»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'p6: say 3;' or rakudo:, or /msg camelia p6: ... | irclog: irc.perl6.org or colabti.org/irclogger/irclogger_logs/perl6 | UTF-8 is our friend!
Set by moritz on 22 December 2015.
00:08 firstdayonthejob left 00:10 bjz left
garu timotimo: interesting, thanks! If I may bug you just a bit further, do you know why this was decided like so? I think it might have something to do with allowing signatures to spot Pair objects, but I'm not sure 00:20
00:21 kyclark_ joined
garu I'd love to hear more about the design process behind the language :) 00:21
grondilu there are the apocalypses 00:22
garu also, is there any way to pass named arguments with spaces in them? 00:23
grondilu: yeah, but I gather they have changed a lot over the years, haven't they?
00:24 garu left
timotimo sorry 00:24
grondilu WIP of a NativeCall interface to Binaryen: paste.debian.net/895092/ 00:27
It seems to be working but literals will likely be an issue.
timotimo i have no idea what bynaryen is 00:28
00:29 garu joined
grondilu wasm C API 00:29
timotimo web asm?
garu (sorry, got disconnected) 00:30
timotimo garu: named arguments are 100% static syntax, in addition there's hash flattening which makes things behave a whole lot different
garu timotimo: interesting! I think I'm bumping into this more often than I hoped for :/ 00:31
grondilu timotimo: webassembly
timotimo right
garu so if I want a method x to accept x(foo => 42, bar => 21) and x('foo' => 42, 'bar' => 21) and my %h = (foo => 42, bar => 21); x(%h), what should the signature be? 00:34
AlexDaniel garu: fwiw, this was a nice post: perl6advent.wordpress.com/2014/12/...ong-right/
garu (assuming I'd want to handle them all as a hash, say %foo) 00:35
timotimo you have to pass |%h 00:36
the signature would accept a foo and a bar named argument
if you want to accept the positional pair objects, too, just accept positionals 00:37
00:37 bjz joined
AlexDaniel m: sub x(+%z) { say %z.perl }; x(‘foo’ => 42, ‘bar’ => 42); my %h = (foo => 42, bar => 21); x(%h) 00:38
camelia rakudo-moar aa86b9: OUTPUT«[:foo(42), :bar(42)]␤[:bar(21), :foo(42)]␤»
AlexDaniel timotimo: ? 00:39
am I missing anything?
timotimo huh? +%z? that's news to me
cool.
AlexDaniel to be honest I don't really remember the difference between * and + 00:40
timotimo well, + gives you "one arg" semantics
that's probably what makes %h work
00:40 kyclark_ left
timotimo i mean passing %h 00:41
garu hmm
timotimo if it's just one argument, it'll be iterated over
garu AlexDaniel: the only difference between your example and mine is your signature was (+%z) and mine was (*%z)
AlexDaniel m: sub x(**%z) { say %z.perl }; x(‘foo’ => 42, ‘bar’ => 42); my %h = (foo => 42, bar => 21); x(%h)
camelia rakudo-moar aa86b9: OUTPUT«Too many positionals passed; expected 1 argument but got 2␤ in sub x at <tmp> line 1␤ in block <unit> at <tmp> line 1␤␤»
garu I definitely need to understand this better :/
00:42 sufrostico left
garu AlexDaniel++ 00:42
timotimo++
thanks!
AlexDaniel see docs.perl6.org/type/Signature#Slur...Parameters and docs.perl6.org/language/functions#...onventions
garu ok, while we're at it, I also have a question regarding return values :) 00:43
AlexDaniel commit: 6c sub x(*%z) { say %z.perl }; x(‘foo’ => 42, ‘bar’ => 42); my %h = (foo => 42, bar => 21); x(%h)
committable6 AlexDaniel, ¦«2015.12,2016.02,2016.03,2016.04,2016.05,2016.06,2016.07.1,2016.08.1,2016.09,2016.10,HEAD»: Too many positionals passed; expected 0 arguments but got 2␤ in sub x at /tmp/7tg9aqOrbv line 1␤ in block <unit> at /tmp/7tg9aqOrbv line 1␤ «exit code = 1»
AlexDaniel hmm, and just for fun
commit: all sub x(*%z) { say %z.perl }; x(‘foo’ => 42, ‘bar’ => 42); my %h = (foo => 42, bar => 21); x(%h)
garu sub x { my $a = 42, my @b = (3,2,1); return $a,@b }; my ($x,@y) = x(); $x.perl.say; @y.perl.say 00:44
committable6 AlexDaniel, gist.github.com/b28c38d0ea1ba2ffe2...3586925fcf
garu In the code above @y becomes [[3,2,1]], I wanted [3,2,1]
what am I doing wrong?
