»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'p6: say 3;' or rakudo:, std:, or /msg camelia p6: ... | irclog: irc.perl6.org | UTF-8 is our friend! | feather will shut down permanently on 2015-03-31
Set by jnthn on 28 February 2015.
dalek ast: 6f1e2f1 | coke++ | S03-operators/assign-is-not-binding.t:
Remove hard skip for test that is passing
 00:04
00:05 vendethiel joined
[Coke] I thought we removed all the parrot fudges. Do we want those gone? 00:09
00:09 erdic joined
[Coke] there are 98 instances left. 00:10
00:11 raiph left 00:13 raiph joined
[Coke] I'll get a patch ready if the concensus is to get rid of them 00:14
00:14 tinyblak joined 00:19 atweiden joined, tinyblak left 00:22 cognominal left 00:28 vendethiel left 00:34 airdisa left
[Coke] finds AnyChar in the spec. 00:50
sorry, the syn
jnthn: Is this still a thing to be implemented or does this predate NFG? the spec test for this (S29-type/declarations.t) dates from 2008. 00:52
t/spec/S12-construction/destruction.t seems like a flawed test. 01:02
01:02 BenGoldberg left, BenGoldberg joined
[Coke] m: say $*OS 01:06
camelia rakudo-moar 56cae1: OUTPUT«opensuse␤Saw 1 call to deprecated code during execution.␤================================================================================␤$*OS called at:␤ /tmp/sPM0lwLhez, line 1␤Deprecated since v2014.9, will be removed with release v2015.9!␤…»
[Coke] m: say $*OSVER 01:07
camelia rakudo-moar 56cae1: OUTPUT«13.2.Harlequin␤Saw 1 call to deprecated code during execution.␤================================================================================␤$*OSVER called at:␤ /tmp/MI4Mrn9Rf7, line 1␤Deprecated since v2014.9, will be removed with release v2…»
[Coke] I'm pulling the trigger on the rakudo.parrot removal; won't mind if someone reverts it. 01:09
dalek ast: 506ffcd | coke++ | / (44 files):
Remove rakudo.parrot fudges

Will have to refudge once parrot support returns regardless.
ast: 4f132bf | coke++ | S02-magicals/perlver.t:
Remove file with gone or going vars.

Rakudo wasn't running this file anyway
[Coke] m: say %=pod 01:12
camelia rakudo-moar 56cae1: OUTPUT«5===SORRY!5=== Error while compiling /tmp/T8o3SZ6YeX␤Variable '%=pod' is not declared. Did you mean '$=pod'?␤at /tmp/T8o3SZ6YeX:1␤------> 3say %=pod7⏏5<EOL>␤ expecting any of:␤ postfix␤»
[Coke] m: say $=pod
camelia rakudo-moar 56cae1: OUTPUT«␤»
tadzik ruoso: I see you're solving some of the problems I did; if I were to build a module to help generate grammars from code, would you be interested in that? 01:14
yoleaux 1 May 2015 05:09Z <skids> tadzik: I sent you a pull request to Grammar::BNF that adds ABNF (popular in modern RFCs)
tadzik oh! Awesome!
xBNF looks like a trendy thing these days :)
skids Also I am currently adapting ruoso++'s slang code to Grammar::BNF 01:15
Almost done.
tadzik cool :)
dalek ecs: 13f1284 | coke++ | S28-special-names.pod:
Track an old change to this var name.
 01:16
tadzik I promise to look at your PR tomorrow, now I'm between drunk and sleepy :)
skids Never a good place to be when attempting to code, but otherwise a fine state of mind :-) 01:17
tadzik well, one of my milestones on GSoC few years back was done under such circumstances 01:18
dalek ast: c3083de | coke++ | S02-literals/pod.t:
Track %=pod -> $=pod change...

  ... but rakudo still can't run this test
tadzik a comfy chair, a fireplace and a few cans of Żubr beer
woke up next morning with code working and tests passing
*shrug*, commit, push, I still don't know how I solved it
skids :-) 01:19
It will be interesting to bench some generated RFC grammars against ugexe++'s pure-perl versions. 01:20
[Coke] jnthn: you just added S04-statements/when.t a little while ago; that exception doesn't exist yet. Do we need a ticket to track ?
tadzik skids: I glanced over your PR, do you have some tuits to put some sort of real-world grammar into the test suite? 01:21
one thing that amused me a lot is when I put the BNF grammar from wikipedia into my tests and thus proved it incorrect :) 01:22
01:22 Ven joined
[Coke] adds some train-based commits. 01:22
tadzik . o O ( TRAIN RAIN ) 01:23
skids tadzik: probably tomorrow or over the course of the next week. I do want to eventually get to doing HTTP header parsing.
01:35 Ven left
skids tadzik: I PRd the slang support just now. 01:36
01:37 tinyblak joined
skids tadzik: RFC grammars are often full of "here's the BNF/ABNF but SHOULD tolerate this on input and MAY tolerate this but only when you are acting in the proxy role" so yeah, fun stuff. There will definitely need to be subgrammars wrapped around for such things. Good thing grammars are subclassable. 01:40
01:45 ilbot3 left 01:47 ilbot3 joined 01:58 bjz joined 02:01 xinyuan_ joined 02:02 Ven joined, BenGoldberg left 02:06 vendethiel joined 02:10 Ven left 02:19 BenGoldberg joined 02:20 noganex_ joined 02:23 noganex left 02:24 rmgk_ joined, rmgk left, rmgk_ is now known as rmgk 02:27 vendethiel left 02:29 gfldex left 02:32 alini left 02:45 mattp|spurs is now known as mattp 03:08 vendethiel joined 03:10 konsolebox left, konsolebox joined 03:12 konsolebox left 03:13 konsolebox joined 03:22 mr-foobar left 03:34 ssqq joined
ssqq p6: grammar A { token sym { <.ident> }; token id { <sym> }; } 03:35
camelia rakudo-moar 56cae1: OUTPUT«===SORRY!===␤Substring length (-3) cannot be negative␤»
geekosaur o.O
ssqq p6: grammar B { token add-char { < $ @ % > }; token id { <+ alpha add-char> }; } 03:37
camelia ( no output )
ssqq p6: grammar B { token add-char { < $ @ % > }; token TOP { <+ alpha add-char> }; }; B.subparse('$'); 03:39
camelia rakudo-moar 56cae1: OUTPUT«Method 'char' not found for invocant of class 'B'␤ in regex TOP at /tmp/tmpfile:1␤ in block <unit> at /tmp/tmpfile:1␤␤»
ShimmerFairy IIRC <sym> is for rule names like token id:sym<stuff> , and then <sym> would match stuff .
ssqq ShimmerFairy: I see, name conflict. 03:40
token name like 'a-b' could not use `<+ token-name>` 03:45
03:45 fhelmberger joined 03:46 Sqirrel left 03:49 raiph left 03:50 tinyblak left, fhelmberger left, tinyblak joined 03:51 ssqq left 03:54 tinyblak left 04:04 xinyuan_ left 04:12 BenGoldberg left, tinyblak joined, Sqirrel joined 04:14 vendethiel left 04:29 larks_ left 04:39 atweiden left
TimToady greetings from Furnace Creek, Death Valley, where today it only got up to 108℉ 04:40
m: say (108-32) * 5/9 04:41
camelia rakudo-moar 56cae1: OUTPUT«42.222222␤»
TimToady so 42℃ or so
timotimo i wouldn't mind if camelia learned to reduce the number of ============= in a deprecation note so that we'll see what should be used instead right here in the channel 04:49
timotimo heads back to bed
Timbus_ 42C.. yeesh. i hope you found shade..
always walk through the shadows in the valley of death.
04:51 ggoebel left
flussence it's 6°C here! Colder than it's been for any day all winter. 05:01
jdv79 of course i get home and no wifi issues - only at the hackathon. figures. 05:26
05:38 jack_rabbit joined 05:39 rindolf joined 05:40 alini joined 05:45 Patterner joined 05:47 jack_rabbit left 05:50 Psyche^ left 05:52 laouji joined
moritz timotimo: or pastebin the untruncated output, and supply a short URL to the pastebin 06:09
06:16 konsolebox left 06:25 domidumont joined 06:27 skids left 06:30 domidumont left 06:31 domidumont joined 06:37 camel___________ joined 06:40 laouji left 06:41 laouji joined 06:42 risou is now known as risou_awy, risou_awy is now known as risou 06:45 laouji left 07:01 konsolebox joined 07:04 rindolf left 07:09 konsolebox left, araujo left 07:15 araujo joined 07:16 diana_olhovik_ joined 07:23 rindolf joined 07:24 laouji joined 07:26 [TuxCM] left 07:27 espadrine joined 07:42 mr-foobar joined 07:44 Alina-malina left 07:46 fhelmberger joined 07:47 darutoko joined
lizmat good *, #perl6! 07:48
m: say gethostname()
camelia rakudo-moar 56cae1: OUTPUT«ns1␤»
07:50 fhelmberger left 07:51 Foxcool_ joined, camel___________ left 07:54 domidumont left 08:04 konsolebox joined 08:07 rindolf left 08:15 Alina-malina joined 08:19 telex left 08:20 telex joined 08:21 Ven joined 08:22 Ven left 08:26 baest_ joined 08:28 baest left 08:36 diana_olhovik_ left
dalek kudo/nom: 065219c | lizmat++ | src/core/Deprecations.pm:
Make deprecation more camelia friendly, timotimo++

