IRC logs for #baserock for Monday, 2015-06-15

*** zoli__ has joined #baserock03:49
*** zoli___ has joined #baserock04:15
*** zoli__ has quit IRC04:18
*** petefoth has quit IRC05:03
*** zoli___ has quit IRC06:14
*** petefoth has joined #baserock06:23
*** straycat has joined #baserock06:38
*** petefoth_ has joined #baserock06:45
*** petefoth has quit IRC06:47
*** petefoth_ is now known as petefoth06:47
*** zoli__ has joined #baserock07:02
*** gary_perkins has joined #baserock07:23
*** gary_perkins has quit IRC07:25
*** rdale has joined #baserock07:26
*** gary_perkins has joined #baserock07:26
*** mariaderidder has joined #baserock07:42
*** ssam2 has joined #baserock08:06
*** ChanServ sets mode: +v ssam208:06
*** mdunford has joined #baserock08:08
*** gary_perkins has quit IRC08:11
*** tiagogomes has joined #baserock08:12
*** jonathanmaw has joined #baserock08:31
*** gary_perkins has joined #baserock08:51
*** lachlanmackenzie has joined #baserock09:22
* SotK wonders why the x86_32 build system for 15.19 on seems to have "devel-system-64b" as its hostname when booted :)09:23
franredyep :)09:24
franredSotK, are you going to look at the x86_32 bit error?09:25
SotKfranred: yeah :)09:25
*** nowster has joined #baserock09:27
franredSotK, I was about to test if with --disable-sse would compile09:27
SotKfranred: well I've not even got to the point of being ready to try a build, so maybe it will be better for you to test if you have the time :)09:29
franredSotK, ok, I'll give it a go09:30
* franred wonders why we try to build a 64bits system as OpenStack is in 32 bits - why does mason32 build the ci.morph cluster where systems with 32 and 64 bits are described?09:42
SotKfranred: what do you mean?09:42
radiofreei thought the 32bit mason was an entirely different vm?09:43
franredradiofree, it is a different VM but some of the systems which mason32 is trying to build are defined as 64 bits09:44
franredSotK, ^^09:44
SotKthat's really weird... what's it trying to build thats 64 bit?09:44
Zara(+1000 to changes 577-579 in principle! I'm avoiding reviewing formally since I don't know morph well enough to know what might get broken.)09:44
franred <-- build_cluster morphology ci.morph09:44
radiofreefranred: well the 64-bit one uses the same cluster, and is smart enough to only build the 64-bit systems09:45
radiofreemaybe mason skips the ones where the arch doesn't match?09:45
franredumm, my bad09:46
radiofreethere's also an arm image in that cluster for example09:46
franredis trying to build swift which is added to systems/devel-system-x86_32-generic.morph09:47
franredand defo mason32 builds it09:47
franredradiofree, SotK, sorry for the noise09:47
ssam2Zara: can you test them and see if they work as described? If so, that's useful info and it makes sense to give a +109:51
ssam2Zara: I sometimes regret that we call this process 'review', since that seems to imply that only line-by-line inspection of the code is useful09:52
Zarassam2: okay, will try them out! :)09:52
ssam2I've seen cases where a patch got inspected, merged, and then it turned out to be broken in some way because nobody actually tested it ;)09:53
ssam2(i'm pretty sure it was my fault in some of those cases, too)09:53
*** mdunford has quit IRC10:04
mwilliams_ctI have a patch (850) with two +1s. is it possible for a +2er to merge or is it still under testing?10:09
*** zoli__ has quit IRC10:13
richard_mawmwilliams_ct: theoretically. If something's had two +1s I have a quick glance in case there's something I see that they've missed, +2 then merge.10:17
SotKDoes anyone know what could cause this?
