IRC logs for #baserock for Monday, 2014-11-03

*** genii [~quassel@ubuntu/member/genii] has joined #baserock01:29
*** radiofree [radiofree@unaffiliated/radiofree] has quit [Ping timeout: 260 seconds]02:02
*** radiofree [radiofree@unaffiliated/radiofree] has joined #baserock02:06
*** radiofree [radiofree@unaffiliated/radiofree] has quit [Ping timeout: 256 seconds]02:39
*** radiofree [radiofree@unaffiliated/radiofree] has joined #baserock02:42
*** genii [~quassel@ubuntu/member/genii] has quit [Read error: Connection reset by peer]06:18
*** abdul_ [~abdul@] has quit [Quit: Leaving]06:19
*** madhu [~madhu@] has joined #baserock06:35
*** petefoth [] has quit [Read error: No route to host]07:21
*** flatmush [] has quit [Read error: Connection reset by peer]07:21
*** violeta_ [] has quit [Read error: Connection reset by peer]07:21
*** flatmush [] has joined #baserock07:22
*** violeta_ [] has joined #baserock07:22
*** petefoth [] has joined #baserock07:22
*** dutch [] has joined #baserock08:00
*** franred [] has joined #baserock08:12
*** fay_ [] has joined #baserock08:12
persianomenclature is hard: while I agree with the desire for the name->chunk mapping to be unique, and to never have a chunk name used as a stratum name, I don't think it possible to determine the correct values in advance.  This is something where appropriate naming should emerge as conflicts develop.08:55
dutch is now known as wdutch08:55
*** abdul [~abdul@] has joined #baserock08:55
paulsherwoodfair enough. but we could settle on a general approach for 'foo and the stuff foo typicall requires'?08:56
paulsherwoodi don't think calling it foo is correct08:56
persiaI'd rather restructure things to allow embedding and dependencies, so "foo" includes the chunk definition directly, and only requires subsidiary files in the special case where the subsidiary is a shared depedency for different things.08:57
persiaOtherwise we end up with N sets of "foo-stratum", an "foo-chunk" for values of foo, stratum, and chunk.08:58
persiaMy understanding of the advantage of the multicomponent specification was to ease system design, so I don't want to have to think more than "I want foo", and get everything (and if I want to know what "foo" means, I want that it one place, excepting when DRY concerns apply).08:59
*** mariaderidder [] has joined #baserock09:05
*** tiagogomes [] has joined #baserock09:07
*** abdul__ [uid49809@gateway/web/] has quit [Quit: Connection closed for inactivity]09:09
*** franred [] has quit [Remote host closed the connection]09:09
*** sherm_ [] has joined #baserock09:11
*** tiagogomes [] has quit [Ping timeout: 255 seconds]09:23
*** tiagogomes [] has joined #baserock09:26
*** sambishop [] has joined #baserock09:28
*** tiagogomes [] has quit [Ping timeout: 265 seconds]09:31
*** aananth [~caananth@] has joined #baserock09:31
*** franred [] has joined #baserock09:32
*** jonathanmaw [] has joined #baserock09:34
petefothAre we09:59
* petefoth starts again09:59
straycatwe are09:59
petefothHave we decoded to deprecate the geological terminology? I’m updating the project Glossary and have reached the word ‘morphology’. I was thinking of changing its entry to deprecated - see *definition*10:02
KinnisonI imagine we need a consensus-like decision on the ML before we deprecate terminology10:03
KinnisonIn the pub last night, richard_maw was still using 'stratum', 'morphology' etc.10:03
petefothI wasn’t planning to chamge ‘stratum’ or ‘chunk’ (at least not yet). But in most of the conversations / emails I notice, we seem to use ‘definition’ more than ‘morphology’10:06
*** ssam2 [] has joined #baserock10:09
Mode #baserock +v ssam2 by ChanServ10:09
ssam2anyone know whether build system autodetection works even if build-commands or something is specified in a chunk morph ?10:12
ssam2I think it only works if there's no chunk morph, right?10:12
petefothFor now I wont use the word ‘deprecated’ - I will make the entry ‘See *Definition*’10:12
ssam2I think 'morphology' could be useful as a technical name for the language/format of the definitions10:13
franredssam2, I think if you don't specify the morph field on the stratum it will try to autodetect the build system10:15
ssam2franred: yes, but what if you do specify the morph field ?10:16
ssam2I think suddenly it stops working. I hope so, because then it explains the problem that someone has mailed me about :)10:16
ssam2by 'stops working' I mean, assumes 'build-system: manual'10:16
KinnisonAutodetection only happens in the absence of a morphology file yes10:17
franredI though that you should specify the build-system if you are using the morph file10:17
