IRC logs for #baserock for Thursday, 2015-10-15

*** gtristan has joined #baserock02:23
*** toscalix__ has quit IRC03:10
*** gtristan has quit IRC06:05
*** gtristan has joined #baserock07:08
*** ssam2 has joined #baserock07:08
*** ChanServ sets mode: +v ssam207:08
jjardongtristan: hey! I was thinking maybe is a better idea to use 'sysusers' to create users we need in a declarative and centralised way:
ssam2we discussed that before & the only problem with that is it needs to run at first boot, and some people have read only /etc07:28
ssam2but there would be way around that i'm sure07:28
gtristanjjardon, I was about to go with:
gtristangnome-initial-setup, if installed, is intended to run at first boot and let you create a user07:29
gtristanwithout any further tinkering on our side :)07:29
gtristan(halfline pointed me to that in #gnome-hackers yesterday evening)07:30
jjardonOh yeah, but that's only for the user login, not special users like gdm, isn it?07:31
gtristanoh for users like gdm and such we have system integration hooks which work07:31
gtristanyou think we should change the regular useradd thing for a systemd'ism ?07:32
jjardonI think we should centralise this instead being a thing in different chunks in your system, and the sysusers thing seems to solve exactly that07:33
gtristan...for systemd targets indeed07:34
* gtristan still refuses to accept the world revolves around systemd07:35
gtristanbut that does look like something to push for upstream packages to install on their own07:35
gtristani.e. it could be nice for gdm, which configured --with-systemd, could install something there, and then we (and distros), could omit stuff from our own scripts07:37
jjardonMmm, true, maybe we should talk with halfline about that07:38
*** mariaderidder has joined #baserock07:56
*** paulwaters_ has joined #baserock08:01
*** bashrc has joined #baserock08:02
*** bashrc has quit IRC08:07
*** bashrc has joined #baserock08:08
*** fay_ has joined #baserock08:09
*** bashrc has quit IRC08:10
*** bashrc has joined #baserock08:10
*** bashrc has quit IRC08:12
*** bashrc has joined #baserock08:12
* pedroalvarez tries GDP on rpi2 with with stable kernel...08:13
pedroalvarezand no, it didn't work08:14
pedroalvarezweston log:
*** toscalix__ has joined #baserock08:15
pedroalvarezgdp-hmi-controller.service is failing..08:15
*** jonathanmaw has joined #baserock08:15
pedroalvarezI wonder if weston not showing anything could be related to gdp-hmi-controller failing08:16
pedroalvarezgdp-hmi-controller[512]: /usr/bin/gdp-hmi-controller: error while loading shared libraries: cannot open shared object file: No such file or directory08:17
pedroalvarezthat sounds like good news. I think I'm just missing a stratum08:17
pedroalvarezor a chunk... but maybe not..08:18
*** Phlogistique has left #baserock08:19
pedroalvarezstrata/glib-common.morph :)08:21
*** toscalix__ has quit IRC08:24
*** edcragg has joined #baserock08:34
*** toscalix__ has joined #baserock08:35
gtristanget prepared for mass additions of lorries08:43
pedroalvarez /o\08:45
gtristanwe dont even have webkit ?08:45
* pedroalvarez hopes these repos are all in git, and are not huge08:46
*** gary_perkins has joined #baserock08:47
gtristanwebkit is a biggy... the rest I'm guessing are relatively small08:49
KinnisonGiven we have qtwebkit, are we not likely to have webkit?08:49
gtristanI didnt see it in the controller page08:49
gtristanbut did notice qtwebkit, so I'm not sure why... perhaps it doesnt really build08:50
gtristanfwiw, I'm trying to satisfy these deps:
gtristanso as to get the experience at first boot of the GNOME system08:50
gtristanwhich... looks pretty neat I think ;-)08:50
gtristanalso I suspect it will satisfy many of the deps we'll need for the apps08:51
*** paulwaters_ has quit IRC08:56
pedroalvarezgdp journal, with some errors in:
pedroalvarezgdp-hmi-controller fails, and so does gdp-hmi-launcher208:58
*** paulwaters_ has joined #baserock09:01
paulsherwoodjonathanmaw: any thoughts? ^^09:06
gtristanummm... looks like definitions is in for a shuffle... is there a reason why, for instance... there are 2 definitions of libicu ?09:07
gtristanone in qt stuff and another in gnome ?09:08
*** franred has joined #baserock09:08
gtristanI'll have trouble building webkit I think, as it should technically sit outside of gnome/, but will require some things also declared in gnome/ (like geoclue for instance)09:08
richard_mawgtristan: a lack of cross-desktop coordination, let's form a working group and call it XDG, then rename it to freedesktop.org09:09
ssam2definitions is definitely in need of a shuffle09:09
jonathanmawpedroalvarez: I see ilm_init is failing, possibly because it failed to connect to a display09:09
jjardongtristan: no, no reason; that should be fixed09:09
ssam2I find that the more use cases we have for it, the more the "strata" package grouping model falls apart09:09
gtristanok I see09:09
ssam2you end up with really specific strata like glib-common because there's multiple things that want to use glib, but also some things that don't want to09:10
jjardonbut is a bit of a problem because you have to put icu in a stratum shared by GNOME and qt strata09:10
ssam2I think we are all in favour of making the model less strict09:10
jonathanmawpedroalvarez: can you confirm that weston is starting with the ivi-shell plugin?09:10
gtristanso as I understan, the naming should be icu-common.morph ?09:10
ssam2that's a good name09:10