any clues? 00:46
AlexDaniel well, let me demonstrate
m: sub x { my $a = 42, my @b = 3,2,1; return $a,@b,42 }; my ($x,@y) = x; $x.perl.say; @y.perl.say 00:47
camelia rakudo-moar aa86b9: OUTPUT«42␤[[3, 2, 1], 42]␤»
grondilu m: use NativeCall; constant size_t := uint32;
camelia rakudo-moar aa86b9: OUTPUT«5===SORRY!5=== Error while compiling <tmp>␤Redeclaration of symbol 'size_t'␤at <tmp>:1␤------> 3se NativeCall; constant size_t := uint327⏏5;␤»
grondilu m: use NativeCall; print size_t;
camelia rakudo-moar aa86b9: OUTPUT«Use of uninitialized value of type NativeCall::Types::size_t in string context.␤Methods .^name, .perl, .gist, or .say can be used to stringify it to something meaningful.␤ in block <unit> at <tmp> line 1␤»
AlexDaniel garu: so one way to make it work is to use $y to hold your variable 00:48
to hold your array*
another way is to flatten it… 00:49
m: sub x { my $a = 42, my @b = 3,2,1; return $a,@b }; my ($x,@y) = x.flat; $x.perl.say; @y.perl.say
camelia rakudo-moar aa86b9: OUTPUT«42␤[3, 2, 1]␤»
AlexDaniel and the sane way to do it does not seem to work, hmmhmm… 00:50
garu AlexDaniel: buy interface-wise I kinda need to return an (Str, Array) I think
I could change things, sure, but I wonder how to make that work 00:51
AlexDaniel garu: return $a,@b already returns array as a second element
that's why @y ends up having the whole array inside its first argument
garu yeah but is there a way to magically flatten it so @y ends up with 3 arguments instead? 00:52
I tried return $a, @b.flat but it didn't exactly do what I wanted :X 00:53
then I learned .flat returns a Seq
AlexDaniel garu: well you can do $a,|@b but I don't recommend you to 00:54
garu why?
hey, it works!
urm... why don't you recommend that?
AlexDaniel well, first of all you said you want to return (Str, Array) and not (Str, array elements here, …) 00:55
but it depends on your needs. Personally I'd much rather let the user of sub x decide which behavior he wants 00:56
garu so you'd return an array as a single argument and get it as my ($x,$y) = x() 00:57
is that right?
AlexDaniel I think so, yes
garu ok, I can do that :)
thanks for the help!
AlexDaniel garu: personally, I find | very scary and I'm not even sure if it works well in this case 00:58
perhaps it works OK in this case, but I don't know…
garu (it did - I mean, it returned that flattened list and my test passed)
but I can try to work it like you said I think
AlexDaniel perhaps others can help :)
garu I want my Perl 6 code to look more idiomatic 00:59
but I find myself doing Perl 5 in Perl 6 :X
(and expecting the same results)
AlexDaniel OK, yes, I am right, | does not work here
garu: if you decide to go this way, then either use ｢return $a, slip @b｣ or ｢return flat($a, @b)｣ 01:00
garu m: sub x { my $a = 42, my @b = 3,2,1; return $a,|@b }; my ($x,@y) = x(); $x.perl.say; @y.perl.say
camelia rakudo-moar aa86b9: OUTPUT«42␤[3, 2, 1]␤»
garu ^^^ what's wrong with that?