Also mention RC-0 as a cutoff time for deprecations.		 08:38
08:39 virtualsue joined 08:43 mj41 joined 08:45 diana_olhovik_ joined
lizmat m: say $*OS 08:49
camelia rakudo-moar 56cae1: OUTPUT«opensuse␤Saw 1 call to deprecated code during execution.␤================================================================================␤$*OS called at:␤ /tmp/zdwQ_nB72L, line 1␤Deprecated since v2014.9, will be removed with release v2015.9!␤…»
08:53 cygx joined
cygx o/ 08:54
found a precompilation bug: gist.github.com/cygx/73df9343bb97208b0534 08:55
BEGIN $*KERNEL in a module closes $*OUT oO
same with $*DISTRO, but eg not $*VM 08:56
lizmat cygx: please rakudobug
that seems like something jnthn can sink his teeth into :-) 08:57
excellent golf!
09:13 Rounin joined 09:34 Ven joined
lizmat m: say $*OS 09:35
camelia rakudo-moar 065219: OUTPUT«opensuse␤Saw 1 call to deprecated code during execution.␤$*OS called at:␤ /tmp/SyplJEvmNi, line 1␤Deprecated since v2014.9, will be removed with release v2015.9!␤Please use $*DISTRO.name instead.␤»
lizmat timotimo: ^^ 09:36
09:36 Ven left 09:48 Akagi201_ left 09:49 Akagi201 joined, RabidGravy joined 09:50 Akagi201_ joined 09:53 Akagi201 left
[Tux] m: say $*OS.VERSION 10:03
camelia rakudo-moar 065219: OUTPUT«Method 'VERSION' not found for invocant of class 'Obsolete'␤ in block <unit> at /tmp/HVhLQhUiyw:1␤␤»
[Tux] m: say $*OS.version
camelia rakudo-moar 065219: OUTPUT«Method 'version' not found for invocant of class 'Obsolete'␤ in block <unit> at /tmp/D6bRPvTsQT:1␤␤»
lizmat m: say $*DISTRO.name.version 10:05
camelia rakudo-moar 065219: OUTPUT«Method 'version' not found for invocant of class 'Str'␤ in block <unit> at /tmp/uf4Thvy8bP:1␤␤»
lizmat m: say $*DISTRO.version
camelia rakudo-moar 065219: OUTPUT«v13.2.Harlequin␤»
lizmat m: say $*OSVER
camelia rakudo-moar 065219: OUTPUT«13.2.Harlequin␤Saw 1 call to deprecated code during execution.␤$*OSVER called at:␤ /tmp/3OBGsMVa8m, line 1␤Deprecated since v2014.9, will be removed with release v2015.9!␤Please use $*DISTRO.version instead.␤»
[Tux] m: $*DISTRO.^methods».name.say
camelia rakudo-moar 065219: OUTPUT«BUILD Str gist <anon> <anon> <anon> <anon> <anon> <anon> <anon> <anon>␤»
10:06 laouji left
[Tux] m: $*DISTRO.perl.say 10:08
camelia rakudo-moar 065219: OUTPUT«Distro.new(release => "13.2", is-win => Bool::False, path-sep => ":", name => "opensuse", auth => "opensuse.org/", version => Version.new('13.2.Harlequin'), signature => Blob, desc => "2015-05-03T12:08:11+0200")␤»
[Tux] thinks release and version are swapped
Harlequin is a nickname for this release, not for the version
lizmat m: say $*DISTRO.release
camelia rakudo-moar 065219: OUTPUT«13.2␤» 10:09
[Tux] the version is 13.2, the release is 13.2 (Harlequin)
10:09 gfldex joined
lizmat m: say /etc/os-release'.IO.lines 10:10
camelia rakudo-moar 065219: OUTPUT«5===SORRY!5=== Error while compiling /tmp/bp_qoAICPg␤Two terms in a row␤at /tmp/bp_qoAICPg:1␤------> 3say /etc/7⏏5os-release'.IO.lines␤ expecting any of:␤ infix␤ infix stopper␤ postfix␤ statement end…»
lizmat m: say '/etc/os-release'.IO.lines
camelia rakudo-moar 065219: OUTPUT«NAME=openSUSE VERSION="13.2 (Harlequin)" VERSION_ID="13.2" PRETTY_NAME="openSUSE 13.2 (Harlequin) (x86_64)" ID=opensuse ANSI_COLOR="0;32" CPE_NAME="cpe:/o:opensuse:opensuse:13.2" BUG_REPORT_URL="bugs.opensuse.org" HOME_URL="opensuse.org/" I…»
lizmat [Tux]: so what should be where in your opinion
[Tux] VERSION_ID => version
PRETTY_NAME => release 10:11
lizmat is PRETTY_NAME portable ?
[Tux] one sec
github.com/abeltje/Test-Smoke/tree...ke/SysInfo and below 10:12
test cases are in t/etc
that is what we use for Test::Smoke identification
github.com/abeltje/Test-Smoke/tree/master/t/etc <= a collection of what OS's gather in dist-info related files 10:13
For Linux, even (very) old versions, use this code in perl5: github.com/abeltje/Test-Smoke/blob...nux.pm#L62 10:14
lizmat ok, that seems like some nice LHF for someone else :) 10:15
lizmat working on is lazy
[Tux] github.com/abeltje/Test-Smoke/blob...nux.pm#L62 <= that one is more up-to-date 10:16
(some distro's have levels of os-release files)
lizmat Belin2015 ? not Berlin2015 ? 10:17
[Tux] github.com/abeltje/Test-Smoke/comm...e520867f4a
10:17 bjz left
[Tux] I did not create that branch name :) 10:17
10:18 bjz joined
[Tux] lizmat, /me is willing to help out there if someone would want me to 10:19
lizmat PR's welcome!
[Tux] I think that t/etc folder will give you a whealth of info
jnthn o/
lizmat jnthn /o 10:20
[Tux] and if anyone has an OS that is not yet listed, I'd be very interested in getting the required info files
lizmat jnthn: work on "is lazy" so far 10:21
[Tux] just execute t/etc/get-lsb.pl
lizmat gist.github.com/lizmat/11ac7b63a2faa074d20e
jnthn wtf, we suddenly have 1,300 tickets in RT?!
lizmat jnthn: trying to adapt the buildplan to shove an empty initiializer in there, however I can't get it empty enough 10:22
my $default := nqp::can($_,'lazy')
?? -> $self, $attr { $type }
!! $_.build;
jnthn: I'm out of ideas atm 10:23
Cannot invoke this object (REPR: Null, cs = 0) :-(
seems it lost !build_closure :-( 10:25
jnthn: wrt to #RT tickets, they weren't added last night, something else changed ? 10:27
jnthn lizmat: Maybe this is a better way: gist.github.com/anonymous/08896f0615a976d6caa1 10:29
lizmat ok, testing that 10:30
jnthn It looks different from setting a default, which is probably good :)
lizmat hehe... 10:32
10:32 game0 joined
lizmat I just found there's a "will build" trait ? 10:32
10:32 game0 left
jnthn That may we what has $.foo = ... desugars to... 10:33
*may be
lizmat ah, ok
yeah, that works and *is* much simpler 10:36
jnthn :)
lizmat althought the: 10:37
$attr.package.^method_table{$attr.name.substr(2)}.wrap(-> \self {
callsame() // $attr.set_value(self, $block())
});
feels a bit like a rabbit out of a hat :-)
[Tux] lizmat, where is the code for DISTRO (in case I would find tuits to come up with suggestions)
lizmat src/core/Distro.pm
[Tux] takk
jnthn lizmat: If you take it apart it's not so bad.
lizmat $attr.name.substr(2) # pretty magic to me
jnthn m: class A { has $.foo }; say A.^attributes[0].name 10:38
camelia rakudo-moar 065219: OUTPUT«$!foo␤»
jnthn m: class A { has $.foo }; say A.^attributes[0].name.substr(2)
camelia rakudo-moar 065219: OUTPUT«foo␤»
jnthn :)
Basically, given the variable name, compute the accessor name
lizmat m: m: class A { has $.foo }; say A.^attributes[0].name.substr(0,2)
camelia rakudo-moar 065219: OUTPUT«$!␤»
lizmat ah, ok
duh 10:39
well, I learned a lot about buildplans and such :-) 10:40
jnthn All the RTs are 'cus somebody but all the Rakudo fudge'd spectests into RT tickets 10:43
lizmat wow
didn't see anything on the ml
jnthn Once I get over the initial shock, it's probably worth it, in so far as having 1,300 RTs will piss me off enough to start processing them. :P
Which we really should do ahead of 6.0 10:44
10:45 mj41 left
RabidGravy having looked at a few bits of that however, the associated PR to roast may have been based on an out of date clone 10:45
10:53 virtualsue left
RabidGravy also, it would appear that there is stuff in S03-magicals/ that would probably be better in S28-named-variables/ as KERNEL DISTRO etc aren't mentioned in S03 10:55
lizmat RabidGravy: it's probably for hysterical raisins that they're in there
RabidGravy er S02 10:56
dalek kudo/nom: bdc66e7 | lizmat++ | src/core/traits.pm:
Introduce Attribute will lazy initializer, jnthn++

Inspired by rjbs++, this will actually first wrap the Attribute composer, and in there, wrap the accessor with code that will call the indicated block once to set the value, then call the original accessor again.		 10:57
lizmat jnthn: I replaced the // with a once block, otherwise if the block returned a type object, it would be called again and again
.tell rjbs "has $.fh will lazy { $file.IO.open }" at your service :-) 10:59
yoleaux lizmat: I'll pass your message to rjbs.
11:00 ggoebel joined
jnthn lizmat: That's...screw it up over instances, no? 11:00
lizmat ah, hmmm... lemme check
grrrr 11:01
you're right
(sorry Grrrr)
11:07 laouji joined 11:12 laouji left
cygx another precompilation bug, this time not golfed: gist.github.com/cygx/5dbbea302e0f536615c5 11:13
11:17 Foxcool_ left 11:19 Foxcool joined
dalek kudo/nom: fd90eb6 | lizmat++ | src/core/traits.pm:
Call lazy code until we have a defined value