ssam2we don't actually have a policy set09:11
jjardongtristan: yep, and dont forget to update all the systems that used to have qt or gnome stratum09:11
jonathanmawor possibly weston failed to launch properly, maybe for pam reasons09:11
jjardonto include the new icu-common as well09:11
gtristanok, so... it will take me some time, and will probably end up in a branch09:11
pedroalvarezjonathanmaw: I thinks so:
jjardongtristan: (yeah, runtime dependencies need to be set manually as well)09:12
gtristanwhich will have lots of commits in it... will have to take care to get gerrit to apply in the correct order09:12
gtristanmaybe it's better for me to hand-off that branch for review outside of gerrit and let me make changes to it before auto-applying commits blindly09:13
jjardongtristan: try to create less commits; the creation of icu and change of the systems/strata can be all go in a single commit09:13
pedroalvarezjonathanmaw: I didn't like the "for pam reasons" :)09:13
pedroalvarezI need to investigate09:13
* gtristan will first focus on getting it all to build, then we'll see about applying09:14
jonathanmawpedroalvarez: the complaint seems to be that it couldn't find /lib/security/pam_{selinux,console}.so09:14
gtristanjjardon, honestly if you guys prefer that, but it sort of messes with the fundamental usefulness of making concise/small commits which can be easily manipulated in the future09:15
jjardongtristan: sure, feel free to push them to a branch and ping here if you want a pre-review09:15
jjardongtristan: completely agree, but seems gerrit workflow is not meant to work very well with lot of small commits09:16
gtristangerrit being on crack should not block us from working properly :-/09:16
pedroalvarezjonathanmaw: both, pam_selinux, and pam_console are "optional" in  /etc/pam.d/ files09:17
pedroalvarezI guess those are like warning msgs, saying that it couldn't find them09:17
pedroalvarezI can be wrong09:17
jonathanmawpedroalvarez: I think optional refers to whether the module passes or fails, that may not extend to missing files09:18
jjardongtristan: yeah; we used to use mailing list, but the idea was to use gerrit so a machine can build the patches before someone review them09:18
jjardonalso it was difficult to keep track of all the pending patches for review09:19
gtristanDoes gerrit actually run a build ?09:20
richard_mawno, there was a plan to have a bot do builds of candidate changes and vote09:20
SotKnot yet09:20
gtristanMailing list is not a good comparison, there exists also proper bug tracking09:20
gtristanDare I say Bugzilla.09:20
gtristan(but I have used others, which do work, there is Jira, etc)09:20
* SotK does not want to use Bugzilla09:21
franredSotK, why?09:21
* paulsherwood does not want to use ****ing jira09:21
* franred was expecting paulsherwood to love it :P09:22
jonathanmawpedroalvarez: in my old GDP system that works, I don't see pam errors like that, but I also don't use weston-launch, I think.09:22
SotKfranred: I don't like its interface very much :)09:22
SotKI would however, rather use Bugzilla than Jira09:22
gtristanyeah, jira was crap - we had that a decade ago at a company I worked at09:22
pedroalvarezjonathanmaw: removed them from pam.d files, but still failing:
paulsherwoodjonathanmaw: do you have a candidate branch for gdp definitions yet?09:24
*** radiofree has quit IRC09:26
jonathanmawpaulsherwood: not yet, haven't finished building. x86 is 319/375, ARM is 305/370. I gather pedroalvarez is further on.09:27
jjardonphabricator seems to be the new cool thing09:27
jonathanmawpedroalvarez: it seems something is doing 'killall weston', but it's after weston had exited, so I'm not sure if it matters.09:28
*** radiofree has joined #baserock09:28
jonathanmawpedroalvarez: at line 453, weston.service exited, which would explain why the gdp-hmi-controller also failed.09:29
pedroalvarezjonathanmaw: I'd say that is happening before I run weston09:29
pedroalvarezhm.. sounds like I don't have to run weston manually09:29
jonathanmawpedroalvarez: I didn't have to09:29
pedroalvarezso this weston.service unit is doing "killall" on  ExecStop09:30
jonathanmawpedroalvarez: it seems so, not that it'd matter09:31
richard_mawjjardon: I've yet to see phabricator as being anything more than annoyingly named09:31
pedroalvarezso this unit is trying to use the drm backend09:31
pedroalvarezwell, weston-launch is09:31
pedroalvarezwhich I don't have with this kernel I'm trying now09:32
jjardonrichard_maw: Not sure what you mean but its being started to be used by some freedesktop projects. And wikimedia as well. Never tried myself though09:32
jonathanmawpedroalvarez: it seems my current jetson is running the drm backend successfully09:34
pedroalvarezjonathanmaw: yes, this is on a RPi09:34
richard_mawjjardon: The name is irritating enough that it's put me off evaluating it myself, and I know some people who dislike it for other reasons.09:35
Zarafrom what I've seen, people are more inclined to use it because it's under active development and they don't have the resources to work on an alternative, than because they like the thing itself.09:36
ZaraI know it's in php. other than that I dislike it on many aesthetic grounds; don't know much more than that.09:37
edcraggpedroalvarez: i thought drm was running? trying a new kernel?09:40
pedroalvarezedcragg: it wasn't very stable, and I mixing non-stable things makes everything difficult to debug09:40
jonathanmawnoooooooo... my builds have just started on qt. looks like it'll be a long day.09:41
