»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg camelia perl6: ... | irclog: irc.perl6.org | UTF-8 is our friend!
Set by sorear on 25 June 2013.
dalek kudo/nom: 5a1e607 | jnthn++ | src/core/MapIter.pm:
Fast-path for common case in MapIter on JVM.
 00:09
kudo/nom: 76c3008 | jnthn++ | src/core/MapIter.pm:
A couple more MapIter optimizations.
kudo/nom: f164c0f | jnthn++ | src/core/ (2 files):
A couple more buffer pre-size optimizations.
00:09 eternaleye left
jnthn Those make a 30s thing into a 7s thing... :) 00:10
00:10 eternaleye joined
jnthn .tell sorear for ^64 X ^64 -> $i, $j { $i + $j } seems to generate a stack overflow, even though METAOP_CROSS (src/core/metaops.pm) is using an iterative, not recursive, algorithm. Could it be something with gather/take? 00:15
yoleaux jnthn: I'll pass your message to sorear.
00:17 eternaleye left
jnthn .tell sorear it does appear to be a huge number of the block that does nqp::continuationinvoke piled up on the stack, having dug a little further. 00:18
yoleaux jnthn: I'll pass your message to sorear.
00:22 eternaleye joined
jnthn 'night, #perl6 00:29
00:31 colomon joined
TimToady the use of $*FATAL in rakudo to define fail semantics looks a bit suspicious to me, unless every single dynamic scope sets it true or false according to its lexical context, which I doubt 00:56
Teratogen kill it with fire! 00:57
TimToady shoots firewords sideways at the spectators 00:59
*works
TimToady goes to fetch Pollo Loco & 01:00
01:01 FROGGS_ joined 01:19 raiph joined 01:32 colomon left 01:42 ztt joined 01:54 colomon joined 02:10 aindilis joined
timotimo jnthn: t.h8.lv/p6bench/2013-07-05_mapiter_improv.html - you made it better! 02:18
02:20 jaldhar joined 02:30 SamuraiJack joined 02:31 jaldhar left 02:34 jaldhar joined
benabik timotimo: That's a lot of graphs. 02:46
labster it sure is a lot of them 02:49
03:01 SamuraiJack left 03:02 SamuraiJack joined
sorear good * #perl6 03:03
yoleaux 00:15Z <jnthn> sorear: for ^64 X ^64 -> $i, $j { $i + $j } seems to generate a stack overflow, even though METAOP_CROSS (src/core/metaops.pm) is using an iterative, not recursive, algorithm. Could it be something with gather/take?
00:18Z <jnthn> sorear: it does appear to be a huge number of the block that does nqp::continuationinvoke piled up on the stack, having dug a little further.
JimmyZ good * sorear 03:16
03:21 colomon left
sorear o/, JimmyZ 03:27
03:32 BenGoldberg left 03:34 colomon joined 03:37 xilo_ left 03:39 xilo_ joined 03:47 colomon left 03:53 SamuraiJack left 03:54 ztt left 04:04 benabik left 04:35 benabik joined 04:39 benabik left, benabik joined
labster r: sub foo { foo(); }; my $a = try foo(); say "alive"; say $!; 04:52
camelia rakudo f164c0: OUTPUT«(timeout)alive␤»
labster is $! infinite looping there? 04:53
04:55 crab2313 joined 05:01 crab2313 left
TimToady I doubt it, but it's probably got a huge stack trace to process 05:13
Teratogen I've done a lot of perl 5 but very little perl 6 05:15
am I going to have to learn a whole new language?
sorear really, Teratogen? this again? 05:16
05:16 kaare_ joined
sorear it's not funny. 05:16
Teratogen it was not meant to be funny 05:19
the more examples of perl 6 I see on this channel the more DIFFERENT it seems to be from perl 56
the more examples of perl 6 I see on this channel the more DIFFERENT it seems to be from perl 5
but... I am going to sleep 05:20
night all!
=D
geekosaur apparently you've not done enough perl5 to understand its limitations 05:21
05:37 mstefanko joined
labster TimToady: Looking at S05:4618 - 4629. Is cat supposed to be join there? 05:38
TimToady no, cat is something that just pretends to be a string 05:39
NYI
sorear labster: cat is lazy join
05:40 raiph left
labster oh, hadn't seen it yet. 05:40
sorear cat map { *~" " }, 1..* # returns a one-sided infinite rope of all the integers 05:41
NYI.
TimToady supposed to make an array or list look like a string good enough to do matching on, without necessarily doing an actual join
sorear will probably be easy enough to do once we have (strict) ropes
labster Cool. Then I can do Dog.cat
TimToady gotta work a Rat in there too 05:42
sorear we have Dogs in the spec now? o_O 05:43
TimToady sure, but they're all meta-Dogs 05:44
labster grep says Dog is in nine specs
TimToady one might say they're all foo dogs.. 05:50
sorear metasyntactic dogs? 05:51
.oO( since orthosyntax stole ... ??? !!!, metasyntax has had to make do with Dog )
 05:52
05:52 xilo_ left
TimToady std: .oO(...) 05:53
camelia std 85cf51d: OUTPUT«ok 00:00 42m␤»
TimToady has to wonder, as a one-time organic chemist, what parasyntax would be... 06:01
labster syntax for parakudo? 06:13
06:18 jaldhar left 06:19 jaldhar joined
sorear that may have been the joke 06:36
moritz good morning
moritz just had pancakes with ice cream and fresh strawberries for breakfast o/ 06:37
sorear good morning moritz \o/ yum
TimToady was referring to relative positions of two radicals on benzene rings: www.mhhe.com/physsci/chemistry/care...es/omp.gif 06:40
06:42 PacoAir joined
moritz reading the postgres docs can be amusing, occasionally. For example www.postgresql.org/docs/9.1/static/...acter.html 06:43
An attempt to store a longer string into a column of these types will result in an error, unless the excess characters are all spaces, in which case the string will be truncated to the maximum length. (This somewhat bizarre exception is required by the SQL standard.)
sorear mmm, punchcardisms? 06:46
JimmyZ moritz: INSERT INTO test VALUES ('12345678'::varchar(5)); won't be in an error 06:47
06:50 kaare__ joined
moritz it is also slightly obscure that 'time' (time of day without date) takes up more storage space than 'timestamp' (time of day + date) 06:51
06:53 cognominal joined
sorear obscure, indeed... 06:53
06:53 kaare_ left 06:58 berekuk left 06:59 Rotwang joined 07:01 berekuk joined, aindilis left
moritz another gem from the pg docs: " conventions before the 19th century make for interesting reading, but are not consistent enough to warrant coding into a date/time handler." 07:01
07:42 domidumont joined 07:43 FROGGS_ is now known as FROGGS, PacoAir left
FROGGS gmorning 07:43
sorear o/ FROGGS 07:44
sorear peeks in bin/org/perl6/nqp/runtime/ and realizes that all but one of the nested classes are his fault 07:52
dalek ast: ec62150 | (Brent Laabs)++ | S05-match/capturing-contexts.t:
RT #117461 Test for Match.lines (which really means Match ~~ Cool)
 08:00
08:02 domidumont left 08:03 araujo joined
labster rn: r: class A { state $x = 0; method inc { ++$x }; }; say A.inc, A.inc, (my $x = A.new).inc, A.inc, $x.inc; 08:19
camelia niecza v24-86-g39ab531: OUTPUT«55555␤»
..rakudo f164c0: OUTPUT«===SORRY!===␤Two terms in a row␤at /tmp/E2iHoBLE4K:1␤------> r⏏: class A { state $x = 0; method inc { +␤ expecting any of:␤ argument list␤ postfix␤ infix stopper␤ infix or meta-infix␤ statem…
labster rn: class A { state $x = 0; method inc { ++$x }; }; say A.inc, A.inc, (my $x = A.new).inc, A.inc, $x.inc;
camelia rakudo f164c0: OUTPUT«12345␤»
..niecza v24-86-g39ab531: OUTPUT«55555␤»
sorear surprised rakudo gets that wrong: 08:21
rn: my $x = 3; (++$x) *= 2; say $x
camelia niecza v24-86-g39ab531: OUTPUT«8␤»
..rakudo f164c0: OUTPUT«Cannot assign to a readonly variable or a value␤ in block at src/gen/CORE.setting:14368␤ in block at /tmp/RNV3r6prvd:1␤␤»
sorear well, "wrong" is arguable I guess 08:22
but in niecza prefix:<++> has always returned a lvalue
because it doesn't allocate any containers - it modifies the argument in place and returns it 08:23
also it predates the read-only scalar container elision optimization
testing with perl5 seems to indicate that both prefix:<++> and postfix:<++> return lvalues o_O 08:24
labster .oO ( kyrie containers )
prefix:<++> as an lvalue seems sane enough. 08:25
sorear anyway, lvalueness means that say() receives an argument list containing five of the same container
which of course all have the same value
labster ah, indeed. 08:26
the whole 'state $x' sharing a variable between instances -- that maybe should be documented :/ 08:27
sorear erm
rn: class A { my $x = 0; method inc { ++$x }; }; say A.inc, A.inc, (my $x = A.new).inc, A.inc, A.inc;
camelia niecza v24-86-g39ab531: OUTPUT«Potential difficulties:␤ $x is declared but not used at /tmp/lDc2fPCk9b line 1:␤------> inc { ++$x }; }; say A.inc, A.inc, (my ⏏$x = A.new).inc, A.inc, A.inc;␤␤55555␤» 08:28
..rakudo f164c0: OUTPUT«12345␤»
sorear you're not seeing any special behavior of state in this example.
it's shared between instances because *it's declared outside the method*
labster S04: Separate clones of the closure get separate state variables. However, recursive calls to the same clone use the same state variable. 08:29
classes aren't closures per se?
sorear At no point do you clone the closure which owns "state $x"
which is the class body sub, which is executed only once in any event
s/sub/block/ 08:30
labster fairy nuff
sorear also my/our/state/anon variables are completely orthogonal to the object system
if you want each instance of an object to have its own version of a variable, use "has" 08:31
labster sorear++ 08:33
that needs to be explained more clearly somewhere
moritz doc.perl6.org/language/objects#Attributes # like this? 08:49
08:51 FROGGS left 08:52 preflex joined, ChanServ sets mode: +v preflex
labster not exactly like that, because that doesn't say what happens with my/our/state in classes 08:55
nor is perlcabal.org/syn/S12.html#Class_attributes all that clear on the subject
lizmat good *, #perl6! 09:01
(still checking on backlog)
meanwhile: for is the same as map, right?
r: sub a { for 1..3 { return $_ } }; say a # works
camelia rakudo f164c0: OUTPUT«1␤»
lizmat r: sub a { (1..3).map( { return $_ } ) }; say a # doesn't work
camelia rakudo f164c0: OUTPUT«Attempt to return outside of any Routine␤ in block at src/gen/CORE.setting:486␤ in block at /tmp/Yl0CJbNEKO:1␤␤»
lizmat is this a rakudobug? 09:02
09:04 iSlug joined
sorear it's a side effect of map being lazy 09:07
and return being lexotic and binding only to 'a'
error message is misworded 09:08
masak good antenoon, #perl6
sorear o/ masak!!!
lizmat n: sub a { (1..3).map( { return $_ } ) }; say a 09:11
camelia niecza v24-86-g39ab531: OUTPUT«Unhandled exception: Cannot unbox a Variable[] from an object of repr P6box[Int32]␤ at <unknown> line 0 (KERNEL map @ 0) ␤ at /home/p6eval/niecza/lib/CORE.setting line 2618 (GatherIterator.reify @ 6) ␤ at <unknown> line 0 (ExitRunloop @ 0) ␤ at /home/p…
lizmat masak! 09:12
09:12 spider-mario joined
dalek kudo/nom: 5c61f12 | (Elizabeth Mattijsen)++ | src/core/Nil.pm:
Add reference to BOOTSTRAP
 09:15
masak so, 18:00Z tonight, then. some more sprintf hacking. 09:18
I'll be away for a while. going to go attent a friend's wedding. but I'll be back in time for 18:00Z :)
does Teratogen have a reset function which makes him come back and ask the same questions again and again? 09:23
(re backlog) 09:24
I'm not upset or irritated or anything. but if we understand Teratogen's reset function semantics better, maybe we can answer his personal FAQs better.
sorear sorear$ ./nqp -e 'my %h := nqp::jvmbootinterop("java.lang.Thread"); say(%h<method:toString()Ljava/lang/String;>(%h<static_method:currentThread()Ljava/lang/Thread;>()))' 09:25
Thread[main,5,main]
clunky, but a start
dalek p: f27b205 | sorear++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/BootJavaInterop.java:
Java interop, part 2
 09:27
p: 4f15d7b | sorear++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/ (2 files):
First working version of java interop coderef generator
tadzik wow 09:29
sorear++
masak sorear++! 09:30
tadzik: hey hey
tadzik: are you planning to forget about tonight's sprintf hacking, too? :P 09:31
tadzik masak: oh, is there one tonight? :) 09:32
sorear masak: you're planning to forget? have fun at the wedding then. 09:33
tadzik life pro tip: if you leave your balcony open, you're going to have a pidgeon's nest in it
sorear next subgoal: callbacks! then attempt to document what exists and how it's intended to be used.
tadzik ...shit
sorear tadzik: type error with sentence. USese::Balcony does not do role "Closeable" 09:34
JimmyZ sorear++ 09:35
sorear autopun? 02:04 < Cale> I tried teaching autodidactism, but it seems people just have to learn it for themselves.
lizmat smells like one 09:36
masak yes. 09:38
sorear: also, bash.org/?3936
lizmat TimToady: stopped trying to build Rakudo with Nil outside of Any
oddly enough, it built the CORE.setting, but not the restricted one 09:39
didn't get to spectest it at all
masak I wonder if perhaps Perl 6 is the biggest spectest suite in history on which language experiments of this magnitude are performed. 09:41
Perl 6's* 09:42
sorear I doubt it
moritz wouldn't be surprised if for example Fortress had a test suite of similar size 09:43
sorear was thinking of C++ 09:44
I know ginormous test suites exist for SQL, but I think SQL is more frozen than C++
lizmat aha, the error "Odd number of elements found where hash expected" while trying to create restricted setting 09:45
is not cause by the creation of the restricted setting, but simply by loading an apparenly incorrectly generated CORE.setting
sorear sounds like a %hash = Nil thing 09:46
not being treated as %hash = ()
lizmat yup, now to find that… :-(
probably not a single one either
lizmat goes spelunking
sorear no backtraces? 09:47
lizmat nope 09:48
masak & 09:55
lizmat happy wedding, masak!