Tried to fix this by introducing an "$!initialized" attribute in the role, but couldn't get that to work, some weird scoping / capturing issues.		 11:30
11:37 airdisa joined 11:42 alini left 11:47 cognominal joined
dalek kudo/nom: c72d670 | lizmat++ | src/core/traits.pm:
Fix message on Attribute will foo
 11:52
ab5tract_ lizmat: cool patches :D 11:53
dalek ast: 1f384ba | lizmat++ | S12-attributes/instance.t:
Add some "will lazy" and "is/will foo" tests
lizmat ab5tract_: those were really jnthn++'s work, /me is just a minion :-) 11:54
11:57 spider-mario joined 12:04 RabidGravy left 12:07 kaare_ left
rjbs lizmat: I look forward to seeing how it got implemented. 12:10
yoleaux 10:59Z <lizmat> rjbs: "has $.fh will lazy { $file.IO.open }" at your service :-)
rjbs lizmat++ jnthn++ # thanks, I think this i a big deal 12:11
lizmat rjbs: github.com/rakudo/rakudo/commit/bdc66e79da and github.com/rakudo/rakudo/commit/fd90eb68fa
rjbs vielen Dank, I'll read this after breakfast; the women of the house have declared we will go get crêpes now
lizmat hmmm.... crêpes .... good choice 12:12
12:15 eli-se joined 12:16 eli-se left
lizmat hopes FROGGS is doing ok 12:16
12:22 tinyblak_ joined 12:23 eli-se joined 12:24 laouji joined, tinyblak left
cygx jnthn: I'm unable to figure out gist.github.com/cygx/5dbbea302e0f536615c5 12:26
jnthn: after compilation, Pointer[type] dies (where type is a CStruct) 12:27
12:28 vendethiel joined 12:32 eli-se left
[Coke] begins to backlog... 12:33
12:33 szabgab joined
dalek ast: 71ea556 | lizmat++ | S02-lexical-conventions/begin_end_pod.t:
Actually test what we say we're testing
 12:33
szabgab hello there 12:36
this might not be a good time to chat here, but at least I leave this info here
[Coke] so, the person who opened all the tickets took something I said at the hackathon and ran as far as they could with it without actually checking with me before they pulled the trigger on anything. I expected maybe ONE more ticket when I woke up this morning. Looks like there is a lot of cleanup to do, I think he opened a ticket for each fudge directive, not for each bug.
so, I'm going to apply his fudge patch (if not already done) and then start consolidating tickets and fudge markers. 12:37
szabgab I am helping the YAPC::EU organizers with some PR (not Pull Request) and I've noticed there are only 3 talks about Perl 6 act.yapc.eu/ye2015/talks/tag/perl6 12:38
(though I only see the accepted talks) It would be great to see some more Perl 6 talks submitted 12:39
jnthn szabgab: I plan to be there, just didn't submit yet.
(It's on my todo list.)
szabgab Also the organizer would like to know if you would like to organize a Hackathon
lizmat I'm also planning to be there, just not planning to do a presentation
szabgab jnthn: great, push it please 12:40
12:40 araujo left
[Coke] is dnmfarrell on IRC? 12:41
12:43 araujo joined 12:45 araujo left 12:46 araujo joined
[Tux] class A { has B $.b is rw; has Str $.s is rw; }; class B { has A @.a; } 12:46
szabgab oh and as I am usually not on IRC, could you just encourage each other to submit talks, please?
[Tux] that is impossible, as B is not defined by the time A is compiled and vice versa 12:47
moritz [Coke]: irclog.perlgeek.de/perl6/search/?ni...ell&q= yes
[Tux] can I force A.b to be of type B later?
12:47 araujo left
[Tux] or in another way 12:47
szabgab Oh and if someone would like to run a Perl 6 course, there are still rooms for that too
12:48 araujo joined, szabgab left
jnthn [Tux]: Stub B first 12:48
class B { ... }
12:48 araujo left
lizmat m: class B { ... }; class A { has B $.b is rw; has Str $.s is rw; }; class B { has A @.a; } 12:48
camelia ( no output )
12:49 araujo joined
lizmat m: cclass A { has B $.b is rw; has Str $.s is rw; }; class B { has A @.a; } 12:49
camelia rakudo-moar c72d67: OUTPUT«5===SORRY!5===␤Type 'B' is not declared␤at /tmp/ZWPzlD_g0v:1␤------> 3cclass A { has B7⏏5 $.b is rw; has Str $.s is rw; }; class ␤Malformed has␤at /tmp/ZWPzlD_g0v:1␤------> 3cclass A { has7⏏5 B $.b is rw; has Str $.s is rw; }; cl…»
lizmat m: class A { has B $.b is rw; has Str $.s is rw; }; class B { has A @.a; }
camelia rakudo-moar c72d67: OUTPUT«5===SORRY!5===␤Type 'B' is not declared␤at /tmp/x4sT0uG0ji:1␤------> 3class A { has B7⏏5 $.b is rw; has Str $.s is rw; }; class ␤Malformed has␤at /tmp/x4sT0uG0ji:1␤------> 3class A { has7⏏5 B $.b is rw; has Str $.s is rw; }; clas…»
[Tux] jnthn, that works. thanks
[Tux] has entangled himself into the swamp of callbacks 12:50
[Coke] anyway, thanks everyone for not freaking out. I'll get RT cleaned up as soon as possible so that we get the benefit of the new tickets, but not the duplication. 12:53
jnthn [Coke]++
12:54 domidumont joined
jnthn [Coke]: Just looking at some of the titles, I can see that some of them rely on out-dated spec. 12:54
[Coke]: Or possibly the design docs need changes :)
Like the AnyChar thing you mentioned before.
[Coke] there's so much stuff that needs review, yes. 12:55
I wish we had a branch of RT we could have done this in. :)
jnthn I suspect some merging will help
e.g. there's a load of ':: NYI'
[Coke] Aye. I just saw "int NYI", even. 12:56
jnthn int is I :)
[Coke] and it's clearly one ticket per fudge, so the total will probably drop by a hundred or so just there.
[Coke] creates an RT view that shows just tickets opened by our new ticketmaniac. :) 12:57
12:58 rurban joined, diana_olhovik_ left
dalek ast: d5c53cf | jnthn++ | S32-hash/delete-adverb.t:
Fix tests sensitive to hash ordering.
 13:00
13:03 alini joined, RabidGravy joined
RabidGravy afternoon 13:04
dalek ast: c8953f2 | jnthn++ | S05-capture/caps.t:
Fix another test sensitive to hash ordering.

Also these are fudged on JVM; maybe they will work now they aren't so sensitive?
lizmat jnthn: confirmed that S05-capture/caps.t: can be unfudged for jvm 13:07
jnthn lizmat: Cool; wanna do it, or shall I?
lizmat I'll do it
jnthn I *think* that we now have clean spectest with MoarVM HEAD
As in, those two were all that needed fixing.
I get one other fail but I think it's 'cus I don't have your lazy patch. 13:08
yes, it was :)
dalek ast: 4099cad | lizmat++ | S05-capture/caps.t:
Unfudge passing tests on JVM

Turns out the testing was relying on hash key order, jnthn++ for noticing		 13:09
lizmat yeah, tests for the will lazy patch are already in roast
jnthn I wasn't expecting it to go into CORE.setting, but I guess we're a good way from a relesae if anybody comes up with a reason not to do it like that. 13:10
dalek ast: a219848 | moritz++ | / (82 files):
Remove # chars from fudge messages

they do not play too well with TODO messages in TAP		 13:11
lizmat fwiw, I think it's a *very* usefult faeature
moritz o/
[Coke] gah. 13:18
moritz: I think the right answer there is to fix the tap output.
esp. given the effort I had put into having a standard way of writing RT #... 13:19
13:19 araujo left
arnsholt Doesn't TAP let you say \# for hashes in the output? 13:20
13:21 Gothmog_ left 13:22 Gothmog_ joined
moritz arnsholt: no idea 13:22
[Coke]: that's unfortunate
[Coke]: how would you fix the TAP output?
[Coke] Digging now. I swear I've already solved this for partcl. 13:23
yes, looks like \# is what I did there. 13:25
Trying to find a spec of some kind to back it up. I'm curious what was breaking with things the way they were, though.
13:26 rurban left
[Coke] we could s/#/SOMETHING ELSE/g, also 13:26
dalek ast: 8b6048b | moritz++ | / (82 files):
Revert "Remove # chars from fudge messages"

