IRC logs for #baserock for Tuesday, 2015-12-15

*** gtristan has quit IRC04:03
*** gtristan has joined #baserock04:29
SotKargh, ran out of space!07:45
gtristansorry, it's *big* hehe07:47
gtristan7GB image now (well, 6GB wasnt enough so I cranked it)07:47
gtristanSotK, I home webkit build completed at least !07:47
gtristanthats a real pain to rebuild07:48
SotKno worries, I think my artifact cache is full of old stuff07:48
* gtristan regularly removes the gnome-system artifact07:48
* SotK notices a bunch of chunk names are appearing as "foo.morph" in his morph log08:03
SotKbut not all of them... (eg. "Building chunk epiphany", "Building chunk yelp.morph")08:04
*** toscalix has joined #baserock08:19
VLetrmxwelcome back SotK :)08:20
gtristanok I think I have spotted a terrible flaw, but I can't be entirely certain08:21
gtristanNow, how do we know what SHA1 a given checkout will clone as a submodule when building ?08:21
gtristanThis is why richard_maw's proposing of multiple source declarations for a given chunk is so important, right ?08:22
VLetrmxi think we're capable of tracking submodules in baserock, it's just not very convenient without multiple sources, and currently i'm not sure we bother to commit submodule updates to fix the sha08:29
gtristanVLetrmx, where is this defined ?08:30
gtristanI.e. how do I know that when building a given SHA1 for module A, I will get a specific SHA1 for it's submodule ?08:31
gtristanthe .gitmodules file is supposed to specify more I guess ?08:33
VLetrmxyou update the submodule, cd into the directory and git checkout the sha you want, then you can commit the subproject at a specific sha afaict08:34
gtristanWhat does this mean  "commit the subproject at a specific sha" ?08:35
*** rdale has joined #baserock08:37
VLetrmxlike this
gtristanWhat is the "common" file ?08:39
gtristansomething understood by git ?08:39
VLetrmxcommon is just the submodule in this case, that file gets generated by git when you run a submodule update08:41
franredgtristan, could not create a branch, and point the submodule to the sha1 that youwant should the problem at the moment? (I think is what VLetrmx is describing anyway)08:44
franredcreate a branch for the repository which contains the submodule08:44
gtristanVLetrmx, I see, so "common" is a file which git generates and can be committed to the git repo along with the .gitmodules file08:48
gtristanright ?08:48
gtristanthen if the file already exists at the given ref of the module, cloning that module at that given ref will automatically clone the submodule at the specified ref ?08:49
gtristanfranred, yes, that is what I didn't understand... *where* one can point a submodule, i.e. where that information is declared08:50
gtristanso basically, we are way off base with the current stuff, none of the submodule fixes we have in place in our baserock/ branches even bother to specify a submodule08:51
gtristansorry, specify a submodule SHA108:51
gtristanwhich means, every time we build anything with a submodule currently, we are just getting whatever happens to be in master for that module08:52
gtristanunless the upstream had *at least* specified a branch or tag for that submodule08:52
gtristanDoes git at *least* handle this automatically when creating a tag ?08:53
gtristanso when I tag a git module which has a submodule, checking out a tag of the parent will get the exact state of the submodule when I tagged the parent module ?08:53
franredgtristan, Im not sure, but I think in definitions we don't specify submodue SHA1. You need to modify the parent repository, pointing implicity the submodule to the SHA1 you want that baserock clones08:54
gtristanfranred, right, but we *do* have such a branch for every, single, module that has any submodules08:54
gtristanin order to at least point it upstream08:54
gtristanhowever we do not specify anything for which SHA1 we want08:55
gtristananyway, I managed to only notice this now08:55
gtristanthanks for clarifying08:55
franrednowster, we do not do anything with submodules at the moment, more than clone them in the specific SHA1 that the parent repository points at. That was why richard_maw proposed the multiple sources declaration08:56
gtristantoscalix, a few weeks ago, you asked if for a given checkout/tag of the definitions repository, we could build *exactly* the same thing in 10 years08:56
gtristantoscalix, I was incorrect to answer that you could - the submodule situation blocks that08:57
gtristanshould not be extremely hard to fix08:57
*** milad has joined #baserock08:58
toscalixgtristan: to me this is a very important point. Maintainability is key for "business" success08:58
*** milad has joined #baserock08:59
gtristanI think it's very high priority too, I think that reproducibility is where baserock shines08:59
gtristanwe should at least get that part perfectly08:59
VLetrmxgtristan, we don't need to change anything anyway, the submodule's sha is tracked when you add it as a submodule08:59
franredgtristan, there are also another case where the multiple source declaration is useful: when a repository requires another repository source code for building08:59
