»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg p6eval perl6: ... | irclog: irc.perl6.org/ | UTF-8 is our friend!
Set by sorear on 4 February 2011.
djanatyn juts submitted my first google code-in task :) 02:22
*just 02:23
diakopter you're a sparticle 03:48
jnthn morning 09:42
moritz \o 09:43
jnthn Quiet here today! 09:45
moritz I'm trying to make my way through the dyncall build system 09:46
it uses a lot of includes 09:47
jnthn moritz: yeah...it seems to be trying to do the Right Thing for lots of different platforms. Though not sure how well that actaully works out ;) 09:51
My hope was that we can mostly just delegate to it.
moritz jnthn: seems there's an easy-ish fix for my build problem
jnthn yay :)
moritz we can just pass BUILD_DIR to make
jnthn aha 09:52
dalek p/nci: 329046c | moritz++ | Configure.pl:
Fix the build on Linux amd64/gcc

Without this fix, "make" in the dyncall-0.6 directory would copy the .a files into a build_out/*/ dir, and thus the nqp ops compilation failed.
I hope this doesn't break the build on other platforms.		 10:50
jnthn yay! 10:51
moritz if it does break the build on other platforms, we might try make BUILD_DIR=. or so 10:52
jnthn Still works fine for me :) 10:55
moritz do you use nmake? 10:56
jnthn yes
...oh, and you didn't change that branch :) 10:57
(branch of the if statement)
moritz that's special-cased in Configure.pl, right
dalek kudo/nci: 1340fa6 | jnthn++ | src/core/Parameter.pm:
Parameter.type should return the type, not its name.
 11:59
p/nci: e11c46a | jnthn++ | src/ops/nqp_dyncall.ops:
Decontainerize passed arguments.
jnthn use NativeCall; 12:01
sub GotValues(int, num32, Str) is native('foo.dll') { * }
GotValues(42, 1.23e0, 'oh hai');
jnthn works \o/ 12:01
moritz \o/ 12:08
moritz do return values work too? 12:16
jnthn moritz: Yes, for integers and floats only so far. 12:17
tadzik mood gorning #perl6 12:43
jnthn o/ tadzik 12:46
dalek : a3341b7 | duff++ | misc/perl6advent-2011/topic-brainstorming:
Add jnthn's topics
 13:44
dalek : ce5b7c9 | jimmy++ | misc/perl6advent-2011/topic-brainstorming:
add nci to topic branstorming
 13:58
moritz should I remove rakudo 2011.07 from the feature matrix? 14:00
there's not too much left that it does that nom doesn't
niecza: die "foo"; CATCH { default { say "ex: $_" } } 14:02
p6eval niecza v11-27-gdaa9117: OUTPUT«ex: foo␤»
moritz nom: do { ENTER { say "foo" } }
p6eval nom e3246e: OUTPUT«===SORRY!===␤ENTER phaser not yet implemented at line 1, near " }"␤»
dalek atures: 154e986 | moritz++ | features.json:
update entries for Temporal and CATCH
 14:03
tadzik moritz: let's wait for nom maybe?
erm, start
grr
star on nom
moritz tadzik: ok 14:04
tadzik nom: class A { method foo { say self.WHAT } }; A.new.foo 14:12
p6eval nom e3246e: OUTPUT«A()␤»
tadzik nom: class A { method foo { say "{self.WHAT}" } }; A.new.foo 14:12
p6eval nom e3246e: OUTPUT«Use of uninitialized value in string context␤␤»
tadzik ^ bug 14:14
jnthn Not a bug. 14:15
The first .gists it. The second .Strs it. 14:16
tadzik mhm
anyone, HTTP::Server::Simple now works on nom 14:17
it could use some tests though, I can only guess what the examples were meant to do
ie should the 4th one just fail with "app missing"
tadzik nom: say so'/my-name' ~~ / '/' (.+) '-' (.+)/ 14:20
p6eval nom e3246e: OUTPUT«Bool::False␤»
tadzik anyway, bailador magically works too
tadzik feels a temptation to clear the ecosystem of modules that he have never seen working 14:23
tadzik shop & 14:24
gfldex rakudo gives me an error without a line number what makes me a little sad 16:00
gist.github.com/1391681
jnthn gfldex: That looks like a crash inside the compiler 16:02
tadzik gfldex: that's because it's a compiler error
jnthn (which just shouldn't happen)
jnthn Whoa. It actually crashes inside the optimizer. 16:03
That's...rare.
jnthn gfldex: Run it with --optimize=off to avoid the issue; I'll track it down later. Filing a ticket would be helpful so it doesn't get lost :) 16:04
tadzik that's probably due to operator defining magic 16:05
multi method postcircumfix:<( )>() {}
jnthn That'd seem like an odd thing to confuse the optimizer.
