IRC logs for #baserock for Thursday, 2016-10-13

*** gtristan has quit IRC04:42
*** toscalix has joined #baserock05:25
*** gtristan has joined #baserock05:31
*** toscalix has quit IRC06:20
*** ctbruce has joined #baserock07:32
*** ChanServ has quit IRC09:14
*** SotK has quit IRC09:17
*** inara` has quit IRC09:17
*** rjek has quit IRC09:17
*** franred has quit IRC09:17
*** dabukalam has quit IRC09:17
*** benbrown_ has quit IRC09:17
*** cosm has quit IRC09:17
*** vgrade has quit IRC09:17
*** juergbi has quit IRC09:17
*** malinus has quit IRC09:17
*** perryl has quit IRC09:17
*** bfletcher has quit IRC09:18
*** ctbruce has quit IRC09:18
*** cyndis has quit IRC09:18
*** _longines has quit IRC09:18
*** persia has quit IRC09:18
*** JPohlman1 has quit IRC09:18
*** persia_ has quit IRC09:18
*** tiagogomes_ has quit IRC09:18
*** anahuelamo has quit IRC09:18
*** jmacs has quit IRC09:18
*** gtristan has quit IRC09:18
*** leeming has quit IRC09:18
*** paulsher1ood has quit IRC09:19
*** bjdooks has quit IRC09:19
*** gary_perkins has quit IRC09:19
*** tardyp has quit IRC09:19
*** radiofree has quit IRC09:19
*** jmacs has joined #baserock10:28
*** anahuelamo has joined #baserock10:28
*** paulsher1ood has joined #baserock10:28
*** leeming has joined #baserock10:28
*** inara has joined #baserock10:28
*** locallycompact has joined #baserock10:28
*** persia_ has joined #baserock10:28
*** radiofree has joined #baserock10:28
*** tardyp has joined #baserock10:28
*** bjdooks has joined #baserock10:28
*** gary_perkins has joined #baserock10:28
*** cyndis has joined #baserock10:28
*** _longines has joined #baserock10:28
*** persia has joined #baserock10:28
*** JPohlman1 has joined #baserock10:28
*** perryl has joined #baserock10:28
*** bfletcher has joined #baserock10:28
*** dabukalam has joined #baserock10:28
*** benbrown_ has joined #baserock10:28
*** cosm has joined #baserock10:28
*** vgrade has joined #baserock10:28
*** juergbi has joined #baserock10:28
*** malinus has joined #baserock10:28
*** rjek has joined #baserock10:28
*** SotK has joined #baserock10:28
*** franred has joined #baserock10:28
*** ctbruce has joined #baserock10:28
*** tiagogomes has joined #baserock10:28
*** ChanServ has joined #baserock10:30
*** sets mode: +o ChanServ10:30
tiagogomesybd is unusable on my system unless I do some hackery. Because it requires running as root, it working directory will be /root/ybd, and the partition where '/' is doesn't have much allocated space comparing with /home.10:31
tiagogomesIs there a way to set the working dir? Maybe setting XDG_CACHE_HOME10:32
pedroalvareztiagogomes: I believe that in the readme is explained how to change that10:35
leeming^ for direct help. make a file "ybd.conf" with the following fields10:36
leemingartifacts, tmp, ccache_dir, gits, jobs (all yaml format)10:37
leemingpointing to the associated directories in your home dir10:37
leemingtiagogomes, ^^10:37
tiagogomesleeming thanks10:39
leemingbut once i get this very messy rebase working, ybd may not require root10:41
tiagogomeshow are you handling extracting artifacts the contain device nodes?10:42
leemingdepends on what you mean by extract?10:45
leemingit should be doing whatever ybd usually does, which is directory based i believe10:45
leeming$SANDBOX/dev , is mounted in bwrap as a special device mount10:45
tiagogomesThe fhs-dirs artifact is a gzip archive that contains devices nodes. To extract this archive, you need root privileges.10:49
tiagogomesSo or you get rid of the device nodes on that artifact, or you have a small wrapper executable with the setuid bit set that is called by ybd and does the extraction.10:50
leemingwould ybd successfully build if it couldnt extract it?10:50
tiagogomesWith current master, it would fail.10:51
leemingthen it does something I do not understand, as it works10:52
tiagogomesPerhaps coz you are running it as root or there isn't a prebuilt artifact for fhs-dirs in the cache.10:54
*** gtristan has joined #baserock11:00
leeminghmm, i will nuke all the caches and try it again then.. and i wasn't running as root11:03
jjardonlocallycompact: paulsher1ood ok to merge this?