petefothssam2: as in ‘the way in which a system or other artifact is defined in a *definition*’ - That does work I think *but* i know/believe that paulsherwood wants (wanted?) to do away with the geological terminology completely10:17
ssam2Kinnison, franred: I thought so10:17
ssam2we may need to communicate that more clearly in our documentation10:18
Kinnisonfranred: yes10:19
* Kinnison is all for requiring that the build system always be declared10:19
Kinnisonand removing autodetection entirely10:19
Kinnisonit'll help speed up the early stages pre-build10:19
ssam2I think we can speed that up without introducing regressions :)10:20
ssam2I may be wrong though10:20
franredKinnison, Im not sure about that, we will end up with loads of *chunk definitions* specifying only the build-system which is not ideal either10:21
franredpetefoth, defined and definition in the same sentence sounds at least weird10:21
Kinnisonfranred: I also think we should be able to embed chunk morphology content into the stratum morphology10:21
jmacsIt's a tautology.10:21
petefothfranred: true - I will rephrase10:22
Kinnisonssam2: My hope is to remove the need to look outside of the definitions repository at all, to know how to build a system10:22
franredKinnison, could work, or add the chunk information from the stsratum into the chunk and only add the chunk name to the stratum10:23
* Kinnison doesn't really want a proliferation of tiny files if we can get away without them10:24
franredbut they are more readable and manageable than have a master piece with all the info which is difficult to read or follow10:25
franredImagine have a stratum, with all chunks, pre-configured/configure/build/{pre,post}install commands, system integration scripts and build dependencies....10:28
richard_mawour current dependency model is lacking in some respects, but we could get to explicit global build and runtime depends (as opposed to the restricted local build-depends we currently have) with a few steps10:28
Kinnisonfranred: Oh those would clearly end up split out10:29
Kinnisonfranred: but most chunks don't need them10:29
richard_maw1.  get rid of stratum artifacts 2.  allow in-line chunks 3.  add stratum runtime-depends10:29
* richard_maw may write an rfc some time this week to describe it10:32
*** Krin [] has joined #baserock10:32
Krinmorning all. 10:33
Krindo we have or know any strata for baserock that have pre-populated a download from maven in their construction? or am i treading on completely new ground with this twisted insanity?10:34
KinnisonCompletely new, unless a non-contributing user has done it10:34
Krinalternativly is there a way of stopping a build part way through so that i can let the build get to the point where all the items have been downloaded from maven and upload said files to the git repo so that they are already pressent and thus cut out maven?10:35
ssam2Krin: there are hacky ways to do either of those things10:35
franredpetefoth, also "definition" is a pretty vague term - definition of what? what does it contain/define? change "morphology" for "definition" does not make the concept easy to understand and make it more undetermined, OMHO10:36
paulsherwoodKrin: would setting build-command: exit 1 help you?10:36
Krini'm also pondering making a script that runs the first time the system is turned on that sets up the zookeeper, so that the maven can be accessed10:36
ssam2Krin: if Maven functions as a packaging system, the best thing to do is probably find a way to import metadata from Maven using the baserock-import framework we've been working on10:37
ssam2I know nothing about Maven, and can't really provide assistance in a few lines of IRC10:37
Krinpaulsherwood, hmm, not sure, because i'd need to get half waty through the build first, and there isnt realy a break in the maven download and the installation of zookeeper, Unless i actualy work with the install script and make a customised version of that, i suppose that could work. hmmm10:37
ssam2KrinL but it might make sense for us to chat about it in person at some point10:37
ssam2paulsherwood: that would cause Morph to exit with failure :)10:38
richard_mawssam2: that generally assumes source for the packages is available, Maven works in a world of binary only packages10:38
ssam2richard_maw: hmm, OK10:38
ssam2Morph generally assumes the source for the packages is available, too10:38
petefothfranred: I take your point: in Baserock, and for the purposes of the Glossary a ‘definition’ is a .morph file in the ‘definitions’ repo. I think that the discussions over terminology are not completed yet (and I will add a note to that effect in the Glossary)10:38