I mean, it's just a method.
nom: my $foo; $foo>>.() 16:07
p6eval nom e3246e: OUTPUT«===SORRY!===␤Method 'isa' not found for invocant of class 'Undef'␤»
jnthn Golfed.
nom: $_>>.()
p6eval nom e3246e: OUTPUT«===SORRY!===␤Method 'isa' not found for invocant of class 'Undef'␤»
jnthn Even further golfed.
nom: $_>>.(1)
p6eval nom e3246e: OUTPUT«Too many positional parameters passed; got 3 but expected 2␤ in sub METAOP_HYPER_POSTFIX at src/gen/CORE.setting:8093␤ in block <anon> at /tmp/eJvWJPFQHn:1␤ in <anon> at /tmp/eJvWJPFQHn:1␤»
jnthn ah 16:08
And that looks like a mis-compile
TimToady Util: my Num() $n = 0; is specced to allow coercion to Num, but nobody implements coercion types yet 16:53
felher Any thoughts on gist.github.com/1391784 ? Otherwise i put it on Rosetta Code, which has no tic-tac-toe-game in Perl6 (or Perl). There is quite a bit of Perl6 stuff in that few lines. Lazy lists, hyper- and reduce-operators, junctions, placeholder-variables, 3-arity-map... 16:55
dalek p/nci: 3587326 | jnthn++ | src/ops/nqp_dyncall.ops:
Fix string passing memory leak.
 16:57
p/nci: 497c48b | jnthn++ | src/ (3 files):
Support user-chosen calling conventions (or at least a few extra ones).
jnthn There's a v2 branch in the Zavolaj repository where I've put what I got done so far of the new NativeCall.pm6. 17:12
moritz jnthn: could you please add host04.appflux.net:5000/dalek?t=freenode,perl6 as a post-commit hook to the zavolaj repo? 17:13
I think we all want to know what's going on in there :-)
jnthn added :) 17:14
Though it missed the initial commits :)
github.com/jnthn/zavolaj/blob/v2/l...veCall.pm6 # the only really interesting thing 17:15
dalek p/nci: 896015d | jnthn++ | src/ops/nqp_dyncall.ops:
Make it so passing a string type object will lead to a null being passed.
 17:20
jnthn There's one thing previous zavolaj never did... :) 17:21
flussence yay! 17:23
jnthn OK, basic struct support next, I guess... :)
vmspb Hi 17:27
How to determine when a minus sign before the number (without whitespaces) is part of the definition of a negative number, and when it is a unary operator?
perl6: sub prefix:<-> ($i) {return 10 * $i}; say -3 + 0; say (-3) + 0; say "-3" + 0; say <<-3>> + 0; say «-3» + 0;
p6eval rakudo e3246e, niecza v11-27-gdaa9117: OUTPUT«30␤30␤-3␤-3␤-3␤»
..pugs b927740: OUTPUT«decodeUTF8': bad data: '\171'␤decodeUTF8': bad data: '\187'␤30␤30␤-3␤-3␤-3␤»
daxim jnthn, pythonistas discover Class::Struct news.ycombinator.com/item?id=3273812
moritz vmspb: what kind of answer do you expect? 17:36
vmspb As defined in the standard? 17:39
moritz then the answer is probably "by reading the spec, and augmenting it where incomplete" 17:44
vmspb ok 17:47
flussence I'm trying to get panda to work... it bootstrap.sh's fine, then complains because ext/Shell/Command.pm isn't installed along with it. 18:01
tadzik bootstrap.sh is fine? News
flussence yep, there's stuff in .perl6/lib and all.
tadzik so tests for Test::Mock and JSON::Tiny pass these days?
