dalek ast: 9c57b43 | lizmat++ | S32-num/rat.t:
Fix now passing tests
 09:56
lizmat (finally *something* on the perl6-dev channel today) 09:57
hmmm... RT #126391 09:58
m: say 42/(.1+.2-.3) # TimToady : how do you feel about this not throwing an exception anymore ? 10:01
camelia rakudo-moar a5c46b: OUTPUT«Inf␤»
dalek kudo/nom: 7afeb68 | lizmat++ | src/core/Rational.pm:
Simplify Rational.new

Part of a cunning plan to handle Inf.Rat better		 10:57
ilmari lizmat: re. (±Inf|NaN).Rat, would it make sense to normalise the numerator to ±1|0? 11:10
lizmat ilmari: that's what it basically does now
which makes us lose 42/0
ilmari but is that meaningfully different from 1/0? 11:11
both the old and the new code in Rationa.new preserve the numerator when the denominator is zero, afaict 11:12
lizmat ilmari: that last change was only to make the next changes possible :-)
spectesting that now
ilmari waits and sees
lizmat m: say Rational[Num,Int].new(Inf,0)
camelia rakudo-moar 7afeb6: OUTPUT«Inf␤»
lizmat it was to allow that 11:13
grrr.... we don't have a way to stub a role, do we? 11:14
ilmari m: say Rational[Int,Int].new(42, 0)
camelia rakudo-moar 7afeb6: OUTPUT«Inf␤»
ilmari m: say Rational[Int,Int].new(42, 0) eqv Rational[Int,Int].new(1,0) 11:15
camelia rakudo-moar 7afeb6: OUTPUT«True␤»
jnthn m: role Foo { ... }; role Foo { }
camelia ( no output )
jnthn Seems OK
Note that you must define it before you compose it into anything
But you can refer to it beforehand as a constraint etc. 11:16
lizmat ok 11:20
dalek kudo/nom: 498d0a4 | lizmat++ | src/core/ (2 files):
Better solution to Inf/-Inf/NaN.Rat

This basically creates a special purpose Rat in which the numerator is a Num rather than an Int. This preserves the normal Failure behaviour of dividing by zero, while keeping the isnanorinf information when coercing to a Rat.		 12:00
lizmat this makes the spectest clean again 12:01
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Simplify Rational.new 12:04
travis-ci.org/rakudo/rakudo/builds/133079966 github.com/rakudo/rakudo/compare/a...feb68bee12
lizmat one could argue that the special Rat should maybe have a denominator of 1
m: say 42/(.1+.2-.3) 12:10
camelia rakudo-moar 7afeb6: OUTPUT«Inf␤»
lizmat (not yet 498d0a4)
(not yet 498d0a4) 12:42
m: say 42/(.1+.2-.3)
camelia rakudo-moar 498d0a: OUTPUT«Attempt to divide 42 by zero using div␤ in block <unit> at /tmp/7Xg5wtvMhZ line 1␤␤Actually thrown at:␤ in block <unit> at /tmp/7Xg5wtvMhZ line 1␤␤»
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Better solution to Inf/-Inf/NaN.Rat 13:06
travis-ci.org/rakudo/rakudo/builds/133091879 github.com/rakudo/rakudo/compare/7...8d0a4ae457
tomboy64 oh joy 13:22
Files built without respecting CFLAGS have been detected: /usr/share/perl6/runtime/dynext/libperl6_ops_moar.so 13:23
:D
psch aha, nqp commit d751661 never was tested against rakudo, it seems 13:54
that's the NoSuchFileException i mentioned yesterday
'cause before we had IO::Path.dir check whether it's moar or jvm and react accordingly 13:55
but since that path changed nqp-j return only a filename instead of the absolute path it returned previously...
hopefully it's as easy as tossing the preprocessor directives from IO::Path... 13:59
otherwise i'd rather revert that commit, unless someone else wants to figure out what IO::Path has to do instead with it
nope, tossing the #?if jvm bits and only running the moar path doesn't work 14:11
apparently nqp-j doesn't support chdir, and never really can support it platform independently
i suppose someone could try porting the nqp-j behavior to moar, so we can use only the #?if jvm path... 14:12
but until that i'll revert the PR that broke it
contingent on "it doesn't get worse", of course :)
dalek p: 8e10572 | peschwa++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/Ops.java:
Revert "Merge pull request #287 from MasterDuke17/master"

