IRC logs for #baserock for Tuesday, 2014-09-23

*** dutch [] has joined #baserock07:16
*** tiagogomes [~tiagogome@] has joined #baserock07:32
pedroalvarezjjardon: looking at the BSP build-depends - no, it shouldn't rebuild  with changes in foundation.07:56
*** tiagogomes [~tiagogome@] has quit [Remote host closed the connection]08:12
*** tiagogomes [~tiagogome@] has joined #baserock08:19
*** violeta [] has joined #baserock08:23
*** jonathanmaw [] has joined #baserock08:24
petefothrichard_maw has documented his attempts to reproduce earlier BAserock releases at No other pages link to that page (even though it is in the 'guides' directory. Shpould we add a link (on the 'Guides' page or elsewhere) or do we leave it as a page that we know about but other users of w.b.o will never find?08:37
petefothpaulsher1ood: :) I'm going to stop asking stuff and 'just do it' from now on!08:43
*** ssam2 [] has joined #baserock08:59
richard_mawpetefoth: ha!09:06
petefothstraycat: I've done a few tweaks on the 'deploy-trove-to' pages. If you have time to check that I haven't introduced any errors, that would be great09:48
liw-orc`I'd make the "guides" page be generated automatically by ikiwiki, actually09:50
*** Krin [] has joined #baserock09:52
petefothliw-orc`: I guess that would ensure that nothing in the 'guides' drectory got missed out. Howvere, given what was said in yetsreday's plannign meeting that maybe maintaining w.b.o should be the responsibility of the community and not somethink that Codethink do, it may have to wait until someone has some upstream time09:53
liw-orc` is now known as liw-orc09:55
paulsher1oodpetefoth: you're paraphrasing a private meeting in public... tssk :)09:59
paulsher1oodliw-orc: your idea seems sensible. is it trivial to do?10:03
liw-orcpaulsher1ood, should be: a quick example (untested): [[!inline pages="page(guides/*)" archive=yes trail=yes]]10:06
liw-orcpaulsher1ood, has the details10:06
paulsher1oodhmmm... that would need quite a lot of tidyup10:08
liw-orcwhat kind of tidyup?10:09
paulsher1oodtitles, ordering, formatting10:10
paulsher1oodit's doable, just not by magic :)10:10
paulsher1oodpedroalvarez: by 'Some users' do you mean me? :)11:49
pedroalvarezactually I meant franred :)11:50
paulsher1ooddid you get to the bottom of th openstack upgrade fail i hit?11:50
pedroalvarezyeah, I'm attempting to fix it11:51
paulsher1oodssam2: are you really ok with jjardon's patch breaking on ARM?12:05
radiofreei wouldn't object, simply because it's unreasonable to ask contributors to also check to see if something builds on arm12:07
radiofreeif we had a public arm build server that could do that however..12:07
jjardonpaulsher1ood: It it helps, nothing is using NetworkManager as a dependency (for now)12:07
ssam2paulsher1ood: I am, because I know that Javier's end goal is getting GNOME working, and there are probably other obstacles to GNOME on ARM12:09
ssam2i'd rather he fixed it but I have no idea what the error means, and I don't want to require the impossible :)12:09
ssam2or at least, not straight away12:10
jjardonssam2: seems a compiler error, so maybe a new gcc will fix it ;) I will take a look when I have some time12:11
paulsher1oodmy concern is that the connectivity stuff is required for genivi... so breaking it on arm would be unfortunate :)12:13
radiofreedo they use networkmanager? or conman?12:14
radiofreelooks like it's conman, so should be fine12:15
jjardonpaulsher1ood: nothing will break, I split connectivity so you can choose between connman and Networkmanager12:15
jjardonso everything that was using connman will kepp using it12:15
jjardonssam2: maybe you could first try to build the nm-0-9-10 branch of NetworkManager in ARM.? Sorry, I do not have any ARM device here to test12:16
ssam2jjardon: running a build, will let you know how it goes after lunch12:22
paulsher1oodjjardon: ah, ok12:27
pedroalvarezhm... I have reasons to think that distbuild and local build are not creating the artifacts with the same artifact cache key12:49
richard_mawit's possible, the distbuild code uses a Mock Artifact12:51