franredgood morning09:00
*** milad has quit IRC09:00
VLetrmxbut if we want to deviate from upstream then we can do so without multiple sources09:00
VLetrmxbye milad :)09:00
*** milad has joined #baserock09:00
VLetrmxhi milad :)09:00
*** milad has quit IRC09:00
VLetrmxbye milad :)09:00
gtristanfranred, I think it's important to not mix that up09:02
*** milad has joined #baserock09:02
gtristanfranred, i.e. if it's really a separate repository, it should be a separate chunk09:02
*** ctbruce has joined #baserock09:02
miladi have problems with vehicle web api09:02
gtristanwe should not have a policy which allows one to build an entire system inside one chunk, just because we have multiple sources capabilities09:03
miladi can install vehicle web api on ubuntu, is there any one who can help me with this09:03
miladsorry, i can't09:03
miladany experience with this api?09:04
franredgtristan, look at this example and
franredrabbit-codegen is required by rabbitmq-server at compile time but it is not required in the system - we have a chunck which copies crap into the system for building another chunk09:05
franredgtristan, ^^09:05
franredmilad, do you want to install vehible web api in baserock?09:06
miladno. actually i want to install it in ubuntu 1509:07
miladi want develop some web apps09:07
miladand i need it09:07
miladthe build scripts are not working09:07
*** edcragg has joined #baserock09:07
miladthe vehicle data simulator and etc are very important to me09:07
gtristanfranred, there are other examples where multi-source is really important though...09:07
gtristanfranred, for instance, there are some really irresponsible upstream maintainers who have wget hardcoded into their makefiles09:08
gtristanand we end up using the tarball because it's just too hard to track09:08
gtristanmulti-source would allow us to factor those out09:08
* gtristan checks the links above09:08
franredgtristan, yep, I've seen that too :/09:08
gtristaneek, that chunk is disgusting09:09
tiagogomes__gtristan  multisources will be also useful to build GCC, as GCC requires access to the GMP, MPFR and MPC libraries09:11
franredgtristan, yep, that is what I mean... but how would you solve that?09:11
franredmilad, not sure if CTtpollard could help you09:12
franredmilad, in any case would be useful if you post the error you are seen09:12
miladfranred, have you worked with this api?09:13
franredmilad, no, I didn't09:13
miladfranred, you know someone who did?09:13
CTtpollardmilad: try asking in #automotive :)09:14
tiagogomes__gtristan, the submodule commit ID is stored in the parent tree, so there is not loss of reproducibility at the moment. However because we don't want to clone repos outside of the trove, currently we have to create a delta in the parent repo to replace the submodule repo URLs. We want a declarative way to do this09:15
*** bashrc has joined #baserock09:17
gtristantiagogomes__, I dont understand how this works09:17
* VLetrmx ← bad at explaining :(09:18
tiagogomes__gtristan what part exactly?09:18
gtristantiagogomes__, when I clone gedit and do 'git checkout SHA1'... how can this possibly effect the state of the libgd submodule ?09:18
gtristanwhich is a different git repository with a different history09:18
gtristantiagogomes__, I can imagine there *can* be some magic when you have a gedit checkout, and run 'git tag release-1.0'... that there is some internal git-foo recording the sha1 of the current state of the submodule09:20
VLetrmx08:59  VLetrmx$ gtristan, we don't need to change anything anyway, the submodule's sha is tracked when you add it as a submodule09:20
gtristanbut even that, I think is doubtful09:20
tiagogomes__gtristan if will affect the libgb submodule, because the submodule SHA1 is stored in the parent repo tree. You can read the submodule SHA1  by doing ` git ls-tree HEAD libgdpath`09:20
tiagogomes__gtristan git internally stores information about directories in a git tree object. The git tree object for the directory where the submodule is checked out, contains the submodule SHA109:22
*** CTtpollard has quit IRC09:35
gtristantiagogomes__, I've presently been asking about this in #git... and now have a bit better understanding09:35
gtristantiagogomes__, thanks for clarifying09:35
gtristanblack magick that it is09:36
gtristanessentially, what it means is that... if you do "git commit" in gedit, without the libgd submodule currently checked out, internally git will record a commit where "the libgd sha1 is the same as the last commit id... has not changed"09:37
gtristanand it also means, one can do a "git checkout <sha1>" from the submodule09:37
gtristanand then git add path/to/submodule09:37
*** CTtpollard has joined #baserock09:37
gtristanand then "git commit" in the parent09:37
gtristanthis should create a commit in the parent repo *even though nothing changed*, and it exists only to adjust the submodule's state in the current HEAD09:38
gtristanor new HEAD09:38
gtristantiagogomes__, so yeah, essentially this tracking just "works for free", however scary it seems to be09:39
