nine I fear I have to rewrite the whole dependency loading code to fix the occasional failures on file systems with coarse time stamps, i.e. RT #128156 09:09
Right now we try to load a precomp file, during loading of the dependencies notice that one of them changed, precompile the source again and load the result. 09:10
Sometimes compilation is so fast that it finishes in the same Instant (with just 2 second resolution on crappy file systems) as the dependency's. Then we declare it outdated though we just precompiled it and try to load from source. 09:12
Problem is that we really don't know which of them is older when the time stamps match, so we cannot rely on the target module being fresher even though we just compiled it ourselves. Someone else could have recompiled the dependency in the meantime. 09:15
All of this was much easier when we locked the whole precomp store during module loading. But a global lock is a luxury we just cannot afford :) 09:16
The solution will probably be to first check all dependencies, then open all dependencies' files, then check again and only if they still check out actually pass them to the CompUnit::Loader 09:18
If they don't check out again, rinse and repeat. 09:19
This will actually make module load times unbounded, but that's just the price for going lockless. 09:24
psch two second resolution on a file system really seems pretty crappy, at a glance :S 09:25
dalek kudo/nom: 05691d4 | lizmat++ | src/core/Rakudo/Internals.pm:
Make EmptyIterator still faster for push-all
 09:27
kudo/nom: beb3c98 | lizmat++ | src/core/List.pm:
Make iteration of Lists about 10% faster

  - by differentiating between reified/non-reified lists
  - by making sure an iterator doesn't take slow, non-reified path again
  - by making smaller, inlineable methods
  - by not allocating if not necessary
  - by using the new EmptyIterator if a list is completely empty
nine On to be fair, the race condition has been there all along. The timestamp issue just exposed it even in a non-parallelized test. 09:29
lizmat psch: that's FAT for you, or are there other ones ? 09:33
if I remember correctly, there were only 5 bits left for the second value
on MS-DOS FAT systems
nine lizmat: seems like on OS X we also only get 1 second resolution of time stamps. Though that could just be a missing patch for MoarVM 09:34
psch doesn't know a lot about file systems
lizmat nine: on OS X, you could use IO.watch 09:35
but I have doubts on the timeliness of it, judging from how long it sometimes takes before a Finder window updates 09:37
nine lizmat: sounds like a possible optimization once I've fixed the algorithm. Because it might turn out that we will recompile unnecessarily in certain cases just to be sure. 09:38
pmurias nine: would just md5ing things be a possible solution on systems with crappy timestamps? 09:58
nine pmurias: indeed an intriguing proposition. That would change the question from "is it up to date" to "is it the same". 10:04
pmurias nine: s/md5/sha1/ ;) 10:11
nine pmurias: yes, yes, yes, I can generate the sha when assembling the precomp file and store it on top 10:17
psch hrm, this is weird 10:25
i've added code to check for valid args for the setcodeobj calls we write
and there's no problem there
which is somewhat inconsistent with the error that pops up :P
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Make iteration of Lists about 10% faster 10:27
travis-ci.org/rakudo/rakudo/builds/133703776 github.com/rakudo/rakudo/compare/7...b3c9868963
psch i suppose that means i'll have to check where exactly that setcodeobj call comes from 10:33
nine psch: look for generated code, not the runtime
psch nine: i'm checking (and dying if there's a problem) during code gen 10:34
the bit commented with "Add code object fixups" in jvm/QAST/Compiler.nqp, to be precise
and if it's not going through there, i really don't know where it could go through 10:35
that's the only place we use setcodeobj in Compiler.nqp
well, except it's a setcodeobj call that gets written to QAST somewhere in e.g. World
...huh, that fits doesn't it 10:36
as in, the problem that we have is that is *doesn't* get a fixup
nine github.com/perl6/nqp/blob/master/s....nqp#L3382 10:38
that's definitely it
psch yeah, that's where my debug code sits
nine the code obj is there during compilation but no longer when loaded from the bytecode file 10:39
the equivalent code on moar doesn't run
but adding a check for is_nested will give you a NPE 10:40
because it's expecting more code objects than it finds
lunch now
dalek p: a855dd0 | (Pawel Murias)++ | src/vm/moar/QAST/QASTOperationsMAST.nqp:
Decontainerize the argument to setwho for consistency with the JVM backend.
