IRC logs for #baserock for Thursday, 2014-10-30

*** persia_ [quassel@ubuntu/member/persia] has quit [Quit: No Ping reply in 180 seconds.]01:07
*** persia_ [quassel@ubuntu/member/persia] has joined #baserock01:21
*** vmeson [~quassel@] has quit [Ping timeout: 240 seconds]01:43
*** vmeson [~quassel@] has joined #baserock01:45
*** De|ta [~arc@] has quit [Ping timeout: 265 seconds]02:38
*** De|ta [~arc@] has joined #baserock02:39
*** VVIBIN [~VVIBIN@] has joined #baserock04:26
*** aananth [~caananth@] has joined #baserock07:42
*** SotK [] has joined #baserock08:03
*** ssam2 [] has joined #baserock08:12
Mode #baserock +v ssam2 by ChanServ08:12
*** tiagogomes [~tiagogome@] has joined #baserock08:23
*** dutch [] has joined #baserock08:35
*** mariaderidder [] has joined #baserock09:08
*** tiagogomes [~tiagogome@] has quit [Ping timeout: 272 seconds]09:08
*** tiagogomes [~tiagogome@] has joined #baserock09:09
*** jonathanmaw [] has joined #baserock09:32
*** franred [] has joined #baserock09:38
*** ssam2_ [] has joined #baserock09:52
*** ssam2 [] has quit [Read error: Connection reset by peer]09:52
*** petefoth [] has quit [Quit: petefoth]09:55
*** jonathanmaw [] has quit [Ping timeout: 256 seconds]10:01
*** jonathanmaw [] has joined #baserock10:14
*** Krin [] has joined #baserock10:16
Krindoes anyone know a way that i can make it so that during the build stage for a system i can add a program to the path manually as it does not do this itself? so for ecample i could just type 'ant' to have the program run rather than '/usr/lib/ant/bin/ant' i tried to add it to a bash profile but i cant seem to get one of those working in baserock... 10:30
KinnisonKrin: I'd add, in the ant chunk, a symlink in /usr/bin10:31
KinnisonKrin: i.e. in install-commands, I'd add:10:31
KinnisonKrin: - ln -sf "${PREFIX}/lib/ant/bin/ant" "${DESTDIR}${PREFIX}/bin/ant"10:32
Kinnisonmaybe need an - install -d "${DESTDIR}${PREFIX}/bin" before that10:32
* Kinnison is sure richard_maw will chime in here and give you better commands too10:32
Krinok Kinnison, thank you :) i did try a bit of PATH radiofree, but every time i mess with that bad things happen. probebly does not help that all these {DESTDIR} & {PREFIX}'s look like place holders to me not actual things i should be typing, need to get my head around that10:34
KinnisonThey're not placeholders10:34
Kinnisonthey're variables morph passes in to the chunk build10:34
richard_mawKinnison: a mkdir -p would be better10:34
Kinnisonfair enough10:34
Krini know that now, but my brain refuses to not se them that way >.< because i dont understand them10:34
* pedroalvarez wonders if it would be possible to say ant to be installed in /usr/bin10:35
Kinnisonnot easily10:35
Kinnisonant, like so many java tools, expects to have its own little playground10:35
Krinant is having it it's way or it's taking it's building blocks and going home :(10:36
ssam2_maybe it's scared because it knows the other unix tools don't like it10:37
KinnisonIt's being marginalised by gradle10:38
KinnisonBut old-school projects still like it10:38
rdaleradiofree: i've built my jetson with the correct mesa, but the clocks demo still isn't working10:42
ssam2_jjardon: I can't see your baserock/jjardon/weston branch, did you push it ?10:42
paulsherwoodKrin: i agree with you that the DESTDIR PREFIX stuff is nasty10:44
Krinthat feels somewhat reassuring paulsherwood :)10:46
ssam2_gobject-introspection failed to build and then succeeded when I retried :(10:54
richard_mawmax-jobs: 1?10:55
ssam2_that's not set10:55
ssam2_i'm debating trying to fix the makefile or just doing a hack to workaround it10:56
ssam2_we'll see how the day progresses!10:56
persiaIf at all possible, please fix the makefile and send the patch upstream.10:57
ssam2_tracking down and fixing parallel build issues in makefiles is sometimes possible :)10:57
persiaSome upstream projects have heard of us, but some have not, and we definitely want to build a reputation for sending patches, rather than a reputation for carrying workarounds (I've seen this cause friction in the past)10:57
ssam2_I agree10:57
persiaHeh, true.  In those cases where it's less possible, jobs=1 works :)10:58
persiaPersonally, I find make -p incredibly useful when trying to debug code.10:58
ssam2_I've not tried that, thanks for the suggestion10:59
radiofreerdale: :\11:03
rdaleis weston 1.4.93 the correct version?11:03
radiofreedoesn't matter, clocks not working is nothing to do with weston11:04