AlexDaniel that's not the right way to test |
let me demonstrate…
m: sub x { my $a = 42, my @b = 3 xx 5; return $a,|@b }; my ($x,@y) = x(); $x.perl.say; @y.perl.say 01:01
camelia rakudo-moar aa86b9: OUTPUT«42␤[3, 3, 3, 3, 3]␤»
AlexDaniel m: sub x { my $a = 42, my @b = 3 xx 99999; return $a,|@b }; my ($x,@y) = x(); $x.perl.say; @y.perl.say
camelia rakudo-moar aa86b9: OUTPUT«Too many arguments in flattening array.␤ in sub x at <tmp> line 1␤ in block <unit> at <tmp> line 1␤␤»
AlexDaniel m: sub x { my $a = 42, my @b = 3 xx 99999; return $a, slip @b }; my ($x,@y) = x(); $x.perl.say; @y.perl.say
camelia rakudo-moar aa86b9: OUTPUT«42␤[3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3…»
garu ah, I was looking into slip earlier but couldn't get it right
AlexDaniel m: sub x { my $a = 42, my @b = 3 xx 99999; return flat($a, @b) }; my ($x,@y) = x(); $x.perl.say; @y.perl.say
camelia rakudo-moar aa86b9: OUTPUT«42␤[3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3…»
garu stil looks right... right? 01:02
AlexDaniel garu: look again
garu: the one with $a,|@b definitely does not look right
garu oops, sorry I was looking at the last example, didn't see it was using flat() 01:03
AlexDaniel my rule of thumb is “never use |”, there's always a better way
garu is it supposed to give the "too many arguments"?
(and if so, why?) 01:04
AlexDaniel garu: actually, use @b.Slip instead of slip @b 01:05
because seems like it works faster…
.tell lizmat fwiw, ｢slip @b｣ seems to work much slower than ｢@b.Slip｣. I have no idea, maybe that's how it should be, but perhaps you might want to take a look 01:07
yoleaux AlexDaniel: I'll pass your message to lizmat.
AlexDaniel .tell lizmat e.g. ｢my @alot = 3 xx 99999; say (slip @alot).perl｣ vs ｢my @alot = 3 xx 99999; say (@alot.Slip).perl｣ 01:08
yoleaux AlexDaniel: I'll pass your message to lizmat.
AlexDaniel garu: I don't know 01:09
01:09 aborazmeh joined, aborazmeh left, aborazmeh joined
garu that's ok. you've been super helpful, thank you! 01:10
AlexDaniel .tell TimToady I was expecting this to work: ｢my ($x, [@y]) = 24, [40, 50, 60]; say @y.perl｣, but right now it is completely bugged. What would be the right result in this case? See #130079 01:21
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=130079
yoleaux AlexDaniel: I'll pass your message to TimToady.
AlexDaniel garu: if you want, you may follow that ticket ↑
I have a gut feeling that it was supposed to work, but I have no idea 01:22
01:22 wamba left
AlexDaniel garu: in any case, “null object” should not appear there. So thanks to you one bug found :) 01:23
01:26 bjz left
MasterDuke m: grammar G { token TOP { "foo" \s* [ \s "bar" ]? } }; say G.parse("foo bar") 01:26
camelia rakudo-moar aa86b9: OUTPUT«Nil␤»
MasterDuke m: grammar G { regex TOP { "foo" \s* [ \s "bar" ]? } }; say G.parse("foo bar") 01:27
camelia rakudo-moar aa86b9: OUTPUT«Nil␤»
MasterDuke m: say $/ if "foo bar" ~~ / "foo" \s* [ \s "bar" ]? / 01:28
camelia rakudo-moar aa86b9: OUTPUT«｢foo ｣␤»
AlexDaniel MasterDuke: are you trying to say that it is a bug? :) 01:29
MasterDuke ^^^ am i crazy for thinking at least the last two of those should match the whole input string?