*** ctbruce has quit IRC09:39
*** ctbruce has joined #baserock09:40
*** gtristan has quit IRC09:45
*** milad has quit IRC09:55
*** locallycompact has joined #baserock09:56
*** ssam2 has joined #baserock09:57
*** ChanServ sets mode: +v ssam209:57
*** jonathanmaw has joined #baserock10:01
*** ssam2 has quit IRC10:05
*** ssam2 has joined #baserock10:06
*** ChanServ sets mode: +v ssam210:06
*** edcragg has quit IRC10:09
*** edcragg has joined #baserock10:30
*** Lachlan1975 has joined #baserock10:42
*** gtristan has joined #baserock11:00
benbrown_SotK, franred: Thanks for the reviews!11:13
*** jonathanmaw_ has joined #baserock11:13
benbrown_I've just received a bunch of emails mentioning failed merges, do I need to do something on my end or will it sort itself out?11:13
tiagogomes__"Submitted, Merge Pending" again11:14
tiagogomes__benbrown_, you may have to bribe Pedro11:15
tiagogomes__or sacrifice a goat11:15
SotKI wonder if using Zuul for merging like OpenStack do will stop this kind of problem11:15
benbrown_tiagogomes__: Thanks, could you provide any additional info on the sacrifice, is there some special ritual involved?11:16
benbrown_Is Zuul the entity I am performing the sacrifice for?11:16
tiagogomes__wrong God11:17
*** jonathanmaw has quit IRC11:17
paulsherwooddoes lorry/lc handle shallow clones yet?11:21
persiaHaving Zuul immediately merge ready branches should reduce the rebases needed as compared to reviewers remembering to press the "merge" button.11:28
ssam2paulsherwood: not that i know of11:28
persiaZuul v3 is nearly ready, with some useful new features for this sort of thing as well.11:29
*** jonathanmaw_ has quit IRC11:31
pedroalvarezHello, God here11:32
pedroalvarezI've received the Goat11:32
pedroalvarezit was tasty11:32
Zarathis was an odd thing to tab into11:32
pedroalvarezI've noticed also that the first merge needed a rebase, breaking the dependencies of the other patches11:33
pedroalvarezSo, I'll use my magic to rebase all of them, and merge them if that's ok11:34
franredpedroalvarez, please, use your magic11:34
* pedroalvarez goes to convert water into wine11:36
rjekI did not realise pedroalvarez was the second coming, but on reflection it explains much.11:37
persiaHis delegation is limited, but within his sphere of authority, his powers admit no equal.11:38
franredpedroalvarez, are you preparing some kind of Dionysus's party?11:38
*** tiagogomes_ has joined #baserock11:39
*** tiagogomes__ has quit IRC11:39
benbrown_pedroalvarez: Thank you, glad you appreciated the sacrifice.11:44
*** tiagogomes_ has quit IRC11:47
*** ssam2 has quit IRC11:55
gtristanhave a few more patches up... but can I get some +1 for this one specifically: ?12:04
gtristanthe g.b.o version has an (irrelevant) delta applied and just need to know what the SHA1 is going to end up being12:05
*** ssam2 has joined #baserock12:09
*** ChanServ sets mode: +v ssam212:09
pedroalvarezouch, just found that libvpx is broken in armv7l12:31
radiofreemight be one of those ones you have to specific CROSS_COMPILE info for, even if the CROSS_COMPILE ends up being /usr/bin/gcc12:38
pedroalvarezor maybe is just because `cat -n` is not supported?12:39
pedroalvarezalthought that link looks like the answer12:40
radiofree"Unable to invoke compiler: arm-none-linux-gnueabi-gcc  -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64"12:41
*** ssam2 has quit IRC12:41
*** edcragg has quit IRC12:43
radiofreepedroalvarez: and
radiofreethose two patches at least get it past configure for me12:46
radiofreemaybe a nicer way of doing it12:46
*** ssam2 has joined #baserock12:58
*** ChanServ sets mode: +v ssam212:58
gtristanah, thanks jjardon12:59
*** jonathanmaw has joined #baserock13:12
gtristanoh crap !13:22
* gtristan saw webkit completed building... so interrupted the build to complete the rest after moving13:22
gtristanbut... artifact creation was not complete !13:22
* gtristan palmface13:22
*** gtristan has quit IRC13:43
*** tiagogomes_ has joined #baserock13:53
*** gtristan has joined #baserock13:54
*** tiagogomes_ has quit IRC14:03
*** tiagogomes_ has joined #baserock14:04
*** edcragg has joined #baserock14:07
jjardonhi, im getting this error in stage2-fhs-dirs when running ybd: any idea what is going on?14:41
SotKiirc `cache.get_cache()` can return a boolean when the thing it attempts to get isn't in the cache14:42
paulsherwoodjjardon: which version of ybd please?14:43
jjardonpaulsherwood: 203edcbba7d2f263b407c833bf86de86f550cb4114:44
jjardonso 15.45 tag14:45
paulsherwoodhmmm. has this worked for you before?14:45
paulsherwoodso new setup... first time, this error?14:46
paulsherwoodand nothing 'interesting' in your definitions?14:47
jjardonah, wait; I have errors before:
jjardonpaulsherwood: yeah, first setup, current definitions14:48
paulsherwoodare you root?14:48
jjardonpaulsherwood: is ybd meant to run as root?14:48
paulsherwoodfraid so14:48
jjardonah, ok. let me try again14:48
paulsherwoodi don't know what the solution is to get to non-root, but it would be lovely if there were one14:50
paulsherwoodjjardon: to speed up your user experience you may want to use the artifact cache...