rdaleweston doesn't work very either11:05
rdalevery well11:05
*** aananth [~caananth@] has quit [Ping timeout: 255 seconds]11:05
* radiofree shrugs11:09
radiofreeworks for me11:09
ratmice________i was thinking of trying to do a gnumake 4 plugin, that preloads some library to every job, and collecting all dependencies from files opened, though I doubt you guys run v4 because of gpl3 :D11:10
KinnisonGiven 'make' rarely ends up on a gplv3-sensitive system anyway, it wouldn't be impossible11:11
*** aananth [~caananth@] has joined #baserock11:17
Krinok, so that works. next, how to get something that during install downloads from maven into the build >.< (thinks this may be the point that he dies) 11:20
KinnisonI fear you will need to find a way to prepopulate the maven cache11:21
KinnisonOr else you'll have to walk the dependencies getting them installed first11:21
robtaylorratmice________: that sounds useful!. I guess you could also do it by stracing.11:22
Kinnisonpreload would be less of a performance overhead than ptrace11:22
Kinnisonalthough it'd catch less11:22
KinnisonI believe a friend of mine did research into using ptrace during package builds/installs to keep track of things11:22
Kinnisonrobtaylor: Rob Bradford did something on those lines I *think*11:23
robtaylorthere was a script for openembedded that analaysed dependancies for licence checking, i think11:23
rjekSomebody I know who did the ARM port of Slackware did something like that for the automatic creation of packages.11:23
rjek(using an LD_PRELOAD, not ptrace)11:23
robtaylor(well, i think it was open, may have just been internal to where i saw it)11:23
rjekHe did a talk about it at FOSDEM a few years ago11:24
rjekBlimey.  A decade ago, not a few years ago.11:25
* rjek feels old.11:25
robtaylorratmice________: Kinnison: think the 'fabricate' tool above may be sufficient?11:26
Kinnisonrobtaylor: for determining if gplv3 source content was accessed as part of building something -- perhaps yes11:27
ssam2_rjek: `checkinstall` on Slackware was amazingly useful11:28
ssam2_basically auto-created a package just from a 'make install' execution11:29
* rjek wonders what Stuart's up to these days.11:29
rjekssam2_: Nod.11:29
* rjek wonders if fakeroot can do something like this these days.11:30
jjardonssam2_: sorry, pushed now!11:38
ratmice________robtaylor: maybe though i'm hoping with a make plugin you can compare the actual dependency graph to the in the makefile11:38
ssam2_jjardon: thanks!11:38
robtaylorratmice________: that would be pretty cool!11:39
ssam2_hmm, I wonder if this parallel make issue is thornier than I thought11:41
ssam2_the required file exists, but the libtool magic required to actually use the required file is not yet complete11:41
robtaylorssam2_: I might be able to help. Point me at the source in question?11:42
ssam2_somewhere in those Makefiles ;)11:43
ssam2_I can paste the build failure somewhere, hang on11:44
robtaylorssam2_: taking a look11:45
robtaylorwow, that *is* special11:47
ssam2_using 'make -j 20' it happens quite often, which is nice11:48
robtaylorssam2_: so its definitely a race?11:49
* radiofree is envious of ssam2_'s computer11:49
ssam2_radiofree: i'm just trying use induce the error, I only have 4 cores in it!11:50
ssam2_robtaylor: yes, it sometimes succeeds too11:50
robtaylorhmm, so this is probably properly solved in automake or libtool 11:53
ssam2_good point11:53
ssam2_in fact!11:53
ssam2_I just saw a build failure from liborc, with the same sort of thing11:53
ssam2_so yeah, probably a bug in libtools autofluff11:54
robtaylorgotta head. i'l have a bit moreof a look later11:55
ssam2_not much has changed recently ... we upgraded Automake 45 days ago, but you'd think the issue would have come up already in that time11:56
ssam2_libtool provides its own m4 macros, and we haven't changed libtool11:56
richard_mawssam2_: biff, I'm afraid I had to manually consume kwargs arguments12:02
richard_mawthough that was back when I was accepting multiple commands, so it may be worth trying again12:03
richard_mawbut there were some fundamental problems with the python parameter parsing that had to be fixed in python312:03
ssam2_I guess *args and **kwargs don't interact well12:04
richard_mawnot stuff you usually encounter, but I wanted to avoid requiring you to pass a list or tuple12:04
ssam2_why, in this case? seems turning cmd into a list would be OK12:05
