IRC logs for #baserock for Monday, 2015-11-02

*** gtristan has quit IRC01:01
*** gtristan has joined #baserock01:36
*** gtristan has quit IRC05:54
*** gtristan has joined #baserock06:51
petefothThe 'CIAT' page on the wiki has a link to which is not currently responding. (apologies of this is a;lready known)07:34
SotKpetefoth: I believe was turned off a couple of weeks ago since it was running on an expensive aws box07:47
petefothSotK: that would explain it :) I shan't mess with the wik page - I'll leave that to the discretion of toscalix who added the links07:53
*** fay_ has joined #baserock08:29
*** tiagogomes has joined #baserock08:40
*** fay_ has quit IRC08:46
*** fay_ has joined #baserock08:47
*** tiagogomes has quit IRC08:49
*** fay_ has quit IRC08:50
*** tiagogomes_ has joined #baserock08:52
*** bashrc has joined #baserock08:59
*** paulwaters_ has joined #baserock08:59
*** toscalix__ has joined #baserock09:07
pedroalvarezmorning all09:08
pedroalvarezinteresting proposal in baserock-dev09:12
tiagogomes_about? (still haven't catch up with email)09:13
petefoth'Axing the Stratum - Enter Runtime Dependencies and Build Flavors'09:15
*** mariaderidder has joined #baserock09:15
pedroalvarezI better don't spoil it :)09:16
pedroalvarezit is long, but interesting09:16
* petefoth is savihg it for lunchtime reading09:19
*** nowster has joined #baserock09:23
*** jonathanmaw has joined #baserock09:33
*** ssam2 has joined #baserock09:42
*** ChanServ sets mode: +v ssam209:42
*** fay_ has joined #baserock09:48
*** CTtpollard has joined #baserock10:02
*** edcragg has joined #baserock10:02
*** petefoth1ringham has joined #baserock10:05
*** petefotheringham has quit IRC10:08
pedroalvarezI had some time to play with docker this weekend, and I think it may be interesting for others, and for the project itself11:14
pedroalvarezI wrote something down, and I'm planning to add it to the wiki:
ssam2it's already there :-)
ssam2but quite out of date11:20
pedroalvarezyes, I took that information as base11:23
pedroalvarezbut I think that developing in your system, without entering in the container with an interactive session, was interesting11:23
ssam2ah, I see11:30
pedroalvarezI also ran morph tests, with no failures :)11:30
ssam2yeah, it's nice to have build-as-an-appliace11:31
ssam2it might be more conventional to provide a HTTP API for building, instead of requiring the user to run a shell command11:31
ssam2so, you run the container, send a HTTP request, it replies with a URL to an image11:32
pedroalvarezthat will be a really slow answer then11:32
ssam2sure. it needs to be a bit more complex :-)11:32
ssam2but it would be great to be able to provide baserock-builds-as-a-service11:32
pedroalvarezyes, my usecase was to make baserock tooling available to anyone that doesn't want to mess up with his system, and doesn't want to create VM's11:33
pedroalvarezbaserock-chroot is nice too11:33
ssam2hmm.. yeah I forget that we still recommend everyone to make a VM11:33
ssam2 probably puts quite a lot of people off11:33
ssam2i certainly wouldn't be arsed with it if I was wondering what Baserock was11:34
pedroalvarezdocker approach sounds simpler if you know what docker is, and if we hide this long docker call in a script.. this could be nice11:34
pedroalvarezalso, if we manage to create a docker image with only the dependencies needed, the docker image should be really small11:35
ssam2docker also works on OS X I believe11:36
ssam2with this hacky boot2docker thing11:36
ssam2wait, that's not it at all11:37
ssam2docker-machine is the thing you use. Which is basically a VM, but automated11:37
ssam2as in, the setup of itis automated11:38
ssam2*it is11:38
pedroalvarezand in windows!11:38
ssam2in windows too? awesome11:39
pedroalvarezbut yes, is a VM11:39
ssam2if it works well, i'd like to replace the current quick-start instructions with suggesting folk use Docker, then11:39
ssam2would be simpler both for us and for users11:39