richard_mawif you can get a copy of a /baserock directories of the artifacts that are calculated differently, it should be possible to track down why they've diverged12:52
pedroalvarezrichard_maw: aha, good approach, I will do that12:53
richard_mawif you need help interpreting the contents of the metadata, let me know12:56
ssam2jjardon: same error with ref nm-0-9-1013:15
jjardonssam2: yeah, seems the code didnt change. Not sure about the cause yet but master and that branch are compiling in ARM, for example:
pedroalvarezrichard_maw: If I use the new version of morph with your "per source building" patches, attr rebuilds13:20
ssam2jjardon: the error seems to be due to 'struct ifla_vxlan_port_range' being undefined13:22
ssam2or badly defined13:22
ssam2this comes from /usr/include/linux/if_link.h13:22
ssam2so it may be to do with the version of linux_api_headers we're using?13:22
richard_mawpedroalvarez: some graphing stuff may have changed, that's not necessarily the problem, the problem is if it's building differently in distbuild to local build13:23
pedroalvarezrichard_maw: in that case, my problem could be that the distbuild system is using a different morph version13:24
pedroalvarezI didn't expect that your changes were causing rebuilds :/13:25
jjardonssam2: mmm, yeah maybe. I remember our current version of linux_api_header is quite old13:25
liw-orcpedroalvarez, in Trove deployment, where does the mason user get created? trove.configure? trove-setup.service? somewhere else?13:26
* straycat thinks it might be worth adding 'autodefrag' as a mount option for /var13:26
pedroalvarezliw-orc: chekcing13:27
liw-orcstraycat, that doesn't sound like a bad idea13:27
pedroalvarezliw-orc: the mason user, or the gitano mason user?13:28
liw-orcpedroalvarez, the unix user mason, not the gitano one13:28
pedroalvarezliw-orc: trove-setup.service calls Ansible. You can find the Ansible scripts in trove-setup. The script which is creating the users is this one:
liw-orcpedroalvarez, I see that: I don't see how it creates the user13:30
*** genii [~quassel@ubuntu/member/genii] has joined #baserock13:30
liw-orcpedroalvarez, is this it:   user: name={{ item }} comment="Trove {{ item }} service" shell=/bin/bash generate_ssh_key=yes ?13:30
liw-orcpedroalvarez, if so, where is generate_ssh_key?13:30
pedroalvarezliw-orc: generate_ssh_key is a flag to indicate that you want the new user with new ssh keys13:31
pedroalvarezso ansible will create the user, and generate ssh keys for the user13:31
liw-orcpedroalvarez, hmm13:31
liw-orcthen we'll have to try to figure out why it isn't happening13:32
liw-orcor rather, didn't happen13:32
pedroalvarezliw-orc: after an upgrade? or a fresh deployment13:32
liw-orcthe unix user exists, but has no home dir13:32
pedroalvarezliw-orc: did the previous system have /home/mason?13:33
liw-orcsuccess; we deluser'd mason, run trove-setup.service, now is worked13:34
pedroalvarezaha, my assumtion is that ansible checks if the user existsn and do nothing in that case13:35
*** doffm [~mdoff@] has joined #baserock14:03
ridgerunnerOK, I'm getting "ERROR: Couldn't find morphology: strata/build-essential.morph". I know I'm supposed to tell Baserock where to find it (and I can see it in <>) but I can't remember how.14:11
ridgerunnerI'm building using
*** genii [~quassel@ubuntu/member/genii] has quit [Remote host closed the connection]14:12
paulsher1oodridgerunner: what is the command you're running?14:13
ridgerunnermorph --verb14:13
ridgerunnerose build distbuild-system-armv7lhf-jetson.morph14:13
ridgerunnerpls excuse the line break.14:13
SotKwhat sha is your definitions checkout at?14:14
paulsher1oodplease cat ~/.ash_history and paste the last few commands somewhere?14:14
ridgerunnerHow do I tell?14:14
paulsher1oodgit log would tell you14:14
ridgerunnergit log -1 gives commit 0bd86c449786be698253d3d247e566cbbc785a6b14:15
paulsher1oodridgerunner: please see my cat request above14:16
ridgerunnertrying it14:16
Kinnisonpaulsher1ood: TTOTD, if you want the last few lines of a file, use 'tail'14:16