pedroalvarezjonathanmaw: we should share our cache09:41
jonathanmawpedroalvarez: do you have any clever shell magic for that?09:41
jonathanmawI think I still have ssh keys for wherever is on09:41
pedroalvarezwell, we are using different branches, but I hope most of it will be similar09:42
pedroalvarezartifact-cache-server =
*** radiofree has quit IRC09:43
pedroalvarezjonathanmaw: in your /etc/morph.conf ^^09:43
edcraggpedroalvarez: i wonder why it wasn't stable, just that driver or something else09:44
tiredcatthanks for the show-build-log command09:44
jonathanmawpedroalvarez: is there a port number after the IP address? it seems I still need to build qtbase09:45
pedroalvarezedcragg: well, it's not complete, although anholt pointed out yesterday it could be a dodgy psu, which may be right09:45
pedroalvarezjonathanmaw: nope, port 8009:45
*** radiofree has joined #baserock09:45
paulsherwoodjonathanmaw: is your current state-of-play?09:46
jonathanmawpaulsherwood: that's what I'm currently using, yes09:47
pedroalvarezjonathanmaw: your branch is not on top of current master09:50
pedroalvarezI believe if you rebase, it will use my cache09:51
jonathanmawpedroalvarez: gdp-rebase-wip or gdp-rebase-wip-rpi?09:54
pedroalvarezyours on top of master, not on top of mine09:54
pedroalvarezmine are a bit of a mess09:54
paulsherwoodjonathanmaw: have you done something to defaults in that branch? ybd keels over there09:55
jonathanmawpaulsherwood: there should only be what the migrations generated, there.09:56
pedroalvarezjonathanmaw: baserock/pedroalvarez/gdp-rebase-wip is my rebase + fixes09:58
pedroalvarezwhich are similar to yours09:59
pedroalvarezthe *rpi one has a lot of patches on top to support rpi deployments etc09:59
jonathanmawhrm, it doesn't look like rebasing my changes onto yours would be a good idea. though I don't have your latest three changes: audio-bluetooth dependency; boost makeflags; audiomanagerdemo depending on audiomanager10:05
jonathanmawdid they cause build failures?10:06
jonathanmawI'm wondering whether I should merge them now, or hope this build finishes today, so I can do proper work, then merge them later.10:07
pedroalvarezaudio-bluetooth build failure10:08
pedroalvarezboost, build speed10:08
pedroalvarezand audiomanagerdemo build failure10:09
pedroalvarezjonathanmaw: no, don't rebase on top of mine.10:09
pedroalvarezbut if you rebase on top of master, you will get the qt chunks from the artifact server I gave you10:10
pedroalvarez(there is a change on build-essential on my branch and not in yours, that's why  we can't use the same cache)10:10
pedroalvarezstill failing  using the fbdev backend├▒
pedroalvarezsystemd[1]: Failed to insert module 'kdbus': Function not implemented10:33
* gtristan has push a couple more patches to lorries & definitions10:38
gtristanlets spoon feed the beast in small mouthfulls10:38
gtristanNOTE: Lorries need to be added *first*, as definitions patches refer to new lorries10:38
pedroalvarezhm.. I merged a couple but then realise that it would be wise to add the gnome/ namespace10:43
pedroalvarezgtristan: do you think that makes sense to put them in the gnome/ namespace?10:45
pedroalvarezadwaita icon theme and gnome backgrounds10:45
pedroalvarezI will do the change before g.b.o lorries them :)10:46
paulsherwoodjonathanmaw: migrations?10:47
pedroalvareztiagogomes: quick review?
tiagogomesI'd had give +3 due the alphabetic order, but gerrit doesn't allow it10:53
pedroalvarezheh :)10:53
gtristanpedroalvarez, you mean in the lorry controller ? gnome/ ?10:54
gtristanI dont know, I asked something about that last week (or before ?) as I found that they didnt really match what was in definitions/strata/gnome10:54
gtristanwhatever you prefer... I will follow :D10:55
pedroalvarezgtristan: it would be good to have all the gnome related  repos using the gnome/ namespace, see;
pedroalvarezas you can see, we failed in some cases, anyway10:56
* jjardon doesnt like that idea: now you dont know if you component is in upstream:<component> or upstream:<namespace>/<component>10:56
gtristanpedroalvarez, indeed, alot is questionable as to what is specific to gnome, perhaps saying: "if it comes from, put it in gnome/" makes sense at the lorry level ?10:57
gtristanclutter/cogl is used outside of gnome (I think webkit had cogl at one point at least)10:57
* gtristan ... afk... helping a friend study french...10:58
pedroalvarezyes.. this is not perfect...10:58
*** zoli__ has joined #baserock10:59
*** zoli__ has quit IRC11:00
*** zoli_- has joined #baserock11:00
ssam2jjardon: it's basically workaround for the fact that has 1000s of lorries11:10
ssam2jjardon: there are better ways to make that bearable, but they require effort upfront to make the mirroring smarter in downstream troves, I think11:11
pedroalvarezjjardon: qt5 to ci?11:13
pedroalvareztoo many systems there already11:14
pedroalvarezsome bugs on morph distbuild making it slow11:14
pedroalvarezmason relying on morph11:14
paulsherwooduse ybd :)11:14
pedroalvarezpeople pushing for other build-tools11:14
pedroalvarezother build tools not really full featured..11:15
jjardonpedroalvarez: as stated there; so it doesn't bitrot and everyone can use the cache11:15
pedroalvarezjjardon: the cache will not be of any use for people using ybd]11:15
paulsherwoodpedroalvarez: i'm interested to know what features are actually needed11:16