This reverts commit a21984810bb4f073291818e6526d9d024c339b98.
I'll try to fix the todo and todo functions in rakudo's Test.pm instead.		 13:27
13:33 rurban joined, rindolf joined 13:34 tinyblak_ left, nbdsp joined
[Tux] What is a more perl6 way to write "for ($r.fields) -> $f { @!fields.push: $f; }" 13:34
$r.fields>> ... 13:35
lizmat @!fields.push: @$r.fields 13:36
??
[Tux] Brrr: 13:37
# expected: (Failure)
# got: (Failure)
lizmat
.oO( not all Failures are alike :-)
 13:38
[Tux] Cannot understand @!fields in this context
on your suggestion
13:39 kaare_ joined
lizmat that's what you were pushing on in your example ? 13:39
[Tux] gist.github.com/Tux/12387de2febd2eda1efc
the commented-out line works as expected 13:40
13:40 mj41 joined
[Coke] I'm trying to keep the giant list of new tickets managed so we can coordinate with the incoming pull request. Please ping me before closing/merging whatever those new tickets. 13:41
lizmat [Tux]: perhaps a Slang::Tuxic artefact ? 13:42
[Tux] ENOFROGGS
13:43 hobbs left
lizmat yeah, I noticed as well... offline for ~ 3 days already :-) 13:43
13:45 hobbs joined
[Tux] so, "is(something-returning-failure, Failure, "Fail");" won't fly, but neither does "like(failing-thing, Failure, "FAIL");" 13:45
what to use there?
moritz isa_ok 13:46
[Tux] \o/ :) thanks
moritz m: sub f { fail 'foo' }; use Test; plan 1; isa_ok f(), Failure, 'stuff'
camelia rakudo-moar c72d67: OUTPUT«1..1␤ok 1 - stuff␤»
moritz or do a throws_ok { use fatal; your test here }, X::The::Exception::Type::You::Except 13:47
*Expect
13:47 rurban left
[Coke] m: my int $a = 2; use Test; isa-ok(2,int); 13:49
camelia rakudo-moar c72d67: OUTPUT«not ok 1 - The object is-a 'int'␤␤# Failed test 'The object is-a 'int''␤# at /tmp/Ir5MYciBud line 1␤# Actual type: Int␤»
lizmat yeah, known issue 13:50
[Coke] is there an RT?
lizmat natives int pretends to be an Int
not sure
[Coke] ok. 124461 for now. let me know if you find another one, I'll merge it. 13:51
ab5tract_ [Tux]: does it work when you explicitly do @($r.fields) 13:52
moritz m: sub f(Mu $a, Mu $b) { say $a.isa($b) }; f 1, int
camelia rakudo-moar c72d67: OUTPUT«False␤»
lizmat moritz: you mean throws_like ?
moritz m: say 4.isa(int)
camelia rakudo-moar c72d67: OUTPUT«False␤»
moritz lizmat: erm, yes
m: say 4.^isa(int)
camelia rakudo-moar c72d67: OUTPUT«0␤»
moritz m: say 4 ~~ int
camelia rakudo-moar c72d67: OUTPUT«False␤»
moritz m: say nqp::istype(4, int)
camelia rakudo-moar c72d67: OUTPUT«===============================================================================␤The use of nqp::operations has been deprecated for non-CORE code. Please␤change your code to not use these non-portable functions. If you really want␤to keep using nqp:…»
moritz m: use nqp; say nqp::istype(4, int)
camelia rakudo-moar c72d67: OUTPUT«0␤» 13:53
jnthn isa-ok(2,int) won't ever work out
[Tux] m: class C { method hash { { :A(1), :B(2) }}}; my $c = C.new;$c<A>.say
camelia rakudo-moar c72d67: OUTPUT«postcircumfix:<{ }> not defined for type C␤ in block <unit> at /tmp/aVQDg9Dhni:1␤␤»
[Tux] why?
jnthn .isa is an object operation
13:53 rurban joined
jnthn And so by the time it sees the thing it's been boxed to an Int object 13:53
[Coke] m: my num $a = 2; use Test; isa-ok(2,num); 13:54
camelia rakudo-moar c72d67: OUTPUT«This type cannot unbox to a native number␤ in block <unit> at /tmp/skYIovXdgi:1␤␤»
[Tux] ab5tract_, no, :(
[Coke] would you say that's the same issue?
jnthn [Coke]: ok(2, num) ain't going to work either 13:55
[Coke] jnthn: ah. Excellent. I can rewrite the test file to do>
jnthn [Coke]: 2 isn't a Num though, so you can't assign it to a num native
moritz [Coke]: no
[Coke]: that's correct, though the error message coudl be better
jnthn m: my num $a = 2e0;
camelia ( no output )
[Coke] m: my int $a = 3; is($a.WHAT, int); use Test;
camelia rakudo-moar c72d67: OUTPUT«not ok 1 - ␤␤# Failed test at /tmp/zDqr2pyog4 line 1␤# expected: (int)␤# got: (Int)␤»
13:55 vendethiel left
jnthn .WHAT is *also* an object-y operation... 13:56
dalek kudo/nom: 6a20f05 | moritz++ | lib/Test.pm:
Test.pm: Escape # in todo() output
[Coke] m: my int $a = 3; say $a ~~int;
camelia rakudo-moar c72d67: OUTPUT«False␤»
[Coke] ok. How can we write the test? ;)
jnthn What is the goal of the test? 13:57
13:57 vendethiel joined 13:58 tinyblak joined
[Tux] I thought giving a class a hash method implicitly gave it postcircumfix:<{ }> 13:58
lizmat no, you need to add AT-KEY and friends
13:59 diana_olhovik_ joined
lizmat m: class A { method AT-KEY($a) { say "foo for $a" } }; A.new<zip> 13:59
camelia rakudo-moar c72d67: OUTPUT«foo for zip␤»
jnthn [Coke]: If it's to check a variable declared is int is really native, you could try sticking a huge value (only storable as a big Int) into an int, and seeing it doesn't round-trip...
dalek p: 8437c84 | jnthn++ | tools/build/MOAR_REVISION:
Get a MoarVM with hash size reduction, opts.
 14:00
kudo/nom: 20e3646 | jnthn++ | tools/build/NQP_REVISION:
Get MoarVM with hash improvements.