09:56 tomaw joined 09:59 grondilu joined
grondilu is trying to compile rakudo-jvm again, this time with a fresh openjdk 10:00
grondilu managed to reach "stage jast" after about ten minutes of compilation
jnthn lizmat: --ll-exception may get you something more helpful 10:02
sorear grondilu: on my machine, rakudo-jvm compiles about three times faster than rakudo-parrot. 10:03
grondilu it worked!
grondilu is about to run perl6 on JVM for the first time!
sorear oh, wait... you said earlier you were on 32-bit D:
jnthn ooh, I see sorear++ patches 10:04
sorear nqp-jvm does not have 32-bit types
it does everything using 64-bit integers
which have to be emulated on 32-bit platforms
I suspect this plays some part in the slowness you see
dalek kudo/nom: ba7f78d | (Elizabeth Mattijsen)++ | src/core/control.pm:
Removed now superfluous class Nil {} definition (now done in BOOTSTRAP)
 10:05
grondilu successfully ran "say "Hello, JVM!"
sorear wait x2. I thought jdk 1.7 dropped support for 32-bit CPUs
grondilu jnthn: awesome work, sir
timotimo jnthn: did you see the cool improvements your fast path caused? 10:06
grondilu tries 'my $sum; for ^10_000 { $sum += rand }; say $sum' and appreciates how fast this is 10:07
jnthn timotimo: Not yet, still backlogging...
sorear grondilu: wait, the setting compilation is several-x slower but that is faster? 10:08
jnthn sorear: Looks like interop is going well ;)
grondilu sorear: yes. I've no idea why 10:09
lizmat jnthn++, it does :-)
grondilu notices that the more perlish 'say [+] rand xx 10_000' is a bit slower (but still faster than rakudo)
jnthn
.oO( rakudo is faster than rakudo? :D )
 10:10
timotimo: Found the link; looking :)
grondilu I meant parrot
sorry
sorear thought thunky xx was a nieczaism 10:11
timotimo no way 10:12
thunky xx is crazy useful!
grondilu any chance nqp-jvm supports 32-bit types in a near future? Compilation worked but it was kind of a pain. 10:13
jnthn iirc, thunky xx got spec'd and Rakudo changed to do it that way.
timotimo what's the simplest way to get around the thunkyness? here's my first idea:
r: my @a = (state $ = rand) xx 10; say @a; 10:14
camelia rakudo ba7f78: OUTPUT«(Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any)␤»
timotimo er, oops
jnthn r: say rand xx 10
sorear grondilu: I have no proof word size is the actual issue.
camelia rakudo ba7f78: OUTPUT«0.126966395971493 0.108979036604193 0.941956717378112 0.661813663912728 0.517164013101123 0.674946465666235 0.892848095431951 0.793093453851188 0.0593905554276262 0.684856679539546␤»
jnthn timotimo: Yes, the graphs seem better now :)
jnthn will have to optimize some more :) 10:15
timotimo r: my $r = rand; my @a = $r xx 10; say @a; @a[2] = 100; say @a
camelia rakudo ba7f78: OUTPUT«0.23256417466061 0.23256417466061 0.23256417466061 0.23256417466061 0.23256417466061 0.23256417466061 0.23256417466061 0.23256417466061 0.23256417466061 0.23256417466061␤0.23256417466061 0.23256417466061 100 0.23256417466061 0.23256417466061 0.23256417466061 0.2325…
sorear grondilu: diakopter thinks the OpenJDK optimizer is a joke compared to the one shipped in the closed-source releases. He might be right.
grondilu ahh. too bad.
sorear grondilu: since the oracle 1.7 releases are 64-only, the only way to test this is to set up a 64-bit openjdk and test performance 10:16
grondilu: do you have any idea how hard it would be to build a 64-bit openjdk on a 64-bit mac?
grondilu sorear: no idea. sorry. 10:17
but openjdk is easy to install, isn't it? 10:18
10:18 colomon joined
sorear o/ colomon 10:19
colomon \o sorear
how goes it?
grondilu me is trying rosettacode.org/wiki/SHA-256#Perl_6 on JVM 10:20
^ wow this actually worked!
noooo 10:21
grondilu feels stupid
I was actually still running perl6 on parrot. Somehow the old executable was not replaced or something.
grondilu runs make install 10:22
colomon java interop! \o/ 10:23
grondilu makes a jakudo link in /usr/local/bin 10:24
timotimo push @b, 0 until (8*@b-448) %% 512; - this seems very wasteful, wouldn't one rather push @b, 0 xx ((8 * @b - 448) % 512)? 10:25
lizmat jnthn: could it be that a bare {} somehow would be generated as {Nil} and thus cause the odd number in hash error? 10:29
grondilu ok now that I can really discriminate rakudo-parrot and rakudo-jvm, I can say that no, rakudo-jvm is not faster on my machine. I tried 'my $sum; for ^10_000 { $sum += rand }; say $sum;'. Got 4 seconds on parrot, 16 seconds on JVM.
sorear jnthn++ # blogt and I missd it
10:30 Rotwang left
colomon Time::Duration went from passing to test failure overnight. 10:30
lizmat colomon: i assume that's jakudo, right? cause I haven't seen any here in any recent spectest on parakudo 10:31
jnthn lizmat: Doesn't sound likely...
colomon lizmat: that's parakudo, it's from the module smoke test
jnthn lizmat: oh...maybe...
colomon: Are the any details on the failure somewhere? 10:32
colomon not stored, but I can try to install by hand...
lizmat jnthn: src/core/operators has " obj.HOW.mixin(obj, $role).BUILD_LEAST_DERIVED({});", note the {} at the end 10:33
this winds up at "sub circumfix:<{ }>(*@elems) { my $x = Hash.new.STORE(@elems); }", and that bombs 10:34
jnthn lizmat: hmm
r: my $a = -> { }; say $a().WHAT
camelia rakudo ba7f78: OUTPUT«Nil␤»
sorear because inside the {} is an empty block, and empty blocks are Nil
jnthn lizmat: Is ^ still meant to be the case?
lizmat jnthn: good question 10:35
sorear jnthn: Would it be appropriate to make a t/qast/01-qast test to ensure that the compiler can cope with QAST::SVal.new( :value( nqp::x("\x[3000]", 30000) ) ) ? 10:37
lizmat r: say {}.WHAT # but isn't this what the {} is about in the BUILD_LEAST_DERIVED context?
camelia rakudo ba7f78: OUTPUT«(Hash)␤»
lizmat it's winding up at the circumfix {} code 10:38
jnthn sorear: Yeah, should test it somewhere
lizmat: The action method for circumfix:<{ }> is what decides if it's a block or a hash.
lizmat: We always parse it as a block, and then look at it and decide if it should be a hash or not 10:39
sorear lizmat: Moreover, the action method involves calling the generated block and passing whatever it returns to circumfix:sym<{ }> 10:40
lizmat hmmm.. that's hidden from the stacktrace, then 10:41
10:46 grondilu left
lizmat put in a little check for @ being @=(Nil) in circumfix, and pop that, and now it builds :-) 10:49
running spectest now 10:50
sorear there's probably a better way 10:54
(you probably think the same)
10:54 rindolf joined
jnthn Yeah, do it in Actions... 10:55
lizmat spectest is definitely not clean :-) 10:57
sorear lizmat: how about the sanitytest?
lizmat sanity test is ok 10:58
sorear jnthn: I was thinking more along the lines of "empty block needs to be () in list context"
jnthn sorear: Well, that's why I was asking what current spec says about what empty block means...
lizmat r: my @a= ->{}; say @a.elems # should say 0, is what jnthn is saying? 10:59
camelia rakudo ba7f78: OUTPUT«1␤»
jnthn oh hah, just realized...I never ported compile time bind analysis, which means the inliner never inlines anything, such as the native types...
(on JVM)
lizmat hmm… it also looks like t/spec/integration/99problems-21-to-30.rakudo.parrot infiniloops 11:00
sorear jnthn: think we might get a small speedup from that?
timotimo *small*? :D
jnthn sorear: Um, it can turn a Perl 6 multi-dispatch into an ladd :)
timotimo jnthn: i'm ready to fire you up ze benchmarks, but only if it's in about 1-2 hours
jnthn timotimo: I'm gonna fix up nextsame et al. first. 11:01
sorear: Probably means we do somewhat better on the native benchmarks... 11:02
sorear: And in all the built-ins that use native types in their impl.
sorear yeah, I kind of wonder if there are enough of those to make a difference 11:05
11:05 woolfy1 joined
sorear I don't actually know what kind of win the optimizer was when it first came it 11:05
in
11:07 Bucciarati joined
[Coke] wonders where is_run is actually defined. Test/Util.pm has lots of protos, but no actual impl. 11:07
doh, it's there, just beyond my scroll. 11:08
sorear jnthn: I'm pretty sure that the gather/take issue is because I stole a concept from Scheme and the papers I was basing it off of implicitly assumed proper tail recursion 11:13
I can see a couple ways to fix this, need to decide which is the most elegant
jnthn sorear: ah..that explains it.
sorear jnthn: every time you take() it seems to add a frame to the bottom of the saved stack fragment that does nothing 11:14
jnthn sorear: Which nicely explains how things that take enough stack overflow... 11:17
11:19 colomon left 11:23 rindolf left 11:28 preflex_ joined, ChanServ sets mode: +v preflex_
lizmat will look at stuff later, now cycling in the searing heat / sun& 11:28
11:29 preflex left, preflex_ is now known as preflex
[Coke] java's exec doesn't handle shell redirection. Trying to figure out if I have to parse (at a minimum) > and 2> myself, or if I can cheat. 11:30
timotimo gist.github.com/timo/2c85b3c62372b72468a3 <- link to source code in the description of the gist; does this ring a bell for any of our jvm-porters? 11:31
11:31 colomon joined
jnthn [Coke]: Well, I guess it's possible to invoke a shell... 11:32
[Coke]: But that may cause other issues...
11:32 berekuk left
jnthn timotimo: There's a spectest that explodes that way, iirc 11:32
timotimo: But it may be completely unrelated.
timotimo fwiw, the == candidate is \a \b, probably comes from $n.Int, which in this case should totally be a valid integer. hm. 11:33
11:34 berekuk joined
jnthn timotimo: Does it do mixins? 11:34
### Cannot call 'Numeric'; none of these signatures match:
* S14-roles/mixin.rakudo.jvm aborted 7 test(s)
That's the spectest that fails in the same way 11:35
sorear [Coke]: you're looking for java.lang.ProcessBuilder. 11:36
oh, you were doing shell, not ``
11:36 colomon left
timotimo i don't think it does 11:37
i don't think it does, but could it have something to do with enums?
jnthn Hm, maybe
timotimo is unable to reproduce it in golfed form so far
sorear (you want ProcessBuilder anyway because exec() doesn't let you NOT automatically redirect output to pipes)
jnthn enum spectests are in quite good shape, though...
timotimo no luck so far. 11:38
[Coke] I don't want processbuilder, but that means I have to parse out > and 2> myself.
s/but/because/ 11:39
I think I have a *nix version working with exec(['sh', '-c', $cmd])
\o/ 11:40
jnthn: mind if I commit the *nix only version to start? 11:41
dalek p: 6602633 | jnthn++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/CallFrame.java:
CallFrame stashing space for args/csd.

Could do this always and save passing them around some, but for now this is just for Rakudo's benefit. Saves us having to savecapture in every frame just for the sake of nextsame et al.
sorear [Coke]: note that if you don't use ProcessBuilder and inheritIO, *unredirected* output will go to the wrong place
[Coke] sorear: where is "the wrong place"? Right now it goes nowhere. 11:42
timotimo r: enum Frob <FOO BAR BAZ>; class Test { has Int $.w; has Int $.h; method foo() { [Frob.pick xx $!w] xx $!h } }; Test.new(:w(5), :h(5)).foo() # does not trigger the error
camelia rakudo ba7f78: ( no output )
dalek kudo/nom: cab99b2 | jnthn++ | src/vm/jvm/runtime/org/perl6/rakudo/Ops.java:
Get nextsame, callsame etc. tests passing.