paulsher1oodjjardon: yup, i've clicked the button11:23
paulsher1oodwould folks consider s/assemblage/stack/ ?11:46
paulsher1oodlocallycompact: ^^11:46
locallycompactNot for my purposes but you're free to do so for the current model.11:49
locallycompactIt actually might make a lot of sense11:49
paulsher1oodi'm just thinking of an easy/memorable name, not too overloaded. it could still stand for everything you want assemblage to be11:50
gtristanWhile I personally find the abbreviation of assemblage more amusing than troublesome, I also find the word to be a bit complicated to digest. I'd also love to take the opportunity to kill off the words "strata"/"stratum"11:51
locallycompactI still like stratum ontologically11:52
locallycompactAlthough, one could get away with substack11:52
gtristansince I think with V10 a stratum only denotes an assemblage that is a part of another assemblage (from the mail), instead of what it actually is, right ?11:52
locallycompactit's an important logical note that resides in something "as something", and doesn't simply exist11:53
locallycompacts/that/that it/11:55
* paulsher1ood proposes 'stack'ain11:55
jmacsStack implies an ordering; there are things at the bottom and top; is that the intention?11:55
paulsher1oodnot my intention, no11:56
jmacsAssemblage doesn't have that meaning, to me11:56
gtristanI understand it as a layer within a system (the system which is more of a stack), unfortunately, they already took the word layer11:57
paulsher1oodtier. but i think it's not correct anyway because assemblage can be a set of layers11:58
locallycompactjmacs, it is however how things are arranged currently, there are things at the bottom and things at the top, because we denote build-depends by name and they form a strict partial ordering11:58
paulsher1ood(as in a deck of cards)11:58
locallycompactwhat I want for assemblages is to infer build-depends by capacity, which would deduce into a stack11:58
leemingare we discussing renaming assemblages?11:59
leemingto use a generic overloaded word?12:00
locallycompactI'm not 'renaming' assemblages because assemblages are totes not finished12:00
locallycompactbut I do think how it appears in V10 as stack-like12:01
* paulsher1ood is proposing to call the thing locallycompact is callig 'assemblages' to 'stack', before the name sticks12:01
gtristanI think that we need to really understand what the big picture is, otherwise when we hit V15 and assemblages mean something else, we'll want to rename it again :-/12:02
paulsher1oodi don't think that's true12:02
paulsher1oodbut then, i think i understand the big picture to some extent12:03
locallycompactassemblages "really mean" what they mean in assemblage theory12:03
jjardonpaulsher1ood: ok to merge ?12:03
locallycompactif it diverges from that I would call it something else12:03
tiagogomesI agree with gtristan. A roadmap would be nice12:03
gtristanI'm not clear on the concrete plan on how one will infer build depends by capacity, I'm not sure how "this layer provides the capacity to be a display server" for instance will be useful concretely12:04
tiagogomesOtherwise the ideas suggested in he mailing list will become a bit forgotten12:04
paulsher1oodtiagogomes: what precisely would you like to see in a roadmap?12:05
paulsher1ood(and same question to others)12:05
locallycompactroadmap: fix the ontology12:06
tiagogomesA list of work items that will lead to a future shape of definitions that addresses the major problems with the current definitions.12:07
gtristanA description of how dependencies will be resolved, once the assemblages idea comes to full fruition. Otherwise: V10 features stand on their own, we name them in the way we feel makes sense with the present V10 features, and anything else is left to future schemes12:07