 11:04
p: eb74bba | (Pawel Murias)++ | src/vm/js/nqp-runtime/ (2 files):
[js] Serialize/deserialize the HLL role.
p: 29186d6 | (Pawel Murias)++ | src/vm/js/ (2 files):
[js] Decontainerize the arguments to a whole bunch of ops.
p: 235f116 | (Pawel Murias)++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/Ops.java:
[jvm] Remove dead code.
nqp: 6f33259 | (Pawel Murias)++ | src/vm/jvm/ (2 files):
nqp: [jvm] Make nqp::reprname decontainerize it's argument.
[Tux] This is Rakudo version 2016.05-28-gbeb3c98 built on MoarVM version 2016.05 11:18
test 20.347
test-t 13.025
csv-parser 34.376
nine Oh the irony! nqp::sha1 can only hash strings, not buffers though SHA1 is clearly defined on a byte level rather than character data. 14:51
timotimo m( 15:23
yeah
nine I guess decoding the buffer from latin-1 should keep it unharmed 15:24
timotimo it'd still be nice if we would just let the sha1 op take bufs, too 15:39
also less overhead i suppose
you wouldn't want to build a hash cracker when you can only work with strings and they'd have to be decoded every single time
nine Meh...with a decode("latin-1") it's not \0 safe 16:52
lizmat I guess it's because nqp::sha predates Buf/Blob implementation 16:58
timotimo yeah, could very well be 17:40
Woodi nine: looks turning off atimes and dirtimes is common optimalization on SSD drives :) isn't atomic move enought ? also repo lock on maintenance work is not uncommon... 18:39
geekosaur that's not so much optimization as it is extending write lifetime 18:44
Zoffix Does anything bad happen if I delete my fork of rakudo? 18:47
I can't be arsed to figure out how to make it up to date with rakudo/rakudo...
geekosaur you lose any non-upstreamed changes you might have 18:48
Zoffix Ah, k. There aren't any.
geekosaur other than that, it seems to do the right thing (I had a trashed fork of another project, nuked it and re-forked, things became sane and nothing appears to have come unhinged/dangling) 18:49
psch alternatively you can add upstream as remote and do something like "git pull upstream && git reset --hard upstream/nom" i think
and then force push to your fork 18:50
actually, i think the pull isn't needed, and a fetch might be enough
'cause pull might want to merge, which makes everything annoying :)
Zoffix Normally git pull --rebase github.com/rakudo/rakudo; works, but there are conflicts this time. 18:51
How long is our deprecation cycle? 18:52
psch pre-christmas it was a half a year i think? then a clean break with christmas, and "it didn't come up yet" now afaik :)
Zoffix m: say from-json '[]' 18:53
camelia rakudo-moar beb3c9: OUTPUT«[]␤»
Zoffix Oh, I thought that was deprecated and giving warnings... never mind
psch m: sub f is DEPRECATED { }; f()
camelia ( no output )
Zoffix (it probably is, but isn't giving warnings) 18:54
Hm, it is: github.com/rakudo/rakudo/blob/3f9f...etty.pm#L7
psch m: say %*ENV.keys.grep(/:i dep/) 18:55
camelia rakudo-moar beb3c9: OUTPUT«(RAKUDO_NO_DEPRECATIONS)␤»
psch m: say %*ENV.grep(.key ~~ /:i dep/)