jjardonssam2: is it really a problem to have 100000 lorries flat? in that way I do not have to check when Im building a system if my component is in upstream:<component> upstream:<some namespace>/<component>11:16
paulsherwoodjjardon: it wouldn't be a problem, except that we already went the other way, and have no policy for deletion11:17
jjardonpedroalvarez: that doesnt invalidate my points; Its actually another different bug: make cache compatible between different tools11:18
pedroalvarezwe should go for a common tool. in my opinion11:18
pedroalvarezpaulsherwood: not sure how many from morph are essential tbf11:18
pedroalvarezbut there are some useful things in there11:19
SotKmorph does a fair bit of stuff that I think should be in a separate tool for interacting with definitions11:21
pedroalvarezI agree11:21
ssam2jjardon: I think it would be useful to be able to mirror *some* of git.baserock.org11:21
pedroalvarezmorph anchor, morph certify11:21
SotKget-repo and the manifest stuff too11:22
pedroalvarezssam2: I agree with you, but I also agree with the point that nothing stops a gnome/* repo becoming essential for all the systems11:23
pedroalvarezSotK: yes11:23
ssam2yeah, the current solution is a bit rubbish, but I think it's better than nothing11:23
SotKI'd like a library for parsing definitions, and then a number of tools which use that library for their own ends (eg build, deployment, manifests, "morph edit" type stuff)11:23
paulsherwoodssam2: i've been considering that kbas could be extended to act as a git mirror11:23
pedroalvarezpaulsherwood: don't fall in the trap of putting too many features to a tool11:24
pedroalvarezSotK: I'd be interested on such a thing11:25
paulsherwoodssam2: isn't it just a matter of allowing clients to retrieve contents from a directory? (which is what kbas already does...)11:25
paulsherwood(and ybd already populates a directory with a subset of repos from a trove11:25
ssam2paulsherwood: that's not at all what lorry-controller does11:26
ssam2maybe we mean different things by 'git mirror'11:26
ssam2I guess l-c is 'git mirror and import'11:26
ssam2but a useful git mirror still needs cgit and git-daemon running, in my opinion11:27
SotKssam2: +111:27
paulsherwoodfair enough... my point was only that a machine running ybd (or morph) already has pulled a subset of g.b.o11:27
paulsherwoodand could serve that content to other machines11:28
ssam2yes, that's true11:28
ssam2it would be nice to make git mirroring more of a 'commodity'.11:28
paulsherwood(reducing load on g.b.o, and being faster for machines on their local network)11:28
ssam2looking at it from the other side, we could try to make it possible to run Trove in a container, and remove morph and YBD's less featureful git caching stuff11:29
ssam2might be impractical, though11:29
* SotK wonders if there is an issue with trusting some random machine's git cache as opposed to a Trove11:29
jonathanmawpaulsherwood: definitions has a directory of migrations for changing the definitions when the format changes.11:30
paulsherwoodjonathanmaw: yup. sadly the DEFAULTS created doesn't include all the defaults ybd cares about (missing build-steps) so i need to amend ybd before DEFAULTS lands in master (or the migration could be improved)11:31
jonathanmawpedroalvarez: any luck with fbdev?11:32
pedroalvarezjonathanmaw: nope, I posted a log before:
pedroalvarezblank screen11:33
jonathanmawgdp-hmi-controller still failing to connect to the display, but no errors from weston, that I saw.11:34
jonathanmawis weston currently running?11:34
pedroalvarezjonathanmaw: yes, but shows nothing11:34
jonathanmawpedroalvarez: quite possibly, we haven't solved the race condition, but it didn't appear on the Jetson, so the gdp-hmi-controller is starting too quickly.11:35
jjardondo baserock people have any preference between smack, selinux, apparmor or TOMOYO ?11:35
pedroalvarezjonathanmaw: any workaround for this?11:36
jonathanmawpedroalvarez: I solved it for yocto by creating a .path unit that starts gdp-hmi-controller when weston is running11:36
Kinnisonselinux is very complex to deal with at the userland level, apparmor is not as full coverage.  TOMOYO I don't know a lot about, and smack seemed fairly new last tme I saw it.  But if I had to pick one, I'd likely go with smack as something new and interesting.11:36
paulsherwoodjjardon: there is AGL discussion on that topic. i'm not sure it's safe to take sides in any case, we probably will end up with examples of more than one11:37
jonathanmawwhere you know weston is running when its socket exists11:37
pedroalvarezoh, annoying I can't restart the unit11:39
jonathanmawso put a gdp-hmi-controller.path unit (which wants weston) in, wanted by, and have gdp-hmi-controller want gdp-hmi-background and gdp-hmi-launcher211:39
*** zoli_- is now known as zoli__11:39
jonathanmawprobably easiest for now to take all of that stuff out of, and see what happens when you start weston manually, then start the gdp-hmi-{controller,background,launcher2} manually11:40
richard_mawjjardon: I've seen how the sausage is made for selinux, and while it's competent as a security mechanism, the tooling for making the rules didn't appear to be traceable. I have no experience with apparmor or TOMOYO, but smack has good support in systemd, its developers appear responsive, and the labelling looks simple.11:40
richard_mawjjardon: however you don't need a big security module to secure a system11:40
