing with nbd booting last weekend23:26
zygamwhudson, it basically can work23:26
zygamwhudson, but needs kernel changes -- but that's okay master kernel can have a better cofig 23:26
zygamwhudson, but those changes could be easily integrated into the generic config23:26
zygamwhudson, sadly currently our kernel configs differ quite vastly between boards (probably an oversight, nothing more)23:27
zygamwhudson, so with master booting from nbd we'd have perfect masters - only bootloaders and kernel binary could be corrupted23:27
zygamwhudson, I have not tried this on stuff like origen23:27
zygamwhudson, reading a post to linaro-dev it seems that origen has amazing 1.2Mbit USB hub so network is ... well... slow23:28
mwhudsonis it origen that does not have on board ethernet?23:28
zyga(link aggregation with serial would be measurable ;-)23:28
zygamwhudson, yeah, and beagle23:28
zygamwhudson, well panda too but AFAIR panda has USB ethernet 23:28
mwhudsonwe don't have beagles in the lab do we?23:29
zygamwhudson, we have 523:29
mwhudsonbeagle xm's have an ethernet port23:29
zygamwhudson, they are off because we cannot script their bootloader somehow23:29
mwhudson(via usb, but still)23:29
zygamwhudson, true23:29
zygamwhudson, I think we have xm's only23:29
zygamwhudson, I still have C4 so I care ;-)23:29
mwhudsoni see :)23:29
zygamay it rest in peace23:29
mwhudsonndb booting would not really help this directly right?23:29
zygait would 23:30
mwhudsononly in that it would make master image maintenance a bit saner23:30
zygaI still suspect we have flakiness in our masters that we never really measured properly23:30
zygamwhudson, different masters/corrupted masters (over time)23:30
mwhudsonbut it wouldn't e.g. make deployment easier, or help with serial truncation23:30
mwhudsoni can see that it might help with general flakiness23:30
zygamwhudson, no that's orthogonal23:30
mwhudsonbut not with the specific flakiness we were talking about 30 mins ago23:31
zygamwhudson, I wonder why cannot we just mount the test image folder 23:31
zygamwhudson, and copy it over something that is better than http at error correction and recovery23:31
zygamwhudson, you are right23:31
mwhudsonzyga: oh, right, replace wget | tar with nfs or something and just untarring a file?23:31
zygamwhudson, can we nfsmount/smbmount and DD the image over from a network filesystem to the SD card partition?23:32
zygamwhudson, right23:32
mwhudsonsure, i don't see why not23:32
zygamwhudson, might be more reliable 23:32
zygamwhudson, without incurring "flag day, rewrite" bit23:32
mwhudsonzyga: experiments needed i guess23:32
zygamwhudson, last call on how I do config for celery?23:32
mwhudsonzyga: whatever is easiest23:32
zygamwhudson, lava-server vs standalone tool from shell23:33
mwhudsoni wonder where we are with dongles23:33
mwhudsonthe one that was made during connect didn't work iirc23:33
zygamwhudson, probably far away23:34
zygamwhudson, still dongles don't solve this ;-)23:34
mwhudsonno not really23:34
mwhudsonthey're in a similar area though23:34
zygamwhudson, NBD might solve enough of this23:34
mwhudsonok, i'm going to have lunch i think23:34
zygamwhudson, well...23:34
zygamwhudson, if we deployed everything on the host then yeah23:34
zygamwhudson, see you23:34
mwhudsonzyga: good luck with celery23:35

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