locallycompactalso have something that allows to visualize the ontology with big descriptions of everything in a web view that makes no reference to syntax12:08
locallycompactlike this but better
jmacs<bikeshedding> Force directed layouts are rubbish </bikeshedding>12:10
gtristanOn topic (I think): some of us agree we need a simple set of words to describe the (hopefully *few*) data types which definitions describe, in the interest of being user friendly. Personally I think accuracy of the word is of less importance than simplicity and recognizability.12:16
gtristanAnalogies are also great. Dog *could* make sense if definitions were Zoo and YBD was ZooKeeper (although probably not Dog, perhaps Cage, where Chunk would be Animal)12:17
gtristanalthough the Zoo theme doesnt seem to be a great fit.12:18
radiofreepaulsher1ood: any chance you could approve this
radiofree(have tested it)12:19
locallycompacttechnical definitions please, metaphors can happen in the future after we are not confused12:19
gtristanI dont feel "stack" to be the best choice, but I would place my vote on "stack", since we do have say... a graphics stack and possibly other stacks which make up a final OS12:20
gtristanI would place my vote on stack in the absence of anything better - and I would hate to rename things in the future, thats a compatibility issue12:21
gtristanwe cant be on a mission to constantly be changing the format :-S12:21
locallycompactI'm happy with stack so long as it only ever means a strict partial order12:23
gtristanIn absence of a clear roadmap of where this is going, I dont think we can guarantee that it will always mean what you want it to. Instead: V10 brings new features which solve real problems12:27
gtristanAnd I think that's the right way to proceed, introducing one change at a time and giving pause before the next.12:28
*** edcragg_ has joined #baserock12:29
locallycompactI want to make assemblages the testbed for capactive dependencies. In that light the cache key logic for assemblages can always be framed as deducing a stack from an assemblage.12:29
locallycompactf: Assemblage -> Stack12:29
locallycompactand I want the mechanism for assemblage capacities to be coeffects12:30
locallycompactfor that to work stacks can only ever be strict partial orders12:30
pedroalvarezstack reminds me to the "full stack" engineers job offers12:33
* SotK looks forward to "add the openstack stack"12:35
gtristanWhat worries me is this: Definitions can not be a testbed for anything, what we have works to accomplish a given task, we can improve on it; and testbed outside of it.12:37
paulsher1oodno-one likes deck over stack?12:37
locallycompactdefinitely not12:37
pedroalvarezI actually don't dislike it12:37
pedroalvarezfor me deck doesn't say much12:38
*** CTtpollard has quit IRC12:39
gtristanWhat describes a group of components with some tight relationship to eachother: Thats what the thing is I think. Its a convenient grouping of components which allows one to disregard to some extent it's internal details.12:39
*** CTtpollard has joined #baserock12:39
SotKgtristan: "group" :)12:40
* SotK runs away12:40
gtristanso what describes that, a stack sort of makes sense because it's software. a group is too ultimately generic, right12:40
paulsher1ood'deck doesn't say much' may be considered an advantage :12:40
* locallycompact wonders if anybody has ever actually googled 'group' and read the definition of a group12:41
* paulsher1ood has12:41
pedroalvareznotes make chords, and then you can make songs12:41
pedroalvarezand the build tool is the band!12:42
pedroalvarezbase rock & roll12:42
SotKpedroalvarez: :D12:42
pedroalvarez(random idea, maybe can be elaborated by someone)12:42
gtristanpedroalvarez, that is already awesome12:43
* SotK knows what a group is also, but I don't see why we have to abide by the mathematical definition of a word here12:43