SotK(on a jetson)10:18
*** mdunford has joined #baserock10:18
mwilliams_ctrichard_maw: well in that case if any +2ers have some time, a quick glance would be appreciated but there's no rush :)10:19
richard_mawSotK: That's a new one.10:19
richard_mawmwilliams_ct: after I've sent a mail to paulsherwood about schemas I'll take a look.10:20
mwilliams_ctthanks richard_maw10:20
richard_mawI had a quick glance earlier, and I'm not sure about your use of IFS, so I'll have to look that up.10:20
mwilliams_ctI'm not sure about it either so any opinions off a shell magi are appreciated10:21
richard_mawtbh I'd shell out to python instead of attempting to parse it in shell10:22
*** zoli__ has joined #baserock10:35
richard_mawmwilliams_ct: I'm 99% sure that IFS doesn't do what I think you intend it to do.10:35
mwilliams_ct1%? good enough for jazz...10:37
mwilliams_ctrichard_maw: it's based roughly on
kejiahu_paulsherwood: openwrt built on armv7 by YBD, and it seems to be functional. But as YBD doesn't do artifact splitting, the resulting image is 10 times larger compare with the morph build10:39
richard_mawmwilliams_ct: I'm not sure where you got the idea that IFS contains a regular expression.10:39
mwilliams_ctrichard_maw: trial and error approach. if it's wrong, I'd much rather fix it ofc. would it be better to comment on the patch so folks playing along at home can join in?10:41
richard_mawmwilliams_ct: I'm in the process of doing so10:42
*** paulw has quit IRC10:42
mwilliams_ctcheers :)10:42
*** paulw has joined #baserock10:43
franredSotK, <-- patch for 32 bit mason - I've compiled Openstack 64 bits with it and it compiles fine10:44
SotKfranred: brilliant, thanks10:45
mwilliams_ctrichard_maw: thanks for taking the time to look through, will fixup. I've added your point about shell variable names to the policies page on w.b.o, though perhaps it isn't appropriate there. might be worth creating a coding standards page at some point10:57
richard_mawpossibly, we haven't really defined a style guide, and generally rely on following the conventions of the rest of the script10:58
richard_mawwe've generally avoided a style guide for shell as we couldn't find anything to crib off, as the google one is for bash and advocates bashisms where we wouldn't be able to use them10:59
* ssam2 would prefer our policy on shell code to be: 'Please use a language that is optimised for readability, instead of one that is optimised for being convenient to type'11:02
mwilliams_ctssam2: I am in no position to discuss that, having not worked on Baserock that long. please feel free to amend, modify or destroy my wiki edit, just trying to be useful :)11:04
nowsterAre there any jbd resources (eg. wiki pages) out there?11:13
SotKnowster: s/jbd/ybd/ ?11:15
SotKnot that I am aware of11:15
SotKI think the only documentation is in in its git repo11:16
nowsterok... and that's on
nowsterah, right.11:18
petefothAnd it's nothing to do with
* SotK is reminded of
*** zoli___ has joined #baserock11:48
*** zoli__ has quit IRC11:51
richard_mawssam2: certainly, the first entry for the style guide should be "If you have an alternative, consider using it."11:54
jmacsI have patch to a system which I don't think will ever be upstreamed; is it acceptable to carry an 11-line patch around in definitions?12:35
richard_mawjmacs: better than many alternatives. Generally we prefer to create a branch with the patch applied, but having the patch in-line in the chunk definition is manageable if it's as short as that12:44
jmacsOK, thanks.12:44
straycatrichard_maw, are you likely to look at anytime soon, it's got 2 +1s and i'd like to a merge it!13:06
franredSotK, radiofree, <-- new version13:06
SotKfranred: thanks!13:06
richard_mawstraycat: I'll try to find some time today to look over it.13:08
straycatrichard_maw, ack, will hold off merge in that case13:08
*** edcragg has joined #baserock13:17
richard_mawstraycat: I didn't find anything I didn't like on my first pass. I'm just going through to see if there's any code-level improvements I can think of. Design wise it looks fine, though I'm not sure about the hinted suggestion that we should make the source resolver read files directly out of the git checkout, but that's a different issue.13:36
