IRC logs for #baserock for Tuesday, 2015-10-13

*** gtristan has quit IRC01:47
*** gtristan has joined #baserock02:19
*** brlogger has joined #baserock02:48
*** inara has quit IRC04:30
*** inara has joined #baserock04:35
*** bjdooks has quit IRC04:42
*** bjdooks has joined #baserock04:42
*** zoli_ has joined #baserock05:25
*** gtristan has quit IRC05:36
*** zoli_ has quit IRC05:53
*** zoli_ has joined #baserock05:56
*** gtristan has joined #baserock06:14
*** zoli_ has quit IRC06:32
*** paulwaters_ has joined #baserock06:39
*** zoli_ has joined #baserock07:15
*** ssam2 has joined #baserock07:18
*** ChanServ sets mode: +v ssam207:18
*** persia has quit IRC07:50
*** persia has joined #baserock07:51
pedroalvarezgood morning baserockers08:01
*** Travis__ has joined #baserock08:05
*** bashrc_ has joined #baserock08:06
*** toscalix__ has joined #baserock08:06
pedroalvarezmeh, the kernel I compiled yesterday doesn't work..08:06
pedroalvarezsame "no drm device found"08:07
*** Travis__ has left #baserock08:07
* persia has trouble remembering whether "Coordinated Greeting Time" or "Universal Greeting Time" is preferred for timekeeping these days08:08
persiaGood morning.08:08
ssam2happy timeless existance everyone!08:17
* paulsherwood is seeing odd errors on moonshot...
paulsherwoodthe actual problem seems to be the mount?08:21
pedroalvarezis it using linux-user-chroot?08:22
pedroalvarezpaulsherwood: seen it before:
paulsherwoodpedroalvarez: wow! thanks08:25
paulsherwoodit is using linux-user-chroot. normally i don't08:25
*** mariaderidder has joined #baserock08:27
paulsherwoodpedroalvarez: sadly my launch script is already doing the recommended magic08:28
pedroalvarezpaulsherwood: I know it is.. but I think it's a really fragile magic..08:28
*** edcragg has joined #baserock08:29
paulsherwoodright so my 'solution' would be to remove linux-user-chroot :/08:29
pedroalvarezI'd try a reboot first... maybe the fact that the script doesn't unmount things at the end is messing things up08:30
pedroalvarezand you have several mounts in the same place, etc..08:30
paulsherwoodreboot didn't fix it - scrapping l-u-c did08:42
*** jonathanmaw has joined #baserock08:46
*** franred has joined #baserock09:06
pedroalvarezok, I think I'm very blocked with getting drm working on rpi2..09:15
jonathanmawhooray, overnight GDP on x86 and ARM both got to qtwebkit. Jetson doesn't appear to be stuck, since top shows a bunch of cc1plus processes09:15
pedroalvarezwe need to start rebasing that monster09:16
jonathanmawcc1plus processes on x86 are much lower down, though, not sure if they've hung09:16
richard_mawlorry-controller's database seems to stay at 87M when jobs more than a day old are cleaned up, though I'd prefer to see if 3 days worth is sufficient, since then you'll still have entries for lorry failures over the weekend09:31
edcraggpedroalvarez: blocked how?09:35
pedroalvarezedcragg: I wish knew...09:39
pedroalvarezedcragg: I'm trying to follow these instructions
pedroalvarezand they didn't work for me, although I'm trying with wayland/weston and not with X09:41
edcraggpedroalvarez: ah ok, so it builds but doesn't work?09:43
pedroalvarezI'm worried about "A port to the downstream 2709 kernel for Raspberry Pi 2 can be found at though this page doesn't have instructions for building it."09:45
pedroalvarezI fear I'm not doing what's needed for the pi 209:45
*** CTtpollard has quit IRC09:50
pedroalvarezpaulsherwood: regarding building armv7 on moonshot, It would be good to validate that the binaries generated can be run in an armv7 board09:52
*** CTtpollard has joined #baserock09:52
paulsherwoodpedroalvarez: agreed. what would be the easiest way to do that?09:55
*** flatmush has quit IRC09:57
pedroalvarezhm.. not sure..09:58