Hashes on r-m now uses (on 64-bit platforms) 8 bytes less per hash, plus 16 bytes less per hash entry.		 14:02
14:03 mohij joined, konsolebox left
dalek kudo/nom: 86b1b2b | lizmat++ | src/Perl6/Actions.nqp:
Make nqp::ops warning a bit more camelia friendly
 14:04
14:06 konsolebox joined 14:08 rindolf left
jnthn away for a bit 14:13
lizmat jnthn: spectest clean for me, not noticing any speadup just yet though 14:16
[Tux] AAAAHRG! after these commits, no more panda. no more panda => no more Slang::Tucic, not Slang::Tuxic no development :( 14:17
$ panda install Slang::Tuxic
===SORRY!===
No candidate found for 'panda'.
14:18 rindolf joined 14:21 rurban left
lizmat hmmm... I'm getting Missing or wrong version of dependency 'src/gen/m-CORE.setting' (from 'lib/File/Find.pm') on bootstrap 14:22
even on a fresh panda clone
[Tux] so the rebootstrap.pl worked for you? 14:23
14:23 nbdsp left 14:24 ir2ivps4 joined 14:25 pmqs joined
[Tux] no idea. I typed "make" :) 14:32
14:33 rurban joined 14:41 laouji left 14:45 airdisa left, cygx left 14:49 airdisa joined 14:52 airdisa left, cognominal left, diana_olhovik_ left
[Tux] t/common.t ..... ok 15:00
# Failed test 'dependencies 1'
# at t/ecosystem.t line 32
# expected: 'some'
# got: 'else'
15:01 rurban left, BenGoldberg joined, rurban joined 15:02 BenGoldberg left 15:04 BenGoldberg joined 15:07 domidumont left, konsolebox left
lizmat down to 2 failing spectest files for my new lazy @*INC work *phew* 15:12
[Tux] without panda I am lost :( 15:21
15:24 skids joined
jepeway m: use Test; my Int $i = 0; isa-ok $i, Str, 'crazy, magic stringification works in my slang'; # not that I can author such a slang, mind you 15:30
camelia rakudo-moar 86b1b2: OUTPUT«not ok 1 - crazy, magic stringification works in my slang␤␤# Failed test 'crazy, magic stringification works in my slang'␤# at /tmp/Ri1kRfnK9S line 1␤# Actual type: Int␤»
jepeway (& greets, #perl6, btb) 15:31
nine waves from his new home
yoleaux 1 May 2015 10:46Z <lizmat> nine: would it be possible for me to get an Inline::Perl5 commit bit ?
lizmat waves back
nine lizmat: if you tell me how I can give you one
lizmat I have *no* idea :-) 15:32
jepeway o/
lizmat moritz might be able to explain?
nine lizmat: you're lizmat on github?
lizmat: I think I found it
jepeway I'm a tad disappointed that diag output doesn't mention the expected type.
lizmat yeah, that's me :-) 15:33
nine lizmat: then you have one now :)
lizmat cool lemme check
skids (backlogging) So... does masak still hold the record for opening RT tickets? :-)
jepeway Seems like it should, whenever the default is overridden.
dalek line-Perl5: a52fd1b | lizmat++ | lib/Inline/Perl5.pm6:
Workaround for "is nodal" breakage, quester++
lizmat now we only need panda to work again 15:34
nine I think I won't backlog this time. Was offline for almost three days. I hope there'll be a weekly tomorrow :) 15:35
jepeway 'cause the default method mentions the expected type, y'see. 15:36
lizmat nine: I hope so too
jepeway s/method/message/
lizmat nine: we just got a new and shiny Moar with many optimisations 15:37
something you probably want to check Inliine::Perl5 against :-)
nine lizmat: on it :) 15:40
15:42 jack_rabbit joined 15:43 zakharyas joined 15:47 hobbs left, fhelmberger joined, rurban left
timotimo could panda rely on hash ordering in that test? 15:50
welcome back, nine
glad to hear your move has succeeded 15:51
15:51 hobbs joined 15:52 fhelmberger left 16:01 diana_olhovik_ joined 16:05 jack_rabbit left
[Coke] m: my bit $a; 16:06
camelia rakudo-moar 86b1b2: OUTPUT«===SORRY!===␤Invalid int size for local/lexical␤»
16:07 jack_rabbit joined 16:09 rurban joined
[Coke] is lowercase junction ever going to be a thing? 16:09
lizmat [Coke] my bit $a is really NYI afaik
[Coke] yah, just a weird error messages. leaving that ticket alone. 16:10
lizmat afk for some dinner& 16:11
[Coke] m: my int $a; $a = 2**100; 16:14
camelia ( no output )
[Coke] in the repl that dies with qast not found.
m: my int $a; $a = 2**100.say;
camelia rakudo-moar 86b1b2: OUTPUT«100␤»
[Coke] m: my int $a; $a = 2**100; $a.say;
camelia rakudo-moar 86b1b2: OUTPUT«0␤»
[Coke] is that the right overflow check? that it's still 0? 16:15
TimToady
.oO(native junctions are, like, diodes and transistors, right?)
moritz m: my int $a; $a = 1 + 2**100; $a.say
camelia rakudo-moar 86b1b2: OUTPUT«1␤»
moritz [Coke]: it's simply that powers of two tend to overflow to 0 16:16
[Coke] TimToady: yah, it didn't make any sense to me either; but there's a test for it.
moritz: well, it's a more valid test than what we have, I guess.
16:23 pmqs left
jnthn I think the lowercase junction thingy can go away. 16:23
As for bit: I added it recently-ish, initially for my bit @vector 16:24
(Which ain't done yet, but at least in progress...)
TimToady "will lazy" ain't Engrish
jnthn hah, I knew you'd have some comment :P 16:25
"will lazily" maybe :P
16:26 baest_ is now known as baest
TimToady most of the other 'wills' are designed to be verbs, but what's a good verb for doing something later? 16:26
sjn postpone? 16:27
geekosaur defer?
...suppose not. hm
procrastinate :p
TimToady otoh, the "later" is already implicit in 'will' 16:28
so 'will init' or so isn't so bad
jnthn 'will be' :)
TimToady * will be, will be...
will get, will start 16:29
don't want init
sjn will do :-P
TimToady start is available again these days, since START went away
nwc10 were weeklies running at the time START went away - I forget why START was able to go away - is there a summary online I can read? 16:31
(I realise that the direct answer to "why did START go away?" might be quite a few lines of repeated typing)
jnthn It didn't really fit the phaser "happens at a weird time" thing, since it happened at that point in the code, just once. So we called it "once" instead. 16:32
nwc10 aha. 16:33
that answer was shorter than I expected.
so, a "once" phaser fires once, at the first time execution reaches that point?
jnthn Well, it's not really a phaser, just sugar for something involving a state variable 16:36
masak good evening, #perl6
vendethiel \o, masak!
jnthn a wild masak appears!
masak <-- actually quite tame 16:37
vendethiel
.oO( that's what all the wild ones say! )
 16:38
masak well, at least the articulate ones.
tadzik: so excited to see what you'll do with github.com/tadzik/synopsebot/pull/6 ! :D 16:39
masak backlogs, wildly 16:40
I just want to say that it's heartening to see so much activity on the channel lately. 16:42
it feels to me that with all the work being applied on rakudo itself, on the test suites, and on the docs, we might actually end up somewhere nice for Christmas.
<3
vendethiel it's not friday, but /hug to everyone indeed :-) 16:43
masak m: say Date.new # and it's *this* Christmas! :D
camelia rakudo-moar 86b1b2: OUTPUT«2015-12-24␤»
16:43 risou is now known as risou_awy, risou_awy is now known as risou 16:44 risou is now known as risou_awy, risou_awy is now known as risou
arnsholt masak: Regarding PARSING_IGNORED in your patch up there, wouldn't something like "regex piece { <info> | . }" do the trick? 16:46
masak hmm :) 16:47
masak tries
arnsholt I think LTM should DTRT in that case
masak let's say that works -- what's the easiest way to "amend" a pull request?
vendethiel git push :P 16:48
RabidGravy :)
masak ah -- to the source branch of the PR?
moritz yes
vendethiel yes
moritz you can even force-push there
masak and does Github DTRT with `git push -f`?
:) 16:49
vendethiel yes
masak goodie.
vendethiel++ moritz++
moritz masak: it even shows comments on replaced patches as such
masak it works \o/ 16:50
masak pushes --force
timotimo jnthn: do we already have an idea why Proc::Async is racy? is that a mvm-level problem or an nqp/rakudo-level one? if you can point me at something, i could probably try to figure it out
TimToady gist.github.com/anonymous/16e85e7c3e1b95ce1577 is a test for Proc::Async written by a craigslist cohort 16:54
discusson at irclog.perlgeek.de/perl6/2015-03-31#i_10370872 16:55
masak tadzik: new and improved! github.com/tadzik/synopsebot/pull/6 :)
(pro tip: a `git push --force` *will* update the pull request. it does not update the OP comment, so if you're expecting that to change, you will be momentarily confused, at least until you start comparing SHA-1s.) 16:56
timotimo i'll have a look with that valgrind tool that checks for race conditions and stuff
jnthn timotimo: I suspect there's problems at a couple of levels; I was pondering spending some time tomorrow on that.
timotimo helgrind*
i bet you wouldn't mind if i fixed it before you :P 16:57
jnthn timotimo: Helgrind is rather noisy :)
True, but I think it's "them" not "it"
timotimo but if it's hard, i may not want to spend any time on it :P
oh my
jnthn And at least some of it is our code-gen, I fear.
timotimo oh?
jnthn For example, our phaser handling isn't thread safe.
timotimo that surprises me a bit
oh, this is about helgrind noise, not proc::async problems?
jnthn In fact, it's not really recursion safe.
timotimo: Until a few days ago we had a bug related to the way we compile blocks inside thunks, which also could have caused thready issues. 16:59
timotimo mhm
well, i'll have a look
jnthn *nod*
There's two tools in valgrind; helgrind and drd
timotimo i may want to try "porting" Proc::Async to nqp so that i'll get much less output?
jnthn Not sure how much that'll help 17:00
I mean, it's not until you kick off a process or two that threads get created
masak .oO( the brave warriors go to valgrind; the craven ones get sent to helgrind )
timotimo i suppose that's fair, but in nqp land the code gen is much simpler and there's much less stuff actually running all in all
jnthn
.oO( callgrind me maybe )
 17:01
TimToady wonders if 'nodal' needs to be split into array nodal vs hash nodal, such that AT-POS gets one, and AT-KEY gets the other 17:02
shower &
masak jnthn: I challenge you to use that as aslide title at some point :P 17:03
nine This is the reason for panda's test failure: 17:05
m: my @a = 1, 2, 3; my @b; say @a; say [(@a (|) @b).list.flat]
camelia rakudo-moar 86b1b2: OUTPUT«1 2 3␤3 1 2␤»
nine github.com/tadzik/panda/blob/maste...tem.pm#L50 17:06
jnthn Sets are unordered, so relying on ordering would be a badass umption. 17:07
m: my @a = 1, 2, 3; my @b; say @a; say (@a (|) @b).WHAT
camelia rakudo-moar 86b1b2: OUTPUT«1 2 3␤(Set)␤»
timotimo food time first :)
jnthn However, our hashes until recently maintained insertion order.
But only on Moar 17:08
timotimo is someone looking at the panda problem already, btw?
jnthn So this may also explain the panda bustage on JVM too...
Which has never had the "oops we keep order" issue
17:09 mj41 left
masak TimToady: my immediate reaction is -- if there's no overwhelming reason to distinguish those, then the API will be a lot simpler with them undistinguished. 17:11
jnthn nine: Do you know if it's a bad test, or wrong code?
nine timotimo: github.com/tadzik/panda/blob/maste...tem.pm#L50
jnthn: unfortunately, I don't know if the order of dependencies is meaningful
masak not dependencies of a given module, no.
at least not as far as I know. 17:12
jnthn yeah
Taht'd be my guess too but I don't know for sure
17:15 liztormato joined
jnthn Hm, on Windows testing File::Find fails, alas 17:15
liztormato has $.a will do { ... }. # like it for the similarity with do { ... } 17:16
masak liztormato: hm, what's the semantics of that? 17:18
liztormato: ISTR `will do` is the way to attach a routine body to a routine, or sump'n.
liztormato Alternate name for "will lazy" 17:19
masak continues backlogging, confused 17:22
jnthn masak: See my gist from earlier today where I sketched a "will lazy" trait in response to a question on channel 17:23
17:23 liztormato left, domidumont joined 17:24 liztormato joined
masak nodnod 17:24
it will all make sense at the end of the rainb^Wbacklog 17:25
liztormato: does the fix at irclog.perlgeek.de/perl6/2015-05-02#i_10537026 suggest that we have insufficient spectest coverage for the List.roll(Inf) case? 17:26
liztormato Yes
masak .oO( make sure this call does not run forever )
jnthn
.oO( They see me .roll-in', they failin' )
 17:27
masak [backlog] should lazy default attributes be part of doc.perl6.org/language/classtut ? did someone file an issue about this? 17:32
s/default attributes/attribute defaults/
jnthn masak: Once we decide what we're callin' the trait
TimToady I guess they can always use .duckmap if they wish to distinguish .[] from .{}
masak I think it's orthogonal to that, no? lazy attribute defaults already exist in Perl 6 and work in Rakudo. 17:33
m: class C { has $.x = "OH HAI" }; say C.new(:x(42)).x; say C.new.x
camelia rakudo-moar 86b1b2: OUTPUT«42␤OH HAI␤»
masak jnthn: that ^
jnthn Oh...
liztormato Thats at object build time 17:34
17:35 virtualsue joined
TimToady anyway, if we're going to encourage people to write 'method AT-KEY' and such, they also need 'is nodal' 17:35
TimToady needs to commute to the city of Lost Wages... 17:36
afk till evening &
masak liztormato: yes. rjbs was asking about that, and didn't find the answer in doc.perl6.org/language/classtut 17:37
masak submits an issue
liztormato To be clear: has $.a = 42 is not lazy but run at build time 17:38
masak oh! yes, you are right. those two are not the same.
which totally motivates having a way to do the lazy thing, too. 17:39
tbh, I'd be fine with `has $.x is lazy = "OH HAI";`
liztormato Otoh, why isn't lazy the default?
masak because that'd be surprising. 17:40
17:42 virtualsue left
masak oh, now I hit the point irclog.perlgeek.de/perl6/2015-05-02#i_10537636 where jnthn++ suggest doing `is lazy` as a module. 17:45
+1
if people want it badly enough, that's the way to prototype it and then eventually core it :)
17:45 virtualsue joined
masak [backlog] by the way, speaking of irclog.perlgeek.de/perl6/2015-05-02#i_10537654 -- is it reasonable to expect `has $.x = !!!` to actually end up emitting a "nice" error message saying something about $.x not being supplied? 17:46
because that... would be... *swoon* 17:47
liztormato Let's prototype it in a module ;-)
17:48 diana_olhovik_ left
jnthn Probably do-able with a little cunning ;) 17:49
dinner &
liztormato masak: what made me put will lazy in the core, is that it is apperemtly a standard feature of Moose 17:51
RabidGravy it is
mst it isn't 17:52
Moose has lazy_build, which is not simple lazy
well, 'lazy => 1' exists
which I guess is equivalent to a trait
the Moo/MooseX::AttributeShortcuts version is much nicer
RabidGravy didn't lazy_build go away?
mst has thing => (is => 'lazy', builder => sub { ... });
no, it's still there, just not really recommended 17:53
largely because I designed it for a very specific case in a web metaframework and then somebody added it to Moose core without thinking about it
hence when I did Moo, I designed a sensible version and supported that
liztormato: I would prototype outside but want it to be cored later 17:56
pretty much all of my attributes these days are either ro+required or lazy w/a builder method
because, well, mutable state is the enemy :) 17:57
masak liztormato: putting `will lazy` in core might be a nice thing -- but it doesn't stand out as obviously so.
liztormato: I would be more convinced if there was a known second use case besides lazy attributes :)
17:57 Isp-sec joined
timotimo hey, hold on, this only just started existing today! ;) 17:58
oh
(backlogged)
liztormato Can't think about another use case just yet 17:59
But the current implementation uses rakudo internals specific code, so it's tied pretty tightly to the core 18:00
Specifically Attribute.compose 18:01
18:02 vendethiel left
liztormato Cycling again& 18:04
18:04 liztormato left, vendethiel joined 18:08 diana_olhovik_ joined
RabidGravy lazy might concievably enable initializing with the result of a method (i.e. no "may not be used on partially constructed objects" if tryin that now) 18:09
(of course you can do that right now with a similarly named method as the attribute) 18:12
18:14 mohij left 18:18 iv left
moritz since the RHS of an attribute initializer is already a thunk (code object), it would make sense to use it as initializer for lazy attributes too 18:19
18:29 zakharyas left
mst one thing that Moo does that's a very useful affordance is that 18:29
has foo => (is => 'lazy', builder => sub { ... });
installs that sub as _build_foo and then calls it by name 18:30
so you can method modifier or replace it in subclasses, roles etc.
18:38 alini left
RabidGravy which is nice 18:39
always a pain having to manually install the builders if you are creating them in e.g. an attribute trait 18:41
18:41 aindilis left 18:46 aindilis joined
masak RabidGravy: sounds a lot like the `builder` thing mst++ just metnioned. 18:48
RabidGravy yes
masak RabidGravy: with the little I've used Moose, I feel there's good use for both `lazy` and `builder`. 18:49
18:49 spider-mario left
RabidGravy but if you had lazy you could do: 18:49
$.a is lazy = $.get_a; 18:50
for instance
or is it
has $.a is lazy with bulder { ..... } 18:51
18:52 telex left 18:54 telex joined
RabidGravy which would be quite good fun 18:54
jdv79 p6 is now taking from Moose? dizzying.
seems kind of warped and/or incestuous 18:55
RabidGravy it's more that it's quite a common pattern and Moose has a good extent example 18:56
extant rather
18:57 mohij joined
jdv79 i'm all for it except i find the p6 stuff tends to be more well thought out than the likes of Moose in terms of holistics. 18:58
masak that's what I'm after here too. 18:59
TimToady++ has said that a feature is seldom there for just one reason.
jdv79 is a heavy Moose user - disclaimer
19:00 vendethiel- joined 19:01 vendethiel left
masak TimToady: dangit, now I see anapests everywhere! :P www.salon.com/2015/04/30/i_secretly..._500_days/ 19:03