This reverts commit 62d0dbad096e9969b2fd082e6b2d020b1daba617, reversing changes made to ca37f08f9a9f633c72e967e6a0a6b9308788db6b.
Rakudo's IO::Path relies on absolute paths returned from nextfiledir, and since we cannot emulate the solution moar uses in IO::Path (because Java does not support changing the CWD, at least not without platform-dependent hacks) we will have to accept this inconsistency until someone rewrites IO::Path. This reverts the failure mode for install-core-dist.pl on Rakudo-j back to
  "setcodeobj can only be used with a CodeRef", from leaking a Java-level
NoSuchFileException.		 14:42
psch now if i just had a clue how to fix the setcodeobj failure mode :| 14:43
lizmat psch++ # sorry for the confusion 14:49
psch lizmat: no worries. i had strongly hoped we could just toss the #?if jvm part of IO::Path, but alas... :) 14:53
dalek kudo/nom: 1ab1fb6 | lizmat++ | src/core/Rational.pm:
Remove now unnecessary test, Brad Gilbert++
 15:00
lizmat wonders where the code for nqp::p6for lives 15:31
psch lizmat: in Perl6::Grammar iirc 15:33
lizmat: ah, no, Actions :)
line 391 on HEAD 15:34
lizmat psch++ 15:35
dalek p: c3122aa | (Pawel Murias)++ | src/vm/js/ (3 files):
[js] Instead of writing Chunk parts to a file and slurping, collect them into an array and then nqp::join them together.
 15:58
kudo/js: bf34614 | (Pawel Murias)++ | src/vm/js/perl6-runtime/runtime.js:
[js] Add the js runtime for the p6* ops.
 16:25
kudo/js: 79abb4e | (Pawel Murias)++ | src/vm/js/Perl6/Ops.nqp:
[js] Fix bug.
kudo/js: bccb2c8 | (Pawel Murias)++ | src/vm/js/ (2 files):
[js] Stub ops.
kudo/js: 1ef5a97 | (Pawel Murias)++ | src/Perl6/Metamodel/BOOTSTRAP.nqp:
[js] Enable the signature binder on the js backend.
rakudo/js: 0ec08b6 | (Pawel Murias)++ | src/Perl6/ (2 files):
rakudo/js: [js] some temporary #if's to work around some misssing nqp-js features.
lizmat decommute& 17:14
brrt for one would love to get rid of dynamic extension ops 18:46
psch brrt: what's that? 18:47
brrt it's basically a set of, or address space for, extension opcodes for MoarVM
stuff like perl6_invoke, perl6_return, perl6_getthis, perl6_stuffthat
psch huh 18:48
i assume that's motivated by your JIT work?
brrt imho, and i'm probably wrong in some important way, it defeats the purpose of having a VM purpose-made for the compiler
and yes, they are - slightly - annoying on the JIT level 18:49
specifically because they assume to be running in an interpreter
tadzik . o O ( perhaps rakudo should fork moarvm :P )
brrt which is a wider problem
there is probably a good reason that i just don't know
:-)
psch i don't think i have a sufficient grasp on the connotations :) 18:50
brrt that is not unreasonable :-) 18:51
basically, the JIT just wants to run code and get on with it
the interpreter, in contrast, looks at a stream of data at each point to decide what to do
we call this stream of data the 'bytecode', although a moarvm opcode is 16 bits wide
when you're running inside an interpreter, you can quite explicitly move about in the program simply by updating pointers 18:53
moreover, you can tell where you are simply by comparing pointers 18:54
now in comparison, JIT compiled code is /operated on/ by the CPU, and while the CPU has a register that tells you where it is, there is a): no way to explicitly read it, b): no way to update it directly 18:55
so all the fancy manipulations of your current position and bounds checks etc.... they are impossible, or at the very least, considerably more tricky 18:56
psch yeah, that sounds quite beyond my understanding... :)
brrt hmm, that means i didn't explain it well :-)
oh well
it is all very possible to deal with these things. it's just not a fun thing to do 18:57
jnthn .tell brrt I'm not that keen on them either, fwiw. If they go away we get to remove the code supporting them from MoarVM too. Alternative is we have API for JITting them. 19:30
skids Those would be needed if a 3rd party (like a rakudo module) wanted to add support for a low-level data type, right? 20:21
jnthn skids: That'd more be custom REPRs, which we don't do at all yet 20:24
[Coke] jnthn: I'm all for not having moarvm support custom opcodes. 20:29
the only goal is to support rakudo, no?
ugexe m: sub foo(--> Array|Str) { }; 20:49
camelia rakudo-moar 1ab1fb: OUTPUT«===SORRY!=== Error while compiling /tmp/WDh4DWFasW␤Type 'Array' is not declared. Did you mean 'array'?␤at /tmp/WDh4DWFasW:1␤------> sub foo(--> Array⏏|Str) { };␤»
jnthn [Coke]: Pretty much 20:55