paulsherwoodactually, i know. i'll have something later today for testing09:58
Kinnisonit'd be handy to double-check that the binaries produced are armv7lhf and not armv8l3209:59
paulsherwoodeasy way to do that?09:59
paulsherwoods/easy/what's the best/09:59
edcraggpaulsherwood: pedroalvarez: i think radiofree posted on the baserock-dev mailing list about a chroot method of doing that a month or so ago10:00
paulsherwoodedcragg: that's what i'm doing, thanks. i wrote it up at
*** mariaderidder has quit IRC10:02
edcraggi know that if you boot a cross-bootstrap system on a later ARM architecture, the toolchain in that system generates binaries for the target architecture natively. i tried it for building armv6 on armv7 and it produced armv6 binaries, not sure if that helps at all10:05
edcraggi think it depends on what the toolchain is set up to output10:05
gtristanjjardon, btw... thanks for cleaning up those patches in gerrit... it saved me time today to not have to think about that10:09
gtristanlooks like gdm did not put up too much of a fight10:09
* gtristan waits 10 minutes to see if the new image boots up directly into gnome-shell10:09
edcraggpaulsherwood: were you asking for an easy way to check binary format? i was using `readelf -a -W <binary> | grep Tag_CPU`10:12
paulsherwoodedcragg: tvm10:12
*** mariaderidder has joined #baserock10:15
edcraggpedroalvarez: what are you using as a starting point for building the kernel from?10:28
pedroalvarezedcragg: these instructions:
pedroalvarezcombined with various other pieces found in the intenet10:33
pedroalvarezthis is how I'm building it atm:
pedroalvarezBut with that kernel, the rpi2 boots, but the screen doesn't stop showing the splash screen10:37
gtristanwell... looks like we just need a valid /etc/pam.d config setup for gdm to autostart, gdm has the option to install some various configs, both redhat/lfs configs crap out for different reasons10:44
gtristansome tweaking of the redhat config gets it to work though, sane is probably to evaluate what we put on the baserock GNOME system and roll our own config10:45
* gtristan heads out for the evening10:46
pedroalvarezgood stuff gtristan :)10:47
richard_mawooh, buildbot has built-in support for spooling up Amazon slaves:
*** gtristan has quit IRC10:50
edcraggpedroalvarez: maybe you need to change something in the DTS? (to support rpi2)10:52
paulsherwoodok so* is serving armv7 things built in an armv7 chroot on aarch6411:19
pedroalvarezpaulsherwood: how can I download an artifact?11:20
pedroalvarezwithout using ybd or any tool? is it possible?11:21
paulsherwoodpedroalvarez: for example: wget
paulsherwoodthere are several builders running, adding more artifacts11:26
pedroalvarezpaulsherwood: these binaries work on a jetson board11:29
jonathanmawbleh, qtwebkit still building, no idea whether it's typical build-times for ybd11:30
pedroalvarezjonathanmaw: is  there any way you can see the live build log?11:31
SotKqtwebkit always takes ages last time I checked11:31
paulsherwoodjonathanmaw: is your branch somewhere? i could get it built on moonshots, see how long it takes there11:31
pedroalvarezplease! :)11:31
jonathanmawpaulsherwood: I'm using, though I had to edit the genivi-demo-platform-armv7lhf system to fix its name before it'd work with ybd.11:32
paulsherwoodjonathanmaw: afaik qtwebkit takes forever... that's not a ybd-specific thing :)11:32
paulsherwoodjonathanmaw:  ack11:33
* jonathanmaw isn't sure how many hours "forever" is11:33
edcraggqtwebkit is the poster child of long build times11:33
paulsherwoodcould you push your change?11:33
paulsherwood(would be a shame for me to build something different)11:34
paulsherwoodjonathanmaw: ^^11:34
jonathanmawpaulsherwood: pushed11:35
paulsherwoodbuilding :)11:39