ssam2so maybe this isn't something Morph is suited for at all?10:39
paulsherwoodssam2: i know. i use it to debug builds all the time. i thought it might help, is all10:40
Krinssam2, i'm starting to think that more and more, i can get the dependancies for zookeeper onto baserock with little trouble, but i'm not so sure about zookeeper itself as part of it's install process is to actualy download things from maven, which obviusly cant be done during the build10:40
Krinnot to my knowledge anyway XD10:40
richard_mawssam2: there's still value in tracing imported binaries, but rather than building, we'd have to import the binaries into repositories and install them into the maven cache somehow10:40
paulsherwoodKrin: i think you may have to override some part of its install process, then10:41
ssam2paulsherwood: actually that is a neat debugging tool :)10:41
paulsherwoodssam2: thank you :)10:41
ssam2Krin: does it download the same things each time?10:42
Krinssam2, yes yes it does. 10:42
ssam2OK. so not too hard to second-guess it a bit10:42
paulsherwoodkey point is it allows you to stop morph at some point in the process, and look around to see what the chunk commands have done, what's available etc10:42
ssam2I remember with the RubyGems work one of the ideas we looked at was importing .gem files directly from into the Morph build process10:42
ssam2and adding a RubyGems server to Trove for those who want to mirror them locally10:43
ssam2it seemed a much better idea to use the source directly10:43
KinnisonKrin: If I were you, I'd be looking to see if ant offers a target for 'download dependencies and stop'10:43
ssam2but given that we can't use the source directly for Maven, I guess revisiting those ideas would make sense10:43
KrinKinnison, i dont even know how i'd look for that, do you mean in the install script? 10:44
KinnisonSadly maven doesn't always provide source content10:44
KinnisonKrin: see if ant has a way to list its targets10:44
KinnisonKrin: e.g. gradle has a 'tasks' command10:44
persiaPlease be careful about how much tools are modified or worked around.  If possible, it's better to discuss strategies with upstream.  The goal is to make sure folk familiar with a given tool can be comfortable with using that tool with the Baserock build system.10:44
Krinalso, oh so thats what maven is! *first google hit = Mars Atmosphere and Volatile EvolutioN*10:44
persiaLack of source is a common case in Java: the idea being to compile once, run everywhere.10:45
Kinnisonpersia: ant, as a direct build tool, seems to be being deprecated in favour of gradle10:45
persiaI've read that, but I'm not sure how widely this is accepted10:45
* persia hasn't been to JavaOne in decades10:45
Kinnisonpersia: all I know is that all the java things I see right now seem to be at, or moving toward, gradle10:48
Kinnisonpersia: I imagine larger projects with investment in maven or ant will not move, or won't move for a while10:48
KinnisonInterestingly, I think gradle can import ant projects10:48
jjardon+1 for inline chunks; almost al of them are tiny and makes move chunks between stratum more difficult (you have to change 3 different files)10:49
persiaI thought everyone liked the *idea* of inline chunks, but it required massive changes to the parser and graph generators.10:50
persiaThere's been some work on this, but I haven't seen anyone show something that worked entirely for that yet.10:51
*** sherm_ [] has quit [Quit: Leaving]11:11
paulsherwoodpersia: i have something that mostly works for that. it just doesn't do anything else yet :)11:14
paulsherwood(ie can parse and build-walk for inline chunks, non-inline chunks, nested content etc, no need for 'stratum' specialness)11:15
jjardonah, ok: didnt know the details: I think the conditionals work is more important though11:17
*** wdutch [] has quit [Quit: Quit]11:37
*** dutch [] has joined #baserock11:42
dutch is now known as wdutch11:42
Krinhmm, it seems that the download part of maven actualy all just goes into a nice neat little folder that other parts of the installation refferance, would it be possible to put said folder into a git repo and have the build, instead of calling 'ant compile_jute' call 'git clone jute_build' ? this seems like far too easy a work around so it's probebly not going to work.11:46
ssam2Krin: it probably makes sense for you to prototype a couple of these approaches and see11:46
ssam2will help for learning too11:46
Krini'm intending to prototype this, i just dont know where to upload the folder to git.11:47
ssam2there's a wide variety of Git servers you can use11:47
