IRC logs for #baserock for Thursday, 2015-08-20

*** zoli__ has quit IRC02:21
*** zoli__ has joined #baserock02:21
*** zoli__ has quit IRC04:13
*** zoli__ has joined #baserock04:15
*** zoli__ has quit IRC04:20
*** petefoth has joined #baserock05:30
*** petefoth has quit IRC05:42
*** paulw has joined #baserock06:24
*** mdunford has quit IRC06:57
*** mdunford has joined #baserock07:15
*** gary_perkins has quit IRC07:34
*** gary_perkins has joined #baserock07:39
*** mariaderidder has joined #baserock07:47
*** bashrc has joined #baserock08:03
*** mdunford has quit IRC08:13
*** mdunford has joined #baserock08:18
*** ssam2 has joined #baserock08:30
*** ChanServ sets mode: +v ssam208:30
*** bashrc has quit IRC08:49
*** bashrc has joined #baserock08:51
ssam2"NameError: name 'True' is not defined"08:52
ssam2that can't be good08:52
ssam2ha! it's because I managed to set __builtins__ to None08:52
ssam2Python certainly is a dynamic language.08:52
*** bashrc has quit IRC09:10
*** mariaderidder has quit IRC09:12
*** bashrc has joined #baserock09:12
*** mdunford has quit IRC09:23
*** mariaderidder has joined #baserock09:26
*** mariaderidder has quit IRC09:29
*** mariaderidder has joined #baserock09:29
*** mdunford has joined #baserock09:38
*** wschaller has joined #baserock10:17
*** mdunford has quit IRC10:38
paulsherwoodhas anyone here tried deploying a baserock system to aws?10:48
pedroalvarezi don't think so10:57
doffmDon't we need PV-GRUB to make aws images work?10:57
doffmI was under the impression we needed to add a modified grub bootloader.10:57
paulsherwooderk. maybe i'll not go that way, then10:58
pedroalvarezI remember someone (Kinnison?) telling me something like that, so you might be right10:58
*** mdunford has joined #baserock10:59
ssam2i guess you could build a normal VM image, boot it, then build and install PVGRUB inside that11:00
ssam2as a first step to getting *something* in AWS11:01
ssam2but we kind of have extlinux hardcoded in the deployment logic, current, which is far from ideal11:01
pedroalvarezhm.. although I believe there are some variables already in place to modify what bootloader you want to install11:02
pedroalvarezalthought there is only one choice atm11:02
*** petefoth has joined #baserock11:17
*** wschaller has quit IRC11:22
*** petefoth has quit IRC11:58
*** petefoth has joined #baserock12:02
*** wschaller has joined #baserock12:10
paulsherwoodnowster: thanks for the splitting pr12:20
paulsherwoodnowster: one thing... why have perl default cmds changed?12:20
nowsterthey haven't... they were missing continuation "\" characters12:22
nowsteroriginally there was a string concatenation, but this didn't get translated into the yaml form12:23
nowsterI blame ssam2 :)12:23
paulsherwoodi think i found a couple of things in the defaults yaml too12:24
paulsherwood(running on clusters/ci.morph)12:24
nowsterAh... is there a DEFAULTS file in your definitions now?12:26
paulsherwoodah, no12:27
paulsherwoodhold on12:27
nowsterah... a fix was missed in rebasing.12:28
paulsherwoodpls send the fix :)12:29
nowsteralready was doing12:29
paulsherwoodnowster: i'm assuming this changes meta format now?12:30
nowsterMeta format is pure yaml12:30
nowsterIt will fail to work correctly if you have existing cached assets12:31
paulsherwooderk. fail to work *at all*?12:32
* paulsherwood has blown away artifacts, but other users might not be so tolerant12:32
*** zoli__ has joined #baserock12:41
nowsterit warns, IIRC, then doesn't split12:46
nowster        app.log(name, 'WARNING: problem loading metadata', metafile)12:46
paulsherwoodyup, saw that12:47
paulsherwoodci.morph using 6 parallel instances of ybd on an aws machine (gits already available) 15-08-20 12:45:45 [303/661/661] [TOTAL] Elapsed time 01:50:5112:48
*** zoli__ has quit IRC12:50
*** franred has joined #baserock13:31
ssam2did you do that in a Baserock reference system or some other OS ?13:31
paulsherwoodit's on a standard AMI13:33
pedroalvarezthat reminds me: why didn't we push towards using baserock in docker?13:59
paulsherwoodi did, but no-one here seemed interested iirc14:00
paulsherwoodbut now you mention it, maybe that would be a way forward on aws14:00
pedroalvarezaha! here we go, you are the man behind baserock in
pedroalvarezI feel like I'm months late for this party, but I also think the party might not be over14:02
paulsherwoodagreed :)14:03
paulsherwoodi feel the same14:03
* paulsherwood notices that baserock has been pulled on docker a few times, even without him waving the flag14:03
pedroalvarezstraycat: patches for lorry merged, if we update the ref in definitions we can upgrade g.b.o in few minutes14:26
straycatpedroalvarez, that would be really cool thanks :)14:30
*** paulw has quit IRC14:34
*** paulw has joined #baserock14:34
paulsherwoodnowster: it seems to be building ok :)14:43
paulsherwoodnowster: have you compared any sizes vs morph results? are they the same?14:49
* paulsherwood vaguely remembers nowster saying something was different14:49