ssam2_callers just need to add [ ] around things12:05
richard_mawit is now, previous versions accepted multiple argv lists, like runcmd does for its pipelines12:05
richard_maw(runcmd needs to do the same tricks as I did)12:05
jmacsWe're going to have to adapt a script which does lots of package management soon - is there a preferred way to adapt package management to Baserock?12:21
jmacsWhat I've thought of so far is pre-installing everything the script needs and then making a 'package management tool' which actually just checks it's there; or we make something that adds new repos, builds and self-deploys12:24
franredjmacs, are you talking about import tool?12:25
jmacsI don't know; I'll go and look that up12:26
straycatIf it's anything like the stuff we've done for gems, python packages etc. Then we generate lorries, morphs etc and add them to baserock.12:26
straycatjmacs, The import tool is a tool to support adding ruby, python and other packages to baserock.12:27
straycatssam2_ has written the bulk of it, so it may be worth asking him12:29
ssam2_jmacs, yeah, I think we should have a chat about this12:31
ssam2_ provides a framework for importing metadata from foreign package systems12:32
ssam2_it's young, though12:32
ssam2_there should be no post-build/deploy config of the system done, I think12:33
ssam2_so if it's possible to create a system that represents what the script would have done, and throw away the script, that's the best approach I think12:34
ssam2_if it's not that simple, perhaps the best thing is for the script to drive the import tool, which would create an appropriate system, and the script would then build and deploy that12:35
ssam2_having a 'check that the software I need is really there' script is an interesting idea, I've not thought about that for Baserock before12:36
ssam2_but it seems (with little knowledge of what you are doing) that it shouldn't be necessary12:36
jmacsI think what we're doing will be more apt-get than gem, but yeah, hopefully we can just make an image with everything installed already12:42
franredjmacs, I can't wait until straycat finishes with the python import tool part... adding all the dependencies looking at the of all the packages is... tedious 12:46
Krinok, i have told the build to download zookeeper from the upstream using ref:master, and with build dependancey of ant, but it's telling me that there is no repo caled zookeeper, yet i can see the zookeeper delta on am i going even more insane?12:51
Krinwill do 12:52
radiofreerepo should be "upstream:zookeper" btw12:52
Krini keep accidentaly typing zookeper too...12:55
radiofreeevery looks fine there, i'll try it here12:56
*** JPohlmann [~jannis@xfce/core-developer/JPohlmann] has quit [Ping timeout: 265 seconds]12:56
*** JPohlmann [] has joined #baserock12:56
*** JPohlmann [] has quit [Changing host]12:56
*** JPohlmann [~jannis@xfce/core-developer/JPohlmann] has joined #baserock12:56
radiofree"Command failed: git clone --mirror -n zookeeper" doesn't look right12:56
radiofreedo you have a morph.conf?12:56
*** Krin [] has quit [Read error: No route to host]12:58
*** Krin [] has joined #baserock12:59
radiofreeKrin: i get a different error13:00
radiofreeERROR: Ref master is an invalid reference for repo git://
radiofreeit clones fine though13:00
Krin... this laptop fails at battery monitoring, i litteraly just checked and it said it had 2 hours remaining and then ran out. 13:00
Krinhuh, interesting13:00
radiofreetry changing the ref to branch-3.513:00
Krinok, i'll try that :)13:01
radiofreebut yeah, copy and pasting what you have works for me (at least cloning)13:01
Kringlad to know i'm not going ENTIRELY insane, just bringing my "this is now broken" magic to the table again :)13:02
radiofreedo you have an /etc/morph.conf13:03
radiofreeis it possible morph doesn't know what to do with "upstream:"?13:03
richard_mawnot really, he's already using it for the ant stuff13:03
Kinnisonzookeeper doesn't have a master ref13:04
Kinnisonit has a trunk ref13:04
radiofreeyeah i suggested changing the ref, however the error looks like it's failing to clone?13:05
Kinnisonthat is a little odd and I'd be running 'git grep zookeeper' and seeing if I accidentally squiffed a repo: elsewhere13:07
Krinwell changing the ref seems to work quite well, just have to work out how to make a morphology that will cope with the insanity now ^.^ 13:08
Krinbut first, lunch time.13:09
pedroalvarezhm.. is not possible to set a different 'init' in the kernel args using KERNEL_ARGS13:39