pedroalvarezi might be able to investigate windows use-case11:42
pedroalvarezLooks like docker-machines automates the creation of VM's... I'm not sure it will be any easier to understand that you need  to install docker-machine, virtualbox, and then get a baserock docker image11:52
pedroalvarezI prefer to give up with windows users :P11:53
ssam2we can just point them to :
ssam2or whatever is the up-to-date doc11:57
straycatso the other day i found i can no longer git clone over https in baserock, it looks as though the certdata2pem script which is used by the ca-certificates chunk to install a bunch of certificates requires python2, but we now have python3 in core.13:20
richard_mawand it's overriding the /usr/bin/python symlink?13:21
straycatit's got #!/usr/bin/python at the top yes?13:22
straycatoh i see your question, iirc yes if there's no python2 installation then python3 will provide /usr/bin/python13:23
* richard_maw would consider that a bug13:23
richard_mawbut… python3 depends on core, which contains python213:24
straycatno core now contains python3?13:24
* richard_maw forgot to reset when he updated master13:24
straycatso i'm just not sure how best to go about fixing this, other than basically putting python2 back into core...13:25
straycatsince things in core want git, which wants curl, which wants ca-certs13:25
richard_mawoh, someone recently swapped out which python was installed by default13:26
persiaCan we do something with six or nine to work around the problem?13:26
* persia would think nine was created specifically for this sort of thing13:27
* richard_maw still thinks having /usr/bin/python → python3 is a bug13:27
persiarichard_maw: Why?  Some people want systems with no python213:28
richard_mawaye, but AFAICT everyone assumes that /usr/bin/python is python2, not python3, at which point I would prefer the more obvious error that the interpreter is missing, than it try to parse it as python313:29
persiaI know of contrary assumers, but I can see your argument, and agree that it is not an uncommon one.13:30
persiaThat said, I do think nine solves most of the problems with /usr/bin/python being python313:30
persiaBut that is because I'm misinformed.13:31
SotKrelated to this:
persiaIt is *six* that solves this problem.  nine solves the other problem (that of assuming /usr/bin/python is python3, and running on a python2 system)13:31
straycatif the solution is going to be to patch that file maybe we can just run 2to3 on it?13:35
persiaAsk upstream what sort of patch they like best.13:36
persiaSo if we do that one, they might adopt it, and then we don't have to redo things later.13:36
straycat"for the time being, all distributions should ensure that python refers to the same target as python2 . " (from SotK's link)13:37
straycatso probably we should also ensure that python points to python213:37
persiaExcept Arch doesn't already (also from SotK's link), so only dual-source compatible python should use /usr/bin/python: python2-only source should use /usr/bin/python213:37
persiaAnd at least some of the folk active in Baserock are known to use Arch, so we are unlikely to have consensus within our community, mirroring global opinion nicely.13:38
straycatmy interpretation of that pep is that arch is broken, hence the creation of the pep, but ok13:39
SotKstraycat: that was how I interpreted it too13:39
radiofreeis arch linux now the arbiter in how gnu/linux distributions should be laid out?13:40
persiaI interpret the rationale for the PEP that Arch was considered broken.  I interpret the text of the PEP as a set of advisories that allow people to ignore whether Arch is broken.13:40
* richard_maw asserts that since python3 is not a superset of python2 and python2 was the first to lay claim to the name /usr/bin/python, that /usr/bin/python can only be python213:40
persiaAnd I rather like the compromise of python3-only using /usr/bin/python3, python2-only using /usr/bin/python2, and dual-interpretable using /usr/bin/python13:41
