IRC logs for #baserock for Thursday, 2015-06-11

*** edcragg has quit IRC00:27
*** zoli__ has joined #baserock05:17
*** zoli__ has quit IRC05:18
*** zoli__ has joined #baserock05:22
*** zoli__ has quit IRC06:48
*** a1exhughe5 has joined #baserock07:24
*** zoli__ has joined #baserock07:31
*** zoli__ has quit IRC07:33
*** mdunford has joined #baserock07:51
*** zoli__ has joined #baserock07:53
*** bashrc has joined #baserock08:07
straycatfranred, assume 838 is tested?08:15
*** mariaderidder has joined #baserock08:15
Kinnisonlast night, I blogged about `curl | sudo bash -` as an installation mechanism: https://blog.digital-scurf.org/posts/in-defence-of-curl-pipe-sudo-bash/ -- I'd be really interested in whether anyone in this community has ideas for how to solve the problem (or knows of projects already on the way to solving it)08:31
*** ssam2 has joined #baserock08:31
*** ChanServ sets mode: +v ssam208:31
rjekFor me, Debian packages.08:32
rjekFrankly, if there isn't a Debian package I look for something else.08:33
* Kinnison nods, but you are probably in the small subset of people who knows they should review scripts, and maybe even could08:33
Kinnison:)08:33
franredstraycat, yes, it is tested and Im gonna test it once again today08:37
rjekKinnison: But your blog post is tl;dr, sorry08:39
Kinnisonrjek: yeah it was a bit wordy08:40
Kinnisonit kinda bimbled out of my fingers while I was catching up on Ted-Ed youtube videos08:40
* SotK read it, it was interesting08:40
*** jonathanmaw has joined #baserock08:50
franredjjardon, biff regading iputils08:51
straycatfranred, cool08:53
*** mdunford has quit IRC08:55
rdaleare .so files which don't start with lib* ignored by ldconfig?08:59
Kinnisonquite likely08:59
rdalei'm finding that a plugin in /usr/lib isn't being found by dlopen(), and it is found in /usr/lib64. but there are no lib64 paths in /etc/ld.so.conf and so i don't know what is giving the lib64 paths to search when dlopen() is called09:01
Kinnison      if (((strncmp (direntry->d_name, "lib", 3) != 009:01
Kinnison    && strncmp (direntry->d_name, "ld-", 3) != 0)09:01
Kinnison   || strstr (direntry->d_name, ".so") == NULL)09:01
Kinnisonthat's part of an early-exit condition09:02
Kinnisonso it wants to start lib or ld- and have .so in the name09:02
Kinnison(or be a hwcap dir)09:02
Kinnisonotherwise it gets ignored09:02
KinnisonConsider the loop at http://git.baserock.org/cgi-bin/cgit.cgi/delta/glibc.git/tree/elf/ldconfig.c#n70909:03
rdaleok thanks09:03
*** zoli___ has joined #baserock09:05
*** zoli__ has quit IRC09:06
rdalewhen i run 'dlh = dlopen("uhttpd_ubus.so", RTLD_LAZY | RTLD_GLOBAL);' under strace i can see it looking in several different lib64 paths including /usr/lib64, but it doesn't look in /usr/lib. so that seems to be a different mechanism to the ldconfig cache09:07
*** mdunford has joined #baserock09:10
*** pacon has joined #baserock09:12
*** lachlanmackenzie has joined #baserock09:12
KinnisonYour rtld might be built with only /usr/lib6409:13
KinnisonTry adding /usr/lib to your /etc/ld.so.conf09:13
rdaleit is in there, and there are no lib64 paths09:13
Kinnisonright, it looks like dlopen doesn't honour ld.so.conf09:14
Kinnisonbut ldconfig does, but will ignore your file for not starting lib or ld-09:14
rdaleyes09:14
KinnisonFrankly running dlopen() on something not by full path and without an RPATH or RUNPATH in your ELF exectuable seems like a losing battle09:15
*** tiagogomes_ has joined #baserock09:15
*** tiagogomes_ has quit IRC09:16
*** tiagogomes has joined #baserock09:17
pedroalvarezten minutes till standup09:20
pedroalvarezgrr09:20
pedroalvarezwrong channel, network and everything09:20
Kinnisonhehe09:21
pedroalvarezgood day everyone :)09:22
richard_mawrdale: if necessary it is possible to set the default search path of the linker when you build the libc, but ld.so.conf and running ldconfig is a more extensible approach09:24
