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 25 January 2016.
04:19 ranguard joined, japhb joined 04:21 sivoais joined 04:31 sivoais joined 04:41 sivoais joined 04:52 sivoais joined 05:02 sivoais joined 05:31 sivoais joined 05:41 sivoais joined 06:23 sivoais joined 06:33 sivoais joined 06:51 sivoais joined 07:01 sivoais joined 07:05 domidumont joined 07:09 domidumont joined 07:11 sivoais joined 07:21 sivoais joined 07:31 sivoais joined 07:41 sivoais joined 07:51 sivoais joined 08:01 sivoais joined 08:11 sivoais joined 08:21 sivoais joined 08:32 sivoais joined 08:44 sivoais joined 08:56 sivoais joined 09:09 sivoais joined 09:22 sivoais joined 09:34 sivoais joined 09:46 sivoais joined 09:58 sivoais joined 10:10 sivoais joined 10:22 sivoais joined 10:34 sivoais joined 10:47 sivoais joined 10:59 sivoais joined 11:11 sivoais joined 11:25 sivoais joined 11:37 sivoais joined 11:51 sivoais joined 12:03 sivoais joined 12:15 sivoais joined 12:27 sivoais joined 12:39 sivoais joined 12:52 sivoais joined 13:06 sivoais joined 13:18 sivoais joined 13:31 sivoais joined 13:43 sivoais joined 13:55 sivoais joined 14:04 FROGGS joined 14:09 sivoais joined 14:19 sivoais joined 16:37 FROGGS joined 17:31 domidumont joined 17:36 cognominal joined
jdv79 ugexe: that trick seems to not have worked either. 18:36
:(
ugexe jdv79: where did you put it?
jdv79 - system( 'panda', '--force', 'install', $dist_dir );
+ system( 'perl6', '-I', 'FakeDir', '/home/jdv/rakudo/install/share/perl6/site/bin/panda', '--force', 'install', $dist_dir );
oops. anyway, is that wrong? 18:37
ugexe ah yeah thats where i was going to suggest :(
jdv79 hmm. ok..
ugexe if its the install the failing maybe it needs to go inside panda itself
the `run $prove-command, "-IFakeDir", $test-file` 18:38
jdv79 is there a PERL6LIB?
ugexe its set in panda somewhere 18:39
jdv79 o 18:40
ugexe do you happen to know what distributions it happens with? 18:42
jdv79 hack seems cranky 18:52
having trouble doing anything atm
i don't think its the same one 18:53
it seems to chug along for a bit and then lock up on a new one each time. i think.
this time it happened to be panda itself 18:55
yeah, panda didn't pass it along 18:57
19:10 Cheery joined
Cheery hi 19:14
www.reddit.com/r/ProgrammingLangua...g_modules/ 19:39
19:44 Kassandry joined
jdv79 ugexe: any idea about where that's happening? 19:46
paste.scsys.co.uk/505098
ugexe jdv79: hmm where what part is happening? 19:52
jdv79 where that second proc is being spawned
its the same thing minus the -I FakeDir and subbing in a hashed path of panda 19:53
ugexe github.com/tadzik/panda/blob/b0710...ter.pm#L22 19:54
jdv79 oh, i was looking no 19:55
no
that's not what spawned the 2nd proc in my paste 19:56
ugexe oh i thought you were asking where one might be getting spawned. lemme look at the paste again
i think thats just the bin wrapper 19:59
as in, if bin/panda is the wrapper then it just launches the precompiled version of panda which is resourecs/<sha1>
jdv79 ah 20:00
20:01 sivoais joined
ugexe github.com/rakudo/rakudo/blob/nom/...ion.pm#L71 20:01
thats where it gets launched
jdv79 thanks 20:02
20:11 sivoais joined 20:15 raiph joined 20:21 sivoais joined 20:32 sivoais joined 20:39 diakopte1 joined 20:42 sivoais joined 20:48 ribasushi joined 20:53 sivoais joined
jdv79 i just saw the lock queue go from 10 to 2. is that a dep chain? this stuff is bizarre. 21:15
i killed some procs. seems to be ok now. maybe its less common that i thought. we'll see. it seems to be chugging along a lot better now. 21:41
i still don't totally get it but whatever 21:42
*than
ugexe hack.p6c.org:5000/v0/release/_searc...a-Selector isn't loading anymore 21:52
jdv79: ^
jdv79 yeah, its weird. something is binding up. es? 21:53
ugexe it seemed to be working until like 10 minutes ago 21:54
jdv79 oh. jeez. its just plackup and there's a req locked up waiting for the cur lock to gen docs
maybe its time to build it out a bit more
so we really need locking? :)
*do
it worked for me now 21:55
ugexe yeah its loading now
jdv79 yeha, the lock queue cleared
yippee
ugexe currently im stuck at there not being a consistent auth to search. For p6c the entire index is local so searching for auth/author + authority/author-only when encountering a :auth<xxx:yyy>. But for searching metacpan we don't have the entire index, so we need to know the field name to search before hand 22:02
as in I can easily create an auth from the information available. but that doesn't tell me which elastic search field i need to search on :( 22:04
this is more a general failing of the ecosystem as a whole though. `auth` should be required and the Distributions "ill just invent an auth since the author didnt specify one" needs to go 22:05
if you just shove in $!author when $!auth is missing then what separates $!author="github:foo" from $!author="foo", $!authority="github" (so $!auth="github:foo") 22:07
22:49 raiph joined
raiph ugexe: want a rubber duck debugging session? i've about no idea what you're talking about but are there really three different things with such similar names: auth, author, and authority? 22:57