19:10 rindolf left 19:11 spider-mario joined 19:14 FROGGS[mobile] joined
FROGGS[mobile] Hi #perl6 19:14
lizmat FROGGS[mobile] o/ 19:15
FROGGS[mobile] lizmat: I'm fine :o)
masak \o
lizmat good to hear!
FROGGS[mobile] just had to build a fence and doors for a friend, and today was birthday party for $son 19:17
and now every bit of my body is acking, and tomorrow is train ride to the GPW in Dresden 19:18
lizmat ah, they come in groups, those birthdays :-)
won't be at the GPW :-( 19:19
FROGGS[mobile] no, it is a belated Party :o)
yeah :/
lizmat are you aware at least some train drivers in Germany will be on strike soon ?
FROGGS[mobile] it will be just lichtkind and me afaik
19:19 zeleiadi joined
FROGGS[mobile] yes, from Tue to Mon 19:20
lizmat hence your travel tomorrow :-)
FROGGS[mobile] and we attempt to get back on Sat ó.ò
nwc10 is not aware of this.
but is now
bother?
FROGGS[mobile] there will be some trains, so I dont care much 19:21
there is always some sort of SCHIENENERSATZVERKEHRRR!!! 19:22
:o)
rurban there's a big car from frankfurt
19:22 zakharyas joined
rurban do you need to come from Berlin? 19:23
What I heard they'll start their strike on Tuesday - Wednesday 19:24
masak 1.3k tickets in RT!? irclog.perlgeek.de/perl6/2015-05-03#i_10538762 -- wow, that snuck up on me as well!
but having a lot of tickets filed on you just means that you're popular.
(and buggy) :P
jnthn .tell lizmat Attribute isn't Rakudo-specific; think S12 mentions it 19:26
yoleaux jnthn: I'll pass your message to lizmat.
masak was gonna say.
lizmat jnthn: true, but Attribute.compose *is*, afaik
yoleaux 19:26Z <jnthn> lizmat: Attribute isn't Rakudo-specific; think S12 mentions it
masak that's probably more an oversight than anything...
lizmat at least until we have a COMPOSE phaser
masak anyway, what feels weird to me is coupling `will do` (or whatever it ends up being called) to attributes only. 19:27
that feels somehow underpowered.
lizmat well, others are for variables only: will leave, will end, etc 19:28
jnthn lizmat: It falls into the same category of much of the rest of the MOP, I think.
masak oh, there actually are a lot of new RT tickets. that explains it.
FROGGS[mobile] rurban: I have three kids with me, a car is no option I fear 19:29
nwc10 FROGGS[mobile]: it might be that I simply can't make it. We're supposed to be on the sleeper overnight on Tuesday. I don't think we have any viable way to make a sane alternative plan.
FROGGS[mobile] yeah :o(
masak lizmat: I didn't get the impression the others are for variables only.
lizmat I'm not sure what "will leave" means for an attribute 19:30
masak lizmat: I've seen no example of it, but I would expect being able to put `will end` on a routine, for example.
lizmat m: { my $a will leave say "bye" = 42; say $a }
camelia rakudo-moar 86b1b2: OUTPUT«5===SORRY!5=== Error while compiling /tmp/0od20F5qIb␤Missing block␤at /tmp/0od20F5qIb:1␤------> 3{ my $a will leave7⏏5 say "bye" = 42; say $a }␤ expecting any of:␤ parameterized block␤»
lizmat m: { my $a will leave { say "bye" } = 42; say $a }
camelia rakudo-moar 86b1b2: OUTPUT«42␤bye␤»
lizmat m: { my $a will end { say "bye" } = 42; say $a } 19:31
camelia rakudo-moar 86b1b2: OUTPUT«42␤»
rurban FROGGS[mobile]: 3 kids? Wife also? there are 2 more wifes there.
lizmat hmmm... guess something's not bothering exeucting the END block?
masak that looks reportable. 19:32
FROGGS[mobile] rurban: aye, and nice :o)
masak anyway, I look forward to `is lazy` provided by a module :)
rurban We have a very nice rooftop terrace in the venue with cafe and perfect overview. kids can stay there 19:33
19:33 domidumont left
FROGGS[mobile] cool... and they want to visit the zoo I think 19:34
19:34 tinyblak left
FROGGS[mobile] but we don't have a concrete Plan yet 19:34
raydiak tried a 'is lazy' module w/the metamodel, but you can't get at the default closure w/out nqp; also tried a version that used a Proxy, but ran into some other problem 19:35
lizmat masak: don't bother, it's NYI 19:36
raydiak .tell adu github.com/andydude/p6-c-parser/issues/4
yoleaux raydiak: I'll pass your message to adu.
lizmat "for some reason exceptions are caught and not rethrown" as a comment in the code
jnthn raydiak: There's a .build method on Attribute
raydiak jnthn: part of the issue was it's a BOOTSTRAPATTR or whatever it's called 19:37
lizmat jnthn: regarding END/INIT phasers: to what Block should I attach those ? 19:38
jnthn lizmat: They don't go on blocks...
lizmat can I get from a PseudoStash to the associated Block ?
ah that would explain
jnthn I think there's an add_phaser method in World iirc
lizmat looks deeper
ah, then the problem is: how to get at World in the trait_mod 19:39
jnthn raydiak: Those only show up in a handful of internal things, not in user code...
lizmat: $*W :P
lizmat aha... hmm...
raydiak jnthn: not sure what I was doing, but it was in my own class; playing with it again now using your .build hint, I'll see how far I get 19:40
FROGGS[mobile] lizmat: you can access $*W because traits are called from the actions directly, and dynamic variables "leak" 19:41
lizmat aha, that make sense :-)
jnthn
.oO( it's a leakture, not a bug... :) )
 19:42
RabidGravy :)
FROGGS[mobile] that's also why Slang::Tuxic can just access %*LANG in EXPORT 19:43
lizmat looks deeper$ 6 ' { my $a will end { say "bye" } = 42 }' 19:45
===SORRY!===
Object of type Block in QAST::WVal, but not in SC
jnthn: ^^
suggestions ?
code in trait: $*W.add_phaser($v.slash, 'END', $block)
jnthn $*W.add_object($block); before it, mebbe 19:46
FROGGS[mobile] $*W.add_object($bl...
lizmat testing
FROGGS[mobile] phone typing ducks
lizmat suckd you mean :-) 19:47
FROGGS[mobile] yeah, that :D
19:48 zeleiadi left, spider-mario left
rurban Wasn't there a plan to support unicode ops also? like → ⇒ ⇔ ≠ ≤ ≥ ÷ I added this now to our perl5: github.com/rurban/perl/commits/unicode-parse 19:49
sjn rurban: you can already make your own unicode operators in Perl 6 19:52
19:52 spider-mario joined
raydiak m: sub trait_mod:<is> (Attribute $a, :$lazy!) { say $a.build.perl }; class Foo { has $.foo is lazy = rand } # is this because the trait is parsed before the default? 19:52
camelia rakudo-moar 86b1b2: OUTPUT«Mu␤»
lizmat sjn: we have them for set operations, why not for "normal" math ops as well ? 19:53
sjn sure :)
jnthn lizmat: It's just a module away ;P
raydiak: Yeah, you'd need to do something at compose time, as in the lazy thing I wrote
lizmat by that logic, we should get rod of unicode set operators and put them in the ecosystem 19:54
*rid
sooner rather than later
DrForr Or resort to trigraphs :)
raydiak jnthn: thanks, will read, haven't backlogged...last tried this a week or two ago 19:55
19:55 mohij left
FROGGS[mobile] gnight 19:55
19:55 FROGGS[mobile] left 19:56 mohij joined 19:57 yqt joined
raydiak backlogs, sees people working on it, goes and does something else :) 20:00
20:01 espadrine left 20:07 mohij left 20:09 Rounin left
dalek kudo/nom: cfdf0a3 | lizmat++ | src/core/Variable.pm:
my $a will end { say "byebye" } now works
 20:10
timotimo is that for when a variable leaves its scope? 20:11
lizmat no, that's "will leave" 20:12
this is when we end the process
vendethiel- lizmat++
lizmat m: { my $a will end { say "bye" } = 42; say $a } END say "byebye"
camelia rakudo-moar 86b1b2: OUTPUT«5===SORRY!5=== Error while compiling /tmp/W246ywT3ty␤Strange text after block (missing semicolon or comma?)␤at /tmp/W246ywT3ty:1␤------> 3$a will end { say "bye" } = 42; say $a }7⏏5 END say "byebye"␤ expecting any of:␤ infix…»
lizmat m: { my $a will end { say "bye" } = 42; say $a }; END say "byebye"
camelia rakudo-moar 86b1b2: OUTPUT«42␤byebye␤»
lizmat $ 6 '{ my $a will end { say "bye" } = 42; say $a }; END say "byebye"' 20:13
42
byebye
bye
vendethiel- Since we lost the "nested phaser" a few days/weeks ago, this "will" thing will most definitely save us :-)
timotimo hum?
vendethiel- mean ENTER { my Lock $l .= new; LEAVE { $l.free } }; <- this used to work but doesn't anymore, right? 20:14
lizmat there is no Lock.free ? 20:15
vendethiel- that's just an exemple of nested phaser.
(about lexical scope etc) 20:16
lizmat ah, but you don't have to have a scope on a phaser 20:17
m: { ENTER my $a = 42; LEAVE say $a }
camelia rakudo-moar 86b1b2: OUTPUT«42␤»
vendethiel- yeah, but in this case, it might be "cleaner" to leave the Lock untouchable from outside the phaser :-) 20:18
lizmat m: { ENTER { my $a = 42; LEAVE say $a } }
camelia rakudo-moar 86b1b2: OUTPUT«42␤»
lizmat m: { ENTER { my $a = 42; LEAVE say $a }; LEAVE say "bye" }
camelia rakudo-moar 86b1b2: OUTPUT«42␤bye␤»
lizmat not the same thing
vendethiel- great. then I'm not sure what I was confusing that with, or which part of the (I think) p6weekly I misread.. 20:19
20:20 diana_olhovik_ left
timotimo if it's from the weekly, it's more likely i explained something poorly than you reading it wrong ;) 20:20
masak vendethiel-: I submitted the LEAVE-inside-ENTER thing not working as an RT ticket, but jnthn rejected it on arguably good grounds. 20:21
vendethiel-: (thrust of the argument was "it's not spec")
vendethiel- aaah, that's probably that then, masak. timotimo++: nope, don't believe that. get some karma instead
masak I could find you the RT ticket if you're interested.
vendethiel- if it's not too much time, I am indeed 20:22
ugexe m: role r1 { has $!rv; submethod BUILD(:$!rv = 1) { say $!rv }; }; role r2 does r1 { has $!rv2; submethod BUILD(:$!rv2 = 2) { say $!rv2 }; }; my $x = "xx"; $x does r2; # is there another way to access self at build time?
camelia rakudo-moar 86b1b2: OUTPUT«Package 'r2' already has a Submethod 'BUILD' (did you mean to declare a multi-method?)␤ in any add_method at src/gen/m-Metamodel.nqp:434␤ in any apply at src/gen/m-Metamodel.nqp:1729␤ in any compose at src/gen/m-Metamodel.nqp:1882␤ in any speci…»
skids ugexe: by abusing attribute default values. 20:23
dalek kudo/nom: 13dc82f | lizmat++ | src/core/Variable.pm:
Some code esthetics
vendethiel- haha, I love how "will begin" is implemented :-)
20:23 spider-mario left, darutoko left
masak vendethiel-: rt.perl.org/Ticket/Display.html?id=123894 20:24
vendethiel- thanks masak
masak (I remembered slightly wrongly, because it wasn't those two phasers. but I think the reply still holds up for your case.)
jnthn masak: Note that { LEAVE say now - ENTER now; sleep 1; } should work out, and is different from { LEAVE { say now - ENTER now } } 20:26
vendethiel- m: my $a = 0 will next { $a++ }; while $a < 50 { say $a }; #let's reinvent C!
camelia rakudo-moar 86b1b2: OUTPUT«5===SORRY!5=== Error while compiling /tmp/CIkcORPWd8␤Two terms in a row␤at /tmp/CIkcORPWd8:1␤------> 3my $a = 07⏏5 will next { $a++ }; while $a < 50 { say␤ expecting any of:␤ infix␤ infix stopper␤ postfix␤ …»
vendethiel- m: my $a will next { $a++ } = 0; while $a < 10 { say $a }; #let's reinvent C!
camelia rakudo-moar 86b1b2: OUTPUT«5===SORRY!5=== Error while compiling /tmp/4H9dQ63DFV␤Variable '$a' is not declared␤at /tmp/4H9dQ63DFV:1␤------> 3my $a will next { $a++7⏏5 } = 0; while $a < 10 { say $a }; #let's␤ expecting any of:␤ postfix␤»
masak jnthn: *nod*
jnthn masak: Rakudo seems ot get the distinctiong right too 20:27
masak nice.
jnthn masak: Sadly, ENTER doesn't work as an rvalue yet
That's one of the tickets open in my browser for a couple of days to look at
But I ended up working on making our hashes use a bunch less memory instead :)
lizmat jnthn: given a &block, how do I get to set $_ inside that block 20:28
that's one of the missing features of will foo { ... } atm
jnthn lizmat: Depends exactly how it's compiled
lizmat: If with implicit, then passing it as an argument when invoking the block should work. 20:29
lizmat it enters as a $block in a trait_mod
jnthn m: my $a = { say $_ }; $a(42)
camelia rakudo-moar 86b1b2: OUTPUT«42␤»
lizmat ah, ok... 20:30
hmmm...
jnthn You can try that. If it doesn't work, we may be able to tweak how will foo { ... } blocks parse...
20:31 spider-mario joined
lizmat $ 6 'my $a will leave { .say } = 42' 20:34
Too many positionals passed; expected 0 arguments but got 1
I guess we need to look at that then
jnthn rule trait_mod:sym<will> { <sym> [ <identifier> || <.panic: 'Invalid name'>] <pblock> } 20:36
Try making it <pblock(1)> at the end
20:36 Foxcool left 20:37 Foxcool joined
lizmat will do 20:38
20:38 vendethiel- is now known as vendethiel
lizmat $ 6 'my $a will leave { .say }; $a = 42' 20:42
(Any)
yeah, it gets the original value of the var
$v.block.add_phaser('LEAVE', { $block($v.var) } ); # too naive 20:43
jnthn Very much so :)
I guess in $v you have the variable's name 20:44
lizmat the Variable object
jnthn Right, but somehow you can get to the name of the symbol
lizmat multi sub trait_mod:<will>(Variable:D $v, $block, :$leave! ) {
yeah, that would be $v.name 20:45
20:45 Foxcool left
jnthn Hm, this is all rather related to the thread safety issues we have with phasers. But for now, for leave/enter at least, we know the caller of the trait is the relevant frame 20:46
So could try something like nqp::atkey(nqp::ctxcaller(nqp::ctx()), $v.name) or so
Which is a cheaper spelling of CALLER::{$v.name} I guess
lizmat will play with that 20:47
20:47 Foxcool joined 20:49 mohij joined
lizmat $ 6 'my Int $a will leave { .say }; $a = 42' 20:50
Lexical with name '$a' does not exist in this frame
guess I need to move a few more frames ?
jnthn oh...argh 20:51
hm, wait, show me le patch
lizmat $v.block.add_phaser('LEAVE', { $block(nqp::atkey(nqp::ctxcaller(nqp::ctx()), $v.name)) } ); 20:52
jnthn yeah, then you're talking about the trait's caller
Oh now, you wrapped it
Hmm
*no
lizmat again, being naive there :-) 20:53
jnthn Yeah, but the wrapping is what I thought you'd forgot
But it looks right
Does --optimize=off help?
lizmat $ perl6 --optimize=off -e 'my Int $a will leave { .say }; $a = 42' 20:54
Lexical with name '$a' does not exist in this frame
nah
jnthn Suspected not.
20:54 Foxcool left
jnthn Ohhh 20:54
The LEAVE thing is processed by something not called directly 20:55
So yes, you do need an extra level
lizmat yeah, trying that now
what is the equivalent of CALLER::CALLER:: ?
jnthn Search for leave_handler or block_exit_handler or so in BOOTSTRAP.nqp for an idea
Just call nqp::callerctx another time :)
ugexe m: my $a = [1,2,3]; my $b = [4,5]; my $c = []; my $d = [$a (|) $b (|) $c]; say $d; # the ordering here seems to be causing pandas current test failure. is it a bug with set or should panda change what it expects order wise? 20:57
camelia rakudo-moar 13dc82: OUTPUT«set(5, 4, 3, 1, 2)␤»
lizmat it shouldn't
jnthn Panda shouldn't expect order there 20:58
lizmat sets are basically a restricted hash and therefore the order is not fixed (anymore)
$ 6 'my Int $a will leave { .say }; $a = 42'
42
yeah!
jnthn Well, it never was on JVM< and Panda has had JVM issues recently too
lizmat: yay :)
20:59 egrep left, egrep joined
lizmat $ 6 'my Int $a will enter { $_ = 42 }; say $a' 21:01
42
aren't KEEP / UNDO / FIRST / NEXT / LAST the same really ?
21:01 TimToady left
masak the latter three may only occur in loops. 21:02
jnthn We won't get it to work on the last 3 I doubt
lizmat yeah, but from the view of scopes, they would be the same, no?
ok
jnthn But KEEP/UNDO are viable with the same technique
As they're just conditioned leaves
lizmat I'll try KEEP / UNDO :-)
masak yeah, KEEP/UNDO are subsets of LEAVE 21:03
21:08 Foxcool joined
lizmat ugexe: merged your PR 21:10
ugexe thanks
lizmat $ 6 'my Int $a will undo { .say }; $a = 42; fail' 21:12
42
:-)
dalek kudo-star-daily: 9f3195c | coke++ | log/ (2 files):
today (automated commit)
 21:13
rl6-roast-data: 7eb6578 | coke++ | / (9 files):
today (automated commit)
masak 'night, #perl6
21:14 zakharyas left
lizmat good night. masak 21:14
RabidGravy yep panda worky again 21:15
21:15 dolmen joined
lizmat looks like FIRST is like ENTER 21:17
$ 6 'for ^2 { my Int $a will first { $_ = 42 }; say $a }'
42
(Int)
RabidGravy nice :-) 21:21
21:22 vendethiel left
dalek ast: 87e6d13 | coke++ | S02-types/declare.t:
RT #124461 - replace flawed test
 21:22
synbot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=124461
21:25 cognominal joined
[Coke] can someone demonstrate a "num" overflow that doesn't happen on Num ? 21:26
21:26 RabidGravy left 21:28 vendethiel joined
[Coke] m: say Num.max; say num.max; 21:30
camelia rakudo-moar 13dc82: OUTPUT«Cannot call max(Num: ); none of these signatures match:␤ (Any:D: *%_)␤ (Any:D: Any $by, *%_)␤ in block <unit> at /tmp/x_Li2GKtJH:1␤␤»
[Coke] m: say num.max; 21:31
camelia rakudo-moar 13dc82: OUTPUT«Cannot call max(num: ); none of these signatures match:␤ (Any:D: *%_)␤ (Any:D: Any $by, *%_)␤ in block <unit> at /tmp/618_W1lICG:1␤␤»
lizmat m: say num.Range.max
camelia rakudo-moar 13dc82: OUTPUT«Inf␤»
[Coke] (locally that says -Inf)
lizmat m: say int.Range.max
camelia rakudo-moar 13dc82: OUTPUT«9223372036854775807␤»
lizmat m: say int8.Range.max
camelia rakudo-moar 13dc82: OUTPUT«127␤»
lizmat m: say bit.Range.max 21:32
camelia rakudo-moar 13dc82: OUTPUT«Inf␤»
lizmat yuck
21:32 bjz left
lizmat m: say int1.Range.max 21:33
camelia rakudo-moar 13dc82: OUTPUT«0␤»
lizmat m: say int1.Range
camelia rakudo-moar 13dc82: OUTPUT«-1..0␤»
lizmat m: say uint1.Range
camelia rakudo-moar 13dc82: OUTPUT«0..1␤»
21:34 TimToady joined
dalek ast: 45e6d35 | coke++ | S02-types/declare.t:
RT #124461 - explicitly check that type boxed.
 21:34
synbot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=124461
dalek ast: 70231d3 | coke++ | S02-types/declare.t:
RT #124467 - native junctions are no more
 21:36
synbot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=124467
dalek kudo/nom: 87b1eea | lizmat++ | src/ (2 files):
Variable will foo &block cleanup