jonathanmawerk! logs show an internal compiler error on both x86 and ARM11:40
paulsherwood(well... building whatever they need to get there, at least)11:40
paulsherwoodfor qtwebkit?11:40
pedroalvarezout of memory?11:40
jonathanmawmost likely11:40
jonathanmawthe fix to that is to have the definitions specify "max-jobs: 1", isn't it?11:42
pedroalvarezor get paulsherwood to build it on moonshot :)11:43
wdutchwho has permission to create repos in g.b.o?12:49
wdutchI need to figure out where CIAT should go and who can help me move it12:49
paulsherwoodpedroalvarez: ^^ ?12:50
ssam2me, pedro, fran and gary, and others all have permissions12:51
ssam2best thing is to send a mail to the mailing list requesting creation of the repos that you need12:51
richard_mawquickest thing if you happen to catch someone with auth is to list them in IRC though12:52
richard_mawwdutch: we could create a subprojects (so baserock/ciat rather than baserock/baserock12:52
richard_mawwdutch: or it could go in baserock/baserock/ciat-foo12:53
richard_maw(assuming it's still multiple repositories)12:53
wdutchi think baserock/ciat makes the most sens12:54
wdutchI'll email the list with the repos I need12:55
richard_mawsure, I'll see about creating the group12:55
richard_mawit's been a while since I needed to do that12:55
* SotK wonders if the repos are going to be added to Gerrit too12:55
richard_mawoh, hm12:56
richard_mawI don't know how that works12:56
ssam2lorry-controller on create the projects automatically12:56
richard_mawfor anything in baserock/ ?12:56
SotKeven for things not in the baserock/baserock group?12:56
*** tom__ has joined #baserock12:59
ssam2hmmm.. might only be for baserock/baserock13:00
ssam2are we going to put stuff outside baserock/baserock now?13:00
*** zoli_ has quit IRC13:00
ssam2here's the lorry-controller config it uses:
ssam2so it should mirror anything under baserock/ except those that are explicitly ignored13:02
richard_mawthat was the plan, though the justification may just have been that we can, we'd have multiple repositories, and we could arrange it as baserock/ciat/orchestration etc13:02
*** CTtpollard has quit IRC13:02
richard_mawah, so baserock/ciat should be fine13:02
ssam2i like baserock/ciat, it's less typing anyway :-)13:02
richard_mawok, ciat-{manager,admin,writer,reader}s groups set up13:03
richard_mawhm, doesn't look like wdutch has an account on g.b.o13:04
richard_mawwdutch: send me a key and I'll add you to the user list13:04
richard_mawand the baserock/ciat groups13:05
pedroalvarezthanks richard_maw :_13:05
richard_mawwdutch: what username do you want?13:07
wdutchrichard_maw: 'williamholland'13:07
wdutchwill this mean I can create repos under baserock/ciat myself?13:09
richard_mawyou could, yes13:09
* richard_maw has given wdutch ciat-managers, so he can delegate user management to wdutch too13:09
wdutchsuch an honour13:10
richard_mawha! I just don't want to be on the hook for adding everyone else who wants to change them too13:10
* SotK wonders how this account creation fits in with the "maintainership" policy on
wdutchI can wait 3 days13:11
tiredcat"they're more like guidelines than actual rules"13:11
* richard_maw forgot about the policy13:11
*** tom__ is now known as CTtpollard13:12
wdutchyar har diddledeedee being a maintainer is alright with me13:12
* SotK has no objection to the addition fwiw :)13:12
wdutchdo what you want 'cos our licence is free13:12
wdutchyou are a baserock13:12
* richard_maw saves the quote for later13:13
* richard_maw queues up some alestorm on his playlist13:14
paulsherwoodhmmm.. 0 15-10-13 00:08:41 [2/19/766] [gnome-shell] Try git clone from git://
paulsherwood(from master definitions iiuc)13:22
paulsherwoodare we going to mirror the gnome stuff13:26