Kinnisonpaulsher1ood: and if you want the first few, use 'head'14:17
SotKridgerunner: strata/build-essential.morph does not exist at that sha14:17
paulsher1oodpaulsher1ood: i know that14:17
paulsher1oodthank you :)14:17
paulsher1oodas you can see, Kinnison, my brain is not entirely here14:18
SotKthat system definition is from after we rearranged definitions, that sha is in the old layout14:18
Kinnisonpaulsher1ood: Aye :-)14:18
paulsher1oodridgerunner: probably best to restart, with a branch of current defnitions14:18
paulsher1oodunless you have a specific reason to go back to 14.29.114:19
ridgerunnerHow current do you have to be? That commit was done by ssam2 on Friday, I think.14:19
SotKthat commit was based on 14.29 (from july14:19
ridgerunnerI'm trying to follow instructions. They're not working.14:20
* straycat tsks14:20
straycatmorph's tests do not pass.14:21
straycatridgerunner, Which instructions aren't working? We should update them.14:21
ridgerunnerMostly, I'm following but adding to them to build for Jetson TK1 using
SotKridgerunner: to make that system definition work with your definitions checkout, you'll need to change all the "strata/foo.morph" to "foo.morph" I think14:22
SotKdepending on how recent your morph version is you may need to change them to just "foo" iirc14:23
ssam2ridgerunner: yeah, what has happened is that since the 14.29 release, the way Morph finds morphologies changed. The system morph you are using dates from after the change14:23
ssam2but the 14.29 and 14.29.1 releases date from before the change14:24
ssam2as SotK says, the old format is to just have 'build-essential' instead of 'strata/build-essential.morph' in the `morph` field, and the same for every other entry in the 'strata' list14:24
ridgerunnerok, a quick check shows  build-es14:25
ridgerunnersential.morph as existing, so I'll try that14:25
ssam2sorry you've hit this, the problem is that we've not done a release of Baserock since making that big change14:25
ssam2so there's new stuff floating round that isn't compatible with the last stable release14:26
ridgerunnerThanks, this makes sense of something that paulsher1ood said yesterday about "getting rid of strata", or words to that effect.14:27
straycatThe tests fail on coverage, the function in question is just calling id(), so I think it's fair enough to put a no cover on it.14:27
* straycat hugs konsole14:29
straycatNow when I open links, they are downloaded and opened in sublime text.14:29
ssam2straycat: is it this ?
ssam2(see the comment under the quoted text)14:30
paulsher1oodssam2: another user just hit the same issue, building fran's gerrit system14:30
straycatssam2, yes14:30
paulsher1oodssam2: i recommended using latest morph. maybe ridgerunner should do the same?14:30
ssam2paulsher1ood: the issue of new system definitions not being usable with the definitions.git from the 14.29.1 release ?14:30
ridgerunnerMoved on (I think) to ERROR: Couldn't find morphology: bsp-jetson-devel.morph14:31
straycatssam2, Are you okay with me pushing that fix?14:31
SotKridgerunner: that probably also isn't in that version of definitions14:31
paulsher1oodssam2: released morph unable to build a current definition14:31
ridgerunnerOK, should I just scrap the whole thing and start again? How far back should I scrap?14:32
ssam2paulsher1ood: that's completely expected.14:32
paulsher1oodnot by said user :)14:32
ssam2whatever instructions told them to use latest definitions should have made it clear they also need latest Morph, I think14:33
ssam2ridgerunner: where did you get the Jetson system morph file? you'll find the other missing files in the same place, presumably14:33
ssam2copying individual .morph files around isn't something we've really considered, I think14:34
ssam2I don't think we should recommend that people do that, at least while we keep making changes to the morphology format14:34
ssam2even if the format was stable, copying around a system .morph without any of its strata is fairly pointless14:34
ridgerunnerssam2: In a link supplied by paulsher1ood :-)14:34