grondilu github.com/grondilu/binaryen-bindings
AlexDaniel m: grammar G { regex TOP { "foo" \s* [ \s "bar" ]? } }; say G.parse("foo bar")
camelia rakudo-moar aa86b9: OUTPUT«Nil␤»
AlexDaniel m: grammar G { regex TOP { "foo" \s* [ \s "bar" ]? $ } }; say G.parse("foo bar") 01:30
camelia rakudo-moar aa86b9: OUTPUT«｢foo bar｣␤»
AlexDaniel m: grammar G { regex TOP { "foo" \s* [ \s "bar" ] } }; say G.parse("foo bar")
camelia rakudo-moar aa86b9: OUTPUT«｢foo bar｣␤»
AlexDaniel MasterDuke: here is a golfed down version 01:31
m: grammar G { regex TOP { "a" || "abc" } }; say G.parse("abc")
camelia rakudo-moar aa86b9: OUTPUT«Nil␤»
AlexDaniel MasterDuke: so this is what happens
01:31 bjz joined
AlexDaniel MasterDuke: top tries to match "a", and it works, so it is happy 01:32
MasterDuke: but .parse for some reason is not, because it wants the whole thing to match… or so it seems?
MasterDuke: right here: github.com/rakudo/rakudo/blob/cb8f...mar.pm#L25 01:36
MasterDuke ah
AlexDaniel $match.to == target.chars ?? $match !! Nil 01:37
so it is a bug indeed…
dalek c: 7ff9021 | coke++ | doc/Type/X/Proc/Unsuccessful.pod6:
remove trailing whitespace
synopsebot6 Link: doc.perl6.org/type/X/Proc/Unsuccessful
MasterDuke m: grammar G { regex TOP { "a" | "abc" } }; say G.parse("abc")
camelia rakudo-moar aa86b9: OUTPUT«｢abc｣␤»
AlexDaniel no, not |
| does longest token matching, if I'm not mistaken 01:38
so it's "abc" || "a" in reality
m: grammar G { regex TOP { "abc" || "a" } }; say G.parse("abc")
camelia rakudo-moar aa86b9: OUTPUT«｢abc｣␤»
AlexDaniel m: grammar G { regex TOP { "a" || "abc" } }; say G.parse("abc")
camelia rakudo-moar aa86b9: OUTPUT«Nil␤»
AlexDaniel m: grammar G { regex TOP { "a" | "abc" } }; say G.parse("abc") 01:39
camelia rakudo-moar aa86b9: OUTPUT«｢abc｣␤»
AlexDaniel m: grammar G { regex TOP { "abc" | "a" } }; say G.parse("abc")
camelia rakudo-moar aa86b9: OUTPUT«｢abc｣␤»
AlexDaniel MasterDuke: will you submit a ticket?
01:41 travis-ci joined
travis-ci Doc build errored. Jan-Olof Hendig 'Fixed another broken link' 01:41
travis-ci.org/perl6/doc/builds/175359360 github.com/perl6/doc/compare/67e69...bc38acdf34
01:41 travis-ci left
MasterDuke well, i'm not exactly sure it is a bug. perl -E 'say "$1" if ("foo bar" =~ /foo \s* (\s bar)? /x)' prints nothing also 01:41
01:41 bjz left
MasterDuke "Method parse only succeeds if the cursor has arrived at the end of the target string when the match is over." 01:43
docs.perl6.org/type/Grammar#method_parse
AlexDaniel m: grammar G { regex TOP { "a" | "abc" } }; say G.subparse("abc") 01:45
camelia rakudo-moar aa86b9: OUTPUT«｢abc｣␤»
MasterDuke m: grammar G { token TOP { "foo" \s* || "foo" \s* \s "bar" } }; say G.parse("foo bar")
camelia rakudo-moar aa86b9: OUTPUT«Nil␤»
MasterDuke m: grammar G { token TOP { "foo" \s* | "foo" \s* \s "bar" } }; say G.parse("foo bar")
camelia rakudo-moar aa86b9: OUTPUT«Nil␤»
MasterDuke m: grammar G { token TOP { ["foo" \s*] | ["foo" \s+ "bar"] } }; say G.parse("foo bar") 01:47
camelia rakudo-moar aa86b9: OUTPUT«｢foo bar｣␤»
MasterDuke m: grammar G { token TOP { ["foo" \s*] || ["foo" \s+ "bar"] } }; say G.parse("foo bar")
camelia rakudo-moar aa86b9: OUTPUT«Nil␤»
AlexDaniel I still think that it is a bug
it could've arrived to the end, but it has no information in the rule itself that it should attempt to do so 01:48
so it bails out and then all of sudden .parse method is unhappy
.subparse is also wrong because we probably don't want it to match partially 01:49
MasterDuke i don't like it, but i'm not quite 100% convinced
AlexDaniel basically the solution is to always put $ in the end of your TOP
if you don't have $ in your TOP then it does not work as you expect…
I mean, sure, we can add another trap to our documentation 01:50
or we can just fix the damn thing, I guess…
I'll submit a ticket then
MasterDuke .ask TimToady do you mind looking at irclog.perlgeek.de/perl6/2016-11-13#i_13557942? is there something wrong or are we just expecting incorrect behavior?