 11:43
sorear [Coke]: into an anonymous pipe that you can read with proc.getOutputStream()
timotimo .o( i think the implementation of xx could be factored out quite a bit )
[Coke] sorear: is shell supposed to display or return any output? 11:44
I thought it discarded output and returned an exit status.
jnthn: what does "say $*OS" say for you? 11:45
jnthn On JVM? "Windows 7" 11:46
sorear [Coke]: It's supposed to allow the subprocess output to pass freely to the parent's standard IO 11:47
11:47 berekuk left
[Coke] r: shell('ls'); 11:48
sorear with no interaction from the parent, so you can do shell("vi", $file)
camelia rakudo ba7f78: OUTPUT«shell is disallowed in restricted setting␤ in sub restricted at src/RESTRICTED.setting:2␤ in sub shell at src/RESTRICTED.setting:7␤ in block at /tmp/_yGLsYxjNx:1␤␤»
sorear shell(%*ENV<VISUAL> // %*ENV<EDITOR> // "vi", $file) :)
11:48 berekuk joined
[Coke] sorear: Ok. I give up, then. 11:49
timotimo gist.github.com/timo/91483e54fa0b20173f5f <- ? 11:50
[Coke] Here is the wrong version: gist.github.com/coke/5939689 11:51
to do it the write way, we need to be able to parse all the shell metacharacters ourselves. I suggest we find a third party library instead.
*right 11:52
sorear Process proc = new ProcessBuilder("sh", "-c", cmd).inheritIO().start()
ought to work
sorear suspects jnthn++ can whip up an if statement involving cmd.exe 11:54
11:54 PacoAir joined
jnthn I guess by looking at System.getProperty("os.name") and then "cmd", "/c/" cmd 11:56
[Coke] wonders if we have any tests for sorear's case in roast. 11:57
11:57 berekuk left
[Coke] sorear++ # that worked. 11:57
sorear jnthn: mainly that I don't know what %*VM<properties><os.name> is on windows, or what the proper syntax for cmd /c/ is 11:58
11:59 berekuk joined
timotimo oh, i may need to nqp::p6bool that ... 11:59
dalek ast: 034ed21 | coke++ | packages/Test/Util.pm:
workaround exec name issue on nqp-jvm
 12:01
p: ef6648d | coke++ | src/vm/jvm/ (2 files):
add (unix only) nqp::shell for jvm

  sorear++		 12:02
kudo/nom: 0eab21a | coke++ | src/core/control.pm:
shell -> nqp::shell
 12:03
[Coke] sorear: there, now it's jnthn++'s problem. :)
jnthn :P
[Coke]: Does that make spectests that depend on it run? :) 12:04
If so, that's quite a few test files we win :D
[Coke] jnthn: at least one runs to completion now, yes.
doing a rebuild with latest changes, spectesting. 12:05
jnthn k
jnthn is looking into why installing a second categorical clobbers the first. 12:07
sorear mixin funs? 12:09
jnthn Could well be.
timotimo my patch above seems to win us a tiny bit of performance 12:11
12:11 berekuk left
timotimo that's probably the jit figuring out everything by itself or something 12:11
12:12 berekuk joined
jnthn Hm, if it's mixin fun it's pretty weird mixin fun 'cus the mro length changes :) 12:14
ohhh
I bet I know.
12:16 birdwindupbird joined
JimmyZ [Coke]: github.com/clojure/clojure/blob/ma...l.clj#L140 looks like support 1>&2 12:17
JFYI 12:18
[Coke] JimmyZ: I have something that looks like it works now. thanks for the pointer, though!
JimmyZ :P 12:21
12:21 berekuk left 12:24 berekuk joined 12:33 ztt joined 12:35 berekuk left, jferrero joined 12:36 berekuk joined
[Coke] r: say 24604 / 26108 12:42
camelia rakudo 0eab21: OUTPUT«0.942393␤» 12:43
timotimo sweet :)
[Coke] r: say .942393-.9348
camelia rakudo 0eab21: OUTPUT«0.007593␤»
timotimo better than nothing ;)
[Coke] I'll regen the gist again after the daily run today.
jnthn How long until the daily run? 12:44
[Coke] it's still manual. 3 kick off at about 12 eastern. rakudo-jvm I then have to run manually 12:45
timotimo did that run already include the fixes to shell? 12:46
[Coke] timotimo: ^^ one? yes. ran it by hand 12:47
jnthn Not that far from 95% :)
[Coke] r: 26108*.9 - 24604 12:48
dalek p: 3fbb512 | sorear++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/ (3 files):
Implement implementClass / callback mechanism
camelia rakudo 0eab21: ( no output )
[Coke] r: 2say 6108*.9 - 24604
camelia rakudo 0eab21: OUTPUT«===SORRY!===␤Confused␤at /tmp/grfwX2fFkW:1␤------> 2⏏say 6108*.9 - 24604␤ expecting any of:␤ whitespace␤»
[Coke] r: say 6108*.9 - 24604
camelia rakudo 0eab21: OUTPUT«-19106.8␤»
[Coke] r: say 26108*.9 - 24604
camelia rakudo 0eab21: OUTPUT«-1106.8␤»
[Coke] (oy)
sorear gist.github.com/sorear/5939789
[Coke] r: say 26108*.95 - 24604 #oy oy 12:49
camelia rakudo 0eab21: OUTPUT«198.6␤»
sorear sleep&
[Coke] yes, excellent plan. see you post nap. :)
zzz
jnthn sorear: !!! :D
sorear++
timotimo looking forward to when there's some sugar for that :3 12:51
12:57 domidumont joined
timotimo turns out the forest-fire benchmark was broken due to my dumbness. 12:59
masak is back from the wedding
jnthn Hm, think I may have got us another 100+ passes
timotimo i'm preparing for another benchmark run. are you going to commit something? 13:03
jnthn yeah, but it ain't performance related
timotimo ok
jnthn I may do the compile time call analysis later 13:04
timotimo at that time i'll be hanging around with my parents :)
13:05 domidumont left
dalek p: 445550b | jnthn++ | src/ (2 files):
Fix handling of generic method names.

On JVM (and MoarVM, for that matter) the name associated with a code object is stored statically, not in the actual code ref itself like on Parrot. This is generally more efficient (saves a field in closures), but we relied on it. This prevented declaring more than one user defined operator in Rakudo on JVM, which busted various tests. On Parrot, we keep the same approach as today.		 13:10
jnthn 100ish more for the run :)
walk etc. & 13:13
timotimo cool :) 13:21
tadzik do we already have a perlbrew-like thing, which can do 'rakudobrew switch jvm' etc? :) 13:22
I know the answer
timotimo well volunteered :)
tadzik aye :0 13:23
timotimo fwiw, the perl6-bench infrastructure is cool. it allows you to specify exact revisions/tags of your stuff to be built on demand. maybe steal a bit of that?
tadzik hacking in the garden, mm
wow, I will look into it
1Mbit internet will be suboptimal for testing, heh 13:25
timotimo at first, you "./bench setup" which will clone rakudo, nqp, perl5, moarvm for you, then you "./bench extract rakudo-jvm/some-branch-or-tag-or-commit rakduo/some-blah" and then you can "./bench time rakudo-jvm/foo rakudo/bar"
tadzik hm 13:26
timotimo make cloning all the shtuff at the beginning optional and it'll already come quite a bit closer 13:27
13:29 berekuk left
tadzik ah, but bench is all Perl 6 13:30
...it tok me 5 minutes to notice that :D
13:32 berekuk joined
timotimo no, it's a bit of perl5, too (timeall, analyze, ...) 13:32
also, that's a good point :|
tadzik well, the ./bench part
where all the logic is
although components.json itself is quite useful
timotimo right. 13:33
i will probably not be able to upload the benchmark results in time. 13:35
13:36 mtk joined
tadzik haha. Bash uses eq for numeric comparisons, = for string comparisons 13:41
affordablehousinginstitute.org/blog...cigar1.jpg
r: say $?VM 13:44
camelia rakudo 0eab21: OUTPUT«===SORRY!===␤Variable '$?VM' is not declared␤at /tmp/p2aep_TGFF:1␤------> say $?VM⏏<EOL>␤ expecting any of:␤ postfix␤»
tadzik jvm reacts in a more spectacular way :P
but just in a repl 13:45
lee_ i've seen that a few times on the jvm repl (infinite scrolling errors) 13:46
tadzik yeah
maybe the jvm repl is not quite there :) 13:47
say $?VM on rakudo-jvm just prints a newline
lee_ mine is a bit out of date, guess I should rectify that 13:48
14:02 xilo_ joined
timotimo gaaah i'll have to wait until late at night to check out jnthns probably to arrive soon patches for native compile-time dispatch ( 14:09
:(
14:16 berekuk left
tadzik gist.github.com/tadzik/5940022 seems to work :) 14:18
timotimo looks good! :)
gotta run now
14:19 berekuk joined 14:29 dmol joined 14:54 dmol left 14:59 rindolf joined 15:00 berekuk left 15:01 berekuk joined, berekuk left 15:21 konundra joined
TimToady .tell masak I suspect it's some form of en.wikipedia.org/wiki/Anterograde_amnesia 15:25
yoleaux TimToady: I'll pass your message to masak.
masak ah. 15:26
TimToady I believe this was even admitted at one point
well, not with those words
.oO(I keep forgetting that I have amnesia.)
 15:29
masak I can't seem to prod the irclogs into remembering, either. 15:30
ah, found it.
irclog.perlgeek.de/perl6/2012-02-10#i_5130347
irclog++ 15:31
TimToady brainlogs++ :) 15:32
moritz masak: you might want to use tina.perlgeek.de/perl6/search/ for searching. It's still in beta, and doesn't have quite up-to-date data, but the search should work much better
masak ooh 15:33
moritz++
moritz (it's powered by Lucy) 15:34
15:34 ajr joined
masak tina powered by Lucy. gotcha. 15:34
15:34 ajr is now known as Guest35475
moritz :-) 15:34
"tina" is just the hostname, and I use that virtual host for experimental stuff
(under the hood I've mostly rewritten the IR clogs; now uses a partially normalized DB schema, Plack, and Lucy) 15:35
15:35 benabik left
masak nice. 15:35
JimmyZ how was ibot6 runing? 15:36
moritz JimmyZ: mostly stable
15:36 Guest35475 is now known as ajr_
moritz but the problem is that the new machine has too litle RAM to compile rakudo 15:36
JimmyZ no memory overflow?
:P
TimToady I'm sure no one has ever told me I have amnesia. 15:37
moritz someone told me I have amnesia, but I forgot who. 15:38
TimToady Dr Who
15:39 benabik joined
masak there's nothing wrong with my memory; I've never forgotten anything, as far as I recall. 15:40
Teratogen so a dyslexic walks into a bra... 15:45
TimToady Lysdexic: someone who can't write due to a broken right hand. 15:55
masak one of the funniest things I ever found in one of my old notebooks (that I had forgotten I had made) was a page with only big, messily scrawled letters spelling out the phrase "OBSERVE HOW I'M TRYING TO WRITE THIS WITH MY LEFT HAND". 16:11
TimToady r: gist.github.com/anonymous/5940348 16:14
camelia rakudo 0eab21: OUTPUT«1..1␤open is disallowed in restricted setting␤ in sub restricted at src/RESTRICTED.setting:2␤ in sub open at src/RESTRICTED.setting:5␤ in sub nonfatalscope at /tmp/CdaZH4SaGB:10␤ in sub fatalscope at /tmp/CdaZH4SaGB:6␤ in block at /tmp/CdaZH4SaGB:13␤␤# Looks …
TimToady well, anyway, that gist is the rakudobug I was carping about last night, where $*FATAL doesn't turn off in non-fatal scopes 16:15
masak runs it locally 16:17
masak submits rakudobug
16:20 lizmat joined 16:22 dmol joined 16:27 FROGGS joined
TimToady hates it when a language sets options using methods like .inheritIO 16:33
diakopter .
masak says the guy who invented $[ :P 16:34
TimToady that's why I use the present tense :)
masak ooh, it was deprecated in 5.12 -- I did not know that. 16:35
TimToady and it was made lexically scope much earlier 16:36
*scoped
masak ooh, I did not know that either.
TimToady in fact, I did that
masak TimToady++ # atoning, partially
TimToady and recall that $[ was really invented for Perl 0, so I could write a2p 16:37
masak oh, because awk is 1-based?
TimToady yup
masak wonders what Dijkstra thought of awk :) 16:38
TimToady well, Henry Spencer though awk was perl without skin cancer :)
*ght
and Perl 0 had no notion of modules at that point (or of lexical scopes, for that matter, or even packages), so a global made sense 16:40
lizmat good *, #perl6!