ssam2if it works and there's nothing hugely wrong, it seems better13:38
ssam2+ to merge it, then do improvements iteratively13:38
richard_mawfine by me, you can look up my comments after merging if you want13:39
franred <-- Im going to merge this if no one has any complaint about it in the next hour (this fixes mason 32 bits)13:44
ssam2I was about to modify ybd to not put elapsed_time in the build log, then noticed Paul already did it last Thursday :)13:47
ssam2in the artifact metadata rather13:47
paulsherwoodssam2: actually, i did it saturday, but time stands still on my vm :)13:53
ssam2that explains why I didn't spot it on friday14:01
ssam2thanks, anyway14:01
richard_mawssam2: I have reviewed, you may want to address my review of DefinitionsRepo.relative_path14:03
ssam2thanks, will hopefully get a chance to look at it this week14:04
richard_maweverything else looks fine to me14:13
richard_mawstraycat: I've finished reviewing if you want to merge it now. I've not +2d as I'm mostly relying on the judgement of other reviewers, but I wouldn't object to being the one to +2 it if you can't or don't want to.14:15
straycatokay, though i wonder if it's worth fixing path_is_outside_repo before merging?14:20
richard_mawI say that's up to ssam2. The cwd thing won't bite anyone for a while as we'd need someone to try and use it in a new way14:22
straycatoh, heh, i just noticed something :/14:26
richard_mawoh, actually I missed something important in the call to path_is_outside_repo14:27
richard_mawit's after a relpath call, so it's already normalised14:27
*** fay_ has quit IRC14:36
franred --> merged14:38
ssam2richard_maw: good point about path_is_outside_repo(), I guess it needs a comment to make that clear14:42
straycatssam2, i've fixed the issues i just commented on, just running tests14:43
ssam2oh, thanks!14:44
*** fay_ has joined #baserock15:05
jmacsI'm a bit stuck with this patch issue: I've made this extra command in configure-commands:
jmacsAnd I've realised my mistake immediately. Thanks.15:14
richard_mawescaped EOF?15:14
jmacsrichard_maw: s/path/patch/15:15
richard_mawprobably a good idea to use <<'EOF' too, so your shell doesn't try to evaluate $(mkinstalldirs) when making the patch file15:17
jmacsGood plan.15:17
richard_mawoh, for the interested parties, it's just been announced that systemd intends to relese 221 this week15:25
* jmacs sighs, no 'patch' in pre-gcc world either15:26
richard_mawhrm, perhaps we should re-enable busybox's patch implementation15:30
richard_mawit doesn't do fuzzy diffs like the GNU version, but if we've got exact patches to apply this it's better than nothing15:30
jmacsOr I could just keep the patched version in another branch15:33
* SotK notices both Masons are failing15:33
jmacsIt's just that that approach isn't so easy when you don't have push access to git.baserock.org15:34
SotK seems pretty weird15:35
jmacsWe just merged a patch for gf-complete, didn't we?15:36
SotKyeah, and the patch should have worked15:37
SotKthe log looks like its morph doing something weird15:37
* SotK notes that mason-x86-64 is failing in a totally different way, failing to transfer the build-system artifact15:40
SotKs/the build-system artifact/some system artifacts/15:41
jmacsrichard_maw: I've tried putting patch back into stage2-busybox. I don't really get why it was removed though; if there's no other patch, what difference does it make?15:51
rdalei don't think we should remove awk either15:53
ssam2jmacs: would it help if we gave you push access to I think you're doing good work around here.15:55
jmacsRight now I'd rather think about how we make this process easier for the general contributor than giving me special privileges15:56
jmacsThanks for the offer, thogh15:56
ssam2ok. with regards to 'patch' in Busybox, it was disabled for a reason15:57
ssam2I think it's because something depends on GNU Patch, and because we have separate /bin and /usr/bin still, if busybox patch is present in /bin, it gets used in preference to GNU patch, and breaks whatever that thing was15:58
ssam2so, it's a fixable issue, but not trivially15:58
* straycat isn't sure why general contributors don't set up their own trove15:58
straycatarguably the resource requirements are a little high15:58