flussence hm... it ran the panda tests with no test failures, I didn't see it run those though. I'll go check manually first... 18:03
apart from a few uninitialised string warnings JSON::Tiny passes 18:04
(with 1 todo!) 18:05
flussence oh, Test::Mock barfed. 18:06
gist.github.com/1391945 18:07
sorear trip&
tadzik flussence: yes, that's expected 18:13
flussence: does everything work if you install panda with ufo? 18:14
flussence nope. `ufo && make` dies while looknig for Shell::Command 18:16
s/ni/in
tadzik hm 18:19
oh, correct 18:20
you'll have to ' PERL6LIB=ext make 'or so
flussence aha, that worked. Still no Shell::Command installed though 18:23
tadzik yeah, you'll have to manually install everything in ext/ :/ 18:26
panda should do that in bootstrap, but bootstrap doesn't work because test-mock and json tests fail
flussence interesting... "Incorrect pre-compiled version of lib/JSON/Tiny.pm loaded" 18:28
rm'ed *.pir, almost works...
tadzik yeah, I know that one 18:29
free beer for anyone who fixes that
flussence aaaand... another error after all that. oh well :( 18:30
oh wait, I'm being dumb
(it's "panda list", not "--list"...) 18:31
that works \o/
tadzik :) 18:32
tadzik any ideas how to make HTTP::Server::Simple a little bit non-blocking to be able to test it? 18:57
moritz IME "a little bit non-blocking" usually doesn't work 19:15
for non-blocking you need quite a different API
tadzik well, any testability is fine 19:16
gfldex is there a nom web repl? 19:22
jnthn Not unless try.rakudo.org runs nom. 19:25
afk for a little bit 19:28
gfldex An error has occured on the server. Error Message: 19:31
Timeout, operation took too long.
that's for An error has occured on the server. Error Message:
that's for try.rakudo.org/
moritz nom: say True.gist 19:47
p6eval nom e3246e: OUTPUT«Bool::True␤» 19:48
moritz nom: say True.perl 19:52
p6eval nom e3246e: OUTPUT«Bool.new()␤»
dalek kudo/nom: d5d5a3f | moritz++ | src/core/ (2 files):
Merge pull request #42 from lumimies/nom

Fix gist/perl for Junction and Bool		 19:53
jnthn back 20:16
Time to make a cup of tea, and hack on structs some. 20:17
japhb hacking++ 20:22
moritz gist.github.com/0e91e92c32570e8fb15c # that's the current spectest summary (of nom branch) 20:23
any idea where the various failures come from?
jnthn t/spec/S04-blocks-and-statements/pointy.rakudo # mls++ pointed out that the test is wrong... 20:24
moritz nom: say ('100.B' .. '102.B').Str 20:25
p6eval nom e3246e: OUTPUT«No applicable candidates found to dispatch to for 'Numeric'. Available candidates are:␤:(Mu, Mu %_)␤␤ in method Numeric at src/gen/CORE.setting:621␤ in sub infix:<==> at src/gen/CORE.setting:2031␤ in sub infix:<==> at src/gen/CORE.setting:2031␤ in method infinite …
japhb I discovered (roughly) why my Str.Numeric changes broke 'my @a = "a".."c";' and I'm still working out the least invasive way to make it work again. But let me just say, the amount of work that gets done for that statement compared to 'my @a = 1..3;' is just crazy.
jnthn japhb: ah, is that the above bug?
japhb Um ... my changes are unpushed; that error is not from me. However, I think that 20:26
jnthn moritz: The S12 ones are new on me.
japhb: ah, ok
I have the range one here...
Or at least, it's in nci branch
japhb the error moritz is seeing is just a more subtle version of the same problem.
jnthn ah, ok
japhb (My changes make the problem more obvious, because they fail() rather than 'return 0' more often.)
jnthn moritz: The two S12 ones pass here 20:27
moritz $ ./perl6 t/spec/S10-packages/use-with-class.t
===SORRY!===
Incorrect pre-compiled version of src/gen/Metamodel.pm loaded
I've seen that one before
jnthn oh hm
moritz I think it goes away if I clean and recompile everything
still it shouldn't happen
jnthn Would really appreciate instructions on reproducing taht one. 20:28
moritz I've run Configure.pl before building my nom
moritz oh! 20:30
I've had a totally unrelated Foo.pir lying in my rakudo home
(from checking the precompilation of fat Rats)
tadzik the .pir stuff is quite broken
moritz deleting that makes the error from use-with-class go away
tadzik flussence observed that today as well
moritz ... and then fails after test 7 with 20:31
You cannot create an instance of this type
ah, that mmight be the Foo.pm
jnthn We really need a --trace-load or something that gets it to dump what modules i'ts actually loading, and from where, etc. 20:32
moritz and we need to get rid of . in the module loading path 20:33
jnthn Very much so.