camelia rakudo-moar beb3c9: OUTPUT«Method 'key' not found for invocant of class 'Any'␤ in block <unit> at /tmp/EUyiVLcsPO line 1␤␤»
psch m: say %*ENV.pairs.grep(.key ~~ /:i dep/)
camelia rakudo-moar beb3c9: OUTPUT«Method 'key' not found for invocant of class 'Any'␤ in block <unit> at /tmp/qVtov65sRF line 1␤␤»
Zoffix Ah
psch m: say %*ENV.pairs.grep(*.key ~~ /:i dep/)
camelia rakudo-moar beb3c9: OUTPUT«WhateverCode object coerced to string (please use .gist or .perl to do that) in block <unit> at /tmp/fiG4AtUwgW line 1␤()␤»
psch grm
m: say %*ENV.grep({ $_ .key ~~ /:i dep/ })
camelia rakudo-moar beb3c9: OUTPUT«(RAKUDO_NO_DEPRECATIONS => 1)␤»
psch there :P
Zoffix m: BEGIN %*ENV<RAKUDO_NO_DEPRECATIONS> = 0; say from-json '[]' 18:56
camelia rakudo-moar beb3c9: OUTPUT«[]␤Saw 1 occurrence of deprecated code.␤================================================================================␤Sub from-json (from GLOBAL) seen at:␤ /tmp/e_bkH_UeV4, line 1␤Please use JSON::Fast, JSON::Tiny or JSON::Pretty from https:/…»
Zoffix Is there some way to find out what stuff is specifically deprecated and when it's supposed to be removed? 18:57
I mean other than grepping for 'deprecated' and looking up commit dates ;/
psch it used to say when stuff was gonna be removed i think 19:01
not sure how that worked, though
Zoffix :/ 19:02
geekosaur that was only done for stuff deprecated before 6.c, and they were all deprecated by release 19:13
I think deprecations that are part of spec are here until 6.d at least
Zoffix k 19:16
FWIW: I'm getting test failures with current rakudo checkout: gist.github.com/zoffixznet/7623d88...d9c4b0ebb9 19:17
geekosaur (actually I'd assume spec deprecations are valid now, marked deprecated in 6.d, removed in 6.e)
psch i think S17 is still suspectable to flapping
Zoffix: do they reliably fail at the same spot, even when run standalone with make or with perl6-m? 19:18
geekosaur S17-supply is certainly still flapping. the S17-lowlevel one I don't recall seeing on linux or os x
oh, I see, from-json. not spec, and no idea how it's being handled 19:19
Zoffix S17 segfaulted on the third try: gist.github.com/zoffixznet/ccf2b5d...c9818da084
psch yeah, i'd think that's the same, known, inconsistent behavior 19:20
unfortunately not enough people who understand moar async around vOv
Zoffix S17-supply/syntax.t just says Aborted on each run: gist.github.com/zoffixznet/3898ea1...4bbe2c9463
Oh... do I need to do some sort of special magic on them, to make them rakudo-specific? 19:21
psch if you ran them with 'make $file' or 'make spectest' there should be a $file.rakudo in the same folder 19:22
i think even .rakudo-moar
that one you can run with perl6-m
Zoffix And t/spec/S32-num/power.rakudo.moar seems to reliably fail the same: gist.github.com/zoffixznet/6e5b6aa...5ab7fb6bc9
I don't have .moar files for the other two tests, so I guess they don't have special stuff 19:23
psch Zoffix: well, considering that "fail" is passing a few TODOs and a few not that's ok :)
TODO tests are expected to not pass, and come up in the summary if they do pass 19:24
Zoffix Oh, ok.