richard_mawjjardon: capabilities and namespaces are also a useful mechansim which is more portable, but if you want to do your dilligence with security, you'll want to layer on as many technologies as makes sense11:41
jjardonrigth, maybe it would be a interesting project for someone to spend some time on this so we can make a choice as a project11:42
richard_mawpaulsherwood: Aye, the initial design for the Baserock systems said we should use smack when it became relevant, but I'm aware of projects that would benefit if Baserock did SELinux.11:42
* richard_maw wonders if the Android SELinux stuff has nicer tooling11:43
* jjardon doesnt know anything about this neither11:43
* richard_maw prefers namespaces to macs11:43
jjardonsmack seems a bit death though: the repo is pointing to gitorious...11:43
pedroalvarezjonathanmaw: if I restart gdp-hmi-controller it works11:44
pedroalvarezso you may be right on the race condition11:44
pedroalvarezcontroller and launcher2 fail, though11:44
jonathanmawwhat does the log say about why controller and launcher2 fail, now?11:45
pedroalvarezsorry, I meant background an launcher211:48
pedroalvarezgetting logs11:48
jonathanmawI wonder if I can track down what error 3004 means for qtwayland11:52
jonathanmawerror 3004 is EGL_BAD_ATTRIBUTE11:54
jonathanmawmy best guess is that the version of egl you're using on the RPi doesn't work with qtwayland 5.4.011:59
pedroalvarezoh well11:59
pedroalvarezthis might be the end of my trip11:59
pedroalvarezI assume egl is part of the kernel12:04
pedroalvarezor bsp12:04
jonathanmawfor RPi, it might be12:04
pedroalvarezalright then... nothing else to do with this12:06
pedroalvarezI'll deploy to a jetson :)12:07
paulsherwoodjonathanmaw: sh -c 'cmake -DCMAKE_INSTALL_PREFIX="$PREFIX"'"'"''12:53
paulsherwoodsh: syntax error: unterminated quoted string12:53
paulsherwoodin genivi-common-api-runtime on x86?12:54
pedroalvarezthose definitions were migrated to version 7.. and they shouldn't have12:56
jonathanmawpaulsherwood: ah, it seems to be a stray ' in the DEFAULTS12:56
paulsherwoodjonathanmaw: is that a bug your end or mine? (ybd maybe...)12:57
pedroalvarezbug on the migrations I'd say12:57
pedroalvarezI'll send a patch for that12:58
paulsherwoodpedroalvarez: could you add build-steps while you're at it?12:58
pedroalvarezpaulsherwood: what is build-steps?12:58
pedroalvarezok, I'll read :)12:59
paulsherwoodit's another lump of data hiding in morph12:59
pedroalvarezI see13:01
pedroalvarezhm.. I don't know if that will make anything else crash :/13:01
pedroalvarezIt shouldn't13:01
*** gtristan has quit IRC13:03
paulsherwoodnever mind13:06
paulsherwood(ie, no need to bother with build-steps)13:07
pedroalvarezI agree with removing logic from the build tool though, but Sam has been thinking about this a lot, and I'd like to check his ideas first13:07
*** mariaderidder has quit IRC13:07
ssam2not sure if putting build steps in DEFAULTS is really worth it -- do we want to change them that often?13:30
ssam2in principle, though, I would still like the definitions format to be a standard, not an implementation detail of any one build tool13:30
ssam2the current migration for definitions version 7 matches does what I proposed definitions version 7 would do13:31
ssam2if you want to change the definitions format further (and please do!!) I think it would be better to think in terms of changing the format specification at
ssam2not patch the migration for definitions v7 to do things which weren't proposed as part of v713:32
ssam2and basically, I mean, propose the change on the mailing list13:32
ssam2maintaining a standard that's independent of the tooling is always a bit more work than treating the format as an implementation detail, but i don't think the process should be more complex than 'propose, discuss, implement'13:33
ssam2I do still think that sequential version numbers for the format are useful.13:34
ssam2thanks for by the way  :)13:34
paulsherwoodssam2: build-steps could be modified to have (eg) license-check, static-analysis or whatever. or maybe to do something completely different from build, eg code analysis13:41
paulsherwoods/modified/modified by users for their use-case/13:43
ssam2but users will have to modify their build tool to understand those anyway13:46
ssam2so there doesn't seem any advantage to moving one part of it into definitions.git13:46
ssam2hmm, I guess all of them are treated the same actually13:46
ssam2I would much rather see it done as a new version of the definitions format, though13:46
paulsherwoodfair enough13:46
paulsherwoodi'm still pondering how to improve this versioning process, but i have nothing positive to offer13:47
*** gtristan has joined #baserock13:53
*** mariaderidder has joined #baserock13:54
richard_mawssam2: build step classes differ in the order in which they are run, and whether MAKEFLAGS is set (I'd also like to prevent DESTDIR being set for anything but install-commands, but that's a new thing), you could fairly easily come up with a declarative format for the sequence of commands required to do a build, which your build tool could use13:55
ssam2I also think that license checking and static analysis doesn't belong in the build tool14:04
* SotK agrees with that14:04
gtristanpedroalvarez, just pushed new versions of those patches, and removed the build-system line for adwaita-icon-theme too14:04
pedroalvarezgood, I'll take a look14:05
* gtristan basically rebased; hit 'git review -R' and held his breath14:05
gtristanseems like it gobbled them up alright :)14:05