paulsher1oodssam2: fixed :)14:35
ssam2paulsher1ood: awesome14:35
ssam2ridgerunner: the Jetson support is not officially released, but as I understand the latest 'master' of definitions.git has necessary stuff for building a Jetson system14:36
ssam2ridgerunner: so my recommendation (which overrides any previous ones I may have made!) is to use latest definitions.git and latest Morph14:36
paulsher1oodridgerunner: careful - i may stop supplying links :)14:36
ssam2ridgerunner: to use latest 'definitions', just do 'morph checkout baserock:baserock/definitions master' instead of 'morph checkout baserock:baserock/definitions 14.29.1'14:36
ssam2to use latest morph, instructions are on the wiki14:36
ridgerunnerpaulsher1ood: Please don't do that :-)14:37
paulsher1oodridgerunner: which specific link led you astray - the jetson one?14:37
ridgerunnerIt probably wasn't the link, more my naive interpretation of what to do with the info it gave. A little knowledge and all that...14:38
ssam2 is pretty cool, by the way. Gives us a target 'developer loop' to work on speeding up, at least14:39
paulsher1oodthank you :)14:39
*** genii [~quassel@ubuntu/member/genii] has joined #baserock14:40
paulsher1ood<10 minutes for build-deploy  with brand new kernel is not too shabby, but i know we can do better :)14:40
paulsher1oodsadly i couldn't get it to work on jetson over the weekend, though14:41
paulsher1ood(the loop that is)14:41
SotKwhat went wrong?14:42
paulsher1oodthen once i hard-coded it, booting into 2: test kept giving me 1: factory14:43
paulsher1oodiirc radiofree had upgrades working, though... so it's probably me as usual14:44
SotKthat is odd, I have successfully upgraded the Jetson on my desk in the past14:44
paulsher1oodas i said, must be me :-)14:45
* paulsher1ood considers repeating the process, to see if anything has improved14:47
paulsher1oodKinnison: do you have an exciting mainline-ish kernel for jetson i could point at?14:48
KinnisonI'm trying to kick off a brain of a Jetson in about 2 minutes14:48
Kinnisonpaulsher1ood: then I'll know14:48
paulsher1oodooh, exciting :-)14:48
paulsher1oodah, but you probably had to fiddle with u-boot too, i guess14:49
radiofreepaulsher1ood: did you set VERSION?14:49
Kinnisonpaulsher1ood: I did have to play with u-boot yes14:49
paulsher1oodradiofree: you mean VERSION, as opposed to VERSIN_LABEL? no14:51
paulsher1oodi set VERSION_LABEL14:51
*** genii [~quassel@ubuntu/member/genii] has quit [Read error: Connection reset by peer]14:51
*** thecorconian [~thecorcon@] has joined #baserock14:52
* richard_maw grumps14:53
radiofreeit's probably VERSION_LABEL14:53
richard_mawCobblerd can't handle importing systems that don't have an initramfs14:54
radiofreeKinnison: it's working?14:57
radiofreecan you check /sys/devices/system/cpu/cpu0/cpufreq/...14:57
radiofreeor something like that14:58
paulsher1oodKinnison: looks promising14:58
paulsher1oodmy jetson is less happy - Warning - bad CRC, using default envionronment14:58
radiofreeKinnison: was that with the "boot from sdcard" trick as well, or is btrfs loading behaving as well?14:58
Kinnisonradiofree: sdcard, I can't get btrfs from u-boot to behave15:02
paulsher1oodradiofree: does bad crc mean my jetson is dead forever?15:04
Kinnisonradiofree: there's no cpufreq type section in there15:05
SotKpaulsher1ood: mine also says that, I've fixed it by reflashing and starting over in the past15:06
paulsher1oodSotK: so there's hope, at least :)15:07
radiofreepaulsher1ood: nah, they are remarkably robust15:08
Kinnisonradiofree: you're welcome to come poke about on it if you want15:09
straycatssam2, Slight confused by this extra field, what uses that?15:14
straycatOr less vaguely put, what is x-products-foo doing, which tool is using that?15:15
ssam2straycat: it's for the import tool I've been working on which was last discussed here:
ssam2the goal of the field is to make it possible for the import tool to go 'I want activerecord, and I see there's a Lorry for which is marked as producing the activerecord Gem, so I'll use that instead of creating a new .lorry file'15:16