yoleaux MasterDuke: I'll pass your message to TimToady.
01:51 araujo left
Xliff Well, now I need to find a testing suite for libxslt. 01:51
yoleaux 12 Nov 2016 14:44Z <raiph> Xliff: Aiui the only thing that makes a data block in P6 a data block is that it's a pod block called 'data'. So the 'foo' and 'qux' blocks don't count. I'll tighten the SO lang to try make that clearer.
Xliff .tell raiph Yes, but __DATA__ blocks aren't always "data" blocks in p5. They can be almost anything but are still referred to as "data blocks". I think we have the intent of the OP clear enough, just splitting hairs over names. 01:52
yoleaux Xliff: I'll pass your message to raiph.
Xliff LOL! Tests done in python! Close enough. 01:53
xmlsoft.org/XSLT/python.html 01:54
01:54 pyrimidine joined
MasterDuke AlexDaniel: hmm, the perl 5 version works if you add a $ anchor at the end 01:54
m: grammar G { token TOP {^ "foo" \s* [ \s "bar" ]? $} }; say G.parse("foo bar") 01:55
camelia rakudo-moar aa86b9: OUTPUT«Nil␤»
AlexDaniel MasterDuke: perl5 version works anyway 01:56
MasterDuke: perhaps you've not noticed because you didn't ｢use warnings;｣, e.g. perl -wE 01:57
MasterDuke what do you mean works? it doesn't capture ' bar' into $1 01:58
AlexDaniel MasterDuke: why would it?
MasterDuke: perl -wE 'say "ok" if "abc" =~ /b/x' # ok 01:59
it does not attempt to match from the beginning, and it doesn't attempt to match till the end
so once ｢foo \s*｣ matches there's no need to backtrack, the solution is just fine 02:00
MasterDuke the behavior i want is "capture the optional bit if it's there, but succeed anyway (without capturing anything obviously) if it isn't" 02:01
AlexDaniel well in this case you probably want [\s+ bar || \s*] 02:03
02:06 pyrimidine left
MasterDuke yup, that works 02:06
02:06 pyrimidine joined
MasterDuke i guess that's the problem with matching vs capturing. just to match it makes sense to bail out as soon as you know either way, but for capturing (and i guess i really mean parsing here) that's not always desired 02:09
02:10 araujo joined 02:11 pyrimidine left 02:15 pyrimidine joined 02:17 bjz_ joined 02:19 FROGGS_ joined, pyrimidine left 02:23 FROGGS left 02:34 pyrimidine joined 02:39 pyrimidine left 02:45 pyrimidine joined 02:46 ilbot3 left 02:47 ilbot3 joined 02:48 amalia_ left 02:49 amalia_ joined 02:53 araujo left 02:54 troys_ is now known as troys 02:55 troys is now known as troys_ 03:17 araujo joined
AlexDaniel .tell TimToady ↑ #130081 03:29
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=130081
yoleaux AlexDaniel: I'll pass your message to TimToady.