* gtristan palmface14:10
gtristanI missed the comment about
pedroalvarezheh, I was going to point that out14:10
gtristanit will fail the applet build but at least not merge conflict14:14
* gtristan is carefully uploading a newer patch14:14
* gtristan has this trick which is probably insane14:15
* gtristan checks out a separate definitions repo, and git format-patch / git am's the commit from the old checkout to the new one14:16
gtristanthis results in git review only uploading one patch and not overwriting stuff14:16
franredpedroalvarez, ssam2, do you know if could be still applied?14:20
pedroalvarezfranred: sorry, I don't have brain left for that patch14:22
ssam2i don't see why not. We can always revert14:23
franredssam2, merged14:25
pedroalvarezI hope we are testing things :)14:25
pedroalvarezjonathanmaw: any luck?14:33
jonathanmawpedroalvarez: the builds have finished, transferring ARM artifacts to my Jetson so I can deploy it, now14:33
pedroalvarezI've found another thing neede, not sure you have spotted it already14:34
pedroalvarezglib-common stratum needs to be in the system14:35
tiredcatfranred, failing to keep up with you >.>14:39
franredtiredcat, ?14:40
tiredcati just mean you're too quick :)14:40
pedroalvarezjonathanmaw: yay, this worked for me on a jetson14:51
paulsherwoodpedroalvarez: what did?14:53
pedroalvarezgdp on jetson14:54
paulsherwoodcool... where is the commit for that?14:55
pedroalvarezpaulsherwood: it looks like the qtwayland is not compatible with the version of egl of the rpi14:55
pedroalvarezthe version of qtwayland*14:55
jjardonfranred: there are curretly are duplicated chunks in definitions (nasm for example): does that patches make the build to break?14:56
tiredcatjjardon, in the same system?14:56
jjardontiredcat: no, globally14:57
tiredcatokay that's fine then14:57
franredjjardon, only if they are in the same system in different chunks (it prevents for a very nasty bug where python report recursive error)14:57
pedroalvarezpaulsherwood: tip of baserock/pedroalvarez/gdp-rebase-wip14:57
jjardontiredcat: nice, I think we should apply it then14:57
franredjjardon, it is applied already :)14:58
paulsherwoodpedroalvarez: how different is that from what jonathanmaw pointed me at earlier?14:58
*** franred has quit IRC15:00
pedroalvarezhe has started to move things to 'automotive/'15:00
pedroalvarezhis branch includes the migration to version 715:01
paulsherwoodi wonder why he did that...15:01
paulsherwood(the migration)15:01
pedroalvarezpaulsherwood: he ran all the scripts in place15:01
paulsherwoodyes, but why? definitions master doesn't have DEFAULTS15:02
pedroalvarezIf you are not in the loop with the versioning of baserock, is an easy mistake15:02
pedroalvarezpaulsherwood: I had the same problem, I rebased on top of master, needing to apply the migrations for all the genivi-demo-platform definitions15:03
paulsherwoodsorry, i don't understand, really. lots of things are confusing me today15:04
jonathanmawerk. it looks like the artifacts I built on the ARM could machine have different cache keys to the stuff I built on Jetson, the Jetson still tries to build qtwebkit15:05
paulsherwoodjonathanmaw: using morph?15:05
jonathanmawpaulsherwood: yes15:05
* paulsherwood doesn't know whether to laugh, cry, or go to the pub15:06
jonathanmawexact same morph version, apparently.15:07
paulsherwoodjonathanmaw: are you confident that morph is building for the same architecture in both cases?15:07
pedroalvarezand that you are using the same commit of definitions?15:07
jonathanmawnot at all, from within the chroot, `uname -a` is reporting armv8l, while jetson reports armv7l15:07
pedroalvarezthat shouldn't matter15:09
pedroalvarezMorph would have complained if you werent building for arm15:09
ssam2i have a hacky branch of morph somewhere that prints all the cache ids (not the sha1s, but the actual contents) for a system15:14
ssam2it might be useful for debugging that15:15
*** toscalix__ is now known as toscalix15:15
toscalixsee you guys tomorrow in Manchester15:15
ssam2I've pushed to morph.git as branch sam/list-cache-keys15:15
*** paulwaters_ has quit IRC15:16
ssam2it patches list-artifacts so you can do `morph list-artifacts $(pwd) $(git rev-parse HEAD) systems/foo-system.morph' to show all cache ids for a system from your current directory15:16
ssam2list-artifacts is old, so it takes a full repo, ref, morph parameter instead of looking at current working directory15:16
*** toscalix has quit IRC15:17
*** paulwaters_ has joined #baserock15:18
jonathanmawpedroalvarez: you have a working thing with artifacts, is that with your branch or mine?15:27
wdutchwhat is 'trove-id' in morph.conf?15:29
pedroalvarezjonathanmaw: use "artifact-cache-server =" :)15:29
jonathanmawpedroalvarez: I apparently am15:30
pedroalvarezthen use my branch, and enjoy them15:30
*** paulwaters_ has quit IRC15:32
jonathanmaw:/ it's building zlib and binutils15:32
jonathanmawI can ping and everything...15:33
pedroalvarezjonathanmaw: `morph --version` ?15:33
pedroalvarezhere: 60c378c55d5d0ef89184b49ae95e445f8de422e315:34
jonathanmawwhich is a commit from September 18th15:34
pedroalvarezwhich was the commit of the latest release15:35
wdutchanybody know what effect trove-id has in a morph.conf?15:36
paulsherwoodwdutch: tells it where to get sources from. default is g.b.o iiuc15:37
KinnisonIt adjusts the default repo aliases15:37
Kinnisonthe default trove-id is 'baserock'15:37