ssam2one of the free git hosting services, like github, might be sufficient while prototyping11:47
ssam2or you could deploy your own Trove, how to do that is documented on the wiki11:48
Krinis there a reccomended fo this comapny? a known one to use or something along those lines?11:48
ssam2Krin: Trove ?11:48
ssam2that's what makes sense for the Baserock project11:48
Krinok i'll look into making my own trove, after lunch though BF wants to meet up in town for noms. so. i shal look after that.11:49
ssam2an OpenStack deployment is the simplest, if you have an OpenStack cloud available11:49
ssam2it's simplest if it has the generic Trove image already there, anyway :)11:50
*** mike [] has joined #baserock11:52
mike is now known as Guest2738911:53
persiapaulsherwood: Indeed.  Yours is one of the things to which I refer :)  But until it actually does builds and populates caches, it's hard to use as a replacement :)11:54
paulsherwoodpersia: agreed. will see what i can do :-)11:55
jmacsWhat's the smallest amount of RAM we can reasonably run a baserock system in?12:07
straycatThe smallest system we have is just a busybox system, so our minimum requirements should be similar to busybox.12:09
straycatI am not awake though12:10
paulsherwoodjmacs: i'm assuming you mean a baserock *Linux* system? it would be possible to use baserock to build a simple RTOS or bare-metal scheduler system for example12:13
jmacsYes, linux12:14
paulsherwoodi think straycat is correct, then. but rjek and others may be able to comment on how small a linux system can be these days12:16
robtaylorKinnison: looks like ssam2 has already looked a bit at scratchbox-in-baserock, might be worth chatting with him as well
paulsherwoodi thought scratchbox was deceased? an ex-project? :)12:18
robtaylorpaulsherwood: Mer (and Jolla) still use it12:19
* petefoth flashes back to his *vetry* early days at Codethink :)12:21
robtaylor(nb scratchbox2, not scratchbox)12:27
paulsherwoodoh, cool12:34
pedroalvarezssam2: yeah! the same patch that you backported to gcc 4.6 to fix big endian, works in gcc 4.7.3 :)12:40
*** wdutch [] has quit [Quit: Quit]12:41
*** thecorconian [] has joined #baserock12:41
*** dutch [] has joined #baserock12:42
*** dutch [] has quit [Remote host closed the connection]12:42
*** thecorconian [] has quit []12:45
*** zoli_ [] has joined #baserock12:49
*** zoli_ [] has quit [Changing host]12:49
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock12:49
robtaylorKinnison: lots of good useful stuff here btw
franredpedroalvarez, \o/12:57
*** jonathanmaw [] has quit [Ping timeout: 264 seconds]13:03
petefothThe old “Baserock System Glossary” mentions ‘the `bootstrap-baserock` script in the Morph source code.’ No sigh of this script in the mproh repo. Has it been superceded / removed?13:19
richard_mawyep, it went the way of the dodo a while back13:19
richard_mawit was before we could bootstrap natively in morph13:19
petefothrichard_maw: ta!13:19
KinnisonIt looks like even if SB or SB2 are in active use, they're not under active development13:27
Kinnisonand considering SB2 has a set of known issues, that is "not good"(tm)13:27
persiaIt's more the lack of anything else folk are happy about to replace them, rather than anything else.13:27
KinnisonI see13:28
*** jonathanmaw [] has joined #baserock13:31
KinnisonSB2 does look significantly more interesting than S B13:32
* Kinnison pokes at it13:32
richard_mawthere's a big section in about how difficult it was for them to handle targets that are of a compatible architecture to the host…13:39
* richard_maw wonders if this is just because they wanted to allow it to not need root13:39
Kinnisonthere's a lot of interesting LD_PRELOAD going on in SB213:39
richard_mawsince presumably you ought to be able to just chroot13:39
* Kinnison was shocked to discover an embedded lua interpreter13:39
richard_mawlots of glibc patches to the linker13:41
madhussam2, I face segfault when I build gdbm package when I "morph build systems/devel-system-x86_64-generic.morph  --verbose"13:43
paulsherwoodmadhu: i assume you're running that on a baserock x86_64 system?13:44
madhuit is same as  but I did "morph build" instead of running scripts/ 13:47
Krinhmm, trying to set up a trove, does anyone have any idea what i should use for my trove host? or if i need to somehow make one for the experiment?13:48
Krini dont think i understand trove much >.<13:49
KinnisonYou're setting up a fresh trove?13:49
paulsherwoodmadhu: are you building just one of the standard baserock system definitions?13:49
Krini dont know what i should be doing, i just want a place to store a folder full of things that i can clone into zookeeper later13:49