  - will init/post/compose will now throw NYI instead of silently doing nothing
  - will enter/leave/keep/undo/first/pre will now set $_ correctly
  jnthn++ for support and suggestions!		 21:41
timotimo can we just write multiple "will" in a row without problems? 21:44
lizmat yes: 21:45
$ 6 'my $a will enter { $_ = "1".IO.open } will leave { .close }; say $a.lines'
%*ENV<RAKUDO_VERBOSE_STACKFRAME>=1; sub a { die }; a
the 2nd line is what is in file "1" :-) 21:46
[Coke] down to 1292 tickets.
lizmat looking at this, I wonder whether will phasers such as "enter" shouldn't write their result to $_ automagically
[Coke] uuuugh. 21:47
lizmat seems this NY hackathon created more work than it solved :-)
21:48 dolmen left 21:49 vendethiel left
timotimo they do what now? 21:50
write their result to $_?
is that only for will phasers or also for block phasers?
lizmat nono, the variable is available as $_ inside the scope
[Coke] lizmat: little bit, yah.
timotimo ah
so you could my $a will enter { .=lc }; OSLT? 21:51
lizmat my $a will enter { "1".IO.open } will leave { .close } # feels more natural
my $a will enter { .=lc } = "FOO"; say $a # na, the initialization happens *after* the enter block is executed 21:52
so I guess the use of "will enter" is limited anyway 21:54
timotimo now to look at the racyness of Proc::Async
21:54 spider-mario left
timotimo how hard is it gonna be anyway ... ~_~ 21:54
21:55 spider-mario joined
jnthn [Coke]: num and Num can store the same value range 21:57
[Coke]: Though I had an idea for testing the decl is effective: write a multi x(num) { 1 }, multi x (Num) { 2 }, and then make sure calling it on a variable decalred as a num calls #1
21:58 virtualsue left
timotimo oh wow. these valgrind thread error detector tools are *really* noisy 21:59
==30166== ERROR SUMMARY: 2464 errors from 104 contexts (suppressed: 4531 from 77)
probably not going to go through all of these tonight
lizmat wishes #perl6 a good night! 22:00
jnthn o/ lizmat
timotimo gnite lizmat
jnthn timotimo: Yeah, well we deliberately race on things like the call count used for specialization. 22:01
timotimo it seems like the DRD thinks our frame invocation is racy ... could be it just doesn't understand the libatomicops thing we use there?
==30166== Conflicting load by thread 2 at 0x09d0fca0 size 8
==30166== at 0x4D874DD: AO_compare_and_swap_full (x86.h:142) 22:02
jnthn That's also possible
timotimo it might have an option for this kind of thing
22:02 Foxcool left
jnthn I've not had time to dig deeply into its analysis model yet 22:02
timotimo The macros DRD_IGNORE_VAR(x), ANNOTATE_TRACE_MEMORY(&x) and the corresponding client request VG_USERREQ__DRD_START_SUPPRESSION. Some applications contain intentional races. There exist e.g. applications where the same value is assigned to a shared variable from two different threads. It may be more convenient to suppress such races than to solve these. This client request allows to suppress such races. 22:04
there's also ANNOTATE_BENIGN_RACE_SIZED and ANNOTATE_BENIGN_RACE_STATIC
ugexe is there a way with phasers to access self to do something like: class XX { has $.var is rw; END { say "var = $.var" }; }; my $q = XX.new(var => 1); 22:05
jnthn ugexe: No, because the phaser is not once per instance. 22:06
ugexe: I'd maintain a list of instances and then have the phaser to over them 22:07
*go over them
timotimo jnthn: how do you feel about adding client requests all over moarvm's code base? 22:11
jnthn timotimo: I generally dislike clutter.
timotimo fair enough 22:12
jnthn Is there not a supressions file approach like in valgrind?
uh, valgrind memcheck I mean
timotimo i haven't read up on that yet
jnthn k
timotimo suppression file lines start with what the suppression is called and then what tools it applies to 22:16
it seems like the method itself is tool-independent, or tool-agnostic or something like that
dalek ast: 5f6d6a6 | coke++ | S02-types/declare.t:
RT #124461 add another verification for num