Kinnisonit also adjusts things like the branches morph uses for temporary refs etc15:37
Kinnisontrove-host is the hostname and defauls to the value of trove-id if trove-host is unset and trove-id is set to non-baserock15:38
Kinnisonit's a non-trivial bit of fiddling15:38
jonathanmawpedroalvarez: and you're able to self-upgrade from a jetson using that cache?15:38
wdutchso for a trove of gbo, the id should be baserock?15:39
Kinnisonyou should just leave it unset15:39
pedroalvarezjonathanmaw: yes15:39
jonathanmawand we're using the same version of morph, and building the same system in definitions...15:40
jonathanmawi.e. we're both using baserock/pedroalvarez/gdp-rebase-wip15:40
jonathanmawsha1 starting 10fa03b15:40
KinnisonIf you're using a 32bit chroot on armv8l64 without somehow telling the kernel to report armv7 then morph is going to build armv8l3215:40
pedroalvarezjonathanmaw: right, the system artifact was missing the latest commit15:41
pedroalvarezjonathanmaw: creating it again15:41
jonathanmawit's not even the system artifact, I'm getting stuck much lower down in the build15:41
Kinnisonyou could try setarch armv7l chroot ...15:42
KinnisonYou might need -B too15:42
pedroalvarezKinnison: shouldn't Morph complain about tryig to cross-compile?15:43
KinnisonI'm unsure, possibly15:44
* Kinnison has, admittedly, not played with that code in about 7 months15:44
paulsherwoodjonathanmaw:* should have (some) ybd artifacts for the branch you directed me to earlier15:47
jonathanmawpaulsherwood: ok, do I need to bother with a kbas-password?15:48
paulsherwoodnot if you're only downloading15:48
jonathanmawpaulsherwood: odd, I seem to be encountering an error building from master, "KeyError: base"15:50
paulsherwoodjonathanmaw: please use latest tag (of ybd)15:51
paulsherwoodi believe so15:52
paulsherwoodsadly, i break master all the time15:52
jonathanmawstill getting KeyError: 'base'15:52
paulsherwoododd. are you using a machine with /src ?15:53
jonathanmawpaulsherwood: yes15:53
jonathanmawit appears to be a separate disk15:53
paulsherwoodplease put 'base: /src' into ybd.conf in the dir containing ybd15:54
jonathanmawpaulsherwood: there is no ybd.conf in the dir containing ybd15:54
paulsherwoodjonathanmaw: ack. please create one15:55
jonathanmawi.e. create a /src/ybd/ybd.conf, with "base: /src" in it?15:55
paulsherwoodyes please.15:55
* paulsherwood hates all software now, not just baserock :)15:56
* pedroalvarez too15:56
jonathanmawfun new error!15:56
paulsherwoodyes. remove that DEFAULTS file you sneaked in :)15:57
paulsherwoodor go to master ybd and accept that you're in the same uncharted waters as i am :)15:58
jonathanmaw>:[ it's trying to build stage1-gcc15:59
jonathanmawand then once it had the git repo mirrored, it grabbed the cache key16:00
jonathanmawbecause that's necessary and I was being pessimistic16:00
pedroalvarezwell that's good :)16:01
paulsherwoodi'm pessimistic too, today :)16:01
* jonathanmaw twiddles while kernel downloads16:01
pedroalvarezwell, I've sent a mega patch, that I don't think is ready to merge:
paulsherwoodwhat's wrong with it?16:02
pedroalvarezwell, there are a couple of things that may not be related with the commit msg16:03
pedroalvarezbut anybody is welcome to pointme to errors that would prevent it to be merged16:03
paulsherwoodcan it break anything that's already there?16:03
pedroalvarezI hope not16:04
jonathanmawwe've added qtwayland to everyone's qt5-tools.morph, we've added some magic to the weston chunk, and we've changed all the GENIVI baseline systems to use input-genivi16:05
pedroalvarezyup, I'm worried about the first 216:06
jonathanmawI think not having qtwayland in qt5-tools may have been an omission when qt5-tools was merged; I think radiofree said as much, but I may be mistaken16:06
pedroalvarezand looks like the magic for weston was created on a different stratum16:07
pedroalvarezI'm going to document the changes a bit on the commit msg16:08
jonathanmawhmm, we already have a weston-gdp chunk which is different from that weston chunk16:08
jonathanmawso we could probably get away with undoing the change to weston.morph16:08
SotKcan it perhaps be split into a few commits (eg one to add qtwayland to qt5-tools, one to do magic, one to use input-genivi and one to add the gdp)?16:09
pedroalvarezSotK: yes, looking into that16:09
pedroalvarezjonathanmaw: do you see the "$" here?
pedroalvarezis that part of any magic I don't know? :/16:13
jonathanmawpedroalvarez: yes. I have no idea what it does there.16:13
* pedroalvarez want's to remove it!16:14
jonathanmawit's a configure-command, so I suppose the best way to find out if it's necessary is to try and build and see if things go wrong if it's removed16:14
richard_mawI'd hazard a guess that it's unintentional16:14
jonathanmawI'll go with the shell wizard's guess16:15
jonathanmawyaml apparently uses $ to indicate scalars, but it's not formatted for it to parse it that way16:16
pedroalvarezIt builds without it.16:18
*** ssam2 has quit IRC16:33
pedroalvarezhere we go:
pedroalvarezplease note, some patches to clean up will follow up16:35
pedroalvarezthere is a plan to move everything related with Genivi to a different folder in definitions16:36
pedroalvarezoh, looks like my first commit is wrong..16:37
jonathanmawpedroalvarez: i.e. it still alters weston-genivi?16:39
pedroalvarezit changes max-jobs to qtwebkit, but a morphology that is not being used :/16:39
pedroalvarezI'll keep the weston-genivi change, I think is good16:40