KinnisonThere's a number of approaches there13:50
madhuyes it is the default devel-system-x86_64-generic.morph file.13:50
Kinnisonperhaps the easiest is to make a branch in the zookeeper repository and add the things directly there13:50
Kinnisonkrin: ^^13:50
Krinso just make a git branch? when i look at git status it shows that no changes are around even though the folder is new...13:51
Krinoh wait it's showing the difference now13:52
paulsherwoodmadhu: strange - my version of that shows 166 chunks, not 151?13:52
petefothKrin: the Trove Refernce Manual on the wiki is based on an original which is pretty old and, while it almost certainly has some useful content, it may be hard to separate that from the outdated stuff. A web user called Michael - tlsa maybe - did a bit of work on it but I don't knw how far he got13:54
petefothKrin: we (I?) should prpobably add some warning text to this effect at the top of the doc13:54
petefothtlsa: can you tell me the current state of the document?13:55
KinnisonKrin: empty folders are not stored in git, so you'd need content in there first13:55
KrinKinnison: so if i add the build file to the git, then morph will clone that file during the build process? hopefully allowing me to bypass the maven download part? petefoth: this time it was an error between keyboard and chair13:56
paulsherwoodmadhu: i think you must be building something else? or maybe using a different trove? if i run that here, gdbm is fetched from cache on git.baserock.org13:56
Krinalso Kinnison, yes the folder has items in it13:56
KinnisonKrin: well I can't be certain, because I've not tried maven, but yes in theory13:57
Krintime to put theory to the test :) *cracks knucles and points menacingly at the computer*13:58
madhupaulsherwood, I did not change the trove,  do you think I am using different definitions13:59
Krinwait no, i needed it seperate so i could pull it in midway through the build... hmm, is there a way of telling it to clone a folder into a location that does not exist until halfway through? or knowing where th file will be copied into so i can move it? it's currently in with the .morph's in the ../strata/zookeeper folder.14:00
paulsherwoodmadhu: so to confirm you don't have a different trove, cat /etc/morph.conf and check no trove-host or trove-id lines active14:01
richard_mawno, you have no control over where morph clones things to, you just get told where it did14:01
richard_mawKrin: but, you could have a separate chunk which provides the files, and then in your zookeeper chunk copy the files where you need them14:01
KinnisonAlso, all cloning (all networking) happens before the build starts14:01
paulsherwoodmadhu: to confirm what commands you actually ran, cat ~/.ash_history14:01
ssam2madhu: that's rather unexpected!14:03
ssam2madhu: what hardware is this on? it's likely to be hardware-specific, I think, because we build gdbm on x86_64 lots and have never seen that14:03
Krinso to make a chunk i would need to push the files up to the baserock.git? they are not source files but neither are they obtainable, they are created midway through the build in a weird maven way >.<14:04
paulsherwoodssam2: i'm mystified so far :) but note that 166 != 15114:04
ssam2Krin: there's a few confusing uses of terminology in that sentence !14:05
ssam2Krin: there's no "baserock.git". We do have a Trove (which hosts Git repositories, amongst other things) running at
ssam2but we have other Trove instances too that mirror the repos from, and can host their own content14:06
* Krin gives up ever learning terminology as every time he think's he's got it it turns out he's saying the wrong thing14:06
ssam2don't give up now ! it makes it difficult to discuss things if we speak different languages14:07
ssam2although we get by with jjardon, I guess ;)14:07
Krincould someone come to my desk so i can talk face to face and show what it is i'm trying to acheive, ether i'm not explaining myself right or i'm failing to understand the responses >.<14:08
ssam2sure, let's have a chat about this14:08
madhupaulsherwood, it does not have trove-host and trove-id in that. BTW thats for locataing the ash_history, it looks like I have done something in busybox in my last session which might lead to this.14:10
madhus/thats for locataing/ thanks for locating14:10
madhuI did "morph edit busybox --verbose", and "git merge origin/master" inside /src/workspace/test-madhu/upstream/busybox14:11
madhulet me revert this, git-status shows me  the updated "ref" and "unpetrify-ref"for busybox in strata/build-essential.morph14:13