Kinnisonrichard_maw: sadly that won't work if his .so isn't named lib*.so or lib-*.so09:25
*** edcragg has joined #baserock09:31
ratmice___in solaris, ld.so will load audit libs from /usr/lib/secure thus you could stuff an audit lib there that would honour ld.so.conf or something09:38
ratmice___anyhow i believe 32/64 bit programs use different default search paths, and don't mix libs, so is it a 64 bit binary 32 bit uhttpd_ubus.so?09:43
ratmice___that is, there are different ld.so's and they have different built in search paths09:43
jjardonfranred: hi! replied09:44
ratmice___that said, I don't see any reason not to, because it can still complain the architecture didn't match which is a much more helpful error probably09:44
franredjjardon, what you suggest is a different patch, that patch only changes the source from where arping is built. I can try to create another patch which includes everything you are suggesting but I think this patch is not the place to do it, though.09:47
jjardonfranred: if you have another example where a chunk name is different from the upstream project, I'd agree with you. But I think is confusing as it is09:50
franredjjardon, do you mean the upstream project or the gbo name of rf the upstream project? (for the first one, I think there are few of them)09:52
franreds/first one/former/09:52
radiofreewhat's the foo for deploying from ybd?09:55
SotKradiofree: `ybd $some_cluster`09:57
radiofreehmm "2015-06-11 10:00:20 [build-system-armv7lhf-jetson] Skipping deployment for armv7lhf"10:00
SotKhmm, it must be detecting the arch wrongly10:01
pedroalvarezjjardon, franred: If we call it 'iputils' then we should install all of them10:01
pedroalvarezwhich is not a bad idea but it will need some testing10:01
pedroalvarezit's already in a good place: 'networking-utils'10:02
franredok10:02
jjardonfranred : if you don't have time for that testingh, fix the sbin -> bin comment and I will remove my -110:03
franredjjardon, I've commented that out too :P10:03
pedroalvarezjjardon: iirc it needs to go to sbin10:03
franredjjardon, I will install iputils and test them10:03
radiofreeSotK: ah, needed to add the arch to the command, thanks10:04
radiofreehmm.. completely failed to create the extlinux.conf10:07
SotK:(10:08
pedroalvarezradiofree: in arm?10:11
radiofreepedroalvarez: this is ybd, and yes10:12
SotKradiofree: what does platform.machine() return in python on there OOI?10:12
radiofreearmv7l10:13
pedroalvarezhm.. don't you need syslinux in the system to be able to deploy rawdisk images?10:13
radiofreehmm.. building from stage1 again10:13
radiofreepedroalvarez: no10:13
SotKah, that's why the auto-detection failed, since ybd uses that to get the arch if it isn't given10:14
jmacsfranred: In my old java patch series, you suggested making java-build a build-dependency of zookeeper. Doesn't zookeeper need java to run?10:16
franredjmacs, in the other way around, make zookeeper build depends on java-build - so you have a java-build stratum which is shared between your java system and the zookeeper systems10:17
jmacsRight, but don't you need java as a component in the finished thing as well as a tool for building it?10:18
franredjmacs, possibly, so you can add java-build in the zookeeper systems as a stratum to install on them10:20
jmacsWhat does that give you that just leaving java-binary as a chunk in zookeeper.morph doesn't?10:20
franredwell, in case any other system needs java-build you just need to add the stratum and if you need to update java-build all the system will get update at the same time10:22
franredif no you will need to update one by one10:22
jmacsWhat kind of update to java-build are you thinking of?10:26
jmacsIt seems quite reasonable to me that the zookeeper stratum could require one version of java-binary and another stratum could require another version10:26
franredjmacs, fair enough, although we try not to keep different versions of the same chunk in different stratum/systems - and if we do we different them with @ I think, like u-boot10:31