pedroalvarezoh, no It is being used16:41
pedroalvarezI think it can be merged as it is, if people here trust we will maintain it16:42
paulsherwood+1 :)16:43
paulsherwoodpedroalvarez: whatabout the one you've abandoned?16:43
pedroalvarezduplicated patches16:44
paulsherwoodok, so you're seeking review for the rest?16:44
pedroalvarezther are 516:44
* paulsherwood will need to build them16:45
paulsherwoodis there a branch anywhere?16:46
paulsherwoodpedroalvarez: ^^16:46
*** mariaderidder has quit IRC16:47
pedroalvarezpaulsherwood:  baserock/pedroalvarez/gdp-rebase16:47
*** tiagogomes has quit IRC16:49
paulsherwood3 15-10-15 00:00:18 [1/51/340] [dbus-python] Starting build of dbus-python.6e1f67646cce327a9d559cc0bea8362945929fc5250e949f3e57ab3e7cc48e7416:50
paulsherwoodshouldn't take too long16:50
pedroalvarezthat's good16:50
paulsherwoodinteresting.... 15-10-15 00:03:02 [11/51/340] [mesa-common-vm] Autodetected build system is NOT FOUND16:52
paulsherwoodpedroalvarez: ^^16:52
pedroalvarezpaulsherwood: I accidentaly duplicated the patch when sending a second version, that's why I abandoned it. But all the changes are included in the other 516:54
*** paulwaters_ has joined #baserock16:55
paulsherwoodpedroalvarez: that's off your branch at g.b.o16:55
pedroalvarezgrr.. sorry, answered to a Paul of some minutes ago16:55
paulsherwoodthese last minute fixings are so exciting, don't you think? :-)16:56
pedroalvarezwhat is that!16:56
pedroalvarezis that ybd failing to autodetect the build system?16:56
jonathanmawybd refused to deploy16:56
jonathanmawso I specified on the command-line that I was using the armv7lhf arcitecture, and it tried to builld eeverything from binutils16:57
paulsherwoodybd refused to deploy?16:57
jonathanmawwhen I try to deploy without specifying the architecture on the command-line, I get the message "15-10-15 00:00:12 [genivi-demo-platform-armv7lhf-jetson] Skipping deployment for armv7lhf"16:57
jonathanmawwhen I add that to the command-line, it tries to build binutils16:58
paulsherwoodhmmm. when you built, did you specify architecture on the commandline?16:58
jonathanmawdid you?16:59
paulsherwoodalways, afaik16:59
paulsherwoodmaybe i should make that non-optional16:59
*** bashrc has quit IRC16:59
paulsherwoodstill, weird... what artifacts were you actually retrieving, is the question? :)17:00
pedroalvarezmy build starts from stage2-make17:04
*** gary_perkins has quit IRC17:06
*** gary_perkins has joined #baserock17:06
*** jonathanmaw has quit IRC17:15
pedroalvarezpaulsherwood: good catch, fixed, force pushed, and patch updated17:19
*** paulwaters_ has quit IRC17:33
pedroalvarezpaulsherwood: heh, my buld started from gcc, I guess it's getting your artifacts17:38
pedroalvarezwhich is good17:38
*** paulwaters_ has joined #baserock17:39
paulsherwoodyup... trouble is, we're still racing for multi-machine. need some kind of multi-machine equivalent of the claim17:39
paulsherwood(so all machines are currentyl building gcc)17:39
*** paulwaters_ has quit IRC17:58
jjardonpaulsherwood: your definitions is quite old :) mesa-common-vm is not needed anymore with new versions of mesa18:18
paulsherwoodjjardon: not mine, pedroalvarez' :)18:18
paulsherwoodit's fixed now :)18:19
pedroalvarezyup, i fixed that18:22
pedroalvarezlooks like there are other things to fix though18:23
pedroalvarezand to be honest, I'm not sure about some of the comments :/ (gtk2 vs gtk3, llvm...)18:24
pedroalvarezI'll try to discuss that tomorrow with jonathan :)18:24
pedroalvarezjjardon: thanks for the review anyway!18:24
jjardonpedroalvarez: gtk2 will not work on wayland, and you do not have x-generic in the system, thats why I asked ;)18:25
pedroalvarezI'll manage to fix and test that tomorrow :)18:42
*** paulwaters_ has joined #baserock19:04
*** Stanto has quit IRC19:08
*** Stanto has joined #baserock19:11
*** Stanto has quit IRC19:18
*** Stanto has joined #baserock19:21
*** paulwaters_ has quit IRC20:30
*** paulwaters_ has joined #baserock21:52
*** tiredcat_ has joined #baserock23:30
*** tiredcat has quit IRC23:30
*** persia has quit IRC23:30
*** persia has joined #baserock23:30
*** nowster has quit IRC23:34
*** benbrown_ has quit IRC23:34
*** nowster has joined #baserock23:37
*** benbrown_ has joined #baserock23:41
*** edcragg has quit IRC23:56
*** petefoth has quit IRC23:56
*** ctgriffiths has quit IRC23:56
*** dabukalam has quit IRC23:56
*** Zara has quit IRC23:56
*** doffm has quit IRC23:56
*** bwh has quit IRC23:56
*** Kinnison has quit IRC23:56
*** radiofree has quit IRC23:56
*** lachlanmackenzie has quit IRC23:56
*** SotK has quit IRC23:56
*** benbrown_ has quit IRC23:56
*** ratmice____ has quit IRC23:56
*** perryl has quit IRC23:56
*** paulsherwood has quit IRC23:56
*** jmacs_ has quit IRC23:56
*** zoli__ has quit IRC23:56
*** drnic has quit IRC23:56
*** rjek has quit IRC23:56
*** tlsa has quit IRC23:56
*** sebh has quit IRC23:56
*** wdutch has quit IRC23:56
*** gary_perkins has quit IRC23:56
*** juergbi has quit IRC23:56
*** JPohlmann has quit IRC23:56
*** gtristan has quit IRC23:56
*** pedroalvarez has quit IRC23:56
*** bjdooks has quit IRC23:56
*** bfletcher_ has quit IRC23:56
*** De|ta has quit IRC23:56
*** petefotheringham has quit IRC23:56
*** nowster has quit IRC23:56
*** inara has quit IRC23:56
*** tiredcat_ has quit IRC23:56
*** richard_maw has quit IRC23:56
*** rdale has quit IRC23:56
*** robtaylor has quit IRC23:56

Generated by 2.14.0 by Marius Gedminas - find it at!