ssam2there are other ways to achieve it, which might be better than the field15:16
ssam2it may be that said information should live in the import tool, but it seems more 'right' that the info should live in the .lorry to me15:17
* pedroalvarez sighs15:18
paulsher1oodoh dear, pedroalvarez rarely sighs15:19
pedroalvarezafter many problems debugging, I found that distbuild doesn't work at all now15:19
Kinnisonpedroalvarez: ? :-(15:19
pedroalvarezit fails when computing the build graph15:21
* paulsher1ood has not touched this...15:21
richard_maw:-( it's the building per-source change isn't it15:22
pedroalvarezhere is the error
richard_mawggrwaajfkhdgkjahdfasj, yes, it is the per-source building15:24
pedroalvarezrichard_maw: I guess so :/ but I'm not blaming you15:24
straycatI blame him, it's all his fault. :)15:24
richard_mawthere's a compatibility commit which makes Artifact.dependencies still work, you could try reverting the removal15:25
jmacsI will be taking offline in about ten minutes to make a final backup before upgrading. Please shout if this is going to derail your work.15:25
straycatrichard_maw, So, is this because the structure of the build graph has changed but we haven't updated the serialisation code to match?15:25
richard_mawstraycat: probably15:26
* richard_maw sighs15:26
* straycat sighs15:26
richard_mawwe really need to get the distbuild tests integrated15:27
Kinnisonrichard_maw: I take it my passing comment about distbuild last week was insufficient?15:27
richard_mawKinnison: it was, since I completely forgot about it15:27
paulsher1oodjmacs: any idea of downtime duration?15:28
richard_mawso, on my shopping list of tasks to do tomorrow are 1. get rid of build-mode: test and 2. integrate distbuild tests15:29
jmacspaulsher1ood: About 90 minutes based on the last time we did this.15:29
* paulsher1ood notes there's a standvy trove at, id baserock-clone15:29
straycatrichard_maw, awesome :)15:29
* straycat wanted to integrate those tests but never got around to it :/15:29
paulsher1oodjmacs: tvm15:29 is going offline now for 90 minutes.15:35
*** CTtpollard [] has joined #baserock15:42
ridgerunnerImproving slowly, build ran for >27 minutes then failed with: "ERROR: Failed to update cached version of repo git://"15:43
CTtpollardg.b.o has been taken down I do believe ridgerunner15:43
paulsher1oodif you're wanting to build/deploy, don't have your own trove, normally use g.b.o and need a trove in the meantime... put trove-host= and trove-id=baserock-clone in your morph.conf15:43
ridgerunnerIs that just a permissions thing?15:43
paulsher1oodridgerunner: more likely your network i think15:43
ridgerunnerok. I don't understand your previous comment. Too cryptic for my little brain, I'm afraid.15:44
jmacsSorry ridgerunner. It should be back up within 90 minutes.15:45
ridgerunnerok, I'll be on the train by then (I hope) but I'll re-run it in the morning.15:45
paulsher1oodridgerunner: see the section on morph.conf. add the above into yours15:45
ridgerunnerDone. It's now thinking about it.15:48
ridgerunnerUrgh: ERROR: Cannot find remote git repository: upstream:binutils-redhat15:50
paulsher1oodridgerunner:byour bugmagentism rivals my own15:51
ssam2double http://15:51
ridgerunnerI shall take that as a compliment :-)15:51
ssam2trove-host shouldn't have http:// at the start15:51
ssam2the error message could be a clearer :)15:52
paulsher1oodmy bad then15:52
ssam2*could be clearer15:52
CTtpollardwill the build I wad currently doing be removed from /tmp automatically?15:52
paulsher1oodif you're using the idea, eventually it will15:53
paulsher1oodnote morph should be working in /src/tmp, not /tmp though15:53
CTtpollardyes sorry /src15:54
pedroalvarezCTtpollard: did the build fail?15:54
paulsher1oodCTtpollard: building an chunk stops network access while building, so the chunk should complete successfully, then the rest of the build may fail if it can't find your trove15:54
ridgerunnerIs it ok to kill a build? Will it recover when re-run?15:55
paulsher1oodridgerunner: yup15:55
paulsher1oodbuilt artifacts should be retained, too15:56
