Fire is step THREE! | github.com/perl6/toolchain-bikeshed | Channel logs: irclog.perlgeek.de/perl6-toolchain/today | useful prior art: metacpan.org/pod/CPAN::Meta::Spec
Set by moderator on 23 March 2016.
02:48 ilbot3 joined
moderator Fire is step THREE! | github.com/perl6/toolchain-bikeshed | Channel logs: irclog.perlgeek.de/perl6-toolchain/today | useful prior art: metacpan.org/pod/CPAN::Meta::Spec
07:16 domidumont joined 07:21 domidumont joined 08:14 FROGGS joined 08:15 domidumont joined
stmuk_ bugs.debian.org/cgi-bin/bugreport....bug=818926 10:13
domidumont stmuk_: I've discussed this issue with nine yesterday. 10:40
nine domidumont: I don't know if I've mentioned this already. The tool-to-be-written is basically gist.github.com/niner/8ad4cbefde16d9494e16 10:46
domidumont nine: you did mention it, but I forgot the URL :-)
nine What's missing is only a build step for modules that need to compile some native library. That depends on us finding a replacement for Build.pm6 files that extend Panda::Builder
domidumont I'm going to test it for simple modules like JSON::Fast 10:47
15:20 stmuk joined
ugexe there are problems with the absoultifying of object. the relative path part is used as 1 of 2 parts, but it only counts the part after `resources/` into the hash 16:02
this means you can have 2 different files with the same id
MyModule/script/xxx, MyModule/resources/scripts/xxx (both hashed with the same root, and if in the same distribution the same Dist.id) 16:03
i worked some of that out but im not sure how well it will merge now 16:04
todays commits for instance seem like that would fit into what i discussed yesterday, the need for an `absolutifier` 16:08
there is lots of redundant stuff going on that should be rolled up into Distribution. for instance CompUnit.new(...ver => $dist<ver>, auth => $dist<auth>, distribution($dist)) 16:09
stmuk does "zef search" work? 16:11
ugexe stmuk: yes, but only exact searches right now 16:15
stmuk ah
ugexe if you --cpan you can use whatever elastic search understands though
stmuk BTW I would have assumed "zef fetch" to have populated the cwd rather than ~/.zef/tmp (or at least printed dest dir) 16:16
ugexe i cant say for sure, but i dont think it will if you disable the LocalCache content storage 16:17
--/cached would disable it
but otherwise it stashes it so if you do `zef fetch "XXX:ver<1>"` and already have it cached, you dont fetch it 16:18
the caching thing is a bit wonky though i agree
stmuk was impressed by uninstall! :) 16:19
ugexe luckily that feature is mostly built in to CURI now 16:22
nine: can i break method.install down into smaller parts? mostly to be `install --> installs source dist --> !install-precomp --> installs precomp (a method like this: gist.github.com/ugexe/c5a74134164bcaabd756) 16:26
let me rephrase: create a `method !precompile-distribution` thats called at the end of `method install`. the pros is it makes method install smaller as well as the possiblity of using a different precompile strategy based on the known topology of a Distribution 16:29
the cons are that piece of code is otherwise inside a lock. ie. calling self!precompile-distribution($myDist) could accidently be done outside the method install lock 16:30
i'd go as far as breaking each step into its own private method `!install-source, !install-bins, etc` leaving `.install` to mostly comments describing the process and using the higher level private methods 16:32
i can do these things, but im assuming im overlooking something 16:33
could probably do things like add-short-name in bulk instead of once per provides too 16:39
since the Distribution would provide all the neccesary info 16:40
16:46 Kassandry joined 17:18 domidumont joined
nine Sounds good! But I've made a lot of changes to install in the relocateable-precomp branch, so maybe it's better to hold off for now to avoid a massive refactoring headache 17:45
ugexe have you had a chance to read github.com/ipfs/specs/blob/master/...s/ipfs.pdf ? it seems to align with what we are doing (sections 3.5 onward) 17:50
nine I think I've read it once...looks awfully familiar 17:59
jdv79 relocatable precomp is a thing now? 18:02
nine jdv79: I merged the first commit today giving us more usefull backtraces and easing some of the packaging pain. 18:03
More to come once I've shaved this massive yak... 18:04
jdv79 i thought jnthn said it was really hard or something like that last i heard about it
wow
nine AFAIK he said that I probably knew better than him how hard it would be and indeed it was not that hard at all :) 18:05
The only absolute paths in precomp files were the source annotations 18:06
jdv79 oh, ok. maybe i'm remembering very outdated or imaginary utterances then . 18:13
[Coke] nine++
MadcapJake ugexe: wrt IPFS, did you see this github.com/whyrusleeping/gx ? 18:44
ugexe MadcapJake: yeah, ipfs seems to be a thing to talk about since the npm shit show the other day 18:55
MadcapJake right, that was my thought too 18:56
i'm honestly surprised that they just caved like that, I'd gander that the perl community wouldn't be so quick to fold but it'd be better to breed the option out completely! 18:59
once relocatable-precomp lands, setting up a gx-perl6 CLI should be really simple (was planning on tackling that sooner but I think I'll wait for that branch to land :P ) 19:00
ugexe whats being worked on will essentially be ipfs 19:03
MadcapJake really? Why reinvent an internet protocol? or is it just over the internet? 19:05
ugexe the protocol/network/filesystem layers are transparent 19:09
IPFS also stands for inter planetary file system, not internet protocol file system 19:11
19:29 FROGGS joined
MadcapJake ugexe: right, but it is described as a p2p hypermedia protocol 19:31
so essentially what you're saying is that Perl6 modules will be a Merkel DAG, right?
ugexe For the most part, yes. Distribution will consist of `method meta`, the data Distribution needs to create a merkel dag(possibly encoded in a file path and not listed in json somewhere), and a `method content($orig-relpath)` capable of translating $orig-relpath based on the dag into an IO::Handle (a common interface to read data from). So perl6 modules themselves are a merkel dag, but they still need 20:01
a Distribution object to imply how to generate the dag (Distribution::Hash -> explitily in the meta already, Distribution::Directory -> grep all files in <lib bin resources>>>.IO (maybe using $meta for hints), Distribution::Tar -> grep all files from run(<<tar --list $module.tar>>, :out).out.lines)
makes it trivial to slap a role onto any Distribution with `method content($xxx) { use HTTP::UserAgent; <do internet stuff>; return $io-handle-to-data; }` 20:07
21:33 MadcapJake joined 21:35 stmuk joined 21:37 pnu joined 22:13 stmuk_ joined

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!