bartolin those passing TODOs should probably fudged specifically for OS X (they pass for me on FreeBSD as well) 19:26
*be fudged
Zoffix So... any idea on how to fix any of this stuff? Or will it go into release anyway? 19:28
psch considering the S17 flappers have been around for probably a year or so... :/ 19:29
Zoffix :/
psch passing TODOs usually only pass platform-specific
well, it *is* moar-dependent async, feel free to dig in :)
Zoffix I'm not ready for that yet :P
bartolin speaking of fudged tests which actually pass: there is skiped block in integration/advent2014-day05.t which works on my machine. could someone else try and run that test file without the "skip"? 19:36
I'm a bit reluctant to change that in roast because it seemed to be a flapper, too
psch seems to hang here..? 19:41
psch waits another minute or so
yeah, still nothing happening 19:43
bartolin: 5 minutes now, still hanging. i'd leave that skipped :) 19:46
Zoffix PR to fix RT#128284 github.com/rakudo/rakudo/pull/777 19:47
rt.perl.org/Ticket/Display.html?id=128284
Oh, w00t! PR #777 :) If only I gave any credence to the existence of lucky numbers 😜 19:48
psch m: my $a = { say "foo"; say "bar" }; say $a.?name 19:50
camelia rakudo-moar beb3c9: OUTPUT«␤»
psch m: say Callable.^can('name')
camelia rakudo-moar beb3c9: OUTPUT«Too many positionals passed; expected 2 arguments but got 3␤ in block <unit> at /tmp/N0KzTkHJ0t line 1␤␤»
psch m: say Callable.can('name')
camelia rakudo-moar beb3c9: OUTPUT«()␤»
psch Zoffix: i think .? doesn't do what you think it does
Zoffix psch, it calls a method if it exists or returns a Nil .... doesn't it? 19:52
bartolin psch: thanks!
psch Zoffix: it does, sorry 19:53
Zoffix m: say "{$_.?name() || $_.^name()}" for sub{}, sub bar {}, *, -> {}, {;}, &[<], &sum
camelia rakudo-moar beb3c9: OUTPUT«Sub␤bar␤Whatever␤Block␤Block␤infix:«<»␤sum␤»
psch actually, i misinterpreted my result for the Block test above :)
Zoffix Ah
psch m: say Nil 19:54
camelia rakudo-moar beb3c9: OUTPUT«Nil␤»
Zoffix psch, it returns an empty string
psch m: say { Nil }()
camelia rakudo-moar beb3c9: OUTPUT«Nil␤»
psch Block.name does?
Zoffix m: my $a = { say "foo"; say "bar" }; say $a.name eqv ''
camelia rakudo-moar beb3c9: OUTPUT«True␤»
psch that is really confusing 19:55
m: say "" // "foo" 19:56
camelia rakudo-moar beb3c9: OUTPUT«␤»
psch well, never mind where that name method comes from vOv
otherwise it looks fine to me, even though i somewhat dislike || for "is this a non-empty Str" 19:57
or "is this 0", for that matter
but that's probably just me :)
psch isn't really programming-minded anymore, apparently 19:58
well, for today only hopefully :P
dalek kudo/nom: 9ff1af5 | (Zoffix Znet)++ | lib/Test.pm6:
Fix RT#128284: warnings on failing Test::cmp-ok

Instead of attempting to stringify the matcher, we try to call .name (that will give us name for named subs) or call ^.name, which will give us the name of the thing, such as Block, Whatever, etc.		 19:59
kudo/nom: 4a7eaa0 | peschwa++ | lib/Test.pm6:
Merge pull request #777 from zoffixznet/fix-RT128284

Fix RT#128284: warnings on failing Test::cmp-ok
Zoffix \o/ 20:00
psch now, considering that passing $op to &cmp-ok as Str is per design, feel free to solve that too ;) 20:01
geekosaur glwt 20:02
Zoffix Wouldn't doing so nullify my RFC to ditch that feature? :P 20:04
psch ...i suppose it would
Zoffix Plus, I don't know how to fix the security issue ¯\_(ツ)_/¯
psch well, it *is* EVAL...
on the other hand it is a test function 20:05
Zoffix m: EVAL "run 'ls'"
camelia rakudo-moar beb3c9: OUTPUT«run is disallowed in restricted setting␤ in sub restricted at src/RESTRICTED.setting line 1␤ in sub run at src/RESTRICTED.setting line 14␤ in block <unit> at EVAL_0 line 1␤ in block <unit> at /tmp/bf7rOdD5sz line 1␤␤»
Zoffix It's protected here...