*** Krin [] has quit [Ping timeout: 256 seconds]14:16
paulsherwoodi'm still confused about the 151 total, though? was your command definitely 'morph build systems/devel-system-x86_64-generic.morph'?14:16
madhuexactly 14:17
petefothIs the word ‘pebble’ (in the Baserock context) now obsolete?14:17
paulsherwoodpetefoth: yes14:17
petefothpaulsherwood: ta!14:17
paulsherwoodpetefoth: the word is now 'Jetson' :)14:17
petefothAnd the phrase ‘staging filler’?14:18
paulsherwoodrichard_maw: i continue to ponder cache_keys...14:19
petefothbut we do still have a ‘staging area’?14:19
paulsherwoodand have arrived at a tail-chasing scenario in my head14:19
madhupaulsherwood, I used "morph branch baserock:baserock/definitions test-madhu baserock-14.40.1 --verbose", is that the reason for 151? 14:20
paulsherwoodideally, it seems to me that i could calculate a notional cache-key for a whole system from a version definitions.git14:20
paulsherwoodmadhu: ah, yes. 14:20
paulsherwoodwe added cloudstuff to devel since then14:21
paulsherwoodrichard_maw: but then it occursr to me that an actual cache-key needs to trace actual tree-refs for things. and that info is not in definitions.git14:22
paulsherwoodso it's not possible to calculate cache-keys without probing the component git repos. is that correct?14:22
ssam2paulsherwood: with the current deisgn of morph, yes14:23
petefothssam2: ‘design’ or ‘implementation’?14:23
ssam2answer unclear, ask again later14:23
richard_mawpaulsherwood: currently, yes. Which is why one of Kinnison's proposed changes is to put information like that in, but also verify that the tree matches the commit when the tree is unpacked for building14:24
paulsherwoodthis then led me to wonder if trove should have an api to tell me tree for given repo/ref14:24
richard_mawit already does IIRC14:24
paulsherwoodah? 14:24
ssam2that's one of the things morph-cache-server does14:24
robtaylorcan someone remind me what cases are optimised for by using the tree rather than the commit?14:24
paulsherwoodnot rebuild for someone tagging the tree14:25
richard_mawrebases and reverts14:25
ssam2and merges14:25
robtaylorpaulsherwood: a tag doesnt change a commit as such14:25
robtaylorssam2: ah, fast forward merge commits?14:25
paulsherwoodthere can be multiple commits leading to the same tree14:25
robtaylorpaulsherwood: there can, just not sure how often that matters14:26
paulsherwoodone of our customers has specifically identified 'dealing better with a merg-based workflow'14:26
ssam2robtaylor: it's the non-fastforward merges where there are new commits that point to an identical tree that you probably already built and tested14:27
robtaylorpaulsherwood: do they mean using merge --no-ff ?14:27
robtaylorssam2: not sure i understand what you mean14:28
*** Krin [] has joined #baserock14:30
* robtaylor rereads that last line a few more times14:30
ssam2robtaylor: I create a branch of definitions, build it, decide that it works. I then send it for review, it gets merged with '--no-ff', and now I have to rebuild 'master' of definitions.git, despite it being the same tree that I sent for review14:30
robtaylorssam2: ah, so you do mean a fast-forward merge14:31
ssam2I guess for definitions.git that doesn't actually matter, because definitions.git is a meta-repo anyway14:31
ssam2robtaylor: I thought 'no-ff' meant 'non-fastforward' ?14:31
robtaylorssam2: --no-ff means 'don't fast forward and create a merge commit' 14:31
robtaylorlets call it a single parent merge for clarity ;)14:32
ssam2The scenario where a merge commit is created is the scenario which that optimisation (using tree sha instead of commit) was done for14:32
richard_mawplus, it helps if you revert your change to the compiler, so you don't have to rebuild the world on a revert14:33
robtaylorrichard_maw: thats quite a low frequency occurance though, probably not worth specifically optimising for14:34
robtaylorssam2: if you have CI, would you care less about optimisation for your specific case there?14:36
paulsherwoodis the morph-cache-server api specified anywhere (apart from in the code)?14:36
ssam2robtaylor: I've not got a strong position on this specific optimisation. CI might well mitigate the need for it14:37
ssam2paulsherwood: no, only in the code (morph-cache-server is considered an implementation detail)14:38
ssam2the code is more or less 1 file of Python, though, it's simple enough14:38
*** genii [~quassel@ubuntu/member/genii] has joined #baserock14:47
madhupaulsherwood, it seems the build is going smoothly now, thanks for the help.14:49
paulsherwoodmadhu: great news! :)14:50
