IRC logs for #baserock for Wednesday, 2016-05-25

*** cosm has quit IRC00:35
*** cosm has joined #baserock00:48
*** gtristan has quit IRC07:04
*** franred has joined #baserock07:12
*** franred has quit IRC07:16
*** ctbruce has joined #baserock07:28
*** franred has joined #baserock07:28
*** gtristan has joined #baserock07:33
*** rdale has joined #baserock07:42
*** fay_ has joined #baserock07:59
*** toscalix has joined #baserock08:01
*** jonathanmaw has joined #baserock08:16
*** _longines has quit IRC08:32
*** edcragg has joined #baserock08:41
*** franred_ has joined #baserock08:54
*** franred has quit IRC08:58
*** ssam2 has joined #baserock09:13
*** ChanServ sets mode: +v ssam209:13
*** Zara has left #baserock09:16
paulsherwoodfolks, in discussions elsewhere i undertook to document some requirements for the kind of deterministic build tooling we are interested in...09:42
paulsherwoodssam2: you mentioned that there's something already on the wiki?09:42
paulsherwoodi don't think i've ever done a table in a wiki before...09:43
ssam2 also has some tables (but because there are links inside the cells, the ikiwiki markdown becomes *super* ugly)09:45
ssam2i'm dubious about 'new user can get started in minutes, not hours' -- i think "days, not months" is more realistic09:47
ssam2if there's a build tool where i could be productive within hours of finding it, we should definitely switch to that right away ;-)09:48
ssam2in fact, rather than commenting on the document, maybe i should just share what i came up with in a similar vein ...09:49
persiassam2: how do you imagine avoiding combinatorial explosion?  I imagined just automating validation so that humans did not care much, but see no way to avoid the underlying issue.10:24
ssam2actually... a perfectly good solution is managing it rather than avoiding it10:24
ssam2e.g. the approach Fedora Atomic currently takes where you build a filesystem tree from RPMs, then commit the exact tree to OSTree10:24
ssam2when you pull an upgrade, it's exactly the same set of bits for everyone because you pull a whole new tree10:25
ssam2so it avoids the combinatorial explosion you get from letting the package-manager do an upgrade one step at a time10:25
ssam2but actually that's not what I was talking about there...10:25
ssam2I was thinking more about the problem with BitBake where you can set variables anywhere and expand them anywhere else and override them in a third place10:25
persiaSure, removing the uncertainty of ordered incremental installs is critical.10:26
persiaOh, that is also bad.10:26
ssam2I should reword that requirement, given that even I managed to misinterpret it :-)10:26
persiaI was thinking of the problem of thousands of commits on thousands of projects of which one picks hundreds.10:27
ssam2right. i don't think there's any way around that10:27
persiaGetting the right set of projects is the first hard thing (as "dependencies" vary in application-specific ways), and selecting commits becomes the next hard problem.10:28
persiaOh well.  I was hoping you had a massive insight, hidden inside a single clause :)10:29
ssam2"copy what other people do" is the best I can offer :-)10:29
persiaI think I prefer the army-of-tobots-trying-everything-in-parallel approach.  Most of the humans being copied are also just guessing wildly.10:31
ssam2if you can find a robot (or tobot) that understands what I want the software to do, great10:33
ssam2beyond "execute without crashing", I think it's a lot of work to educate them on how to evaluate stuff10:33
persiaSome things robots can evaluate easily: does this combination compile?  Does it link?10:51
ssam2sure, that's the easy part10:52
persiaWith a bit more effort: are symbols resolved in a syntactically compatible way?  Are other classes of dependency resolved in a syntactically valid way?10:52
persiaYes, but the amount of human effort currently applied to deal with these is large.10:53
ssam2i updated my list a bit ...
persiaIf we further assert that any project should have a branch for users (maybe master, maybe current-release, whatever), then the robots can try all the commits in the branch to detect mist issues, and present the human with the solder job of behaviour specification and implementation selection.10:57
paulsherwoodssam2: i like your list10:57
persiaI also.  I think I have more opinions, but want to think a bit.10:59
ssam2feel free to evolve it. although I doubt we'll ever agree on one list of requirements to rule them all :-)11:02
paulsherwoodssam2: well, as a first pass, we'd have them in a git repo and accept pull requests11:03
rdalepoint 16. should be '..has the latest releases..' not have11:04
persiaI'd rather review merge candidates: it is difficult for multiple parties to collaborate on a pull request.11:07
SotKpersia: +111:08
ssam2rdale: ta11:08
ssam2make sure to update and to point at said Git repo instead of listing old requirements :-)11:09
paulsherwoodgtristan, ssam2 - any thoughts on christoph's 'image won't boot' using the ybd+aboriginal approach?11:28
* paulsherwood is excited someone else is trying it11:28
* gtristan reads emails...11:32
gtristanssam2, when you tried it... did you build with CROSS_COMPILER_HOST=x86_64 ?11:37
* gtristan is curious because he mostly tested with the i686 host...11:37
ssam2i think i did, yeah ...11:41
ssam2presumably that message means the kernel is totally broken11:42
gtristanit looks like kernel wont boot in qemu indeed11:43
* gtristan suspects that if this is a message from ./ or the aboriginal-controller scripts, that it may be a too old qemu ?11:47
paulsherwoodgtristan: maybe reply asking christoph to discuss on irc?11:51
gtristanpaulsherwood, already sent11:56
*** gtristan has quit IRC12:32
*** gtristan has joined #baserock12:43
*** cosm has quit IRC13:02
*** cosm has joined #baserock13:14
*** CTtpollard has quit IRC14:00
*** CTtpollard has joined #baserock14:08
*** anahuelamo_ has joined #baserock14:55
*** anahuelamo has quit IRC14:55
*** rdale has quit IRC15:08
*** rdale has joined #baserock15:15
*** franred_ has quit IRC15:38
*** jonathanmaw has quit IRC16:07
*** ctbruce has quit IRC16:24
*** janderJLR has joined #baserock16:41
*** fay_ has quit IRC16:48
*** toscalix has quit IRC16:57
*** edcragg has quit IRC17:07
*** rdale has quit IRC17:16
*** ssam2 has quit IRC18:05
*** janderJLR has quit IRC23:41

Generated by 2.15.3 by Marius Gedminas - find it at!