jjardonpaulsherwood: yeah thanks, I'm using it14:53
paulsherwoodah, so you are :)14:54
paulsherwoodlet me know how it goes :)14:54
* jjardon is using the x86 morph cache, arm7hf morph cache and arm7hf ybd cache at the moment :)15:09
jjardonpaulsherwood: seems is working now running as root, thanks!15:10
paulsherwoodgreat :)15:11
jjardonbtw, vc4 (rPi GPU driver) is almost there:
pedroalvarezjjardon: ta! I've been following his work for weeks, but stop doing it15:35
jjardonpedroalvarez: this looks useful then:
*** franred has quit IRC16:17
*** Lachlan1975 has quit IRC16:31
*** Lachlan1975 has joined #baserock16:33
*** ssam2 has quit IRC16:44
*** ctbruce has quit IRC16:53
*** edcragg has quit IRC17:02
*** rdale has quit IRC18:03
*** jonathanmaw has quit IRC18:05
*** toscalix has quit IRC18:14
* SotK clears enough space to upgrade to the GNOME system18:29
*** Lachlan1975 has quit IRC18:40
gtristanGuys, it's late, I thought I was coming home to a built webkit and trying out gnome calendar, but...18:52
gtristanFirst, they moved over to github18:53
gtristanThen, the new devs are amongst those who want to waste everyones time converting stuff to cmake18:53
gtristanand finally this (I commented on that link)18:53
gtristanfor some reason lorry wont eat up the new tarball properly, either18:54
gtristanon the other hand... besides being really annoyed with ical atm... I am excited that somebody else on the planet will actually boot a baserock GNOME image other than me !18:55
gtristanSotK, let me know how it runs :D18:55
gtristanpedroalvarez, yes... I will take any chance I can get to rant against the cmake zealots18:55
* SotK will do, if it ever boots :)18:58
gtristanheh, it should !18:58
gtristanI sure hope it does18:58
SotKok, its booted but gdm didn't start18:59
SotKand I can't log in with my pre-upgrade password18:59
gtristanwell, this is some kind of btrfs rollover atomic overwrite everything upgrade thing ?19:00
gtristanis a password expected to survive that ?19:00
* gtristan has never tested an upgrade of the GNOME system19:01
SotKit normally does19:02
SotKis there some default password I could expect it to have been set to?19:02
gtristanthe root is passwordless19:03
gtristanby default19:03
* gtristan notices it's now 4am, while it was certainly only 3am just moments ago :-/19:04
paulsherwoodsleeeepppp, gtristan19:06
gtristanI will go to sleep :)19:06
* SotK sees "Failed to start Authorization Manager" when trying to boot19:48
SotKbut can't read the journal because logging in isn't working :(19:48
SotKI have a /etc/passwd.rej in my gnome system, I guess that is a sign of something going wrong?20:18
SotKalso a /etc/group.rej20:19
* SotK fixes the merge conflicts and boots into gdm20:25
SotKI can't log into the mysterious "Linux User" though, nor can I log in as root with my old password20:27
persiaIt may be that the merge, even though fixed, still has issues.20:36
persiaTry mounting the filesystem from something else, chrooting, and setting a password.20:36
* SotK tries it20:40
SotKdidn't seem to work :(20:47
persiaOdd.  I'd expect using the passwd from the target filesystem to reset the password in the target filesystem would create a shadow that contained the right magic to cause the login from the target filesystem to be able to authenticate.  Hrm.20:54
*** jjardon has quit IRC23:23
*** jjardon has joined #baserock23:26

Generated by 2.15.3 by Marius Gedminas - find it at!