03:36 cyphase left 03:43 noganex joined 03:46 noganex_ left 03:52 kyclark_ joined 03:54 pyrimidine left, pyrimidine joined
garu AlexDaniel: are you still there? I'm found an "issue" with our previous signature solution 03:57
03:59 AlexDaniel left
garu AlexDaniel: sub x(+%z) gets x("foo"=>1) and %h=(foo=>1); x(%h), but not x(foo=>1) 04:00
also, I'm getting an array of pairs both times... how can I coerce it to a hash? 04:01
or is this one of those "I'm still thinking in Perl 5" moments? 04:02
04:02 kyclark_ left 04:05 kyclark_ joined, pyrimidine left, kyclark_ left 04:06 pyrimidine joined 04:11 pyrimidine left, cyphase joined 04:13 pyrimidine joined 04:14 skids joined 04:15 aborazmeh left 04:16 cyphase left 04:27 cyphase joined, pyrimidine left 04:36 cyphase left 04:41 aborazmeh joined, aborazmeh left, aborazmeh joined, cyphase joined 04:43 mr-foobar left 04:44 mr-foobar joined 04:48 zacts left 04:54 araujo left 04:56 araujo joined 04:58 araujo left, araujo joined 04:59 troys_ is now known as troys 05:00 bjz_ left, araujo left 05:02 pierre_ joined, araujo joined, araujo left, araujo joined 05:07 araujo left 05:17 bjz_ joined 05:18 pierre_ left 05:22 aborazmeh left, Cabanossi left 05:25 w4and0er96 left, timeless left, pierre_ joined, Spot__ left 05:26 Cabanossi joined 05:41 timeless joined, Spot__ joined, w4and0er96 joined 05:44 canopus left 05:45 bjz_ left 05:47 canopus joined 05:51 bjz joined 06:03 bjz left 06:04 troys left, zacts joined 06:12 bjz joined 06:18 zrr joined 06:24 bjz_ joined 06:25 bjz left 06:48 wamba joined 07:05 p6_nb joined 07:10 labster left 07:18 p6_nb left
grondilu On docs.perl6.org/language/nativecall#Structs it is written that native functions must pass and returns structs by reference. So how to deal with function that do it by value? 07:18
I tried to ignore the problem and got segmentation fault.
lizmat . 07:20
yoleaux 01:07Z <AlexDaniel> lizmat: fwiw, ｢slip @b｣ seems to work much slower than ｢@b.Slip｣. I have no idea, maybe that's how it should be, but perhaps you might want to take a look
01:08Z <AlexDaniel> lizmat: e.g. ｢my @alot = 3 xx 99999; say (slip @alot).perl｣ vs ｢my @alot = 3 xx 99999; say (@alot.Slip).perl｣
lizmat .tell AlexDaniel see github.com/rakudo/rakudo/commit/37d0e46614
yoleaux lizmat: I'll pass your message to AlexDaniel.
lizmat commute to T-Dose&
grondilu that is how would you translate struct foo { int n; }; struct foo increase(struct foo bar) { struct foo result = { bar.n + 1 }; return result; } 07:21
07:21 lizmat left, djbkd joined
geekosaur grondilu, I suspect you can't, sanely. struct passing and returning still has weird and platform specific corner cases in C. (it will work if you use the C compiler, but figuring out how to do it from outside is often an exercise in futility even when you can inspect said compiler's source) 07:23
this is a common shortcoming in foreign function interfaces and the workaround is almost invariably to defer it to C by making a wrapper that uses pointers and calls the real one 07:24
07:24 labster joined
geekosaur for example some platforms may quietly rewrite it to use pointers (e.g. stick the value on the stack, then a pointer to that stack value which is used as the real parameter; and for the return make extra space on the stack and either pass that as a hidden parameter or make sure said space is just above the return address and the function returning the struct has to offset everything else by that), while some others may do so only if the struct do 07:28
esn't fit into a machine register
07:30 rurban joined
geekosaur for a long time K&R C struct return values made a function non-reentrant (it used memory allocated next to the function itself, like pre-Algol activation records) 07:30
although ANSI disallowed that
hm, lemme rephrase that. K&R C permitted that, older pcc (what most old-timers think of as "cc") did it that way but newer ones switched to stack allocation to get rid of the "hidden" non-reentrancy 07:33
07:34 BenGoldberg left 07:40 darutoko joined 07:49 BenGoldberg joined 08:01 BenGoldberg left 08:06 rurban left, firstdayonthejob joined 08:07 bjz_ left
grondilu geekosaur: it so happens I was considering writing a C wrapper. I guess I'll do. 08:08