locallycompactbecause it's ubiquitous in every field of science, that's wy12:43
paulsher1oodpedroalvarez: lol12:43
SotKwhen I hear group I first think of the layman definition, i.e. "look at that group of people"12:44
gtristanI do prefer analogies naming, but nobody cares about strata and morphologies, that was a weird choice12:44
* SotK likes the geology metaphors, but knows he is alone :(12:45
locallycompactThe actual "thing" that this is a series-parallel partial order, it's what I modelled V10 on12:45
gtristanWe have: chunk | strata | system | cluster... system is the best name we have, cluster implies that as a superset of systems one might be deploying a cluster of systems, perhaps a render farm12:49
gtristanchunk is.. alright.. strata well.. I dont know12:50
gtristansubsystem ?12:50
pedroalvarez"system" has a lot of meanings too12:50
gtristanTo be completely boring and bland ?12:50
SotKsubsystem is already a part of clusters12:51
* gtristan facerock12:51
gtristanI think we should just make pedroalvarez name everything12:52
gtristanHe has the best imagination so far12:52
* SotK agrees12:52
pedroalvarezwe might get away by just creating random names12:52
gtristanand the result will be better than everyone having a say :)12:52
*** ChrisPolin has joined #baserock13:03
tiagogomestbh, I would just leave the names as they are.13:14
pedroalvarezI was thinking that I like the word "tree"13:15
pedroalvarezto define what a system currently is13:15
tiagogomesostree is taken13:16
pedroalvarezi said just tree13:17
paulsher1oodwe'd confuse this with git trees, i fear13:18
locallycompactAnd also it isn't a tree, trees only capture one relationship, this has two - containment and dependency13:18
pedroalvarezroots and fruits13:19
SotKmaybe we should extend the geology metaphor and call them "formations"13:19
leemingfruit n nuts?13:20
pedroalvarezlocallycompact: anyway, true. I just meant what a built system is. A tree.13:20
leeminga hedge.. that is a flat tree right?13:21
jjardonpaulsher1ood: can we have this in?
pedroalvarezjjardon: quick question to understand gitlab better13:29
pedroalvarezin the discussion tab of that MR, there are like 4 different blocks13:30
pedroalvarezare they different versions of the MR?13:30
jjardonpedroalvarez: yes13:34
* SotK wonders if there is a way to see a diff between versions that he can't find13:35
jjardonyou can compare differences between versions here: similar as you can in gerrit13:35
SotKjjardon: lovely, thanks13:36
pedroalvareznot bad13:37
* tiagogomes finds that gitlab UI is pleasant to use13:38
pedroalvarezbit annoying that the diff between versions include all the commits13:40
SotKhmm, those diffs don't seem right actually13:40
SotKI see the same basic changes in all the options13:41
tiagogomesmaybe the new versions were just a rebase on master?13:42
SotK(eg I see the change in in the diff between v4 and v3, and also in the diff between v3 and v2)13:42
leeminggitlabs UI is a little minimal at times13:50
* SotK notices the diff content wraps if the columns are too narrow in side-by-side mode :(13:52
*** gtristan has quit IRC14:05
*** gtristan has joined #baserock14:11
*** franred has quit IRC14:14
*** franred has joined #baserock14:27
radiofreehi, i'm getting an error building (and occasionally cloning) ansible14:48
radiofreefirst time round cloning the repo seemed to fial14:49
radiofree/sbin/ldconfig.real: /usr/lib/ is not a symbolic link14:50
pedroalvarezradiofree: jjardon hit the same problem this week14:50
pedroalvarezradiofree: are these problesm in gitlab CI or also locally?14:51
radiofreewell, it's running locally on a machine here14:51
radiofreevia gitlab14:51
radiofreein a docker container if that makes a difference?14:53