SotKrichard_maw: what is the state of your strip-commands branch now?10:33
richard_mawSotK: ssam2 had some concerns about my change to make objects out of things. AIUI that's the only controversial change.10:34
richard_mawSotK: I don't have time to rework that individual change at the moment, so I can drop that and send a version without it, if you want it to be merged sooner.10:35
SotKthat would be very useful, thanks!10:35
*** bashrc has quit IRC10:38
*** pacon has quit IRC10:40
*** pacon has joined #baserock10:40
*** bashrc has joined #baserock10:41
richard_mawSotK: Updated the series to move the objects change to a separate topic10:42
SotKrichard_maw: great, thanks10:42
jmacsThe arrangement of strata and systems doesn't seem to follow any logic to me. I can just keep resubmitting random versions until everyone gives it a +1, though.10:43
*** straycat has left #baserock10:46
franredjmacs, what are you concerns about the arrangements of strata and systems?10:47
*** ssam2 has quit IRC10:56
*** ssam2 has joined #baserock11:07
*** ChanServ sets mode: +v ssam211:07
*** tiagogomes has quit IRC11:48
*** tiagogomes has joined #baserock12:09
*** gary_perkins has joined #baserock12:41
ssam2richard_maw: what are the .cmxs and .node files mentioned in the code in https://gerrit.baserock.org/#/c/782/5 ?12:46
ssam2both seem to be un-googlable12:46
richard_mawssam2: I followed the lead of what Debian does. I'll find you a link12:47
richard_mawssam2: https://anonscm.debian.org/cgit/debhelper/debhelper.git/tree/dh_strip#n120 https://anonscm.debian.org/cgit/debhelper/debhelper.git/commit/dh_strip?id=476465a81dcd871c5b916a65017b4f15de5206df and https://anonscm.debian.org/cgit/debhelper/debhelper.git/commit/dh_strip?id=f89b688d5eb629f0e6a1e374ff5a471d62020f6412:49
ssam2ok, thanks12:49
franredbuilding iputils I got the following error: http://paste.baserock.org/diyiveqiki12:51
franredwhich openssl are we building?12:51
pedroalvarezfranred: our definitions are pretty readable :)12:54
franred.....12:54
franredok, I can change the question, we are not building gnu-openssl12:55
franrednevermind I can change USE_GNUTLS=no and ping6 will compile without errors12:57
richard_mawwhat's gnu-openssl?13:01
franredrichard_maw, http://en.wikipedia.org/wiki/GnuTLS, ? https://gitlab.com/gnutls/gnutls/13:03
franredrichard_maw, hold on...13:05
franredrichard_maw, looks like gnutls has some openssl headers which iptools refers to - so there are no gnu-openssl but these headers are in the repo I posted before13:14
franredwhich..... ewwww13:14
franredfor iputils13:15
*** pacon has quit IRC14:02
jmacsfranred: Now I've modified zookeeper definitions, would you be happy if zookeeper-server and zookeeper-client build, or do you want a deployed test of zookeeper?14:02
franredjmacs, if you are confident your changes do not break them, it will work for me14:07
jmacsI can't think of any way it could break it :)14:07
franredcool :)14:08
*** edcragg has quit IRC14:09
*** edcragg has joined #baserock14:14
*** edcragg has quit IRC14:28
*** edcragg has joined #baserock14:35
SotKDoes anyone else have time to take a look at the "Moving writeexts.py into definitions" email thread on baserock-dev? :)15:31
jjardonHi, I realized several packages depen on git in build time to set its version (because the use of the m4_esyscmd([build-aux/git-version-gen .tarball-version]) in configure.ac) ; to fix this, do we want to patch every chunk to add the "echo "[version]" > .tarball-version" hack, or should we make depend this chunks on git?15:34
jjardon(note this will only affect chunks in the core stratum)15:34
paulsherwoodSotK: i could +1 it here, but i'm biassed. i fear we need to get ripsum to revoke his -1 in any case?15:34
SotKpaulsherwood: indeed we do, but other folks reviewing it in the meantime would be useful too15:35
jmacsIs git.baserock.org on a bit of a go-slow today? (I mean actual git, not http)15:35