richard_mawpedroalvarez: you ought to be able to, but I don't recall whether I hooked that up in the initramfs13:40
radiofreeinit=/sbin/init is a default morph argument13:41
richard_mawit should be handling it13:41
richard_mawwith an initramfs at least, the rightmost init= will be used13:42
* pedroalvarez will research 13:42
persiaOn maven: wasn't there some discussion about handling ruby dependencies on the list?  Can the same class of architecture be used to parse maven?13:51
ssam2_I hope that's the right approach for Maven, but I don't know enough about Maven to comment13:59
*** genii [~quassel@ubuntu/member/genii] has joined #baserock14:03
persiaAs I understand it, the main issue with maven is that it is designed to work on binary blobs, many of which are not sourceful, which complicates things.14:06
persiaOh, and the magic of jars and JAVA_PATH means that many of the filesystem issues encountered with other systems do not apply, so maven doesn't handle them at all.14:07
ssam2_can anyone remember the exact cause of the 'clone: Invalid argument' error when running a Morph build14:15
ssam2_I know it's some missing kernel feature14:15
ssam2_but I forget exactly what14:15
richard_mawkernel doesn't have namespace support built in14:15
ssam2_do you remember the config flag needed ?14:15
ssam2_maybe just CONFIG_NAMESPACES14:16
richard_mawyou'll need at least that one14:16
richard_mawyou'll also want network namespaces14:16
richard_maw(unqualified namespaces means the mount namespace)14:16
ssam2_I'm struggling to find which flag is required for that14:18
ssam2_CONFIG_NET_NS. got it14:18
*** VVIBIN [~VVIBIN@] has quit [Ping timeout: 256 seconds]14:23
*** flatmush [] has quit [Quit: Leaving.]14:54
*** flatmush [] has joined #baserock14:54
*** aananth [~caananth@] has quit [Ping timeout: 256 seconds]15:11
*** dutch [] has quit [Quit: Quit]15:16
*** dutch [] has joined #baserock15:18
*** aananth [~caananth@] has joined #baserock15:20
*** aananth [~caananth@] has quit [Ping timeout: 250 seconds]15:42
*** aananth [~caananth@] has joined #baserock15:50
ssam2_my 'make' problem seems to vanish if I build on a filesystem that isn't backed by binds/FUSE15:51
ssam2_but ! the check I added to libtool to see if the compiler output really exists after the compiler exits continues to fail15:52
ssam2_also, I can't find a testcase that finds a bug in bindfs, so far15:52
KinnisonOh dear15:53
richard_mawkdbus has been sent for review now:
aananthssam2_: I have a question on "Deploy Trove to KVM". I have sent you an email. Can you please reply?16:04
richard_mawa couple of comments asking why it's needed, a lot saying that it should be checking capabilities at open time, rather than binding or message sending time, a couple complaints that the delegating of busses to containers can't work, despite evidence to the contrary, and complaints that it can't be checkpointed16:06
richard_mawoh, and complaints that it's not opt-in on the sender side as to what metadata is sent with the messages16:07
ssam2_aananth: sure, one sec16:10
ssam2_aananth: I've replied now16:19
aananthssam2_: Thanks. I have replied. I am just copying the question I raised in the email again here: I did not quite understand the last point "It's actually better to have a separate disk for /home, so that the Trove data is stored separately." Are you referring to the /home directory of the newly created trove VM? 16:34
*** aananth [~caananth@] has quit [Quit: Leaving]16:36
*** tiagogomes [~tiagogome@] has quit [Ping timeout: 245 seconds]16:54
*** jonathanmaw [] has quit [Quit: Leaving]17:03
*** perryl [] has quit [Ping timeout: 258 seconds]17:42
*** ssam2_ [] has quit [Quit: Leaving]17:43
robtayloroh, in other news, source for four-times-faster qemu released
paulsherwoodooh :)17:47
jmacs"Read the documentation in qemu-doc.html" says the README. qemu-doc.html is in .gitignore, of course17:47
Kinnisonpresumably it's a generated output17:47
*** dutch [] has quit [Ping timeout: 250 seconds]18:03
*** mariaderidder [] has quit [Quit: Ex-Chat]18:08
*** franred [] has quit [Quit: Leaving]18:14
*** dutch [] has joined #baserock18:16
*** Krin [] has quit [Remote host closed the connection]18:59
*** dutch [] has quit [Quit: Quit]19:24
*** VVIBIN [~VVIBIN@] has joined #baserock20:19
*** VVIBIN [~VVIBIN@] has quit [Ping timeout: 244 seconds]20:24

Generated by 2.15.3 by Marius Gedminas - find it at!