Think that's a confused error :)
Heuristics don't always go so well... :) 20:56
brrt jnhtn: you mean like, insert expr fragments for extension ops? we could do that, but it is not very high on the priority list 21:02
oh well, too tired today :-) 21:08
AlexDaniel ugexe: rt.perl.org/Ticket/Display.html?id=125902 21:17
m: my Int; 21:18
camelia rakudo-moar 1ab1fb: OUTPUT«===SORRY!=== Error while compiling /tmp/bbnrsiWAAY␤Malformed my␤at /tmp/bbnrsiWAAY:1␤------> my Int⏏;␤»
AlexDaniel hmmm, the error message is different now
ah, it is mentioned in the ticket
bisect: use Test; exit 1 if throws-like { EVAL ‘my Str where 'foo' $test’ }, X::Syntax::Malformed 21:30
bisectable AlexDaniel: (2015-12-25) github.com/rakudo/rakudo/commit/07fecb5
AlexDaniel bisect: use Test; exit 1 if throws-like { EVAL ‘my Str where 'foo' $test’ }, X::Syntax::Malformed; exit 0 21:31
bisectable AlexDaniel: (2015-12-25) github.com/rakudo/rakudo/commit/07fecb5
AlexDaniel meh 21:32
ah right, the change was before 2015.12 21:34
tomboy64 how far along is the node.js backend? 21:40
psch tomboy64: blogs.perl.org/users/pawel_murias/ is the latest status report 21:47
tomboy64: there's been a few more nqp-js commits since then, but i gather if anything notable happened pmurias++ would have blogged again 21:48
tomboy64: note though that it's not so much intended as a node.js backend, but as a javascript backend - note.js is currently just the easiest way to try and make it run 21:49
tomboy64 :) i see
so "experimental pre-alpha" describes it best, huh?
psch fwiw i'm somewhat curious how the performance difference between e.g. asm.js rakudo-moar and native rakudo-js in a browser turns out 21:50
...i think that's a bit loaded. i'd call it "in development" :)
tomboy64 :p
hmmm. in a browser? 21:51
be careful what you wish for :p
psch nah :P
imagine rakudo.js with similar perseverance as jquery.js \o/
the world would be a better place :P
tomboy64 :D 21:52
tomboy64 bites his tongue
psch yeah, i know
the better solution would of course be convince the w3c that every browser needs a perl6 compat scripting engine :P
+to 21:53
tomboy64 HAH
psch in all seriousness, i don't know what would be the best mandated HLL for browsers 21:54
i do know that ECMAScript is weird as hell
tomboy64 well, ECMAScript was an attempt to standardize browser-js
psch and tacking lots of expressive features onto it doesn't help the fundamental weirds that are around
yeah, but that's the thing 21:55
tomboy64 the problem is that js in itself is aimed at hobby-devs
psch there's something around that's weird and broken
and then the commitee comes and says "we need these extra features to make it good"
tomboy64 with all kinds of stuff that's supposed to make "programming" easier - but instead makes the whole shebang more complicated to get it right 21:56
and introduces all kinds of errors
psch exactly, because of a wonky foundation
i mean, sure, Perl 6 has its share of WATs as well
tomboy64 dash was a good concept - that never took off
psch but they're almost exclusively hidden in obscure features
and the ones that aren't - like binding vs assignment - are easy to explain
...although i suppose binding could be argued to be an obscure feature already vOv 21:57
tomboy64 WAT?
psch yes, after gary bernhardts talk
www.destroyallsoftware.com/talks/wat
it's a good talk
not even 5 minutes :) 21:58
well, unless "binding is an obscure feature" was a WAT to you :) 22:02
AlexDaniel well, there are also some hopes for WebAssembly 22:05
tomboy64 and extraordinarily entertaining 22:07
really enjoyed that, thank you
psch there's a few more free talks by him, they're all good 22:08
not quite as humorous as the WAT talk, but still good
lizmat m: my @a = ^5; @a[2]:delete; $_++ for @a; dd @a # continuing from this afternoon, will look at this tomorrow 22:20
camelia rakudo-moar 1ab1fb: OUTPUT«Cannot call postfix:<++>(Any); none of these signatures match:␤ (Mu:D $a is rw)␤ (Mu:U $a is rw)␤ (Int:D $a is rw)␤ (int $a is rw)␤ (Bool:U $a is rw)␤ (Bool:D $a is rw)␤ (Num:D $a is rw)␤ (Num:U $a is rw)␤ (num …»
lizmat good night, #perl6-dev 22:24
psch m: say ++Any
camelia rakudo-moar 1ab1fb: OUTPUT«Cannot call prefix:<++>(Any); none of these signatures match:␤ (Mu:D $a is rw)␤ (Mu:U $a is rw)␤ (Int:D $a is rw)␤ (int $a is rw)␤ (Bool $a is rw)␤ (Num:D $a is rw)␤ (Num:U $a is rw)␤ (num $a is rw)␤ in block <uni…»

Powered by Cro and the Raku® Programming Language.

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

Thank you!