paulsherwood(not submodules)13:26
paulsherwoodi know gtristan requested some lorries... maybe he missed some?13:27
pedroalvarezI think this is a submodule case13:27
SotKall the repos in that stratum appear to be "upstream:foo"13:27
paulsherwoodnope...  repo: gnome:*13:28
paulsherwoodoh, maybe i'm wrong13:28
paulsherwoodmultitasking here13:28
pedroalvarezit is a submodule case :)13:28
pedroalvarezthey are lorried, so I guess we only missed to create a branch for gnome-shell to point to upstream:13:30
paulsherwoodwdutch: what's up with the red on arm-base-build on ciat?13:38
* wdutch will have a look13:39
SotKits been red for days, log suggests out of disk space13:39
*** zoli_ has joined #baserock13:40
pedroalvarezCTtpollard: thanks for pointing me to that meta-raspberrypi thing, I think I had some progress :)13:40
CTtpollardpedroalvarez: cool :)13:41
* wdutch can't recall the IP of the arm-slave13:44
* edcragg wonders what pedroalvarez found in the yocto meta-thing13:44
wdutchthat was quicker than looking it up, thanks pedroalvarez :D13:44
pedroalvarezedcragg: they have some patches13:45
edcraggyeah, looks like i2c and sound?13:45
pedroalvarezone of them called HULK_SMASH.patch13:45
pedroalvarezidk... this is still not working, but at least the it can find the "drm device"13:46
edcraggcool, progress...13:47
pedroalvarezweston log looks ok, but screen is blank13:48
paulsherwoodjonathanmaw: qtwebkit is not building for me13:49
jonathanmawpaulsherwood: qtwebkit seems to be building for me on Jetson, but my x86 VM seems to be hanging when running qmake13:50
paulsherwoodsub-Source-WTF-WTF-pro-make_first-ordered.... that's two WTFs13:52
richard_mawusr/include/c++/4.9.2/bits/stl_relops.:67: Parse error at "std"13:53
richard_mawhmm, blow away the hardlink cache?13:53
paulsherwoodi could... but any thoughts on how this could happen?13:53
richard_mawa crash while creating the artifacts, improper shutdown of the node13:54
richard_mawneutrino bit-flipping on filesystems that don't have built-in integrity checking13:54
paulsherwoodthis is the kind of unpredictability we're trying to remove13:54
richard_mawwe can rule out a crash while creating the artifact, since YBD does atomic artifact creation doesn't it13:56
paulsherwoodi believe so13:56
richard_mawit could be that another build modified files in the staging area, hence why I recommend against running without linux-user-chroot protection whenever possible13:57
paulsherwoodi've been ctrl-c-ing lots of these builds... and i'm now mixing and matching artifacts from jonathanmaw's jetson build and the moonshot13:57
*** zoli_ has quit IRC13:57
paulsherwoodi guess i should run the whole thing from scratch13:58
richard_mawhm, unless you ^C'd a copy between artifacts, or if the kbas artifact upload/download isn't atomic, I'm not sure how that could happen13:58
paulsherwoodme neither13:59
paulsherwoodrichard_maw: but you're saying that the problem is in the c++ ?13:59
richard_mawI'm guessing it's a partial file, though if you can check "usr/include/c++/4.9.2/bits/stl_relops.:67" to see whether it looks truncated, it would confirm the hypothesis14:00
pedroalvarezwhat am I missing if my screen goes blank when booting after a few lines of kernel messages?14:01
pedroalvarezit sometimes come back for the login prompt14:02
pedroalvarezsometimes not14:02
wdutchthat du command took hella-fortnights14:03
wdutchshould I just remove all the contents of /src from the arm-slave?14:04
* richard_maw doesn't know how well ybd copes with that manner of cache cleanup, but we build from low enough in the stack that it shouldn't be too much of a problem14:05
pedroalvarezFailed to process Wayland connection: Connection reset by peer14:12
pedroalvarezfailed to create display: Connection reset by peer14:12