Hard to distinguish between between num/Num
  jnthn++		 22:18
synbot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=124461
timotimo jnthn: what about teaching the valgrinds about how the fixed size allocator works so that we'll be pointed to the place where a chunk of it gets "allocated", rather than where a page is added 22:19
jnthn timotimo: I usually just switch the FSA into debug mode if I want that so far. 22:20
timotimo helgrind seems to report writes to any place inside a page as conflicting
jnthn timotimo: Where it degenerates to using malloc for all
Ah
timotimo ah, neato
there's a -D for that?
jnthn you just tweak an 0 to a 1 at the top of fixedsizealloc.c
But feel free to make something more convenient
timotimo OK
maybe later 22:21
22:22 kaare_ left 22:28 rurban left
timotimo so one of the races seems to be spesh possibly dropping in a spesh candidate while other threads could be invoking the frame 22:31
i think we race that intentionally? 22:32
jnthn I think we do something to protect the installation of 'em too 22:34
timotimo huh ... interesting. reentrantmutex_lock (a read) is apparently conflicting with initialize_mutex (a write)
there are interesting things to see in here
but the chances that i'll figure out what's wrong with the Proc::Async implementation with this tool are ... slim
22:40 mohij left 22:46 risou is now known as risou_awy, risou_awy is now known as risou
timotimo the gen2 seems to trip up helgrind, too 22:49
22:53 rarara joined
timotimo Address 0x9ea49f0 is 9,824 bytes inside a block of size 12,288 alloc'd - this is all so useless :( 22:54
rjbs perl6 has --doc to show program docs and -M which has to resolve a module, but I don't see a way to get module docs. Am I missing something? 23:05
timotimo oh, that's a good point
23:09 yqt left 23:12 vendethiel joined 23:23 spider-mario left 23:28 Isp-sec left 23:34 gfldex left, vendethiel left
rjbs m: my %h; %h<a> = 1; %h.perl 23:51
camelia ( no output )
rjbs m: my %h; %h<a> = 1; say %h.perl
camelia rakudo-moar 87b1ee: OUTPUT«{:a(1)}<>␤»
rjbs What's the trailing <> mean?
timotimo "decont" 23:56
23:56 mule joined, mule left
timotimo i'm hoping to find something in valgrind that'd let us annotate a traceback that includes MVM_interp_run with the moar bytecode traceback 23:57
rjbs I see that "decont" has already been complained about recently, so I will be mostly quiet. 23:59

Powered by Cro and the Raku® Programming Language.

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

Thank you!