TimToady because that's all we had...
lizmat glad to be back in a cool place :-)
TimToady Q: Why are you banging your head against the wall? A: Because it feels so good when I stop. 16:41
lizmat r: my $a=Hash; say $a; $a<b>=1; say $a.perl # just before I left, I realized that this already works:
camelia rakudo 0eab21: OUTPUT«(Hash)␤{"b" => 1}␤»
lizmat aka, somewhere a "new" is done 16:42
masak lizmat! \o/
FROGGS I should try banging my head against Wall next time :o) 16:43
jnthn o/ lizmat 16:44
16:44 crab2313 joined
lizmat so I decided to rig circumfix {} to do a Hash.new only if parameters are specified 16:44
otherswise, just assign the type 16:45
s/assign/return
jnthn Could we not rig it to just not pass Nil along? :)
TimToady r: my $a; say $a; $a<b>=1; say $a.perl
camelia rakudo 0eab21: OUTPUT«(Any)␤{"b" => 1}␤»
TimToady you don't need the Hash even
jnthn lizmat: That feels kinda wrong, fwiw.
r: say {}.WHAT 16:46
camelia rakudo 0eab21: OUTPUT«(Hash)␤»
jnthn r: say {}.DEFINITE
camelia rakudo 0eab21: OUTPUT«True␤»
lizmat I think the Hash makes it cleaner 16:47
anyway, now I got the following behaviour:
$ perl6 -e 'my $a={}; say $a.defined; $a<b>=1; say $a'
False
("b" => 1).hash
16:48 ztt left
lizmat isn't that what we wanted for uninitialized hashes ? 16:48
TimToady that's an empty hash, not an undefined hash
jnthn {} is ...that.
lizmat ok, so we don't want it there, but if we say: 'my %h", then the %h should not be defined, right? 16:49
jnthn lizmat: All it needs, afaict, is setting a "it's empty" flag in the "elsif $stmts == 0 {" branch, then not looping over @children further down if that was the case. 16:50
timotimo t.h8.lv/p6bench/2013-07-06.html more plots
lizmat jnthn: will look for that
timotimo weird, i think the trim_string benchmark b0rked again 16:52
masak oh, how time flies. sprintf hackathon in 1h and a bit. 16:55
TimToady timotimo: looks like
16:55 atroxaper joined
FROGGS ohh 16:56
masak: I'll join late
16:56 atroxaper left
masak ok. 16:56
TimToady whoever wrote that blue-line program must have gone totally insane over performance and scaling... :) 16:57
FROGGS I've set up a new server for that blue-line program at work yesterday 16:58
dalek p: 121892f | jnthn++ | src/vm/jvm/runtime/org/perl6/nqp/ (5 files):
Give in and move name into CodeRef itself.

After finding another obscure bug that boils down to the same issue, and with the !reify_name fix feeling fragile if mixed with pre-comp, it's much easier to go this way for now.
p: 8645a09 | jnthn++ | src/ (2 files):
Revert "Fix handling of generic method names."

This reverts commit 445550b24d7952a9a3ca440e76347474cf79715d.
FROGGS our other servers run 5.10 still :/
dalek kudo/nom: 4e86c80 | jnthn++ | src/vm/jvm/runtime/org/perl6/rakudo/Ops.java:
Chase latest NQP change.
 16:59
masak I see three things we can focus on in today's sprintf hackathon: (a) dropping the our implementation into nqp-JVM. (b) figuring out what the spectests want that we don't have yet. (c) working on those missing features.
(c) is fully dependent on (b). (b) is only sorta-kinda dependent on (a). 17:00
jnthn That fixes S12-methods/multi.t...
masak I'll work on (a) if nothing unexpected happens.
TimToady jnthn: does/could the for ^$max {...} optimization figure out that ... doesn't use $_ and, say, substitute a countdown?
lizmat masak: why nqp-JVM only ?
why not nqp, period ?
TimToady since 'while --$countdown' tends to be fast, especially in native types (assuming native decrement is available) 17:01
masak lizmat: because I believe that'll lose us passing spectests on Parrot.
jnthn TimToady: There's no optimization of that at present, and you'd need to be sure that $max is not bigint...
lizmat: I think we should do that eventually, but it should handle everything we already do first.
lizmat masak: isn't that the whole wrong reason, and a very short term as well?
ok, as long as we see that's the end goal :-) 17:02
jnthn lizmat: Whereas with nqp-jvm we've nothing to lose by having a not-done-yet version.
TimToady: Which makes it a bit harder to do static analysis of...
masak lizmat: definitely the long term.
FROGGS lizmat: I guess it is just an #?if jvm, and will be used for all backends when it supports all features
masak lizmat: I'm just *very* against regressing on passing spectests.
lizmat: that just isn't done. it hurts people. it hurts the ecosystem. 17:03
FROGGS because no packend will ever support printing NaN in %d
(or what the rt report what about)
lizmat ok ok ok :-) just wanted to make sure we're on the same page regarding the future here :-)
TimToady if $_ isn't used, then counting down from 2**64-1 will probably take a long time anyway, if you cap it at that
masak no-one would be happier/prouder than me if sprintf were the standard across all backends. that's why I started implementing it in pure nqp. 17:04
so yes, of course that's the long-term goal. ;)
TimToady I'm guessing nobody would ever notice a cap at 2**64-1 :)
jnthn timotimo: It appears that something I did unbusted S14-roles/mixin.rakudo.jvm
TimToady: That's, uh, sick. But you're right. :D
timotimo: Which is the test that exploded with the Numeric method error
timotimo: You may want to try a forest fire again 17:05
TimToady see earlier note about "insane"
jnthn Hm, seems that fix I just did incidentally fixed a few things... 17:07
aww, but not S06-multi/type-based.t 17:08
FROGGS maybe next time *g* 17:11
lizmat coming back to '{}' versus 'my %' I don't really see a difference 17:17
the fact that we see {} as an empty (defined) hash, feels like a p5ism to me
.oO( wonder what TimToady's thoughts are on that )
 17:18
17:18 domidumont joined
TimToady I already said, I thought. {} and [] are simply empty, not undefined 17:18
lizmat argh, yes, sorry for the noise 17:19
TimToady r: say ().defined; # this too 17:20
camelia rakudo 4e86c8: OUTPUT«True␤»
17:20 patspam joined
dalek kudo/nom: 4e9635e | (Elizabeth Mattijsen)++ | src/core/Hash.pm:
Remove superfluous call to STORE, "new" handles this also
 17:20
p: bc36181 | jnthn++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/Ops.java:
Make nqp::shell work on Windows.

Means that the spectests that spawn a sub-process can now be run on Windows also.
jnthn ...apart from they seem to fail under make spectest... 17:22
17:23 patspam left 17:24 patspam joined 17:28 birdwindupbird left
masak yes, there's a discontinuity there between item-size variables and container-size variables. 'my $s' starts out undefined until assigned to. 'my %h' starts out empty-but-defined. 17:30
I guess it even makes sense in a way. there are no "empty scalars", so the only way they can show their emptiness is to be undefined. but with containers it doesn't make much sense to be undefined, only to then fill it with a container -- might as well start your existence as an empty container. 17:31
jnthn oh, maybe it did.
TimToady masak: it doesn't make sense to start with an empty container if you're immediately going to rebind it to something else; consider a function parameter like (@shaped[100][100]) that accepts an array of that shape, but it would not make sense to allocate @shaped and then bind over the top of that 17:35
one could argue similarly for: my @shaped[100][100]; ...; @shaped := @some_100x100_array 17:37
autoviv just becomes an all-or-nothing affair in that case 17:38
this would eliminate the discontinuity, though it might make optimizing harder 17:39
masak aye. 17:40
TimToady s:g/'[100][100]'/[100;100]/
we'd like 'my @shaped[100][100]' to make some static guarantees that the optimizer can depend on 17:41
masak I wasn't thinking of shaped arrays at all -- just trying to rationalize the discontinuity between scalar containers and unshaped array/hash containers.
TimToady or that we can at least point the finger back at the user if they violate it
masak preferably at compile time.
TimToady "static"
masak *nod* 17:42
TimToady though yes, we might not be able to detect the erroneous code at compile time
darn it, I wrote [100][100] again :)
TimToady wonders if we should allow that in the declaration, or maybe we do 17:43
std: my @shaped[100][100] is Blurfl; 17:44
camelia std 85cf51d: OUTPUT«ok 00:00 43m␤»
TimToady I guess we do
std: my @shaped[100]{Bool}[42] is Blurfl;
camelia std 85cf51d: OUTPUT«ok 00:00 44m␤»
TimToady so we can intermix array and hash shapes 17:45
std: my @shaped[*]{*}[*]; # probably see a lot of shaped arrays that merely do Array vs List like this 17:49
camelia std 85cf51d: OUTPUT«ok 00:00 44m␤»
TimToady in that case it's still autoviv at every level, arguably including @shaped itself 17:50
17:50 mstefanko left
lizmat jnthn: I guess this is *not* what we were going for: $ perl6 -e 'my @a=->{}; say @a[0]' 17:50
Block.new()
r: my @a=->{}; say @a[0]
camelia rakudo 4e86c8: OUTPUT«Block.new()␤» 17:51
TimToady looks correct to me :)
jnthn lizmat: Looks fine
lizmat: It's only the {} form that becomes hash
*a
lizmat r: my @a={}; say @a[0] 17:52
camelia rakudo 4e86c8: OUTPUT«().hash␤»
17:52 mstefanko joined
jnthn Right. 17:52
TimToady because why would you write an empty closure anyway? maybe for a callback that does nothing, but {;} is not an onerous burden in that case
lizmat hmmm… seems I've lost the plot, then :-( 17:54
r: sub a {}; say a; my @a=a; say @a 17:55
camelia rakudo 4e86c8: OUTPUT«Nil␤␤»
lizmat jnthn: then what did you mean by "[18:50:04] <jnthn> lizmat: All it needs, afaict, is setting a "it's empty" flag in the "elsif $stmts == 0 {" branch, then not looping over @children further down if that was the case." ? 17:56
17:57 mstefanko left
[Coke] kicks off the rakudo.jvm run 17:57
masak so, what is this new 'is Blurfl' trait I keep hearing about? :P
TimToady it's a trait that fails if the thing to the left doesn't parse as a shaped array :) 17:58
17:58 aindilis joined
TimToady std: (my @shaped[*]{*}[*]) is Blurfl; 17:59
camelia std 85cf51d: OUTPUT«===SORRY!===␤Two terms in a row at /tmp/NfepoOlEkG line 1:␤------> (my @shaped[*]{*}[*]) ⏏is Blurfl;␤ expecting any of:␤ feed_separator␤ infix or meta-infix␤ infixed function␤ statement modifier loop␤Parse failed␤FAILED 00:00
..4…
TimToady std: my @shaped[*]{*}.[*] is Blurfl;
camelia std 85cf51d: OUTPUT«===SORRY!===␤Two terms in a row at /tmp/iLisXTqOpK line 1:␤------> my @shaped[*]{*}.[*] ⏏is Blurfl;␤ expecting any of:␤ feed_separator␤ infix or meta-infix␤ infixed function␤ statement modifier loop␤Parse failed␤FAILED 00:00
..46…
masak ok. sprintf hackathon begins. 18:00
TimToady ^^^ a place we distinguish [] from .[]
masak TimToady: yes, and S09 mentions that, methinks.
18:00 mstefanko joined
[Coke] jnthn++ nqp:shell for win 18:04
tadzik for the win! 18:06
TimToady my @shaped[100;100] = 0 xx *; # allowed? or should we require something resembling (0 xx *) xx * 18:07
masak QAST::OperationsJAST.map_classlib_core_op('sprintf', $TYPE_OPS, 'sprintf', [$RT_STR, $RT_OBJ], $RT_STR, :tc);
jnthn: ^ guessing I want to do .add_core_op here instead?
TimToady: the conservative thing to do is to require something of the same shape. 18:08
[Coke] masak: instead of?
masak [Coke]: instead of that line, which uses JVM's sprintf.
[Coke] masak: (yes, you'll need that to put in a hook for nqp::sprintf
TimToady I think there's some constraint that we initialize a multidimensional object with a LoL that every dimension must limit either in the decl or in the initializer, but both can't be * 18:09
[Coke] masak: yes. though I only know how to point to a java method, not a perl6 one.
masak it's implemented in nqp.
and I see examples of .add_core_op in that file, don't worry.
TimToady so my @shaped[100;100] = lol (0 xx *) xx * is allowed 18:10
masak biggest question right now is where in the file hierarchy to place the sprintf code...
std: lol 42
camelia std 85cf51d: OUTPUT«ok 00:00 42m␤»
TimToady but my @shaped[*;100] = lol (0 xx *) xx * is not allowed
masak wasn't aware 'lol' was a listop, too 18:11
TimToady but my @shaped[*;100] = lol (0 xx *) xx 100 is allowed
masak TimToady: makes sense.