And probably lib too
moritz aye
jnthn We do need a way to tell the test suite to use a local Test.pm over any installed one.
Well, a general way to say "take a local one over an installed one"
Trying to run the test suite with an installed Test.pir will, of course, fail epicly. 20:34
We *can* loosen this a bit, if we can start naming serialization contexts after a hash of the source.
Rather than a timestamp of when the compilation happened. 20:35
dalek ast: aed8452 | moritz++ | S04- (2 files):
fix a test, rakudo unfudges
 20:36
japhb How would I dump a stack trace from within the code I'm debugging? 20:40
Basically, I'm trying to figure out what the call chain was that led to something being called bogusly.
moritz an easy hack is try { die }; print $!.backtrace 20:42
nom: try { die }; print $!.backtrace
p6eval nom d5d5a3: OUTPUT« in block <anon> at /tmp/riwobDpo2g:1␤ in block <anon> at /tmp/riwobDpo2g:1␤ in <anon> at /tmp/riwobDpo2g:1␤»
jnthn moritz: could Backtrace.new() just me made to give you one from the current position? 20:43
japhb moritz: thanks for the idea!
moritz jnthn: yes, it can be made 20:44
just needs some fiddling, I guess
nom: say pir::getinterp__P().backtrace 20:45
p6eval nom d5d5a3: OUTPUT«Method 'backtrace' not found for invocant of class 'ParrotInterpreter'␤ in block <anon> at /tmp/homoBgBpqC:1␤ in <anon> at /tmp/homoBgBpqC:1␤»
moritz I just need to remember how to make parrot emit one :-)
moritz ah, I can just use the cheat that I told japhb about 20:48
it won't be very efficient, but I don't think it's a very common operation either
japhb One hopes not ... 20:50
moritz I have a small patch that makes Backtrace.new() DWIM, just need to check that it actually compiles :-) 20:52
dalek kudo/nom: de3997a | moritz++ | src/core/Backtrace.pm:
make Backtrace.new() DWIM
 21:01
dalek p/nci: 5899327 | jnthn++ | / (3 files):
Start to stub in a CStruct REPR.
 21:10
p/nci: 9f0fc01 | jnthn++ | tools/build/Makefile.in:
Add some missing makefile deps.
p/nci: c7250db | jnthn++ | src/6model/reprs/CStruct.h:
Sketch out more of how CStruct probably needs to end up looking once it supports the bits it'll need to.
p/nci: fabd74f | jnthn++ | src/ops/nqp_dyncall.ops:
An (untested) shot at getting the native call handling side of passing a CStruct in place.
felher rosettacode.org/wiki/Tic-tac-toe#Perl_6 21:23
japhb Hah! I think I found the problem: github.com/rakudo/rakudo/blob/nom/...nge.pm#L31 21:41
That == ends up coercing both sides to Numeric, and if the coercion of $.max results in Failure, then boom. 21:42
jnthn oh!
good find
japhb++
japhb thanks. :-)
What's the most performant fix? 21:43
jnthn most performant?
bind $.max into a lexical, then nqp::istype($max, Numeric) && $max == $Inf 21:44
japhb Will cause the least loss of performance.
jnthn well, that's probably fastest.
It saves calling .max twice
and nqp::istype is the fastest way to do type checks
japhb What about replacing == with something that checks identity (I'm fuzzy on those ... === or =:= ?) 21:45
jnthn Sounds a little risky 21:46
japhb Why? (I have a feeling the answer to that will Teach Me Something Important.)
Oh wait, 21:47
jnthn === will need to return an ObjAt at some point 21:47
japhb You could have gotten an Inf from some other way than assignment to $Inf, right?