08:09 pierre_ left 08:10 kent\n left 08:13 kent\n joined 08:17 skids left 08:18 kurahaupo joined 08:29 djbkd left, djbkd joined 08:30 djbkd left, rurban joined, djbkd joined 08:31 djbkd left, djbkd joined, Actualeyes joined, djbkd left 08:32 djbkd joined, rurban left 08:37 djbkd left 08:38 kurahaupo left 08:42 canopus left 08:43 bjz joined, djbkd joined 08:48 canopus joined 08:50 rurban joined 08:57 rurban left 09:09 toolforger joined 09:15 rindolf joined
toolforger I just saw nqp/docs/jvminterop-goals.md mention an "NQP class loader". Does anybody know why NQP wants its own class loader? 09:15
09:16 rindolf left 09:21 rurban joined 09:23 rurban left 09:24 rurban joined 09:28 rindolf joined 09:38 RabidGravy joined
moritz toolforger: is this maybe about loading NQP classes from Java land? 09:43
10:06 labster left
toolforger No idea 10:10
you don't need that in Java land, unless you want to achieve something specific 10:11
such as the ability to unload classes
or having a different set of permissions 10:12
moritz you mean Java has native support for loading NQP classes? 10:13
toolforger it has native support for loading an arbitrary byte array as a class 10:15
10:15 rurban left
toolforger assuming that this specific byte array conforms to the rules for class bytecode 10:16
i.e. it must typecheck at the bytecode level, and a few other things
That's why NQP is using asm.jar, it has functions that deal with the nitty-gritty details of creating a bytecode array that the JVM will accept as a class 10:17
10:18 bjz_ joined, bjz left
toolforger You can also make a class loader if you want to load a class blob from a data source that's configured differently than what the standard classloader offers 10:19
though the standard classloaders in Java land are pretty versatile, they can load from .class files, from zip archives (.jar usually), and from URLs 10:20
Unfortunately, they cannot load from nested zips, that's one of the major reasons why people write classloaders
10:20 pierre_ joined 10:24 bjz joined, bjz_ left
toolforger NQP's class loader does not do anything of that kind though, all it does is define a wrapper function that prefills two parameters 10:25
no new functionality at all
10:25 cibs left
toolforger so no new data sources, the other things might still happen, that's done in the runtime, not in the classloader itself 10:26
toolforger goes to check where org.perl6.nqp.runtime.ByteClassLoader is used 10:27
10:27 cibs joined
[ptc] moritz: I've noticed some links to repos on modules.perl6.org give 404 page. The reason seems to be that the author has deleted their GitHub account. How does one handle such situations? 10:28
should the modules be removed from the modules list, or should the code be maintained somewhere else?
I've noticed another user had forked at least an earlier version of the code, so the code still exists in some form
*give a 404 page 10:29
10:31 espadrine joined 10:35 rurban joined, rurban left 10:36 seatek left
RabidGravy there's a community modules account for just such a purpose I think 10:37
El_Che RabidGravy: if the specific module has a license that allows that 10:38
RabidGravy if the module was already being used as a dependency then it probably does want to be preserved
toolforger Hm. NQP does nothing that requires a classloader, unless it's buried somewhere in callback code. 10:39
10:40 lizmat joined
toolforger Nope. None of the things you do with a classloader, not even there. 10:42
Whom should I ask about NQP-on-JVM specifics? I might still have missed something. 10:43
[ptc] RabidGravy: they all belong to the user "carbin", hard to tell if they are used as a dependency in something though 10:44
RabidGravy you could probably do something with the projects.json, actually I think I have a little script here that does the reverse dependencies 10:45
'ang on
[ptc] oh cool!
and here was me wishing for MetaCPAN for perl6, so I could get the reverse deps :-) 10:46
the list I have is: p6-unix-privileges, p6-jdf, p6-fcgi and p6-fcgi-psgi 10:47
RabidGravy a cursory scan of the depended on things would suggest nothing depends on those 10:50
bartolin toolforger: maybe psch or arnsholt. I seem to remember that you had a discussion with them about some jvm/interop stuff before 10:54
toolforger yep
10:54 djbkd left 10:55 djbkd joined
[ptc] RabidGravy: so should I submit a patch to the modules list to remove these modules? 10:56
[ptc] hates seeing code go to waste though...