psch m: "run 'ls'".EVAL.say
camelia rakudo-moar beb3c9: OUTPUT«run is disallowed in restricted setting␤ in sub restricted at src/RESTRICTED.setting line 1␤ in sub run at src/RESTRICTED.setting line 14␤ in block <unit> at EVAL_0 line 1␤ in block <unit> at /tmp/p5No4LGFRS line 1␤␤»
Zoffix m: use Test; cmp-ok '', '~~>;warn run "ls"; <z', '', '';
camelia rakudo-moar beb3c9: OUTPUT«Perlito␤dalek-queue␤evalbot␤evalbot.log␤foo␤lib␤log␤mbox␤nqp-js␤p1␤p2␤p6eval-token␤perl5␤rakudo-j-1␤rakudo-j-2␤rakudo-j-inst␤rakudo-j-inst-1␤rakudo-j-inst-2␤rakudo-m-1␤rakudo-m-2␤rakudo-m-inst␤rakudo-m-inst-1␤rak…»
psch oh hm 20:06
actually that's quite easy to fix in that specific case i think
Zoffix Well volunteered! :P
psch m: my $x = '<'; say ::("&infix:<$x>") 20:07
camelia rakudo-moar beb3c9: OUTPUT«No such symbol '&infix:<<>'␤ in block <unit> at /tmp/f7U8AXJNyy line 1␤␤Actually thrown at:␤ in block <unit> at /tmp/f7U8AXJNyy line 1␤␤»
psch m: my $x = '<'; say ::("&infix:[$x]")
camelia rakudo-moar beb3c9: OUTPUT«Use of uninitialized value <element> of type Any in string context␤Any of .^name, .perl, .gist, or .say can stringify undefined things, if needed. in any canonicalize_pair at src/Perl6/World.nqp line 4331␤Use of uninitialized value <element> of type A…»
psch m: my $x = '<'; say ::("&infix:<<$x>>")
camelia rakudo-moar beb3c9: OUTPUT«Use of uninitialized value $x of type Any in string context␤Any of .^name, .perl, .gist, or .say can stringify undefined things, if needed. in code at /tmp/v0XBTsvRYJ line 1␤No such symbol '&infix:<<<>>'␤ in block <unit> at /tmp/v0XBTsvRYJ line 1…»
psch well, maybe not :P
also kinda silly that the actual symbol isn't found, but it doesn't parse error
m: my $x = '=='; say ::("&infix:<$x>")
camelia rakudo-moar beb3c9: OUTPUT«sub infix:<==> (Mu $?, Mu $?) { #`(Sub+{<anon|52917568>}+{Precedence}|51437744) ... }␤»
Zoffix m: my $x = '<'; say ::("&infix:«$x»") 20:08
camelia rakudo-moar beb3c9: OUTPUT«Use of uninitialized value $x of type Any in string context␤Any of .^name, .perl, .gist, or .say can stringify undefined things, if needed. in code at /tmp/cBZsGiWNCQ line 1␤sub infix:«<» (Mu $?, Mu $?) { #`(Sub+{<anon|52917568>}|67963888) ... }␤»
Zoffix m: BEGIN my $x = '<'; say ::("&infix:«$x»")
camelia rakudo-moar beb3c9: OUTPUT«sub infix:«<» (Mu $?, Mu $?) { #`(Sub+{<anon|52917568>}|52845664) ... }␤»
psch oh, right, that again
Zoffix How come the == worked?