pedroalvarezjmacs: check now15:37
jjardonrdale: you made a patch to remove the git dependency from bison, so maybe you have an opinion on that?15:37
pedroalvarezthere are sometimes some `cvsps` processes around in g.b.o..15:38
pedroalvarezjust killed them15:38
rdalejjardon: yes i think we should remove git dependency where possible even if creating a .tarball-version seems a bit hacky15:38
jmacsDespite what I said, I was actually trying to "git clone http://git.baserock.org/..."15:39
jmacsgit protocol works fine, but http doesn't appear to get anywhere15:39
jjardonI was working towards that goal (so we can move git out of core), but maybe others think different?15:39
pedroalvarezjmacs: seems to be working for me15:40
* jmacs shrugs - I have a checkout anyway15:40
franredrdale, jjardon, IMHO it is very hacky to have to add the version manually, is any plan to move git out of core? why you don't want to build depend on git in the core stratum chunks?15:42
jjardonyeah, I was of the opinnion that git should be out of core, but as I'm writing more and more patches to add the version manually I'm not sure anymore15:44
paulsherwoodradiofree: looks like you found a bug in ybd arm arch? was it fixable?15:45
wdutchbashrc: could I change the commit message of gerrit.baserock.org/#/c/161 to give more detail?15:45
bashrcwdutch: yes15:45
wdutchcool :)15:45
*** a1exhughe5 has quit IRC15:46
franredjjardon, I saw quite a lot today, I've +1 some of them but I start thinking is not a good approach, I prefer that the version get autogenerated someway15:46
wdutchgerrit question: does clicking the Edit Message button change the commit message proerly or just in the web thingy?15:46
radiofreepaulsherwood: yes, i'm about to send a patch15:46
radiofreeit's not just ARM, it's every deploy15:46
franredwdutch, it changes the commit message not just in the web15:47
paulsherwoodoh. i'm out of my depth again, then15:47
SotKradiofree: what was the problem?15:47
radiofreewasn't passing deploy_defaults to the extension15:47
SotKoops, I should probably have noticed that15:51
wdutchcool, but it doesn't tell you the line length :(15:54
wdutchfranred: could you merge these please? gerrit.baserock.org/#/c/161 gerrit.baserock.org/#/c/162 gerrit.baserock.org/#/c/16315:56
franredwdutch, you can always check that in the COMMIT_MSG15:56
* wdutch will use git review and commit --amend in future15:57
franredwdutch, done15:59
wdutchthanks15:59
franredwdutch, 164 too?15:59
radiofreeyou can't have subsystems within subsystems right?15:59
radiofreewithin a single cluster16:00
wdutchfranred: yes please, it was conflicting with one of the others and I wasn't sure how long gerrit takes to process things16:00
SotKradiofree: you can16:00
radiofreeah16:00
radiofreehmm16:00
SotKybd doesn't handle that case properly yet (for building at least) I think16:01
* SotK is working on a patch to make it so16:01
radiofreeok, well i wont try and fix that now then16:01
franredwdutch, done16:02
paulsherwoodradiofree, SotK - thanks for the patches!16:15
paulsherwoodSotK: do you want me to revert?16:21
*** mariaderidder has quit IRC16:50
*** jonathanmaw has quit IRC16:57
*** zoli___ has quit IRC17:04
*** zoli__ has joined #baserock17:07
*** tiagogomes has quit IRC17:13
*** mdunford has quit IRC17:17
pedroalvarezwow, OpenStack Kilo in baserock has been merged17:25
pedroalvareztime to share this video I guess: https://vimeo.com/13042222517:25
radiofreecongrats!17:30
*** edcragg has quit IRC17:36
radiofreeeek NameError: global name 'definitions_schema' is not defined17:39
*** ssam2 has quit IRC18:27
*** gary_perkins has quit IRC18:43
*** lachlanmackenzie has quit IRC18:45
*** zoli__ has quit IRC22:14
*** zoli__ has joined #baserock22:18
*** zoli___ has joined #baserock22:21
*** zoli__ has quit IRC22:23
*** edcragg has joined #baserock22:48
*** edcragg has quit IRC23:30

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!