TimToady nr: say lol (0 xx 5) xx 5
camelia niecza v24-86-g39ab531: OUTPUT«===SORRY!===␤␤Undeclared routine:␤ 'lol' used at line 1␤␤Unhandled exception: Check failed␤␤ at /home/p6eval/niecza/boot/lib/CORE.setting line 1502 (die @ 5) ␤ at /home/p6eval/niecza/src/STD.pm6 line 1147 (P6.comp_unit @ 36) ␤ at /home…
..rakudo 4e86c8: OUTPUT«===SORRY!===␤Undeclared routine:␤ lol used at line 1. Did you mean '&log'?␤␤»
TimToady been specced for a long time 18:12
nr: say ((0 xx 5) xx 5).lol
camelia rakudo 4e86c8: OUTPUT«ListIter.new()␤»
..niecza v24-86-g39ab531: OUTPUT«Unhandled exception: Unable to resolve method lol in type List␤ at /tmp/xP4XcTkLzX line 1 (mainline @ 4) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4579 (ANON @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 4580 (module-CORE @ 576) ␤ at /home/p…
lizmat I guess that qualifies as LHF
TimToady though the fact that we have to write the (0 xx *) xx 100 above probably means our xx operators are backwards
you'd like to say 100 xx (* xx 0) instead 18:13
lizmat maybe we need a reverse version of the "xx" operator, something like "xx" ? 18:14
TimToady :D
lizmat :-)
TimToady nr: say 'xx'.flip
camelia rakudo 4e9635, niecza v24-86-g39ab531: OUTPUT«xx␤»
TimToady there you go
hmm 18:15
masak r: say 5 Rxx "OH HAI"
camelia rakudo 4e9635: OUTPUT«OH HAI OH HAI OH HAI OH HAI OH HAI␤»
TimToady nr: say 3 Rxx 'It's true!'
camelia rakudo 4e9635: OUTPUT«===SORRY!===␤Two terms in a row␤at /tmp/VXFljlRaKq:1␤------> say 3 Rxx 'It'⏏s true!'␤ expecting any of:␤ postfix␤ infix stopper␤ infix or meta-infix␤ statement end␤ statement modifier␤ …
..niecza v24-86-g39ab531: OUTPUT«===SORRY!===␤␤Confused at /tmp/limiDH6GgD line 1:␤------> say 3 Rxx 'It'⏏s true!'␤␤Parse failed␤␤»
masak I win. 18:16
TimToady I said hmm first though
masak ;)
TimToady sometimes this language scares me :) 18:17
masak fwiw, I did the apostrophe-inside-singlequote-string thing yesterday... :)
(in code I was writing)
TimToady it's one of the reasons I usually use "" instead
lizmat but if we will revert 'xx's meaning, a *lot* will need to be changed around
masak yes no please don't do that.
TimToady lizmat: nobody is suggesting that :)
notice me not suggesting it very hard 18:18
masak panic oh noes aaaaaah oh ok. :P
lizmat :-D
masak .oO( masak is only allowed to panic 255 times )
18:18 pmurias_ joined
TimToady 2**64-1 18:18
lizmat after that, panic really strikes
TimToady maybe Jesus really meant we should forgive our brother 70 ** 7 times 18:20
masak "turn the 471st cheek"
er, "490th" 18:21
TimToady nr: say 70 ** 7; say 70 R** 7
camelia rakudo 4e9635, niecza v24-86-g39ab531: OUTPUT«8235430000000␤143503601609868434285603076356671071740077383739246066639249␤»
masak oh, ** :)
TimToady let's aim for 70 R** 7
timotimo jnthn: i got the numeric thing in xx due to a sillyness i made mostly; forest fire actually does run
just not very much
masak TimToady: right, but after that, I think the brother's actual malice has been firmly established. 18:22
TimToady let's hear it for Bayse
masak .oO( Bayes-less accusations ) 18:23
18:24 raiph joined
TimToady no wonder I couldn't make Bayseian look right... 18:24
timotimo forest fire is b0rked somehow. there's forest, but no fire
TimToady does the forest grow? 18:25
raiph <TimToady> but my @shaped[*;100] = lol (0 xx *) xx 100 is allowed
did you get that wrong?
TimToady raiph: no, the order is backwards because xx varies in the opposite order of subscripting
hence the remark about it being backwards
raiph ah, gotchya 18:26
TimToady we're supplying 100 to the * and * to the 100
just can't do * in both places, unless we allow lazy dimensions somehow (brane asplodes)
lizmat
.oO( sixties colours abound )
 18:27
TimToady or we say that a coinciding * means that we must wait till one is autovivified to initialize the child 18:28
maybe that works out
but really, initializing only leaves is the bailiwick of 'is default' I suspect 18:29
raiph what's the term for when there's an ideal (semantic+syntax) and something less than ideal but easier to implement that will be forward compatible? 18:30
TimToady hack?
raiph heh
TimToady stopgap?
raiph i meant something that's good planning 18:31
TimToady a partial emulation?
a subset?
moritz incomplete implementation?
raiph i don't like incomplete
TimToady subset implementation then 18:32
lizmat suboptimal
TimToady a realistic implementation :)
raiph eg, don't allow * for both decl and init in same dimension for now; but this is forward compatible with allowing it later
so we can formalize an approach toward 6.0.0 18:33
TimToady a deconjectural implementation :)
lizmat now can build rakudo with Nil inheriting from Mu, thanks to jnthn++'s pointers
TimToady a conservative implementation
moritz Nil outside Mu?
TimToady now that we've decided not to?
lizmat hehe
yes, I know we decided not to... 18:34
masak yes, "conservative implementation"
lizmat call me stubborn :-)
masak it's "conservative" because it hedges in such a way that switching to the better way won't break things.
lizmat I just wanted it to at least build
raiph yes, i like conservative
TimToady we might still put Failure outside of Any, if the binding failure message can be taught to throw when it sees "got Failure, expected Any"
lizmat so coming back to Nil, should it maybe inherit from Any then? 18:35
TimToady that's what it does if you don't say "is Mu'
*"
lizmat I just want to run the spectest with Nil inherting from Mu just once, so I can gist it for future reference 18:36
TimToady but putting Failure outside of Any also makes it impossible to say $x.WHAT === Failure
lizmat to me, putting Failure outside of Any, doesn't make much sense 18:37
as anything out of Any basically has different shades of nothingness 18:38
TimToady well, it traps a lot more accidental uses than the method overrides that are currently supplied with Failure
lizmat and a failure is something
TimToady my $x = thisfails(); throws, while my Mu $x doesn't 18:39
lizmat and I would consider that a feature
just as, if Nil was outside of Any,
TimToady most attempts to bind Failure would cause a binding failure, hence the need for binding failure to notice Failure and throw it instead
lizmat you could only store Nil if it was Mu 18:40
TimToady same deal, but Nil is supposed to be failsofter than Failure
lizmat as in: my $a=1; $a= Nil # sets default
my Mu $a=1; $a=Nil # sets Nil
please note that neither of the above fail 18:41
TimToady if we wanted Nil to be failsoft outside of Any, we'd have to do things like the autothreader in Junction does to patch things up
r: say Nil + 0 18:42
camelia rakudo 4e9635: OUTPUT«use of uninitialized value of type Nil in numeric context in block at /tmp/vX0MZtiQYr:1␤␤0␤»
TimToady this sort of thing becomes fatal if Nil is outside of Any, unless we "autothread" it to replace Nil with 0
lizmat $ perl6 -e 'say Nil + 0' 18:43
Nominal type check failed for parameter '$a'; expected Any but got Nil instead
<Internal error while creating backtrace: Nominal type check failed for parameter 'lhs'; expected Any but got Nil instead in sub METAOP_TEST_ASSIGN:<||> at src/gen/CORE.setting:14373
TimToady yes, like that
but Perl likes undefineds to warn and then convert to 0 or ''
if Nil is outside of Any, does the dispatch failover automatically call Junction::AUTOTHREAD? 18:45
lizmat for the record: diff and spectest result of Nil being outside of Any: gist.github.com/lizmat/5940829
TimToady: I'm not deep enough in that to answer that yet 18:46
TimToady how many extra spectests fail? (niecza failed about 26 more)
most of them in .trans 18:47
but I think that's because Nil failsoft semantics are not well tested in roast 18:49
18:49 colortiger joined
masak colortiger: oh hai :) 18:50
lizmat the last test infiniloops, so I had to abort it, hence no summaries
TimToady well, one could kill the subprocess
*kill only 18:51
18:51 colortiger left
masak bye bye, colortiger. 18:51
TimToady 色虎, o/
dalek rl6-roast-data: 2426ce4 | coke++ | / (5 files):
today (automated commit)
masak TimToady: 呵呵 18:52
lizmat 94.68%
TimToady: will rebuild and spectest again :-)
masak TimToady: it's kinda cool that we both studied different languages having the same half-impossible character set, and now I get (some of) your kanji jokes without having to go via Google Translate. :) 18:53
TimToady we've long speculated other "autothread" semantics for the Each type; fallback recovery semantics for Nil and Failure are similarly possible 18:54
masak: well, at this point I actually know more Chinese characters than Japanese 18:55
I just don't know how to pronounce them
raiph gist.github.com/TimToady/1828667 # is this the right OKness proposal for lizmat to read?
GlitchMr fail 'Cannot .unshift an infinite list' if @elems.infinite; 18:56
TimToady she doesn't have to read it really; we're just coming at similar ideas from a different direction
GlitchMr Why unshifting infinite list would fail.
IMO, it should replace list with that infinite list.
TimToady it should just unshift into the reified part, I suspect, and leave the infinite part alone 18:57
masak TimToady: "I know more Chinese characters than Japanese; I just don't know how to pronounce them" -- that's how my knowledge of Chinese characters feels these days :/ 18:58
GlitchMr github.com/rakudo/rakudo/commit/68...00103f2baa
Seriously?
pop() in like every programming language is RW.
masak GlitchMr: that one was later reverted. 18:59
GlitchMr ok
lizmat indeed, and TimToady later specced "is parcel"
TimToady but that's not what the 'is rw' means there anyway
masak GlitchMr: and if you're saying pop() is rw in every programming language, I don't think you understand what it does.
TimToady it's talking about the return value, not the array
GlitchMr Oh, I see.
lizmat but pop itself is not a left value, right? 19:00
TimToady it's not 'self is rw'
GlitchMr But not sure if I understand.
Why `pop` `is rw` then.
TimToady it's sort of assumed that an object is mutable anyway
GlitchMr Why I would want to assign to return value of `pop`?
(I'm not talking about @array.pop[0] = 42, that still should be fine) 19:01
TimToady because it's like shift, and shift is what drives the mutability of 'for @x { s/foo/bar/ }'
GlitchMr Oh, right
TimToady I think that was the upshot of the conversation; correct me if I'm wrong 19:02
masak no, that sounds right.
GlitchMr With aliases it makes sense, I guess.
TimToady presumably 'for reverse @x {}' could optimize to pops 19:03
masak Perl 6 doesn't realy have the concept of aliases. it's just binding.
GlitchMr @array[0] := $value
Looks aliasy enough for me. 19:04
dalek kudo/nom: 6319760 | (Elizabeth Mattijsen)++ | src/Perl6/Actions.nqp:
Changes suggested by jnthn to make sure empty {} doesn't generate Nils
kudo/nom: 4e29215 | (Elizabeth Mattijsen)++ | src/core/Hash.pm:
Slight cleanup for circumfix {}
masak QAST::OperationsJAST.add_core_op('sprintf', -> $qastcomp, $op { ... }); # what do I put inside the block to make it call the 'sprintf' logic? 19:07
or is that the wrong question to ask? should I be using something other than .add_core_op for this?
lizmat TimToady: wrt unshifting to infinite lists, should that be allowed? 19:08
TimToady didn't I just say that?
masak jnthn: ping 19:09
lizmat [20:57:16] <TimToady> it should just unshift into the reified part, I suspect, and leave the infinite part alone
TimToady r: my @array = 1..*; @array.unshift: 0; say @a[^5]
camelia rakudo 4e9635: OUTPUT«===SORRY!===␤Variable '@a' is not declared␤at /tmp/s5fS7QCI07:1␤------> ay = 1..*; @array.unshift: 0; say @a[^5]⏏<EOL>␤ expecting any of:␤ postfix␤ statement end␤ statement modifier␤ statement modi…
TimToady r: my @a = 1..*; @array.unshift: 0; say @a[^5] 19:10
camelia rakudo 4e9635: OUTPUT«===SORRY!===␤Variable '@array' is not declared␤at /tmp/mCJsB4h5kw:1␤------> my @a = 1..*; @array.unshift: 0⏏; say @a[^5]␤ expecting any of:␤ postfix␤»
TimToady r: my @a = 1..*; @a.unshift: 0; say @a[^5]
camelia rakudo 4e9635: OUTPUT«0 1 2 3 4␤»
lizmat I guess the "I suspect" made it unsure for me
TimToady seems like it works already :)
lizmat r: (1..5).unshift: 0 19:11
camelia rakudo 4e9635: OUTPUT«Cannot call 'unshift'; none of these signatures match:␤:(Any:U \SELF: *@values, Mu *%_)␤ in method unshift at src/gen/CORE.setting:1486␤ in block at /tmp/77RQQ034Ah:1␤␤»
lizmat r: (1..*).unshift: 0 19:12
camelia rakudo 4e9635: OUTPUT«Cannot call 'unshift'; none of these signatures match:␤:(Any:U \SELF: *@values, Mu *%_)␤ in method unshift at src/gen/CORE.setting:1486␤ in block at /tmp/r1D3TN5vyM:1␤␤»
TimToady I doubt that Range has an unshift
lizmat indeed
r: (1..*).list.unshift: 0
camelia rakudo 4e9635: ( no output )
lizmat r: say (1..*).list.unshift: 0
19:12 crab2313 left
camelia rakudo 4e9635: OUTPUT«(timeout)» 19:13
TimToady someone is being overly eager, seems
lizmat r: my @a=1..*; say @a.infinite
TimToady oh, the say is :)
camelia rakudo 4e9635: OUTPUT«True␤»
TimToady r: say ((1..*).list.unshift: 0)[^5] 19:14
camelia rakudo 4e9635: OUTPUT«0 1 2 3 4␤»
TimToady there you go
so already works
lizmat which is odd, as the code says:
multi method unshift(List:D: *@elems) {
fail 'Cannot .unshift an infinite list' if @elems.infinite;
TimToady perhaps slurpy arrays don't know that 19:15
oh the @elems is just the 0
means you can't say [email@hidden.address] 1..*'
since it would presumably have to prepend the reified list with something non-reified 19:16
arguably it could be allowed if the reified portion were empty, but it seems of little utility to allow it 19:17
lizmat updated gist.github.com/lizmat/5940829 with complete test + summaries 19:18
TimToady but it's perfectly okay to unshift real things into the reifiedlist
[Coke] gist.github.com/coke/5879701 updated. yay, new errors. 19:19
lizmat brb 19:23
[Coke] jnthn: someone fixed a bunch of "Nominal type check failed" 19:25
19:27 kaare__ left
[Coke] r: say 12+68+1+123+56+30+131+130+22 # number of rakudo-jvm tests aborted due to sprintf 19:30
camelia rakudo 4e2921: OUTPUT«573␤»
masak ooh, that's an attractive prize. 19:31
I'm currently blocking on not knowing enough about nqp internals.