pedroalvarezTrace/breakpoint trap (core dumped)14:12
pedroalvarezfull weston log here:
jonathanmawSotK: Would it be accurate to say that at its basic level, your partial deployment patch is about putting a rootfs that only contains the required chunk where a write extension would normally expect a whole system?14:26
* paulsherwood wonders why it needed ostree14:27
* SotK tries to remember what his partial deployment patch does14:27
SotKpaulsherwood: it didn't14:27
paulsherwoodisn't it simplest just to untar it in the right place?14:27
paulsherwood(assuming we don't care about reproducibility at that point)14:28
SotKpaulsherwood: that is what it does14:29
jonathanmawpaulsherwood: untarring the artifact into the right place for the write-extension to run, yes. untarring it directly to the target, less so (what if the difference was a file moved or deleted?)14:29
paulsherwood(says it depends on your ostree work)14:30
richard_mawthe implementation depended on it because it was worked on top of the ostree work14:30
paulsherwoodah, ok14:30
SotKsince that didn't get merged I reworked it to not use the ostree work14:31
paulsherwoodah :)14:31
SotKjonathanmaw: I think it just unpacks the artifact and its dependencies and passes that to the write extension14:31
SotKI think it errors when using a write extension that isn't tar or sysroot though14:32
*** franred has quit IRC14:32
* SotK would explicitly recommend not using it :)14:33
jonathanmawpaulsherwood likes to live dangerously14:33
SotKalso configuration extensions aren't run, for obvious reasons14:33
*** franred has joined #baserock14:35
wdutchrichard_maw: did you add my key? 'ssh ls' says Permission denied14:39
* richard_maw forgot that user add doesn't read the sshkey from stdin14:40
richard_mawtry now14:40
*** zoli_ has joined #baserock14:40
jonathanmawpaulsherwood: the sysroot deployment will work for us if we don't care about moved or deleted files (and have direct access to the target's rootfs from the deploying machine)14:45
jonathanmawan ssh-rsync extension which can handle partial deployments would be ideal, but I don't think we have time for that.14:46
* SotK wonders what horrors jonathanmaw and paulsherwood are up to14:47
jonathanmawSotK: partial deployments, because full deployments might require downloading 8GB of artifacts from the wider Internet.14:48
jonathanmawI don't think the Internet's that good, even in Seoul14:49
*** toscalix__ is now known as toscalix14:49
SotKYou may end up downloading a fair bit anyway if your chosen chunk has a lot of dependencies14:51
pedroalvarezwell, this is a re-work for ybd, we could fetch only one chunk and deploy that14:53
paulsherwoodwhat is 'fetch'14:53
SotKthat would be an even more horrific hack, but would be doable14:53
pedroalvarezpaulsherwood: well, I was thinking on the same use-case that jonathan described, downloading the artifacts from internet14:56
pedroalvarezthat's why I used "fetch", but of course if the artifact is present local, then it doesn't need to14:57
paulsherwoodok, i understand14:57
paulsherwoodi'm starting to think it needs to be yb, and yd14:57
pedroalvarezI like that idea, tbh14:58
pedroalvarezI think more people around might like it too14:58
wdutchrichard_maw: works, ta14:59
ssam2i think build and deploy should be treated separately, they actually have very little in common15:01
* SotK agrees15:01
* toscalix agrees15:02
richard_mawwhether configuration extensions should be distinct from builds is another question15:05
*** edcragg has quit IRC15:06
ssam2richard_maw: you mean, should we remove the 'configuration-extensions' field from system.morph files? If so, I think yes15:06
richard_mawssam2: I was thinking more that if we had parameterisable builds, a lot of configuration extensions will be the same shape as the system integrations15:07
