IRC logs for #baserock for Monday, 2016-02-01

*** gww has joined #baserock02:13
*** juergbi` has joined #baserock05:48
*** wdutch has quit IRC05:49
*** juergbi has quit IRC05:49
*** wdutch has joined #baserock05:56
*** franred has joined #baserock06:09
*** ctbruce has joined #baserock08:49
*** paulw has joined #baserock08:51
*** bashrc has joined #baserock09:03
*** toscalix has joined #baserock09:10
*** tiagogomes has joined #baserock09:43
*** jonathanmaw has joined #baserock10:02
*** zoli_ has quit IRC10:45
*** bfletcher has quit IRC10:45
*** locallycompact has joined #baserock10:55
*** rdale has joined #baserock11:25
*** locallycompact has quit IRC11:36
*** rdale has quit IRC11:43
*** rdale has joined #baserock11:48
*** zoli_ has joined #baserock11:57
*** bfletcher has joined #baserock12:03
*** rdale has quit IRC12:31
*** toscalix_ has joined #baserock15:00
*** toscalix has quit IRC15:02
*** toscalix_ has quit IRC15:31
*** gtristan has joined #baserock15:41
*** toscalix_ has joined #baserock16:01
*** toscalix_ has quit IRC16:18
*** toscalix_ has joined #baserock16:18
*** franred has quit IRC16:59
*** ctbruce has quit IRC17:18
*** toscalix_ is now known as toscalix17:38
*** jonathanmaw has quit IRC17:50
*** bashrc has quit IRC18:01
*** toscalix has quit IRC18:24
*** rdale has joined #baserock18:41
*** nowster has quit IRC18:49
*** juergbi` is now known as juergbi19:13
paulsherwoodso, transitive dependencies...19:24
paulsherwoodif i try to build asciidoc from definitions, i believe the dep tree includes both python 2.7 and python 3.419:25
*** rdale has quit IRC19:26
paulsherwoodso it appears that depending on what order the dependencies are installed in the sandbox, asciidoc fails or succeeds19:26
paulsherwoodmy head is hurting too hard to understand whether an order is implied in the yaml or not19:28
paulsherwoodbut in other contexts i've shuffled orders without a problem19:29
persiaThe fix is probably to tell asciidoc to use a specific python, rather than /use/bin/python or `env python` (but I am guessing based on similar issues, without investigating)19:40
paulsherwoodack19:41
persiaThe common problem being that often the most recently installed python claims /usr/bin/python19:41
persiaAnd `env python` usually finds that first in the path19:41
paulsherwoodyes, that's what's happening. what i'm unclear about is whether the order should be made deterministic, or assumed to not overlap19:43
* paulsherwood prefers that order is not deterministic, since randomizing has led to performance improvments and exposed some assumptions19:43
paulsherwoodbut others differ in this, i'm sure19:44
persiaNot indicating a syntax version is a bug in asciidoc, if it cannot use either python2 or python3.  If it can, then some set of dependencies is insufficiently precise about needing something built with some specific version of python.19:46
paulsherwoodhttp://paste.baserock.org/dohecusayi19:47
paulsherwoodthat's what happens with python319:48
paulsherwood(i believe)19:48
persiaActually, my memory of definitions (an older format now) is that the toolchain version was not captured in the archive name, such that if one builds a system with two options for some toolchain (python2+python3 or glibc+musl), there us no way to say that one depends on some library built with one or another if them.19:48
persiaThat error makes me think asciidoc does not work with python3, and so needs to specify /usr/bin/python219:50
persiaBut I may be wrong.19:50
paulsherwoodaccording to https://github.com/asciidoc/asciidoc/issues/8319:51
paulsherwoodit requires python219:51
persiaRight, so change the code to use the version-spevific interpreter for the quick fix (just an adjustment to the !# line)19:55
paulsherwoodand submit as a patch to upstream?19:55
persiaThat should work in an environment with both pythons unless asciidoc depends on a library that is not built against python219:55
persiaYes.  From issue 83, upstream would clearly prefer a python3 port patch, but the workaround should work for today.19:56
paulsherwoodhasn't been a release for a long time... we're using last tag, from 201319:57
persiahttps://sources.debian.net/src/asciidoc/8.6.9-3/debian/patches/code-filter.py-python-path.dpatch is the relevant patch in Debian, but that assumes default-python2.  It may be a sensible start for your patch.19:59
persiaThere are a number of other useful patches in Debian you may want, as long as you are branching anyway.20:00
persiaIssue 83 seems to imply that the long-term plan is to abandon and deprecate asciidoc, which is not promising in terms of landing the change upstream.20:01
paulsherwoodhmmm... # nasm wants asciidoc and xmlto for generating its docs20:30
persiaAs python libraries or shell utilities?  If the latter, that should be unaffected by the patch (and if asciidoc is replaced, easy to patch in nasm and xmlto)20:46
paulsherwoodi was quoting from definitions. anyway, after reordering my code, problem goes away :)21:34
persiaYou are hiding it :)21:37
persiaThe bug remains (or there would not be the Debian patch I referenced)21:38
persiaThat said: glad to hear your yak is trimmed enough to move again21:39
paulsherwood:)21:40
*** fay_ has joined #baserock22:13

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!