* paulsherwood sometimes wonders what 'is considered an impelementation detail' actually means14:52
ssam2paulsherwood: i doubt I use it in a consistent way. Here I meant "the user isn't meant to interact with it directly"14:53
ssam2hence, I feel less bad for it lacking any kind of API specification14:54
paulsherwoodoh, i see :)14:56
* paulsherwood gives up on trying to understand the 'simple enough' python :)14:56
pedroalvarezpaulsherwood: are you trying to achieve anything particular?14:57
pedroalvarezin particular*14:58
paulsherwoodpedroalvarez: just get tree from a ref for a given repo, without having to clone the repo14:59
paulsherwoodi was assuming this was impossible, til someone said that morph-cache-server has an api14:59
pedroalvarezwho told you that?15:00
pedroalvarezI mean, I think that morph-cache-server can't do that15:00
pedroalvarezI can be wrong, though15:00
paulsherwoodpedroalvarez: see 14:25 in backscroll15:00
* pedroalvarez opens the morph-cache-server code15:03
ssam2it's this:
ssam2I guess you do have to understand the Bottle framework a bit for the code to make sense15:04
pedroalvarezI was wrong :)15:06
* ssam2 tries and fails to manually construct an example request for the endpoint15:07
ssam2oh, because it's on port 808015:08
ssam2resolves master of definitions.git to commit SHA1 and tree SHA115:08
paulsherwoodyou did that manually? :)15:08
ssam2with a little help from Python's urllib.urlencode()15:09
ssam2i have not memorised the URL escape sequences :)15:09
paulsherwoodwhat do they mean? 15:10
paulsherwoodbut thanks, i can hack what i want from that :)15:11
paulsherwoodincidentally, what did you do exactly to create that?15:11
* paulsherwood is always on the lookout for cool tricks :)15:12
ssam2import urllib; urllib.urlencode(('repo', ''))15:12
ssam2no, sorry, not that one15:13
ssam2urllib.urlencode({'repo': ''})15:13
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]15:34
*** aananth [~caananth@] has quit [Quit: Leaving]15:38
ssam2can anyone clear up what the following means in a Trove's journal ?15:53
ssam2the Trove seems to work fine15:54
paulsherwoodpossible break in attempt?16:00
richard_mawit may just be missing reverse-dns entries16:00
*** Kinnison [~dsilvers@gateway/shell/pepperfish/x-hhzujylmdusfoekc] has quit [Ping timeout: 255 seconds]16:08
*** Kinnison [~dsilvers@gateway/shell/pepperfish/x-ipohiqaowqtrqqyr] has joined #baserock16:08
robtaylorssam2: it 's your ./etc/hosts16:09
*** zoli_ [] has joined #baserock16:10
*** zoli_ [] has quit [Changing host]16:10
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock16:10
robtaylorssam2: looks like you've got two entries for ct-training (or maybe one entry for ct-training and another for ct-training.localdomain/ct-training.<something in your search>16:11
robtaylorssam2: basically the message means - <ip address> maps to  <foo> but when i reverse-dns <foo> I don't get the same ip address16:12
ssam2robtaylor: ok, thanks ...16:14
ssam2seems like we have a problem with the defaults for Trove deployments then16:14
pedroalvarezif this is running in our datacentred tenant i think that the reverse-dns doesn't work16:14
ssam2ah, OK16:15
ssam2that may well be true for Aananth too (who sees the same message)16:15
pedroalvarezeasy to check in python: import socket; socket.getfqdn("IP ADDRESS")16:17
ssam2that's useful, thanks16:18
paulsherwoodmaybe this goes in 'Other common errors' ?16:18
ssam2this trove doesn't have a /etc/hosts16:18
ssam2maybe if it did it'd be happier16:18
ssam2although in fact, reverse DNS does seem to work here ...16:19
ssam2>>> socket.getfqdn('')16:19
ssam2>>> socket.getfqdn('')16:19
ssam2>>> 16:19
ssam2this is without even adding a /etc/hosts file16:20
pedroalvarez>>> socket.getfqdn("")16:20
ssam2so maybe it's conflicting entries instead like Rob says16:20
ssam2pedroalvarez: ah, true16:20
pedroalvarezbut indeed, I don't know if in this case is using the public ip or the private one16:21
ssam2actually the log shows: 'Address ::1 maps to ct-training, but this does not map back to the address - POSSIBLE BREAK-IN ATTEMPT!'16:21
*** Krin [] has quit [Remote host closed the connection]16:21
ssam2I believe ::1 is the IPv6 localhost16:21
Kinnisonit is16:21
ssam2but, I can resolve that to localhost in Python too ...16:21
Kinnisondoes ct-training on that box resolve to ::1 ?16:22
robtaylorssam2: localhost != ct-training16:22
ssam2robtaylor: heh! good point16:22
ssam2now I understand :)16:22
franreddoes someone knows if nbd is in the kernel tree? I mean if it is an out of tree kernel module or not?16:33
Kinnisonnbd is in-kernel I think16:33
Kinnisonthere's some userland apps too16:34