And it would not be the same object.
jnthn And even what it does today is probably no cheaper than the ==
But more important is...yes, that. :)
japhb ah, OK.
japhb had the singleton-ness of $Inf locked in his brain, and forgot that it was merely one possible way to get an infinite value. 21:48
colomon_Droid o/ 22:09
jnthn o/ colomon_Droid
colomon_Droid happy Thanksgiving!
jnthn To you too :) 22:11
colomon_Droid sorear: ping 22:12
jnthn colomon_Droid: sorear mentioned he was leaving for a trip earlier; I got the impression he's away a couple of days 22:13
colomon_Droid Darn
jnthn++
Anyone out there know how to create a System.Byte in niecza? 22:14
colomon_Droid Doing this sort of development without a proper internet connection is a real pain 22:15
masak dear #perl6, hello. 22:18
colomon_Droid o/ 22:19
Tene hi masak 22:20
masak I'm having the busiest week of the year this week. that's why I've been a bit absent lately.
jnthn masak's doing all the work at $dayjob while I'm just bumming about at home hacking on NCI :P 22:21
masak yeah; slacker :P 22:22
Tene I had my busiest week of the year two weeks ago; I was on-call during switch upgrades in the cluster that went poorly (ran into a bug causing repeated switch failures). 22:27
arnsholt OTOH, we all heart jnthn for making NCI =D 22:28
masak ♥ 22:32
jnthn didn't really heart reading through a disassembly of generated in-memory assembly code trying to work out why his Win32 API calls blew up 22:33
Turns out they fixed the Win64 call conv alignment bug in SVN, so I just need to grab us the latest at some point :)
dalek p/nci: 02dca35 | jnthn++ | src/6model/reprs/CStruct. (2 files):
Steal some bits from P6opaque and twiddle with them to start getting things into a more CStruct-ish shape.
 22:40
p/nci: 878aafe | jnthn++ | src/6model/reprs/CStruct. (2 files):
Get allocationy bits of CStruct in place, or at least for the basic cases.
masak perl6: my %h = foo => 0, bar => 42; say so %h{ 'foo' | 'bar' } 23:05
p6eval rakudo de3997, niecza v11-27-gdaa9117: OUTPUT«Bool::True␤»
..pugs b927740: OUTPUT«*** No such subroutine: "&so"␤ at /tmp/PXzZmjvt_A line 1, column 34 - line 2, column 1␤»
masak perl6: my %h = foo => 0, bar => 42; say %h{ 'foo' | 'bar' }.perl
p6eval rakudo de3997, niecza v11-27-gdaa9117: OUTPUT«any(0, 42)␤»
..pugs b927740: OUTPUT«\undef␤»
jnthn ooh, basic struct passing nearly worked
japhb jnthn, would there be any performance advantage to binding 'self 23:09
erg
jnthn, would there be any performance advantage to binding 'self' to a lexical and using $self instead for an expression that involves several uses of self? 23:10
jnthn japhb: No
japhb: self *is* stored as a lexical :)
japhb jnthn, ah, good.
jnthn japhb: And looked up as one :)
So you'd just have another variable and spend time doing a binding, and win nothing ;) 23:11
japhb I suspected so, but as my machine is currently recompiling Rakudo, I couldn't check directly. :-)
nodnod
jnthn yeah, mine too
jnthn thinks he founds his off-by-16-bytes error
japhb Bah. I hate it when I find a bug, but have to go AFK and can't fix it immediately. :-( 23:15
dalek p/nci: c14163d | jnthn++ | src/ops/nqp_dyncall.ops:
Handle cstruct as one of the types we now recognize.
 23:15
p/nci: ceaf986 | jnthn++ | src/6model/reprs/CStruct.c:
Fix struct offset computation; add the attribute lookup bits we need for the basic int/num attributes we support thus far.
jnthn OK, here's the state of NCI now: gist.github.com/1392493 23:16
jnthn Still loads to do on this struct stuff, but it's easier to hack towards something that works more when there's something that basically works :) 23:16
jnthn The top file is C, which I compile to a DLL. The second thing is the Perl 6 binding to it. Note the "is repr('CStruct')" on the class. It's a normal class, but it builds a C structure underneath. Quite limited so far; in the end it'll do the struct of sturct thing, etc. 23:18
tadzik supercool :)
jnthn++
masak good night, #perl6 23:32
jnthn Updated gist.github.com/1392493 to show one thing that wasn't clear before 23:34
felher yeah, jnthn++ 23:39
jnthn sleep time & 23:50

Powered by Cro and the Raku® Programming Language.

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

Thank you!