straycatOkay, in any case I'll go play around with 2to3 and see if I can get our ca-certs back >.>13:42
* SotK likes that compromise too13:42
* richard_maw dislikes the compromise because it breaks existing programs13:42
straycatyeah, there is that13:42
persiastraycat: The easy fix is to add the character "2" to the file.13:42
straycatprior to python3 you could safely assume that python meant python213:42
straycatpersia, no because python2 is not in core13:43
persiarichard_maw: Yes, but given the agenda of the interpreter folk to get other authors to move to the new interpreter, it is least bad.13:43
persiastraycat: Ah, hrm.  Yes.  Maybe ask upstream then, as all other solutions are uglier.13:43
* straycat nods13:43
*** toscalix__ has quit IRC14:55
*** toscalix_ has joined #baserock14:55
pedroalvarezthis fix needs could do with more eyes on it:
straycatpersia, seems the upstream already tried to move to python3 but then reverted it
pedroalvarezlooks like it's incomplete15:13
persiastraycat: Excellent: now we know their preferred approach (and some of the pitfalls)15:14
pedroalvarezI have been wondering if anyone will miss gerrit's old change screen whenever we upgrade15:19
ssam2no luck flashing the device so far15:23
ssam2oops, wrong channel15:23
*** straycat_ has joined #baserock15:30
*** edcragg_ has joined #baserock15:30
*** edcragg has quit IRC15:35
*** straycat has quit IRC15:35
jjardonpedroalvarez: are we upgrading gerrit? yes! avatars!15:53
pedroalvarezI want to15:54
persiaCan we haz 2.11?15:54
persiaAnd  Storyboard upgrade too?15:54
pedroalvarezbut as I said, they remove the old change screen15:54
persiaThat even has the useful bugfix :)15:54
pedroalvarezwhich is?15:54
persiaThat is why we have gertty15:54
pedroalvarezoh, lovely15:55
persiaI think it was
pedroalvarezI want to: create a system with latest gerrit, import our database backup, migrate the database, and validate the result15:56
pedroalvarezpersia: I think I've never experienced that error15:57
persiapedroalvarez: It only happens when you upgrade a gerrit with lots of large repos.15:59
persia(or at least, that is what I heard)15:59
*** inara has quit IRC16:59
*** straycat_ is now known as straycat17:01
pedroalvarezContinue thinking about upgrading gerrit, I just realised that merging master of definitions on infrastructure.git will be a pain17:02
pedroalvarezmost of the systems will be broken, I'm sure17:02
*** mariaderidder has quit IRC17:04
*** gunnarx has joined #baserock17:06
*** gunnarx has quit IRC17:11
*** gunnarx has joined #baserock17:12
ssam2i would update to the last release, rather than master17:13
ssam2but yeah, lots of stratum rearrangement which leads to broken systems downstream due to lack of runtime deps ...17:13
jjardonHi,  What is the baserock way to add support to new ARM boards? is there a generic arm rootfs offered somewhere?17:14
radiofreethere used to be17:14
pedroalvarezfirst of all you need the board :)17:16
pedroalvarezis this a supported architecture?17:16
radiofreeany armv7 board is "supported architecture"17:16
rjekAny? :)17:16
* rjek <- pedant17:17
radiofreewe can massively aid the process of adding additional supported architectures if we provide a rootfs17:17
pedroalvarezmy question was mainly if this was an armv7 board17:17
rjekjjardon: What CPU/SoC is on the board in question?17:17
rjekI think the baseline we assume is ARMv7-A with VFPv3-D1617:18
pedroalvarezthis gerrit upgrade is already giving me a headache17:19
jjardonrjek: ifc6410: Snapdragon 600 APQ806417:20
pedroalvarezheh, I wanted to avoid updating infrastructure.git fork, but looks like their definitions are in version: 117:22
rjekOh, old17:22
radiofreetiagogomes_: the rootfs is tested in the same way as the jetson one17:23
rjekjjardon: ARMv7-A, VFPv417:23
rjekjjardon: Will probably work17:23
radiofreeit's essentially a jeston image without the bsp17:23
rjekI cna't find out if it's D16 or D32, but either way I think we target the lesser17:23