Kinnisonwhich I don't think are in the kernel tree16:34
pedroalvarezKinnison: does that mean that by default it should be installed in every baserock system?16:34
pedroalvarez(if it's in-kernel)16:34
KinnisonOh, it's in-kernel-tree16:34
Kinnisoni.e. we could build it in if we wanted16:34
pedroalvarezfranred: I think is: CONFIG_BLK_DEV_NBD16:37
KinnisonSo we've not *built* it16:38
pedroalvarezfranred: yeah, you have to enable it in the kernel morphology16:39
Kinnisonconfirms pedroalvarez's belief of CONFIG_BLK_DEV_NBD16:39
ssam2at the bottom of it says 'To make the Trove use the changes you make, push them to the local-config repository on the Trove.'16:44
ssam2so if I push a proxy.conf to local-config/lorries.git does that get used straight away automatically ?16:45
KinnisonIt should16:45
ssam2I pushed an invalid proxy.conf as a test and I don't see any errors16:45
ssam2but it maybe that it silently ignores errors, rather than that it's missing the update16:45
KinnisonCheck l-c's source?16:45
ssam2I shall16:45
pedroalvarezworth noting that node doesn't build in big-endain16:53
franredKinnison, pedroalvarez, cheers16:53
pedroalvarezwait, I'm wrong, node is not in this system I think16:54
*** franred [] has quit [Quit: Leaving]16:57
pedroalvarezit actually is16:58
* pedroalvarez removes it for now16:58
*** ssam2 [] has quit [Quit: Leaving]17:08
*** dutch [] has joined #baserock17:10
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]17:19
pedroalvarezI'm trying to call the baserock-installer script as the init script (init=/myscript in the kernel args)17:25
pedroalvarezI don't know what to do when the system is installed17:26
pedroalvarezI think that I should reboot, and I tried to do that, but something is different when I run this script as an init script, and I can't reboot the system17:26
straycatWhy can't you reboot?17:28
pedroalvareznot sure, it fails, I was trying to do it with['reboot')]. I'm trying now with os.system('reboot')17:31
pedroalvarezand this time it fails with: Failed to talk to init daemon.17:32
pedroalvarezwhich makes sense17:32
* straycat nods17:32
*** dutch [] has quit [Quit: Quit]17:34
straycatreboot -f maybe?17:34
pedroalvarezI think that `reboot..` talks to the init deamon, and then the init daemon decides to reboot17:35
straycatI think using -f skips over that init stuff iirc17:36
pedroalvarezthen I will try17:36
straycatIt's not a safe way to bring the system down though, you probably need to sync and unmount stuff first.17:37
*** hari_ [uid49867@gateway/web/] has quit [Quit: Connection closed for inactivity]17:37
pedroalvarezstraycat: that just worked, thank you17:39
*** jonathanmaw [] has quit [Quit: Leaving]17:54
*** mariaderidder [] has quit [Quit: Ex-Chat]18:02
pedroalvarezthis is scary18:06
radiofreeis that u-boot?18:10
radiofreeyes it is make: *** [u-boot] Error 13918:12
radiofreeyou don't need to recompile that be?18:12
*** Guest27389 [] has quit [Quit: Leaving]18:20
pedroalvarezradiofree: i guess so. I will reuse the armv7lhf one18:23
pedroalvarezI was doing it just copying the existing system, changing the architecture, and morph build. 18:27
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock18:51
jjardonalmost there and now I need a new version of sqlite :/ time to some fresh air19:15
pedroalvarezjjardon: for gnome? 19:17
pedroalvarezThat should be easy, but it will trigger a full rebuild 19:18
jjardonpedroalvarez: evolution-data-server, yes19:18
jjardonpedroalvarez: we do not track the git repo but seems we only lorried a tarball19:18
jjardonand also: full rebuild for a new version of sqlite? crazyness :)19:19
pedroalvarezNot full, but I believe that is in core. Don't ask me why 19:20
jjardonpedroalvarez: python19:21
jjardonwell cython19:21
* jjardon would split python from core19:21
pedroalvarezjjardon: I want to see that running! :) 19:26
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]20:00
*** cosm [] has joined #baserock20:02
cosmhi, it's been suggested I might find people hacking jetson tk1s here20:04
cosmis anyone familiar with the memory controller?20:04
cosmI've replaced the RAM on my board with 1 GB chips, but I'm having trouble finding stable settings for a 4GB configuration20:07
pedroalvarezHi cosm! radiofree has been playing hard with the tk1, maybe he knows something... 21:10
robtaylorhey cosm 22:09
robtaylordon't think anyone's been hardware hacking them yet here, i'm afraid22:09
robtaylorcosm: tried steve warren? that's the sort of thing he might know22:10
*** genii [~quassel@ubuntu/member/genii] has quit [Remote host closed the connection]23:05

Generated by 2.15.3 by Marius Gedminas - find it at!