jmacsstraycat: On what?15:59
straycatbut other than that having your own trove can be quite liberating15:59
ssam2is where Busybox patch was disabled15:59
radiofreestraycat: how much RAM does a trove require?16:00
straycatdepends on the number of lorry-controller minions you have running, but given a default of 4 then I guess you'd need an upper bound of 2GB per minion so that would be 8GB16:01
jmacsssam2: But there is no GNU patch in the staging chroot which gcc gets built in. Are there things outside build-essential that use stage2-busybox?16:01
jmacs(Definitions says no)16:01
jmacsssam2: That patch applies to "proper" busybox, and it makes sense there, but it was disabled in stage2-busybox too with a similar comment.16:03
jmacsEither way, my copy of gcc seems to have built successfully with busybox's patch.16:05
ssam2jmacs: the stage2-busybox config is just a clone of the busybox config, I think16:09
ssam2so I think you could reasonably enable patch in stage2-busybox16:09
paulsherwoodjmacs: don't suppose you'd be interested to try toybox instead, while you're there? :)16:10
richard_mawpaulsherwood: let's not creep the scope16:11
jmacspaulsherwood: I'm happy to give it a go, but it does take about an hour to do builds since everything depends on busybox16:13
paulsherwoodjmacs: understood. cloud vms are the way to go :-)16:16
jmacsSure. As I said to ripsum though, bear in mind some contributors might not be sponsored. My personal cloud VM would be much slower at building baserock.16:18
paulsherwoodjmacs: i think baserock upstream may be able to offer you some time on vms? x86 or arm?16:24
jmacsx86 would be my preference for now16:25
*** nowster has quit IRC16:25
*** jonathanmaw has quit IRC16:30
*** franred has quit IRC16:32
*** nowster has joined #baserock16:36
*** gary_perkins has quit IRC16:37
*** mariaderidder has quit IRC16:39
paulsherwoodpedroalvarez, ssam2 ^^ ?16:47
KinnisonI think pedroalvarez is on holiday16:47
ssam2we have a devel system at Datacentred which you could have access to16:56
ssam2it's not amazingly well isolated from other systems there, but i trust you not to vandalise anything :)16:56
ssam2Adam will be using it to do a release tomorrow though, so it may not be much faster16:57
ssam2we could also instantiate a fresh VM for you if our sponsors agree that it's worth the extra cost :)16:57
paulsherwoodssam2: i think i'm able to speak for a sponsor? +1 from me16:59[0:]: errno=Connection refused17:02
*** bashrc has quit IRC17:03
ssam2in what context? seems fine from here17:04
nowsterssam2: sandboxlib doesn't seem to want to install cleanly17:05
paulsherwoodmay be me, i guess17:05
ssam2nowster: how so?17:05
nowsterI get an /usr/lib/python2.7/site-packages/UNKNOWN-0.0.0-py2.7.egg17:05
ssam2using './ install' ?17:05
ssam2do you have 'pbr' installed?17:05
nowsteryes, and I think so17:05
ssam2jmacs:  if you email your public key to then I will set up an x86_64 devel system for you in the cloud, then17:06
ssam2may have to wait til tomorrow though17:06
nowstersite-packages/pbr-1.0-py2.7.egg-info/  exists, as does the site-packages/pbr/ dir17:06
ssam2right. I don't know then17:06
nowsterhowever, /usr/lib/python2.7/distutils/ UserWarning: Unknown distribution option: 'pbr'17:07
ssam2that sounds suspicious17:07
jmacsssam2: OK, that sounds great. Thanks!17:08
nowster"import pbr" works in python.17:08
ssam2I get 'usr/lib/python2.7/site-packages/sandboxlib-0.3.0.post6-py2.7.egg-info' from 'python ./ install' with commit 776019ae8df349cb35fdeb2c37e2e8561bcad31b of git://
nowstersame commit here17:09
nowsterpbr is 1.017:20
nowstersetuptools is 7.0dev17:20
*** mdunford has quit IRC17:30
*** ssam2 has quit IRC17:52
nowsterok... pbr seems to want to hook into setuptools before it calls distutils.setup()18:14
nowsterand somehow the hooks aren't getting run18:15
paulsherwoodthis is just proof of the difficulty of installing 'packages' i guess :/18:20
paulsherwoodnowster: what's your actual error?18:21
paulsherwoodjjardon: do you now much about gnome-builder?18:48
*** lachlanmackenzie has quit IRC18:58
*** zoli___ has quit IRC19:31
*** zoli__ has joined #baserock19:34
*** zoli__ has quit IRC19:59
*** edcragg has quit IRC23:34

Generated by 2.15.3 by Marius Gedminas - find it at!