*** gunnarx has quit IRC17:27
*** edcragg_ has quit IRC17:27
jjardonrjek: pedroalvarez  so is there a rootfs somewhere to start working on this? or is there any "baserock way" Im not aware of?17:28
rjekjjardon: Unpack one of the images on ?17:29
* rjek doesn't actually know the answer17:29
radiofreei download a jetson image, mount it, copy out the rootfs (from /systems/factory/run), then boot that sd card17:29
*** edcragg has joined #baserock17:29
radiofreewhich isn't as easy as "extract this rootfs to an sdcard"17:29
rjekYeah, a .tar.gz of the rootfses for both x86-64 and armv7l would be nice17:30
rjekSo you can chroot into them easily, say17:30
radiofreeyes, a chroot is nice for the aarch64 stuff we have17:31
rjekWith qemu-user, you can chroot into ARM chroots on x86, which can be very handy17:32
radiofreei wouldn't want to use that to build though17:32
rjekNo, but it's handy for quick investigations17:32
radiofreeand running passwd17:33
rjekradiofree: passwd can be given the path to passwd/shadow to modify, IIRC17:33
*** jonathanmaw has quit IRC17:36
jjardonrjek, radiofree; ok, thanks17:50
radiofreebasically the current situation is less than ideal17:51
ssam2completely agree, the fact that releases still take someone a day to do also sucks17:57
ssam2adding an arm rootfs to the mix wouldn't make the release process significantly worse17:58
rjekWhen was the last release made, OOI?17:58
radiofreeshouldn't add anything to the build time either?17:58
radiofreeit's just the jetson image without a bsp17:58
rjekCould probably be done on itself to save uploading 400MB twice17:58
radiofreeyes it would be trivial to write a script to do it from a jetson.img17:59
rjekOr simply put an script on for people to use18:00
radiofreethat's still more of a ballache than just providing a rootfs18:01
radiofreeyou couldn't do it in windows for example18:02
rjekIs a tar.gz much use on Windows?18:02
rjekYou won't be able to extract it to anything useful18:02
radiofree7zip can open it18:02
radiofreeoh yeah18:02
rjekYeah, but not extract it losslessly18:02
rjekNo dev nodes or hardlinks18:02
radiofreealso extracting it to a non-vfat sd card18:03
* rjek nods18:03
*** bashrc has quit IRC18:03
ssam2rjek: last full release was 15.34 i think -- week 34 201518:07
*** nowster has quit IRC18:07
ssam2and yes, build time would be the same adding a chroot, just more moving parts to remember about18:08
*** toscalix_ has quit IRC18:12
*** ssam2 has quit IRC18:13
*** edcragg has quit IRC18:16
jjardonabout the rootfs discussion, Ive sent this in case wants to take a look:
*** toscalix_ has joined #baserock21:09
persiaRegarding rootfs: in my imagination, we have a simple way to derive a system with an alternate kernel, and a simple way to include a blob, such that instead of using a rootfs with other tools, a developer simply grabs the blob kernel, and generates a system using Baserock.21:28
persiaWhen that works, the developer then tries with some known hah of some kernel tree, and slowly moves towards complete alignment with upstream.21:29
*** gunnarx has joined #baserock21:32
*** gunnarx has quit IRC21:32
*** gunnarx has joined #baserock21:32
*** inara has joined #baserock21:55
*** CTtpollard has quit IRC22:32
*** flatmush has quit IRC22:32
*** tiagogomes_ has quit IRC22:32
*** paulwaters_ has quit IRC22:32
*** fay_ has quit IRC22:32
*** petefoth has quit IRC22:32
*** gary_perkins_ has quit IRC22:33
*** paulwaters_ has joined #baserock22:33
*** tiagogomes__ has joined #baserock22:33
*** CTtpollard has joined #baserock22:33
*** fay_ has joined #baserock22:33
*** gary_perkins has joined #baserock22:34
*** petefoth has joined #baserock22:34
*** flatmush has joined #baserock22:34
*** gunnarx has quit IRC22:37
*** toscalix_ has quit IRC22:45
*** gunnarx has joined #baserock22:56
*** gunnarx has quit IRC22:56
*** gunnarx has joined #baserock22:56
*** gunnarx has quit IRC23:05

Generated by 2.15.3 by Marius Gedminas - find it at!