IRC logs for #baserock for Thursday, 2014-09-04

*** flatmush [] has quit [Ping timeout: 245 seconds]03:50
*** flatmush [] has joined #baserock03:51
*** fay_ [] has quit [Ping timeout: 240 seconds]03:53
*** fay_ [] has joined #baserock03:54
jjardonHi, Im trying to compile a new version of linux-api-headers (the one in baserock/morph branch), but I always get this error:06:05
jjardon"/ trap: line 113: ERR: invalid signal specification"06:05
jjardonseems to be a limitation in ash, so I upgrade busybox to latest master but the problem persist. Did anyone have this problem before?06:07
*** tiagogomes [~tiagogome@] has joined #baserock07:38
*** gallit [] has joined #baserock07:52
*** franred [] has joined #baserock07:55
*** ssam2 [] has joined #baserock08:19
*** jonathanmaw [] has joined #baserock08:32
richard_mawjjardon: that's a new one, don't we usually build linux-api-headers from baserock/build-essential?08:48
jjardonrichard_maw: yeah, I upgrade the version we build to the one in baserock/morph branch because I need a newer version and also to be in sync with the kernel we are building at the moment08:50
richard_mawaiui we don't want to sync the headers exactly to the current kernel version, since we need to be able to build userlands compatible with older kernels08:55
richard_mawthat, and the uapi change made building the headers require more tooling08:56
*** jjardon [sid723@gateway/web/] has quit [Ping timeout: 240 seconds]09:04
*** jjardon [sid723@gateway/web/] has joined #baserock09:04
ssam2I feel we need to work on a better way to specify splitting rules, in the long term10:05
ssam2I'm trying to come up with a way of saying "put /usr/lib/ruby/[version]/gems/* into -libs but put /usr/lib/ruby/[version]/gems/doc/*" into -doc"10:06
ssam2it's all possible with regexps, but the default rules are going to end up looking like a sea of complicated regular expressions that nobody will ever dare to touch10:06
ssam2might be easier if the order in which the rules were applied could be defined, but that'd probably make other things more complex10:07
ssam2also it'd be nice if I still had tea to drink instead of having drunk it10:08
liw-orcssam2, I know how to solve one of those things in about seven minutes!10:09
ssam2you've cracked time travel ?10:10
richard_mawssam2: back when I was trying to specify the split rule behaviour, we pondered adding `exclude: [ regular_expression ]` to the rules10:11
liw-orcno, but I make a new cuppa in seven minutes, I think10:11
richard_mawso that could cover not wanting docs to go in -libs10:11
ssam2richard_maw: that might be clearer than a regexp that excludes docs, indeed10:12
richard_mawthough if we provide a way to say we're completely overriding the defaults, you could specify the order yourself to do it10:12
ssam2well, this *is* the defaults :)10:13
richard_mawgenerally though, I'd like to move the defaults out of the morph code and into the definitions repository10:14
richard_mawmy suggested idea of having morphologies being able to inherit and override fields means we could move the complexity into the build-system morphologies10:15
ssam2yes, that's probably an improvement, although lots of work and probably some additional code complexity10:18
ssam2having settings like that be per-definitions-repo would make it practical to have a definitions repo "per OS"10:19
ssam2so for example we could have a different definitions repo for Android systems10:19
ssam2with different defaults for layouts, etc.10:20
ssam2might make sense, anyway10:20
richard_mawalso, build-system specific default splitting rules would eliminate the number of empty chunk artifacts10:20
ssam2I wonder if it could be as simple as this: have two sets of rules, one "specific" and one "catch-all"10:22
ssam2chunks could then add to the specific rules, which would override any catch-all rules10:23
*** dabukalam [] has joined #baserock11:13
*** franred [] has quit [Quit: Leaving]11:36
*** franred [] has joined #baserock11:37
*** ssam2 [] has quit [Ping timeout: 260 seconds]11:46
*** ssam2 [] has joined #baserock11:46
*** juergbi` [] has joined #baserock11:49
*** SotK_ [] has joined #baserock11:49
*** juergbi [] has quit [Ping timeout: 260 seconds]11:50
*** SotK [] has quit [Ping timeout: 260 seconds]11:50
franredin a configuration extension can I do chroot "$ROOT" su - "$username" -c "application_which_I_want_to_run_as_$username" ?11:51
franredIm trying to deploy a system with this as a configuration extension and Im getting: "ERROR: gerrit.configure failed with code 1: su: Cannot drop the controlling terminal"11:52
liw-orcthat might work11:53
liw-orcthere's interesting issues about dependcies (has the user been created?) but we'll solve them if you encounter them11:53
franredliw-orc, that is what is failing... but I don't know why11:53
franredyes the user was created before11:56
franredthis is what Im trying to do:
franreds/trying to do/doing/11:57
franredand morph deploy complains about it:11:58
franredERROR: gerrit.configure failed with code 1: su: Cannot drop the controlling terminal11:58