*** edcragg has joined #baserock15:08
richard_mawthough I suppose the main difference is that configuration extensions are allowed to interact with the host system15:08
ssam2I think I understand ... no easy answer, as different configuration extensions exists for different reasons15:09
ssam2some should probably turn into system integration commands, some of them are actually deploy-time config for specific systems15:10
ssam2but i'd rather we used some kind of config management tool for the latter, rather than ad-hoc scripts15:10
pedroalvarezwe could try ansible against a chroot of another architecture and see what happens15:14
ssam2I would actually like there to be some Ansible modules like 'baserock_vm', which would hide the build/fetch process totally15:21
ssam2so deployment instructions don't need any kind of special .morph file15:22
ssam2same for any other config management tools people like15:22
ssam2e.g. for baserock_gerrit, we could delete the current cluster .morph file that deploys it to OpenStack, and add a line to the ansible playbook that configures it that says 'get me a build of the gerrit-system-x86-64.morph and start it up in OpenStack' as the first step15:23
ssam2not quite sure how that would work with upgrades, though... i haven't thought it through fully15:23
ssam2i guess you don't really upgrade things in OpenStack, you just deploy the new one then delete the old one :)15:26
*** paulwaters_ has quit IRC15:35
pedroalvarezyay!!!!! drm backend worked on the Pi2!15:35
CTtpollardpedroalvarez: awesome!15:35
pedroalvarezit doesnt look faster than the fbdev backend though15:37
pedroalvarezno no, it's faster15:39
pedroalvarez[    3.801039] [drm:vc4_hdmi_bind] *ERROR* Failed to get ddc i2c adapter by node15:48
pedroalvarez[    3.808361] vc4-drm soc:vc4@0x7e4c0000: failed to bind 3f902000.brcm,vc4-hdmi (ops vc4_hdmi_ops): -51715:49
pedroalvarezI think that's why I stop seeing things through hdmi15:49
*** flatmush has joined #baserock15:52
*** ssam2 has quit IRC16:18
*** zoli_ has quit IRC16:24
*** edcragg has quit IRC16:32
*** edcragg has joined #baserock16:34
pedroalvarezjonathanmaw: if I make sure that genivi-baseline works on a rpi2, does that mean that genivi-demo-platform will work too?16:38
jonathanmawpedroalvarez: Not necessarily. the GENIVI baseline systems are mostly to demonstrate that it builds, the GENIVI Demo Platform is where we usually find out if various parts work.16:40
pedroalvarezMy goal is to put GDP on the rpi2, but it's not in master16:41
*** mariaderidder has quit IRC16:42
pedroalvarezoh, your branch rebases cleanly on top of master16:45
*** zoli_ has joined #baserock16:46
*** toscalix has quit IRC16:49
pedroalvarez.. I'm having problems with the migration scripts:
*** bashrc_ has quit IRC17:03
*** jonathanmaw has quit IRC17:05
*** zoli_ has quit IRC17:07
pedroalvarezalso wayland-gdp fails to build17:17
pedroalvarezconfigure: error: Package requirements (wayland-server >= 1.5.91 pixman-1 >= 0.25.2 xkbcommon >= 0.3.0) were not met:17:18
pedroalvarezNo package 'xkbcommon' found17:18
pedroalvarezthis is after rebasing it on top of master17:18
*** edcragg has quit IRC17:26
*** franred has quit IRC17:40
*** tiagogomes_ has quit IRC17:54
*** zoli_ has joined #baserock18:35
*** zoli_ has quit IRC18:45
*** zoli_ has joined #baserock18:46
*** zoli_ has joined #baserock19:29
*** zoli__ has joined #baserock19:30
*** zoli_ has quit IRC19:30
*** zoli__ has quit IRC19:54
*** zoli_ has joined #baserock19:55
*** brlogger has joined #baserock21:00
*** zoli_ has quit IRC21:16
*** tiagogomes has joined #baserock22:55
*** brlogger has joined #baserock23:28

Generated by 2.15.3 by Marius Gedminas - find it at!