FROGGS masak: I'm making #-option of %X work right now
masak FROGGS++
jnthn [Coke]: Yes, that was me yesterday ;) 19:32
19:33 domidumont left
[Coke] the list of aborts is getting smaller! 19:33
I was able to generate it much quicker today.
dalek kudo/nom: 7d7df4a | (Elizabeth Mattijsen)++ | src/core/List.pm:
Remove useless fail in unshift
[Coke] here's a new one, jnthn : java.lang.RuntimeException: No such attribute '$!do' for this object 19:34
jnthn masak: pong 19:35
[Coke]: Is list updated?
[Coke] jnthn: aye
jnthn ah, I see it
Yeah, though I don't think it's a regression, I think it's that we didn't get that far into that test before. 19:36
[Coke] jnthn: right, yah. Fix one failure, get another.
masak jnthn: see my question at irclog.perlgeek.de/perl6/2013-07-06#i_7298362
[Coke] it used to block on shell() 19:37
jnthn masak: Hmmm
masak: I think the best bet is to install the sprintf implementation using nqp::bindhllsym('nqp', 'sprintf', ...) 19:38
masak: And then make a QAST::Op.new( :op('call'), QAST::Op.new( :op('gethllsym'), nqp/sprint SVals here ), ...push children of op here... ) 19:39
lizmat WWZ& 19:40
masak the 'and then make' part being the contents of that block?
jnthn Of the add_core_op one? Yes 19:42
Then the only question is where to put the sprintf impl :)
Maybe in src/core/sprintf.nqp, and make sure it goes inside a block so that it doesn't leak out as a function.
And then just include it in the JVM build for now. 19:43
(by adding to Makefile-JVM.in or whatever it's called)
masak *nod* 19:45
requesting clarification on "make sure it goes inside a block so that it doesn't leak out as a function".
how then will I be able to call it with the :op('call') thing? 19:46
jnthn 'cus you're stashing it in the "secret" guts-y namespace thing
That's what the gethllsym thing is for.
masak could you link me to an example of this so I can see how it's done? 19:48
jnthn I guess ModuleLoader.nqp uses the mechanism 19:49
And then somewhere in NQP::World it does the other part (the lookup)
masak oki 19:50
that should be enough to get me started. thank you.
Teratogen poor 777
19:52 dmol left, dmol joined
FROGGS masak: I'm adding now (just locally) all sprintf tests from roast 19:55
but I guess next important things are %f and %e/%E
diakopter Teratogen: a 777? 19:58
19:58 berekuk joined
masak FROGGS: yes, I think so. 19:59
FROGGS: will be interesting to see the results of the sprintf/roast collision.
FROGGS: you're essentially doing (b), then ;)
FROGGS I'll gist it 20:00
right :o)
(a) is up to you :o)
20:01 xilo_ left
masak *nod* 20:01
20:08 djanatyn joined
[Coke] jnthn: I golfed the duplicate method error to this snippet: 20:09
'a;b,c,' ~~ m/(<.alpha>) +%% (<.punct>)/;
seeing if I can further golf it.
20:10 ajr_ left, xilo_ joined
masak hopes juxtaposing '+' and '%%' like that doesn't become the norm 20:12
Teratogen diakopter, 777 crashed at SFO
perigrin Reports say most of the people are safe though. 20:15
diakopter hunh; tail hit the rocky edge of the runway/shore
perigrin has never landed at SFO but can see how that could happen 20:17
[Coke] jnthn: '' ~~ m/'a' +%% ('b')/;
rn: say '' ~~ m/'a' +%% ('b')/;
masak actually I think I'm less against juxtaposing '+' and '%%' than I am against introducing whitespace between the atom and the postfix '+'...
camelia rakudo 7d7df4, niecza v24-86-g39ab531: OUTPUT«False␤»
sorear good * #perl6
masak sorear! \o/ 20:18
[Coke] sorear: ^^ there's the duplicate method culprit.
masak [Coke]: why do you write `'a' +%%` and not `'a'+ %%` ?
[Coke]: you must have a different mental model of what those operators do :)
jnthn masak: I suspect [Coke]++ was golfing it from a test... 20:19
sorear Teratogen: any survivors?
diakopter sorear: seems all but two
Teratogen there are reports of two deaths
but it is all up in the air right now
a lot of people were seen coming down chutes
jnthn
.oO( I'm pretty sure the plane is not still up in the air right now... )
 20:21
FROGGS masak: pasting the result of running roasts test too is sort of boring, the first 20 passes, then one kills the rest :o) 20:22
jnthn [Coke]: Turns out it can be golfed to nqp -e "/'a' +%% ('b')/" 20:23
TimToady masak: one writes it that way to allow whitespace after the left side in a rule 20:25
masak FROGGS: could you maybe note down the killer in some list somewhere, and comment it out?
TimToady doesn't matter otherwise 20:26
masak TimToady: right. this isn't a rule.
FROGGS it was %i, I post the out in a sec
TimToady nevertheless, if people get used to +%%, they'll right it more correctly for rules, on average 20:27
*write
jnthn %% is the one that means "and we can match another one afterwards", right?
TimToady it means they can come out even :)
jnthn yeah
jnthn can see what's wrong but...hm, how to fix that...
FROGGS masak: gist.github.com/FROGGS/0fdcdde74679e512f097 20:28
masak looks
FROGGS masak: the commented out tests do not parse 20:29
(in the diff around line 100)
masak FROGGS: wow, that still looks like really good news!
FROGGS yes, indeed
masak jnthn: could you please have a look at gist.github.com/masak/5a34e6893a80a5420c2d ? 20:30
FROGGS mostly e, f, g
TimToady also, since %% requires a quantifier, it's arguably better to put the quantifier with it, to avoid the \w+ %% ',' confusion
with \w +%% ',' it's more obvious what's wrong
masak FROGGS: e, f, g are challenging but by no means impossible.
FROGGS yeah, I'm going for %f now
TimToady so I think +%% is a good thing, on average 20:31
masak TimToady: ok, you convinced me.
jnthn masak: "nqp/sprint SVals here" was an instruction to you :P
TimToady and I like it better than <foo> + %% <bar>
masak jnthn: yes, I realize that :) just wondering what it should be :)
jnthn two QAST::SVal nodes
masak ok, hold on.
jnthn (for nqp, then sprintf
sorear rn: say 5 Rxx rand 20:32
camelia rakudo 7d7df4: OUTPUT«0.70025693867089 0.70025693867089 0.70025693867089 0.70025693867089 0.70025693867089␤»
..niecza v24-86-g39ab531: OUTPUT«0.77211467026365677 0.77211467026365677 0.77211467026365677 0.77211467026365677 0.77211467026365677␤»
FROGGS no cookie for both
masak jnthn: like this? gist.github.com/masak/5a34e6893a80a5420c2d 20:33
jnthn masak: right
masak hm, sorear++ just found a bug in two implementations ;)
rn: say rand() xx 5
camelia niecza v24-86-g39ab531: OUTPUT«===SORRY!===␤␤Unsupported use of rand(); in Perl 6 please use rand at /tmp/dnlJsPVO8F line 1:␤------> say rand⏏() xx 5␤␤Parse failed␤␤»
..rakudo 7d7df4: OUTPUT«===SORRY!===␤Unsupported use of rand(); in Perl 6 please use rand␤at /tmp/Y8U3jpzN6N:1␤------> say rand⏏() xx 5␤»
masak rn: say rand xx 5
camelia niecza v24-86-g39ab531: OUTPUT«0.62266399973196163 0.18375288144860086 0.37745106424086311 0.96518965482953456 0.73068906819945623␤»
..rakudo 7d7df4: OUTPUT«0.181609341499524 0.340130289589194 0.252906549579116 0.617402868939354 0.385966828776859␤»
masak right. 20:34
masak submits rakudobug
I'm just assuming that if infix:<xx> thunks its lhs, then infix:<Rxx> should thunk its rhs.
sorear masak: it's a very long-standing behavior of niecza that weakly macroey operators lose their macroness when metaopped 20:35
[Coke] masak: jnthn is right, I don't expect that to do anything except not generatea weird java bytecode issue.
sorear masak: because metaops require a coderef 20:36
masak sorear: that sounds like an argument against this being a bug.
sorear strongly macroey operators like := can't be metaopped at all
rn: my $x; 5 R:= $x; say $x
camelia rakudo 7d7df4: OUTPUT«Nominal type check failed for parameter 'op'; expected Any but got Mu instead␤ in sub METAOP_REVERSE at src/gen/CORE.setting:14378␤ in block at /tmp/qy3FLcnjuj:1␤␤» 20:37
..niecza v24-86-g39ab531: OUTPUT«===SORRY!===␤␤Cannot reverse the args of := because list assignment operators are too fiddly at /tmp/zIuIfbFJwJ line 1:␤------> my $x; 5 R:=⏏ $x; say $x␤␤This macro cannot be used as a function at /tmp/zIuIfbFJwJ line …
masak submits LTA rakudobug
20:39 domidumont joined
sorear done with backlog 20:39
20:41 domidumont left
masak jnthn: why the two SVals 'nqp' and 'sprintf', by the way? 20:43
jnthn masak: They're the operands to look up the thing you installed with nqp::bindcurhllsym 20:44
FROGGS like nqp::sprintf, no?
jnthn masak: oh, wait 20:45
You do need gethllsym, not getcurhllsym, there
masak that sounds... circular.
ok.
jnthn And just sprintf
uh, dammit
You need both args
But gethllsym
Otherwise it won't work out in Rakudo...
masak still 'nqp' and 'sprintf'?
jnthn yes
masak I confess to being utterly confused as to the distinction between gethllsym and getcurhllsym. 20:46
sorear why are we using an op here and not a global function?
masak: do you remember how Parrot's namespace worked? 20:47
We're emulating that.
jnthn sorear: Installing stuff with one of these is as close as you get to a global function.
Well, but we don't have any nesting besides HLL name and symbol.
masak sorear: my goal tonight is to bind nqp:sprintf (an op) to a global(-ish) function (on JVM only).
jnthn masak: cur means "current HLL"
masak I guessed that. 20:48
sorear masak: the global namespace in parrot is segregated by HLL name
20:48 pmurias joined
sorear nqp;sprintf is a different symbol from perl6;sprintf 20:48
masak ok.
jnthn Grr. That %% bug is hard to fix by building a different AST... 20:49
sorear today: document jvmbootinterop and/or explain the design principles to jnthn 20:50
jnthn The easiest fix is to fiddle with QAST::Block code-gen to not spit out the same block twice...
TimToady is just a little distracted by the plane crash at SFO... 20:52
masak aww. got 'Can not invoke this object' during compilation of NQPCORE.setting 20:55
20:56 rindolf left
masak that probably means I embedded things wrongly in src/core/sprintf.nqp 20:56
jnthn [Coke]: Trying a patch for S05-capture/caps.rakudo.jvm now, looks promising... 20:59
masak: prolly
20:59 FROGGS left
jnthn masak: Which compilation, ooc? 20:59
stage1 or 2? 21:00
masak stage1
jnthn ok, then it almost certainly really is inside what you added
masak would it help if I gisted it?
jnthn I can look 21:01
masak gist.github.com/masak/d04ffe773bdfdc07d7e3
cognominal that would be nice, when $<foo> is an array to have an array method .ast that did : .map: *.ast
masak, what is knowhow? 21:02
dalek p: ad036ee | jnthn++ | src/vm/jvm/QAST/Compiler.nqp:
Don't spit out same QAST::Block twice.