jjardonpedroalvarez: the strange thing is that is always the ansible repo the one that fails to build14:53
pedroalvarezcan you giveme a timestamp of a recent clone failure?14:55
radiofree~22minutes ago14:57
radiofreecan't be more exact sorry14:57
pedroalvarezlet'ssee if I can see something weird in g.b.o.14:57
pedroalvarezradiofree: I can't see anything related15:06
pedroalvarezeither network problems, either the build tool15:06
locallycompactcan we bring gcc 5.4 into gcc-tarball?15:15
pedroalvareznot 6?15:17
locallycompactI thought there are problems with 615:17
pedroalvarezthere are, but all problems can be fixed? :P15:17
locallycompactI am hitting this bug bootstrapping from 6 on arch into 5.3 in baserock
pedroalvarezlocallycompact: send a patch to the lorry file to upgrade15:18
SotKspeaking of lorries, can we upgrade g.b.o to understand yaml lorries soon? :)15:19
pedroalvarezwhy are you looking at me?15:19
* SotK is happy to do it, but doesn't know if he has the permissions to do so, or if there is some process around upgrading the infra these days15:20
pedroalvarezSotK: is it just an upgrade to lorry? or also to lorry controller?15:21
SotKlorry and lorry-controller15:21
pedroalvarezright.. then I think I could try to upgrade test that it works in there, and then upgrade g.b.o.15:22
SotKthat seems sensible, sorry for creating work15:36
leeminglong overdue (sorry) but here is an implementation of bwrap in sandboxlib ->
tiagogomesno gitlab PR instead?15:43
pedroalvarezhm... pbr fails to build?15:44
pedroalvarez:) nope, g.b.o can't be upgraded15:47
SotKwhats up with pbr?15:49
pedroalvarezpython3-core pbr15:49
pedroalvarezmay be trivial to fix,
SotKit worked at the weekend when I built a trove15:52
SotKI wonder if the python3 update broke it15:52
* SotK will look later15:52
SotKthough that would be a weird way for it to break...15:54
paulsher1oodjjardon: do you think dropping python2 is acceptable? will any users be affected by this ?15:55
pedroalvarezSotK: upgraded to latest, and it works15:56
jjardonpaulsher1ood: yes, I think so15:56
pedroalvarezSotK: maybe a built-in library is not included anymore?15:56
* paulsher1ood would prefer that ybd continues to be runnable with python2... but assumes this is too much to hope for16:01
pedroalvarezhow do I trigger gitlab ci? by sending a MR?16:05
pedroalvarezhow do I do that also?16:05
pedroalvarezoh, just found the fork button16:05
*** franred has quit IRC16:10
*** CTtpollard has quit IRC16:14
*** ctbruce has quit IRC16:20
pedroalvarezeverything fails:
tiagogomesAt least they are consistently failing16:23
jjardon_matrixpedroalvarez fork from current master16:24
pedroalvarezhm.. gitlab's or g.b.o ?16:26
pedroalvarez(not sure if gitlabs carries something)16:26
jjardon_matrixGitlab is a mirror16:26
jjardon_matrixExactly the same on both16:27
jjardon_matrixSo simply rebase your branch and force push it16:27
pedroalvarezgreat, no need to MR to trigger ci16:30
pedroalvareznicer than github+travis16:31
pedroalvareznot sure if it's my network, but gitlab is slow to navigate16:33
leemingno, you just need a .gitlab-ci file or somethjing16:33
leeminggitlab is slow16:33
paulsher1oodwe're working on that, i believe16:34
pedroalvarezto improve the speed of the web ui?16:34
tiagogomesyes, my main complain about gitlab so far16:34
paulsher1oodoh, if you mean the webui, we'd need to self host16:35
leemingtheir web ui can be slow at times, as well as the unknown wait queue for runners16:35
* paulsher1ood thought you meant runners16:35
jmacsI've also found gitlab's site rather slow, on a different project16:36
* SotK has always found it slow also16:36
tiagogomes"We'll be deploying GitLab 8.13.0-rc2 shortly". rc?17:06
edcragg_rc2 at that17:10
*** tiagogomes has quit IRC17:16
*** gtristan has quit IRC20:37

Generated by 2.14.0 by Marius Gedminas - find it at!