psch well, the ::() lookup itself is runtime
so i suspect that we really don't install &infix:<<>, because we couldn't even parse it anyway 20:09
Zoffix m: my $x = '<'; say ::("&infix:«$x»")
camelia rakudo-moar beb3c9: OUTPUT«Use of uninitialized value $x of type Any in string context␤Any of .^name, .perl, .gist, or .say can stringify undefined things, if needed. in code at /tmp/AD4OLKgg4Z line 1␤sub infix:«<» (Mu $?, Mu $?) { #`(Sub+{<anon|52917568>}|73985008) ... }␤»
Zoffix m: my $x = '=='; say ::("&infix:«$x»")
camelia rakudo-moar beb3c9: OUTPUT«Use of uninitialized value $x of type Any in string context␤Any of .^name, .perl, .gist, or .say can stringify undefined things, if needed. in code at /tmp/QbpwRhao_k line 1␤No such symbol '&infix:«==»'␤ in block <unit> at /tmp/QbpwRhao_k line 1…»
psch see, it warns, but finds it
Zoffix oh
wtf
m: my $x = '=='; say ::("&infix:<$x>")
camelia rakudo-moar beb3c9: OUTPUT«sub infix:<==> (Mu $?, Mu $?) { #`(Sub+{<anon|52917568>}+{Precedence}|53330048) ... }␤»
psch yeah, double interpolation
m: my $x = '=='; say ::("&infix:<<$x>>")
camelia rakudo-moar beb3c9: OUTPUT«Use of uninitialized value $x of type Any in string context␤Any of .^name, .perl, .gist, or .say can stringify undefined things, if needed. in code at /tmp/MhnVYdLjll line 1␤No such symbol '&infix:<<==>>'␤ in block <unit> at /tmp/MhnVYdLjll line 1…»
psch ...probably, something like that 20:10
m: my $op = '&infix:<<>'; say ::($op)
camelia rakudo-moar beb3c9: OUTPUT«No such symbol '&infix:<<>'␤ in block <unit> at /tmp/kqEHnTJzve line 1␤␤Actually thrown at:␤ in block <unit> at /tmp/kqEHnTJzve line 1␤␤»
psch m: my $op = '&infix:[<]'; say ::($op)
camelia rakudo-moar beb3c9: OUTPUT«No such symbol '&infix:[<]'␤ in block <unit> at /tmp/25Vr57bheU line 1␤␤Actually thrown at:␤ in block <unit> at /tmp/25Vr57bheU line 1␤␤»
psch m: my $op = q{&infix:['<']}; say ::($op)
camelia rakudo-moar beb3c9: OUTPUT«No such symbol '&infix:['<']'␤ in block <unit> at /tmp/8jkZPafY7u line 1␤␤Actually thrown at:␤ in block <unit> at /tmp/8jkZPafY7u line 1␤␤»
psch gives up vOv 20:11
Zoffix I don't get what you mean by double interpolation.... What difference is there between m: my $x = '=='; say ::("&infix:«$x»") and m: my $x = '=='; say ::("&infix:<$x>")
psch m: my $x = "foo"; say <$x>; say <<$x>> 20:12
camelia rakudo-moar 4a7eaa: OUTPUT«$x␤(foo)␤»
nine Woodi: we don't use atime, but we surely need mtime. Also I was not talking about locking the repo on maintenance but on read
psch i'm really just getting more confused right now, honestly
the "" interpolate the $x already :/ 20:13
bartolin psch, Zoffix: is this related: RT #127284
rt.perl.org/Ticket/Display.html?id=127284
psch bartolin: probably somehow, yeah 20:14
Zoffix m: sub &infix:«$op» { say $^a + $^b }; 2 $op 4
camelia rakudo-moar 4a7eaa: OUTPUT«===SORRY!=== Error while compiling /tmp/Ab3Fo4SH5q␤Missing block␤at /tmp/Ab3Fo4SH5q:1␤------> sub⏏ &infix:«$op» { say $^a + $^b }; 2 $op 4␤ expecting any of:␤ new name to be defined␤»
Zoffix m: sub infix:«$op» { say $^a + $^b }; 2 $op 4 20:15
camelia rakudo-moar 4a7eaa: OUTPUT«===SORRY!=== Error while compiling /tmp/lNbcZ1arYQ␤Colon pair value '$op' too complex to use in name␤at /tmp/lNbcZ1arYQ:1␤------> sub infix:«$op»⏏ { say $^a + $^b }; 2 $op 4␤ expecting any of:␤ postfix␤»