Fixes S05-capture/caps.t in Rakudo; [Coke]++ for golfing.
jnthn masak: method sprintf($format, *@arguments) { 21:03
Should that not be sub?
masak: And then in the nqp::bindcurhllsym('sprintf', sprintf); in the same scope
And should probably be &sprintf
dalek kudo-js: 643e321 | (Pawel Murias)++ | runtime/serialization.js:
Closure deserialization without binding contexts.
 21:04
kudo-js: 2bceb50 | (Pawel Murias)++ | / (2 files):
What we call an outer context is really a caller context.
kudo-js: 2c2c39c | (Pawel Murias)++ | / (4 files):
Move out Ctx to a seperate file. Create closures with an outer lexical, static variables still missing.
kudo-js: b9dc21c | (Pawel Murias)++ | / (4 files):
Fix bugs.
jnthn masak: I'd also make it a "my module sprintf..." or so and remove the sprintf:: on the grammar/actions decl
21:04 FROGGS joined
sorear TimToady: In what sense is .inheritIO like $[ ? I see it as the Java named argument faking pattern 21:05
o/ FROGGS
masak jnthn: oki 21:06
FROGGS hi sorear
sorear wonders if jnthn has seen the thread-starting gist
masak jnthn: ok, that makes more sense.
jnthn: I changed 'sub' to 'method' because I was basing things off ModuleLoader.
21:06 pmurias_ left
masak cognominal: 'knowhow' is a low-level MOP object in nqp. 21:07
jnthn sorear: Yes!
masak: Ah...
sorear: Didn't try running it though... :)
masak wat, nqp, you don't know what a "modeul" is!? oh... 21:09
jnthn :P
masak .oO( it's Korean! ) 21:10
[Coke] ah, we will probably also get the FormatConversionException errors with a working sprintf
jnthn yeah
masak we'll... get them?
jnthn well, get rid of... :P
masak yeah, that's the idea.
21:10 bruges_ joined
masak bruges_: aloha. 21:11
jnthn: updated sprintf.nqp and with added stacktrace: gist.github.com/masak/d04ffe773bdfdc07d7e3 21:12
seems it fails during parse.
jnthn class sprintf::Actions # did you meant to keep the sprintf:: there? 21:13
hmmm 21:14
masak: Where did you add it, ooc?
masak hm, no; I'll remove the sprintf:: there too. 21:15
jnthn Near to the end?
masak jnthn: I added a line in tools/build/Makefile-JVM.in
yeah, after testing.nqp
jnthn hmmm 21:16
masak before YOUAREHERE.nqp
jnthn Thing is that NQPRegex, the code objct for NQP regex-y things, is also declared in the NQP setting.
Wondering if we're hitting a "can't do this yet" thing... 21:17
Oh...
...we don't actually have QRegex compiled at this point either.
That can't help...
masak: Let's put it in the HLL lib instead.
We know grammars/actions work in there :)
masak ok. 21:18
src/HLL/sprintf.nqp ?
jnthn wfm
masak seems to work much better :) 21:21
jnthn++
jnthn :) 21:22
hm, think I may have figured out why version.t explodes... 21:24
masak ok, I got all the way to stage2 this time ;) 21:29
jnthn ooh
And then? :)
masak "Method 'jast' not found" during compilation of src/stage2/gen/QRegex.nqp
jnthn ...what?! 21:30
oh, OK, NFA.nqp mentions nqp::sprintf, so I guess that's our first use of it
Can I see the QAST::CompilerJAST patch? 21:31
masak gist.github.com/masak/5a34e6893a80a5420c2d
jnthn woo, I have a passing verstion.t 21:32
oh, oops
yeah, you need to compile that QAST 21:33
$qastcomp.as_jast(QAST::Op.new( ... ))
Also, after :op('call'), add a :returns(str),
The problem is that something registered with add_core_op needs to return JAST, not QAST, even if it is just desugaring. 21:34
masak ah. 21:35
:returns(str) like that? not :returns('str') or something? 21:36
jnthn Like that; it expects a 6model obj
Well, a real type
masak I see :returns($TYPE_STR) elsewhere in that file. 21:37
jnthn This is one of the places QAST and 6model intersect in a deeper way :)
jnthn looks
JAST::Method.new( :name('hllName'), :returns($TYPE_STR), :static(0) )
masak: That's a JAST node, not a QAST one ;)
FROGGS masak: %f is working for the given tests... gist.github.com/FROGGS/0fdcdde74679e512f097 21:38
I wonder how we detect bigint stuff... 21:39
masak FROGGS: nice! 21:40
yes, I haven't dug into bigint matters at all.
FROGGS I'm not sure we can say: hey, this arg is a bigint, lets use abs_I instead of abs_i 21:41
jnthn I wonder if we should provide a way to configure the bigint types
Then at some point in the setting nqp::gethllsym('nqp', 'sprintf_bigint_reg')(Int) or so 21:42
FROGGS sounds nice... what does it do?
jnthn I dunno, I ain't the one implementing sprintf :P 21:43
FROGGS hehe
jnthn Maybe you keep an array of known bigint types and just see if the thing you have matches that type...
masak wat... build completed!
jnthn Or just make it a scalar
wat wat
masak $ ./nqp 21:44
> sprintf("OH HAI")
java.lang.NullPointerException
hehe.
FROGGS hehe
jnthn nqp::sprintf
masak oh!
FROGGS wait for it...
jnthn Not sure if nqp's repl auto-prints...
masak > say(nqp::sprintf("OH HAI"))
OH HAI
\o/
jnthn \o/
FROGGS \o/
jnthn masak++
FROGGS masak++
masak I really didn't expect that.
any test I can do to make sure it's really ours?
I guess anything with %f... 21:45
FROGGS %B?
masak > nqp::sprintf("%f", 4.2)
'f' is not valid in sprintf format sequence '%f'
yes, it's ours.
FROGGS cool
masak > say(nqp::sprintf("%B", 1000))
1111101000
ok, so. I'll commit and push this, yes? :) 21:46
jnthn go for it, imho
dalek kudo/nom: 516d49f | jnthn++ | src/vm/jvm/Perl6/Ops.nqp:
Fix // compilation, thus fixing version.t.
jnthn There's 35 more from version.t, and maybe a few more from elsewhere 21:47
dalek p: 2057ed3 | masak++ | / (3 files):
add nqp-native nqp::sprintf for the JVM backend

The plan is to extend this implementation until it passes all of the roast spectests that the Parrot backend passes. Until then, it's JVM-specific, though, because currently the JVM implementation will benefit from having this one rather than what it had before.		 21:48
masak FROGGS: I didn't check whether you have pushed commits since I copy-pasted. checking now.
FROGGS I have not I guess 21:49
masak yes, it seems I missed 3b345540
feel free to port it over.
FROGGS ohh, k
masak hm, any reason I should keep my github repo at this point?
maybe I should delete it?
jnthn masak++, FROGGS++, (others who helped)++
FROGGS masak: go ahead
masak deletes 21:50
FROGGS hmmm, is there a sprintf.t? 21:51
masak FROGGS: nope.
jnthn installs the updated NQP and builds a Rakudo against it
masak FROGGS: we might want to port that over, too.
FROGGS: but my time is up for today, realistically.
FROGGS yeah, makes sense if it is an nqp-native
I'll add it 21:52
masak FROGGS++
FROGGS :o)
masak FROGGS: remember that it's JVM-only for now.
FROGGS yeah...
jnthn There's a t/jvm 21:53
We can move the test file later when it comes standard on all impls
FROGGS that was what I was looking for right now :o) 21:54
thanks
btw, how do I proper make clean?
jnthn There's a make clean target, no?
FROGGS yes, but it doesnt help
jnthn if it's improper I suggest properizing it 21:55
masak git clean -dfx, perhaps?
FROGGS main class 'nqp' wasn't found or something like that
jnthn Oh, did you re-Configure?
FROGGS ahh, git clean.... I try that
jnthn 'cus a file was added...
masak yeah.
FROGGS ohh, yeah, that too
masak rakudo is better than nqp at detecting when you have an outdated Makefile
jnthn masak: Did you trying building Rakudo against this?
...no, you didn't :(
FROGGS :o( 21:56
jnthn --stagestats uses nqp::sprintf.
Apparently, a not-working feature of nqp::sprintf.
FROGGS which one?
I'll add it :P
jnthn Cannot stringify this in statement (src/stage2/gen/NQPHLL.nqp:2099)
FROGGS hmmm
masak oh, that's interesting.
jnthn: doesn't Rakudo fix a version of nqp?
and build against that? 21:57
jnthn oh...nqp::join
masak: For Rakudo JVM we're always doing HEAD
masak oh!
sorry, I didn't realize.
jnthn masak: Otherwise it'd be bump all the way...
np
FROGGS jnthn: we should use join() ?
masak should I revert the nqp commit for now?
FROGGS still compiling...
21:58 spider-mario left
jnthn masak: nah, let's fixing it 21:58
@pieces.push: infix_x(padding_char($st), $st<size>.ast - nqp::chars($st.ast)) if $st<size>;
...huh, nqp supports that syntax?
nqp: my @a; @a.push: 123; say(@a[0])
camelia nqp: OUTPUT«123␤»
jnthn wow!
I didn't know that!
FROGGS *g*
masak NQP: Quiter than you'd think! 21:59
jnthn tries a fix
dammit, no 22:04
Stage %-11s: %7.3f 22:07
That's the format
masak oh, we don't do %f yet.
FROGGS I can push %f support in a sec
jnthn ok
masak \o/
feel free to take any action necessary to make Rakudo-JVM continue to build.
I'm going to bed.
'night, #perl6
jnthn FROGGS: Can test if it works with --stagestats fwiw 22:08
FROGGS: Even with just nqp
FROGGS k
gnight masak
k, I see the fail on an unpatched sprintf 22:09
jnthn: so should I use join() instead of nqp::join()? 22:10
I remember that you mentioned it last week
sorear masak++ FROGGS++ et-al++
jnthn FROGGS: I'm not sure that's the actual problem here, though
FROGGS: That explodes with a different error.
FROGGS k
jnthn Here it really is finding a something it doesn't know how to stringify. 22:11
Meaning calling join(...) won't help as it just stringifies all the bits...
sorear um 22:12
PIRT? fresh_p()? what is this
jnthn huh? Where?
sorear looking at jnthn's defor patch
FROGGS fresh prince?
jnthn sorear: The -'s, removing commented out code?
sorear yes 22:13
jnthn sorear: The JVM Ops.nqp started out as a copy of the Parrot Ops.nqp all commented out
sorear: And I uncommented/translated
sorear what's that used for? surely we couldn't have passed 92% without a working //
jnthn This was one of the last pieces, if not the last one.
It "worked"
It was using isconcrete
Which works for most cases
But not for things you actually need to call .defined on. Like a Failure. 22:14
btw, it's 94% now :P
sorear jnthn: is PIRT QPOST? 22:16
jnthn Yes
Its main difference is it collapses to a string ASAP rather than keeping a gazillion nodes around in memory.
Well
"ASAP"
Once a QAST::Block is done. 22:17
22:17 iSlug left
FROGGS ahh 22:18
jnthn suddenly wonders if the interop stuff is gonna be efficient enough to use ASM directly from QAST::Compiler ;)
Though it may not be convenient :)
FROGGS jnthn: in HLL::Compiler it gets an arrayref
jnthn FROGGS: nqp::sprintf normally takes an array no? 22:19
oh!
FROGGS no
jnthn sub sprintf($format, *@arguments)
yeah :)
FROGGS :o)
jnthn It should be @arguments :)
That's how the op is expected to be :)
FROGGS k
then it should now be fine on my box... we'll see 22:20
22:22 snoopy joined
FROGGS This representation can not unbox to a native str 22:22
:/ 22:23
sorear jnthn: should be. there's no reflection going on, it basically just generates sub foo($a,$b,$c) { nqp::box_j(whatever(nqp::unbox_j($a), ...)) }
jnthn neat
FROGGS: OK, *that's* the join/nqp::join error :)
FROGGS yeah, already changed && compiling
jnthn
.oO( a much better approach than changed & compiling )
 22:24
lizmat is back from WWZ
jnthn There's a Zth world war?
lizmat Brad Pitt's newest flick
jnthn oh!
jnthn hadn't heard of it :)
perigrin based on a book of the same title.
FROGGS haz kids, so no cinema for poor FROGGS 22:25
lizmat the way I understand it, is that the movie has not a lot to do with the book anymore
perigrin Kids enjoy seeing movies too.
FROGGS jnthn: works
perigrin lizmat: yeah I've heard it more as "based on the title of a book of the same name."
sorear jnthn: OTOH, it would basically kill any chance of ever supporting cross compiling again :(
dalek p: 74575f5 | (Tobias Leich)++ | src/HLL/sprintf.nqp:
impliment %f and changed param to be non-slurpy
 22:27
22:28 dmol left
FROGGS jnthn: "use Test" is implicit? 22:29
sorear FROGGS: there's no Test in nqp, ok() and plan() are in the setting 22:30
jnthn sorear: oh, good point
FROGGS sorear: k, thanks
jnthn FROGGS: Does --stagestats work now?