nowsterthe metadata will be different14:54
paulsherwoodbut all the other contents the same?14:54
nowster /etc/os-release  wasn't made14:54
nowsterand for some reason morph missed a file!14:55
paulsherwoodwhat is /etc/os-release please?14:56
*** paulw has quit IRC15:03
*** paulw has joined #baserock15:04
*** mdunford has quit IRC15:14
nowsterIt's a file that morph makes.15:16
pedroalvareznowster: what file missed morph?15:16
nowsterFor some reason nsswitch.conf didn't make it to the morph tarball15:17
* pedroalvarez notices that nsswitch.conf is in a separate artifact 'glibc-nss'15:19
nowster...which is included by build-essential-minimal15:24
paulsherwoodquite :)15:27
*** mdunford has joined #baserock15:28
nowsterjmacs: an artifact created by build-essential15:29
* pedroalvarez confirms that nsswitch.conf is not present in the minimal system15:31
nowsterIt is if built by ybd!15:33
jmacsThe "products" section is new to me, but apparently it's been in there a long time15:37
paulsherwoodjmacs: i ignored it for ybd til recently, but for tiny systems we need to remove lots of stuff15:38
paulsherwoodnowster: what's the size of one of the typical openwrt systems?15:39
jmacsThat's fine, my only concern was that "essential-minimal" is a tautology15:39
paulsherwoodi think build-essential is a misnomer, myself15:40
pedroalvarezwell, the nsswitch.conf is in the right chunk artifact, but not in the stratum artifact :/15:41
pedroalvarezit is in build-essential-runtime instead15:42
paulsherwoodybd's parallel instances are still tripping over each other occasionally :(
pedroalvarezok, I think I don't understand the syntax of stratum splitting15:46
paulsherwoodnowster: any idea what the gcc error might be? (bearing in mind that only one instance fails on it)15:51
paulsherwood([3/662/662] [stage1-gcc] Error:)15:52
pedroalvarezoh well, I have to say that ybd and morph are right when building build-essential-minimal15:52
pedroalvarezan that is the actual definition of build-essential what is wrong :)15:53
pedroalvarezglibc-nss is being included in b-e-minimal **and** in b-e-runtime. This creates some conflicts in the definition, and depending on the build tool you will have different results15:54
paulsherwoodi think that the syntax needs to be simplified15:54
nowsterpedroalvarez: right... with morph one will claim over the other... with ybd, both can claim it?15:54
ssam2I'm impressed that nobody spotted until now that the syntax actually allows you to specify the same thing in two different places15:55
nowsterpaulsherwood: You'd have to look in /src/artifacts/
nowsterah.. it's this:15:56
nowstermake[2]: x86_64-bootstrap-linux-gnu-ranlib: Command not found15:56
paulsherwoodnowster: the tail of that (200 lines) follows the error msg15:56
nowsterLooks like stage1-gcc doesn't have a build-depends on something15:57
paulsherwoodhow could that happen for one, but not an other?15:57
nowster(possibly binutils)15:57
nowsterpaulsherwood: build order randomization15:57
nowstermorph implicitly builds in order of definitions15:58
paulsherwoodthing is i've never seen this on a single instance... only parallel ones15:58
paulsherwoodi'm not discounting your theory, just a bit skeptical15:58
nowsterhmm... gcc BUILDS ranlib15:59
nowsterthat probably implies linker error in that case15:59
paulsherwoodwould the full log help?16:00
nowsterprobably not16:00
nowsterHmm.... 0 15-08-20 14:07:04 [1/662/662] [stage1-binutils] Bah! I raced and rebuilt stage1-binutils.dac85c48b260aa710b7910ab871bf5b5cce1e27910bf090b4b60bf2dc35fbbbb16:03
paulsherwoodnowster: that's expected16:03
nowsterand it means?16:03
paulsherwoodit means that 6 instances start, all of them try to build stage1-binutils, only one of them gets to publish the artifact16:04
paulsherwoodi think it's quite elegant16:05
*** jonathanmaw has quit IRC16:05
nowsterdo they get separate build trees?16:05
paulsherwoodi see 61 instances of racing, in 662 jobs it's not too bad. room for optimisation i'm sure, but certainly this leads to a faster overall result than the same machine running the jobs serially16:07
paulsherwoodso i don't *think* the problem is that, but it might be16:07
nowsterso... this['sandbox'] is unique to each instance?16:08
nowster(ie. one can't overwrite the entry in defs for another instance?)16:08
*** lmackenzie_2015 has joined #baserock16:17
*** mariaderidder has quit IRC16:22
*** paulw has quit IRC16:32
*** franred has quit IRC16:39
*** bashrc has quit IRC16:56
pedroalvarezHere there is a draft for a possible release this week:
pedroalvarezAll comments are welcome17:30
*** ssam2 has quit IRC17:40
*** wschaller has quit IRC18:02
*** lmackenzie_2015 has quit IRC18:17
SotKpedroalvarez: looks fine to me at a glance, though I've not been keeping up with changes so don't know if anything is missing19:13
*** persia_ has quit IRC19:28
*** radiofree has quit IRC21:17
*** radiofree has joined #baserock21:19

Generated by 2.15.3 by Marius Gedminas - find it at!