ugexe raiph: yes. auth is meant to be $owner (no one uses this) + $authority unless $auth is explicitly declared 22:58
raiph and "author"? 22:59
ugexe: is "author" basically an alias for "owner"? 23:00
ugexe author is the author name, so i assumed this could be pretty much any value
author = "Joe Blow", owner = "joeb90". so owner is more like a cpan/github id 23:01
github.com/rakudo/rakudo/blob/nom/...ion.pm#L14 23:03
raiph m: my ($auth, $author, $owner, $authority) = (Nil, 'Joe Blow', 'joeb90', 'CPAN'); say $auth // $author // $authority 23:09
camelia rakudo-moar f99bdb: OUTPUTĀ«Joe Blowā¤Ā»
ugexe so why is this a problem? because say i want to do `install "XXX::YYY:auth<github:foo>"`. The author of XXX::YYY does not have the auth field declared, but has author=foo and authority=github. so to find this module on metacpan i need to search for 2 fields, but the way the request will likely be written (:auth<a:b>) you never know if you can split on a colon or not because author/authority use 23:10
arbitrary data including colons
raiph bbi15 23:11
ugexe it uses the author field, so maybe the author has "h:f:e" 23:12
authority=heh:heh
all i know is the $auth. the delimiter in s22 is a colon, which clearly doesnt work if they can also be used in author and authority 23:13
lizmat as far as I remember, there can be multiple authors (human names) 23:14
ugexe so i get :auth<Joe Blow>. is this the author? or the authority?
lizmat the authority is a key that indicates who or what is responsible for the module
github:lizmat 23:15
might be such an authority
or cpan:ELIZABETH
ugexe this is simply relative to splitting the fields to search on metacpan
so i *might* have the auth but have to search the authority and author fields
if authority and author can have colons, and if auth may also contain no colons, there is no way to determine what the author and authority are 23:16
lizmat I'm not sure what you mean, but then again, I'm very tired and really should go to bed
ugexe: S22:205 states "Optional. A list of (mnemonic) names of people who work / have worked on this 23:18
distribution. For informational purposes only."
ugexe turning :auth<github:XXX> -> query.pl?author=XXX&authority=github, but considering the problem of how Distribution.auth is generated
lizmat the string "github:lizmat" *IS* the authority
ugexe right, i agree. but thats not how its implemented right now github.com/rakudo/rakudo/blob/nom/...ion.pm#L14
lizmat then maybe we need to look at that 23:19
ugexe that was my point :)
lizmat but I'm too zonked out right now to figure out the ramifications
I should be back online in about 10 hours from now
ugexe alright 23:20
lizmat ugexe++ fot persevering :-) 23:21
raiph ugexe: I'm persuaded that if you have to support splitting a string on a delimiter, and the string is composed of two substrings either of which can also contain the same delimiter, you're screwed. You already knew that, but now lizmat and a rubber duck know too. :) 23:29
jdv79 so you can or can't do what you want now with mc6? 23:32
even if in a hacky way
jdv79 was eating
its kind of a double edged sword that the es queries are exposed like that 23:34
ugexe you can. just not requests for exact identities like `install XXX::YYY:auth<...>:ver<...>` or `github:xxx:XXX-YYY:*`
jdv79 too much surface area for my taste but its mc...
ugexe version will work, but not auth
jdv79 is it not there?
maybe we need to gen it on index 23:35
ugexe the problem is some distributions use auth, some use author, some use authority
jdv79 or default it i mean. same thing.
auth is the first choice right?
ugexe but a user just wants XXX:auth<xxx:yyy>
yeah
raiph jdv79: I think ugexe is saying we can't make the current spec / code work reliably; explanation above
jdv79 we can just make that work for mc based on what we get
ugexe you could, but i think it should be round-tripleable anyway 23:36
jdv79 raiph: yeah but we can paper over it or now:)
yeah. Distribution.pm should probably change i guess
ugexe so if i want to grep a bunch of local meta files for :auth<> i dont have to guess which field to use if auth is not actually defined 23:37
jdv79 how come auth isn't a field in s22? 23:38
i thought it was
ugexe design.perl6.org/S22.html#auth 23:39
jdv79 that's not a field
oh, this is where p5's provides syntax will do it 23:40
we'll wedge it all in there. forgot about that:
:)
ugexe the fields are content storage and owner
jdv79 those aren't META6 fields - just concepts
that info should go in the provides - where each entry is a hash
like p5
right? 23:41
i think we've talked about this
i'll munge it as best i can for now with mc. tomorrow. 23:42
raiph .tell nine I thought you and maybe others might find it fun to chat with htuhola/Cheery: www.reddit.com/r/ProgrammingLangua...g_modules/ either at the reddit or maybe here or #perl6 23:44
ugexe yeah. if a request for :auth<whatever is here> always points to the same required field it should work
jdv79 ugexe: github.com/jdv/cpan-api/issues/3 23:50
oh look - another lock queue pileup. woohoo! 23:51

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!