ridgerunnerGreat. That means I can catch a sensible train and do kernel patch work on the way home rather than leaving my laptop here :-)15:56
rjekIIRC, if you kill a build then when you restart it will continue from the beginning of the chunk it was building when you cancelled it.15:57
rjek(ie, any work done towards building the current chunk will be lost, but any completed chunk builds will not.)15:57
liw-orcrjek, yep15:58
ridgerunnerSounds good enough to me :-)15:59
paulsher1oodin other news my jetson came back from the dead ;)15:59
* richard_maw resists making ghost jokes15:59
*** dutch [] has quit [Quit: Quit]16:00
ridgerunnerA good day today, thanks guys :-)16:00
straycatssam2, +1 for your RubyGems lorries16:06
straycatNot well today, going offline. o/16:07
CTtpollardty for the helps folks16:07
paulsher1oodssam2: i've seconded straycat on the ml16:11
*** genii [~quassel@ubuntu/member/genii] has joined #baserock16:22
*** tiagogomes [~tiagogome@] has quit [Quit: Leaving]16:34
pedroalvarezjmacs: is up and running?16:44
jmacspedroalvarez: Just!16:45 is now up and ready to use. Thanks for your patience.16:45
paulsher1oodjmacs: is it updated?16:45
* paulsher1ood wants the moon on a stick as always :)16:46
jmacspaulsher1ood: No. The upgrade will be running tonight, and we'll need to do one more reboot to see the update tomorrow.16:46
paulsher1oodjmacs: ok tvm16:46
CTtpollardre-pointed morph.config and it's now building from gbo, ty16:48
paulsher1oodyay :)16:48
*** jonathanmaw [] has quit [Quit: Leaving]17:02
CTtpollardif I've built the dev system, and the web system. should I expect linux to be already cached to some extent?17:08
ssam2CTtpollard: depends. Morph caching of built chunk artifacts is 'all or nothing'17:09
ssam2it decides whether it can use a cached artifact based on the source and all of the chunks that were built before the one your building17:09
CTtpollardI'm building the gerrit system atm, and it's been hanging on cloning linux for a while17:09
ssam2hmm, it should definitely not be cloning it again17:10
ssam2unless you changed your trove-host ... which you did :)17:10
ssam2morph doesn't do anything smart about 'is this linux.git from Trove A the same as linux.git from Trove B?'17:10
ssam2it just assumes they are different repos17:10
ssam2perhaps it could be made smarter, but perhaps it'd be more trouble than it'd be worth17:11
CTtpollardI thought cloning would be excessive if it already was cached and was expecting a pull instead that's all17:11
CTtpollardI suppose switching troves makes sense17:11
* jjardon just built a genivi-baseline-system without any X11 stuff (only 175 chunks)17:17
paulsher1ooddoes it work? :)17:17
CTtpollardwould it be faster to force it to git clone, or leave it to fetch the tarball for linux?17:21
paulsher1oodtarball is faster i think17:22
paulsher1oodshouldn't take all night... should be 10-15 mins iirc17:22
paulsher1oodmaybe your network is slow17:22
*** ssam2 [] has quit [Remote host closed the connection]17:23
*** Krin [] has quit [Remote host closed the connection]17:23
jjardonpaulsher1ood: yes ;)17:24
paulsher1oodjjardon: push a branch and i'll follow you overnight? :)17:25
jjardonpaulsher1ood: The branch is waiting for some stuff to be merged and need some cleanup but sure17:26
paulsher1oodjjardon: do you have somethings stuck in review?17:27
jjardonthe mesa upgrade (already have a +1) and the dbus-glib move (I sent this today at lunch time)17:28
paulsher1oodok i'll try those, too17:29
jjardonpaulsher1ood: jjardon/only_wayland in my gitorious repo (this branch include the dbus-glib and the mesa upgrade stuff)17:30
*** violeta [] has quit [Ping timeout: 250 seconds]17:37
paulsher1oodhow would i verify that dbus-glib is working?19:08
* paulsher1ood confirms that jjardon/only_wayland builds, and runs and has wayland :)20:04
*** thecorconian [~thecorcon@] has quit [Quit: Leaving.]21:52
*** genii [~quassel@ubuntu/member/genii] has quit [Read error: Connection reset by peer]22:02

Generated by 2.15.3 by Marius Gedminas - find it at!