richard_mawwe also generally recommend against configuring systems with chroot, since it means that you can only deploy them from an architecture that can run the code11:58
liw-orcthe chroot probably doesn't have a fully populated /dev and that makes su unhappy11:58
liw-orcrichard_maw, good point11:59
franredrichard_maw, I don't expect to deploy this from a different architecture11:59
ssam2franred: does seem more and more like you should use Ansible to do this :) but I guess that'd delay having a working Gerrit instance11:59
richard_mawtrove.configure is currently an offender, I'd generally not accept any configuration extensions in morphlib that weren't architecture independent11:59
pedroalvarezssam2: hm.. for now he only has to run a command, but yes, if the configuration is going to be big as in trove-setup, I'd recommend the use of Ansible12:00
liw-orcfranred, you should be able to use the useradd command to add a user, without chroot; I don't know if you'd be able to run the gerrit jar, might be better to do that on first boot12:00
juergbi` is now known as juergbi12:01
franredummm, I think about this...12:01
ssam2useradd isn't idempotent, I think12:02
* franred goes for lunch12:02
ssam2I guess you can just ignore a certain exit code of useradd to achieve basically the same thing, though12:02
* franred tries adduser before but he hasn't any luck, so he copies something from vagrant configuration extension12:03
liw-orcfranred, adduser is a wrapper around useradd; useradd is a lower level tool that might work better12:03
jjardonrichard_maw: Seems linux-header compiled just fine after removing ERR from that script. AFAIK we are building something linux >= 3.15 but the headers are older than 3.8, any idea what other stuff need to be done for the uapi stuff?12:06
richard_mawjjardon: no, the only other issue I recall was that it required perl as a pure build dependency for splitting the headers, which caused issues for the cross-build command, since it prevented us building an appropriate kernel with morph12:08
richard_mawbut we decided that was out of scope anyway for initial bringup, since you also need custom deployment instruction12:08
jjardonrichard_maw: ok, let me try to finish the build of the whole system tonigh and I will let you know if it worked or not12:15
*** gallit [] has quit [Quit: Leaving]12:36
*** liw-orc [] has quit [Quit: Coyote finally caught me]13:00
*** liw-orc [] has joined #baserock13:03
franredpedroalvarez, ssam2, configuration shouldn't be big in terms of commands, just add an user and install gerrit using a custom config file.13:14
franredbut we will see13:15
franredrichard_maw, what do you suggest to do to replace my chroot command?13:15
richard_mawfranred: unless you have java on the host machine, either a systemd unit or an ansible script13:19
richard_mawbtw, your shell quoting of the command to run is incorrect13:19
richard_mawyou need to escape embedded interpolations13:20
richard_mawyour double quotes inside the command to run are wrong13:20
richard_mawthey work because you happen not to have spaces in any of the strings13:21
franredoh, I see, I will change it13:22
*** jonathanmaw [] has joined #baserock14:23
franredliw-orc, richard_maw, to use useradd for adding a new user to my system I need to use chroot so I suppose that I can not use it14:26
richard_mawdoes the useradd command in busybox support -R/--root?14:27
franredrichard_maw, useradd supports -R yes14:29
*** 17SAA4G6P [] has joined #baserock14:33
*** 17SAA4G6P [] has quit [Ping timeout: 244 seconds]14:33
richard_mawhm, I don't have a useradd in baserock systems I have to hand14:33
richard_mawif in --root mode it doesn't need to exec any binaries or load any libraries from the chroot, then we'd have a winner14:34
liw-orcI assume useradd is in the shadow chunk14:35
franredrichard_maw, odd I have it in my baserock system14:36
richard_mawyep, it comes from shadow14:36
richard_mawhm, -R makes it do the chroot, and if we built it with PAM it would need to dlopen libraries AIUI14:42
richard_mawcompletely independent of the libc using /etc/nsswtich.conf, which requires loading libraries14:42
richard_mawin a very specific set of circumstances it might work14:43
richard_mawbut in general, you need to make this happen on first boot14:43
franredrichard_maw, thanks, you give me a better idea about how to do it14:48
*** paulsherwood [] has quit [Read error: Connection reset by peer]16:03
*** benbrown1 [] has quit [Read error: Connection reset by peer]16:03
*** SotK_ [] has quit [Read error: Connection reset by peer]16:03
*** liw-orc [] has quit [Read error: Connection reset by peer]16:03
*** petefoth [] has quit [Write error: Connection reset by peer]16:03
*** petefoth [] has joined #baserock16:05
*** benbrown1 [] has joined #baserock16:05
*** SotK [] has joined #baserock16:05
*** paulsherwood [] has joined #baserock16:05
*** liw-orc [] has joined #baserock16:14
*** jonathanmaw [] has quit [Quit: Leaving]16:28
*** tiagogomes [~tiagogome@] has quit [Quit: Leaving]16:32
*** ssam2 [] has quit [Quit: Leaving]16:48
Kinnison_ is now known as Kinnison18:33
*** straycat [] has quit [Remote host closed the connection]19:03

Generated by 2.14.0 by Marius Gedminas - find it at!