FROGGS jnthn: yes
22:31 dmol joined
sorear jnthn: that said, cross compile doesn't quite work at the moment anyway. -target=jast output is missing the calls to nqp::deserialize - haven't looked closely. 22:31
jnthn sorear: Probably that jast isn't recognized as a precomp target 22:32
timotimo you people seem to be haling fun :) 22:36
i'm glad
jnthn: you didn't get around to making compile-time inlining yet afaict?
jnthn timotimo: no 22:39
Won't get to it today
Did fix some tests though :)
timotimo yeah, it's all good :) 22:40
i'm just reasonlessly fixated on benchmarking and performance improvements for some reason ;)
jnthn It's nice to make things faster, but also good to keep chewing through the spectest fails. 22:41
timotimo has the 95% mark been reached in the mean time? last i saw we were a bit past 94% 22:42
jnthn Though I said on my blog the aim was for 95% by the next Rakudo compiler release, and I think we're gonna make that.
timotimo sounds doable indeed
jnthn I'm not sure whether we've made it yet. We stand a good chance of doing so in the run tomorrow; if this sprintf stuff works then we'll certainly make it.
We may even hit 96% or so 22:43
timotimo that would be swell :)
jnthn FROGGS: I'm...not sure it gives quite the right output ;)
Stage start : 0
timotimo but the compiler release is still a week into the future, no?
jnthn Not sure
jnthn checks
sorear next rakudo compiler release? I think we're likely to get to 95, yeah :)
jnthn 2013-07-18 Rakudo #66 tadzik
tadzik yep
jnthn Over 10 days 22:44
FROGGS jnthn: what is the expected? 0.0?
timotimo 0.001 or so i think
sorear erp. I was thinking of Niecza Irregular Release Day (-29)
jnthn FROGGS: I think it is meant to always get the same numbe of decimals.
FROGGS okay, will check and fix
sorear 15:07 < jnthn> Stage %-11s: %7.3f
jnthn Anyway, confirm that Rakudo builds again now
FROGGS++
FROGGS ohh, that is good to know 22:45
sorear fF The double argument is rounded and converted to decimal notation in the style [-]ddd.ddd, where the number of digits after the decimal-point character is equal to the
precision specification. If the precision is missing, it is taken as 6; if the precision is explicitly zero, no decimal-point character appears. If a decimal point
appears, at least one digit appears before it.
in %7.3f, the precision is 3
jnthn Well, S02-literals/listquote-whitespace.rakudo.jvm passes :)
sorear so \d+\.\d**3
timotimo tadzik: is there something i can try to do to make this compiler release already be compatible with panda?
FROGGS yeah, have to pad with 0s
timotimo because that would be ridiculously cool
jnthn And S02-literals/quoting-unicode.t 22:46
sorear padding on the left with spaces to a minimum of 7 chars
timotimo "remember last month when our jvm port could do nothing? this time we'll give you a module installer to go with your spiffy new rakudo-jvm"
sorear FROGGS: If you're padding on the right with zeros, you're doing it wrong
jnthn ok 71 - Can't quote a string with a snowman and comet (U+2603 and U+2604)
bwahahah!
timotimo :D
jnthn That still goes down as one of the best Rakudo bugs ever. :)
FROGGS sorear: yeah, need to think about it
timotimo jnthn: it would be amazing if someone actually stumbled upon that bug by accidentally having a string where they expected ... what exactly? 22:47
tadzik timotimo: is it not compatible in some way?
sorear FROGGS: %f should basically do round($num * 10**$prec); format as %0{$prec+1}d; insert a . $prec spaces from the right if $prec > 0
timotimo i dunno, there's probably something missing i guess?
dalek p: 706fb51 | (Tobias Leich)++ | t/jvm/06-sprintf.t:
added tests for nqp::sprintf
 22:48
tadzik timotimo: you mean on the jvm part?
jnthn FROGGS: On gist.github.com/coke/5879701, first 3 test files that failed in sprintf now pass
FROGGS awesome
jnthn FROGGS: S32-str/sprintf.rakudo.jvm is the first that fails
'i' is not valid in sprintf format sequence '%i'
sorear sorear$ ls t/jvm
01-continuations.t 06-sprintf.t
what happened to 2 thru 5?
FROGGS jnthn: that patch is in the queue
tadzik timotimo: so far it's lacking %*CUSTOM_LIB, that's what I encountered 22:49
jnthn asbra! S32-temporal/Date.t passes! :)
FROGGS sorear, it will be moved to t/hll soonish
timotimo tadzik: i mean anywhere i might be able to chop off a LHF ;)
FROGGS so I took the named it would have in t/hll
jnthn S32-temporal/DateTime-Instant-Duration.rakudo.jvm fails 1, which may or may not be sprintf related
22:50 woolfy joined
sorear
.oO( Instant Perl 6! )
 22:50
timotimo just add jnthn
tadzik timotimo: well, um, running rakudo-jvm from outside the source directory would be a astart :)
jnthn S32-temporal/DateTime.rakudo.jvm passes now, S32-temporal/calendar.t also, S32-temporal/local.t too 22:51
22:51 woolfy1 left
jnthn !! 22:51
timotimo hmm, don't really know how to do that
jnthn So it's only sprintf.t left from that section that aborts
dalek p: 3ad22fa | (Tobias Leich)++ | src/HLL/sprintf.nqp:
add %i, a synonym for %d
22:51 raiph left
FROGGS jnthn: but sprintf.t fails because of bigints ant rats 22:51
sorear jnthn: how many passes are we up to? 22:52
jnthn And in other news, S02-names-vars/fmt.rakudo.jvm passes.
FROGGS: integration/advent2009-day02.t is the other one needing sprintf attention.
FROGGS k
jnthn FROGGS: So, sprintf.t and that one are the ones to look at :)
sorear: lemme check
sorear wonders if jnthn uses the t/spec/test_summary --archive 22:53
jnthn sorear: I dunno about extra passes, but looking at Coke's gist, we're now running 470 tests we previously aborted.
sorear \o/
jnthn sorear: No, I just ran the ones from [Coke]++'s gist manually, out of curiosity how we did on them.
tadzik timotimo: I suppose it's a matter of perl6.class looking for other classes
so there's some CLASSPATH adjusting in order
jnthn This could mean that we're going to, for the first time, fall below 1000 failing tests. 22:54
tadzik at least that's my guess
timotimo isn't that ... yeah classpath
sorear there is no perl6.class
timotimo perl6.jar then?
jnthn (less than 1000 behind Rakudo on Parrot)
tadzik well, whatever lives in ./perl6
timotimo at the moment i'm very close to falling asleep unintentionally 22:55
jnthn timotimo: Suggest finding a horizontal surface, preferably a padded one :)
sorear right now, nqp-runtime.jar, rakudo-runtime.jar, asm-4.1.jar, jline.jar, and asm-tree-4.1.jar need to be in the application classpath for nqp/rakudo to function
timotimo looks at how intellij idea starts itself 22:56
sorear all other files are searched for using @*INC, not the classpath
I've been planning to improve the classpath handling in nqp
timotimo that seems a bit overkill.
sorear in particular we're currently referencing the application class loader in quite a few places where we ought to be referencing the nqp-runtime.jar class loader 22:57
22:58 raiph joined
timotimo i don't even know what a classloader does 22:59
sorear so what are *you* trying to accomplish, timotimo? 23:01
my goals for the classpath improvements are 1. one-jar deployment 2. applet mode 23:02
jnthn I think timotimo wants a make install so he can then try and get Panda working ;)
sorear oh
15:50 < tadzik> timotimo: well, um, running rakudo-jvm from outside the source directory would be a astart :)
jnthn I think you already can do that...
sorear I thought we already had that?
jnthn :)
tadzik we do? :)
jnthn 'cus the .bat/.sh has absolute paths. :)
tadzik > .rakudobrew/jvm/perl6
Error: Could not find or load main class perl6 23:03
or am I missing something
sorear sorear$ cat perl6
#!/bin/sh
exec java -Xms100m -Xbootclasspath/a:.:/Users/sorear/dl/rakudo-jvm/../nqp-jvm/install/nqp-runtime.jar:/Users/sorear/dl/rakudo-jvm/../nqp-jvm/install/asm-4.1.jar:/Users/sorear/dl/rakudo-jvm/../nqp-jvm/install/asm-tree-4.1.jar:/Users/sorear/dl/rakudo-jvm/../nqp-jvm/install/jline-1.0.jar:rakudo-runtime.jar:perl6.jar -cp /Users/sorear/dl/rakudo-jvm/../nqp-jvm/install perl6 "$@"
note the relative paths to rakudo-runtime.jar and perl6.jar
jnthn ah
tadzik it uses relative paths here
exec java -Xms100m -Xbootclasspath/a:.:nqp/install/nqp-runtime.jar:nqp/install/asm-4.1.jar:nqp/install/asm-tree-4.1.jar:nqp/install/jline-1.0.jar:rakudo-runtime.jar:perl6.jar -cp nqp/install perl6 "$@"
jnthn That is probably fixable 23:04
sorear that's probably because you passed a relative path to --with-nqp
jnthn ah
tadzik yes 23:05
timotimo ooooooh 23:06
is it that easy to fix?
jnthn mebbe :) 23:08
sorear: I'm thinking about sleeping soon; anything you need from me before I go?
jnthn noted that "make jnthn understand X" was a goal ;) 23:09
dalek ecs: 47003e0 | (Elizabeth Mattijsen)++ | S99-glossary.pod:
Add wfm
 23:11
sorear jnthn: nothing quick 23:16
jnthn ok :)
sorear jnthn: if you feel like reading a half-written infodump before bed, I can gist what I have... 23:17
jnthn sorear: Nah, it'll likely make more sense to me when it's done and I've slept ;) 23:18
sorear gnight then :D 23:19
23:20 pmurias left
jnthn 'night o/ 23:20
lizmat night jnthn! 23:22
23:29 colomon joined 23:33 dmol left
colomon o/ 23:34
anything exciting happen today opn #perl6?
sorear colomon: masak's sprintf stuff was merged, and I committed java->p6 callback support 23:36
colomon \o/
sorear we're passing "a few hundred" more tests with sprintf and other misc fixes
I have a little nqp demo that subclasses java.lang.Thread, implements run, and starts it
colomon \o/ 23:37
sounds like a very good day
sorear I'm not done.
colomon \o/
sorear Although you did just miss jnthn. 23:38
colomon I may be just starting. :)
don't really have anything else on for the night other than drinking wine with my uncle.
labster sounds exciting today, sad I missed it. I had to drive to $dayjob and pick up a new computer. It's the same model as lizmat's new laptop, so I'm very happy with it. 23:44
And hopefully, it will run the spectests faster
lizmat spectest run a ~7 minutes 23:45
labster nice
colomon which one, jvm or parrot? 23:46
23:46 thou joined
sorear yow 23:48
my 1-year-old work laptop takes nearly half an hour at TEST_JOBS=4 for jvm 23:49
23:49 augustus joined
colomon started now 23:49
sorear remembers dons in #haskell purloining a 16-core work machine in 2007 for work on Data Parallel Haskell
lizmat colomon: I only do parakudo for now 23:51
augustus hello, I have some questions creating and using sub-languages, can anyone tell me if it still requires hacking the rakudo source as in this article: strangelyconsistent.org/blog/extend...x-of-perl6
FROGGS augustus: yes, I think so 23:52
augustus: since you can't access %*LANG from perl6 land right now, if you could, you could mixin rules
sorear r: say %?LANG.WHAT 23:53
camelia rakudo 516d49: OUTPUT«===SORRY!===␤Variable '%?LANG' is not declared␤at /tmp/3aW7o0ysM7:1␤------> say %?LANG.WHAT⏏<EOL>␤ expecting any of:␤ method arguments␤»
augustus do you know if any work has been done in this area for common languages like SQL, XML, HTML etc? or is this work that needs to be done from scratch?
sorear r: say nqp::getlex('%?LANG').WHAT
camelia rakudo 516d49: OUTPUT«Can only use get_what on a SixModelObject␤ in block at /tmp/Qjb3ajFCps:1␤␤»
sorear r: BEGIN say %*LANG
camelia rakudo 516d49: OUTPUT«===SORRY!===␤Nominal type check failed for parameter '$got'; expected Any but got Perl6::RegexGrammar instead␤»
FROGGS augustus: I am working on a Perl5 slang, but I think there is no SQL, XML or (X)HTML slang out there atm 23:54
a slang only makes sense if you perform action on the token you get IMO, so why make a HTML slang at all? 23:55
actions*
augustus FROGGS: once I translate the standard grammars out there into Perl6 notation, how hard is it to make a "slang"
23:55 lizmat left 23:56 lizmat joined
FROGGS augustus: not very hard, since my Perl 5 slang already is a module (just with minor rakudo-hacks) 23:56
sorear augustus: a slang is only useful if you want to embed XML directly in your perl 6 code, without quotes
augustus: this seems marginally useful at best
augustus FROGGS: yes I understand this. I am interested in making a Perl6 version of JSP type stuff
FROGGS ahh, and I should mention that the Perl 5 slang is actually a nqp-module, not Perl 6 23:57
augustus I'm not sure how much of a difference there is with NQP, I haven't worked with it at all
FROGGS well, when you think of the rakudo compiler, this sits on top of nqp 23:58
so if you have an nqp module, you dont have access to the high-level bits 23:59
augustus meaning that "slangs" have only NQP not full perl6?

Powered by Cro and the Raku® Programming Language.

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

Thank you!