Zoffix m: sub infix:<$op> { say $^a + $^b }; 2 $op 4
camelia rakudo-moar 4a7eaa: OUTPUT«6␤»
Zoffix OK. For some reason I thought «» in this case was just a different way to write <>
m: sub infix:q{$op}| { say $^a + $^b }; 2 $op 4 20:16
camelia rakudo-moar 4a7eaa: OUTPUT«===SORRY!===␤Internal error: find_var_decl could not find $_␤»
psch *that* is neat
but yeah, the colonpair after routine_decl doesn't allow different quoting constructs 20:17
as in, it's really a colonpair, not a colon followed by a Str
Zoffix m: sub infix:<\<\<\<> { say $^a + $^b }; 2 <<< 4 20:18
camelia rakudo-moar 4a7eaa: OUTPUT«6␤»
b2gills m: my $op = '+'; say CORE::{"&infix:<$op>"}(3,4) 20:19
camelia rakudo-moar 4a7eaa: OUTPUT«7␤»
bartolin heh, RT #126646 is also fixed, Zoffix++ 20:21
Zoffix m: use MONKEY-SEE-NO-EVAL; my $op = '<'; $op.subst-mutate: '<', '\<'; my $matcher = EVAL "&infix:<$op>"; $matcher(3,4).say; $matcher(4,3).say
camelia rakudo-moar 4a7eaa: OUTPUT«True␤False␤»
Zoffix So that's the fix for the Str form of cmp-ok
Zoffix won't submit it, because of wanting Str form to be gone :P 20:22
Well, part of the fix. > is needed too 20:23
psch i wouldn't really consider it fixed while it still has EVAL, actually
Zoffix Is there even a way to avoid the EVAL, while keeping Str? 20:24
psch m: my $x = '<'; say quietly ::("&infix:«$x»")
camelia rakudo-moar 4a7eaa: OUTPUT«Use of uninitialized value $x of type Any in string context␤Any of .^name, .perl, .gist, or .say can stringify undefined things, if needed. in code at /tmp/rxyY2JVyLD line 1␤sub infix:«<» (Mu $?, Mu $?) { #`(Sub+{<anon|61760624>}|66583376) ... }␤»
psch hrm, that's not how that warnings goes away :/
but yeah, without that uninit warning that form would be right
'cause that's a runtime lookup, not an EVAL
m: my $x = '<'; say quietly ::('&infix:«$x»') 20:25
camelia rakudo-moar 4a7eaa: OUTPUT«No such symbol '&infix:«$x»'␤ in block <unit> at /tmp/SX20Q2JPSf line 1␤␤Actually thrown at:␤ in block <unit> at /tmp/SX20Q2JPSf line 1␤␤»
psch ...really weird
m: my $x = '<'; say quietly ::('&infix:«\qq[{$x}]»')
camelia rakudo-moar 4a7eaa: OUTPUT«sub infix:«<» (Mu $?, Mu $?) { #`(Sub+{<anon|61760624>}|64543632) ... }␤»
geekosaur ?
psch ah, there you go
m: my $x = '<'; say ::('&infix:«\qq[{$x}]»')
camelia rakudo-moar 4a7eaa: OUTPUT«sub infix:«<» (Mu $?, Mu $?) { #`(Sub+{<anon|61760624>}|70859600) ... }␤»
psch so, that form, with < > if the op contains « or » 20:26
Zoffix Ah, k
psch m: my $x = '<'; say ::('&infix:«\qq[$x]»')
camelia rakudo-moar 4a7eaa: OUTPUT«sub infix:«<» (Mu $?, Mu $?) { #`(Sub+{<anon|61760624>}|73292640) ... }␤»
psch well, *that* form :)
don't need {} inside qq for a scalar 20:27
Zoffix What if it contains both > and »?
psch m: my $x = '<'; say ::('&infix:[\'\qq[$x]\']')
camelia rakudo-moar 4a7eaa: OUTPUT«No such symbol '&infix:['<']'␤ in block <unit> at /tmp/MmAxrmkdTf line 1␤␤Actually thrown at:␤ in block <unit> at /tmp/MmAxrmkdTf line 1␤␤»
psch :/
then everything is broken /o\
m: say &infix:['<']
camelia rakudo-moar 4a7eaa: OUTPUT«sub infix:«<» (Mu $?, Mu $?) { #`(Sub+{<anon|61760624>}|68303488) ... }␤»
psch ah 20:28
Zoffix ?
psch m: say &infix:«==»
camelia rakudo-moar 4a7eaa: OUTPUT«sub infix:<==> (Mu $?, Mu $?) { #`(Sub+{<anon|61760624>}+{Precedence}|34816160) ... }␤»
psch m: say ::('&infix:«==»')
camelia rakudo-moar 4a7eaa: OUTPUT«No such symbol '&infix:«==»'␤ in block <unit> at /tmp/nwWRSu5Krx line 1␤␤Actually thrown at:␤ in block <unit> at /tmp/nwWRSu5Krx line 1␤␤»
psch Zoffix: well, we only *install* one form of the &infix ops for runtime lookup, but the grammar cheats when it parses it at compile time
Zoffix: note that «==» also doesn't nominally exist 20:29
m: say CORE::{&infix:«==»}
camelia rakudo-moar 4a7eaa: OUTPUT«Sub+{<anon|61760624>}+{Precedence} object coerced to string (please use .gist or .perl to do that) in block <unit> at /tmp/fpmM9ei7yf line 1␤Nil␤»
psch m: say &(CORE::{&infix:«==»})
camelia rakudo-moar 4a7eaa: OUTPUT«Sub+{<anon|61760624>}+{Precedence} object coerced to string (please use .gist or .perl to do that) in block <unit> at /tmp/qN0gdtVvPM line 1␤Nil␤»
psch anyway, yeah
b2gills say ::{'&infix:«<»'}; say ::{'&infix:<+>'} 20:30
m: say ::{'&infix:«<»'}; say ::{'&infix:<+>'}
camelia rakudo-moar 4a7eaa: OUTPUT«sub infix:«<» (Mu $?, Mu $?) { #`(Sub+{<anon|61760624>}|60447488) ... }␤sub infix:<+> (Mu $?, Mu $?) { #`(Sub+{<anon|61760624>}+{Precedence}|52712400) ... }␤»
psch m: say ::('&infix:«==»')
camelia rakudo-moar 4a7eaa: OUTPUT«No such symbol '&infix:«==»'␤ in block <unit> at /tmp/SX6n176EIX line 1␤␤Actually thrown at:␤ in block <unit> at /tmp/SX6n176EIX line 1␤␤»
psch ...i already did that
psch better heads bedwards o/ 20:31
b2gills For some reason if there is a < it has to be «» otherwise it has to be <>
psch b2gills: 'cause that's the literal lexical names
b2gills m: say ::('&infix:<==>')
camelia rakudo-moar 4a7eaa: OUTPUT«sub infix:<==> (Mu $?, Mu $?) { #`(Sub+{<anon|53470544>}+{Precedence}|42791312) ... }␤»
psch b2gills: the compile look-up cheats, but the runtime on doesn't
b2gills: hence why &infix:«==» prints &infix:<==> 20:32
..."hence why" reads like i should really go to bed :P
o/
b2gills m: say -> $op { ::{'&infix:<' ~ $op ~ '>'} // ::{'&infix:«' ~ $op ~ '»'} } for « + < » 20:34
camelia rakudo-moar 4a7eaa: OUTPUT«-> $op { #`(Block|55970128) ... }␤-> $op { #`(Block|55970200) ... }␤»
b2gills m: say -> $op { ::{'&infix:<' ~ $op ~ '>'} // ::{'&infix:«' ~ $op ~ '»'} }($_) for « + < » 20:35
camelia rakudo-moar 4a7eaa: OUTPUT«sub infix:<+> (Mu $?, Mu $?) { #`(Sub+{<anon|53470544>}+{Precedence}|47724080) ... }␤sub infix:«<» (Mu $?, Mu $?) { #`(Sub+{<anon|53470544>}|55459632) ... }␤»
travis-ci Rakudo build failed. peschwa 'Merge pull request #777 from zoffixznet/fix-RT128284 21:20
travis-ci.org/rakudo/rakudo/builds/133778427 github.com/rakudo/rakudo/compare/b...7eaa06b5e3

Powered by Cro and the Raku® Programming Language.

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

Thank you!