10:57 djbkd_ joined 11:00 pmurias joined, djbkd left
RabidGravy just checking harder 11:00
pmurias toolforger: jnthn is the most familiar with the JVM backend
toolforger thx 11:01
What's the best way to contact him? I see him in chat but he hasn't noticed it seems. 11:02
RabidGravy [ptc], there's something weird going on there though 11:06
toolforger Oh, I also see Arne Skjærholt in the git log. That's arnsholt here, right?
right, that's his mail address
RabidGravy the github user 'carbin' was only created on thursday
[ptc] RabidGravy: that's weird. 11:07
RabidGravy I'll poke a bit more in a bit, 11:08
[ptc] when I searched for one of the modules I found e.g. this: github.com/nbrown/p6-jdf
11:08 pierre_ left
[ptc] where the user was actually "Carlin" 11:08
what if the new "carbin" user is actually a new user? And has taken the name after the old one deleted their account? Clutching at straws, I know, but... 11:09
11:13 djbkd_ left 11:14 djbkd joined 11:18 djbkd left
pmurias toolforger: usually jnthn is much more active during the week 11:19
toolforger good to know
even if it's unfortunate
pmurias toolforger: what would using the builtin class loaders directly give us? 11:31
toolforger Less class confusion
Classes are not necessarily unique across classloaders
e.g. if you have classloaders A and B, you could have MyAwesomeClass loaded twice 11:32
this becomes nasty if objects cross classloader barriers
e.g. in callbacks
or generally in APIs 11:33
e.g. MyApi.doSomething(MyClass m) lives in classloader A and gets handed a MyClass object that lives in classloader B 11:34
You will get a ClassCastException telling you that org.whatever.MyClass is not compatible with org.whatever.MyClass 11:35
because the error message does not tell you the classloaders, because classloaders often do not have a human-readable name
pmurias toolforger: looking at the JVM backend it seems to have two custom class loader classes 11:36
toolforger google for "classloader hell" for more stories about this
I saw only one, ByteClassLoader
I seem to have missed the other one - which is it?
pmurias IgnoreNameClassLoader
which seems to load serialized objects from a different source 11:37
toolforger: isn't ByteClassLoader just a convinience wrapper as it just delegates? 11:38
toolforger yes
I'd recommend making it a function on some other class, possibly even a static function 11:39
this also avoids instantiating your own classloaders
pmurias toolforger: the ByteClassLoader itself doesn't seem to load anything 11:42
toolforger Indeed it doesn't
pmurias it only seems to be a subclass inorder to gain access to a protected method 11:43
toolforger though you'd have to check where it's instantiated and how the instances are used to see whether it's used for access control etc.
Oh right, the called defineClass is protected 11:44
11:44 cibs left
RabidGravy [ptc] actually something like that is highly probably 11:44
raw.githubusercontent.com/nbrown/p.../META.info 11:45
pmurias toolforger: it's always created like new ByteClassLoader(getClass().getClassLoader())
toolforger OK, then all instances of ByteClassLoader are chained to the class loader of the calling code 11:46
For access control etc., one would have to check how these ByteClassLoader instances are used
RabidGravy github.com/nbrown/p6-jdf
11:46 cibs joined
pmurias IgnoreNameClassLoader seems like it always loads Perl 6 classes only, so it shouldn't cause mixup with Java classes 11:46
toolforger: access control? 11:47
toolforger sandboxing
that's what application servers and browser plugins need
nqp shouldn't need that 11:48
but you never know
Hmm... I see that java.lang.ClassLoader itself does not offer a way to turn a byte array into a class 11:49
pmurias I'm not a Java programmer so I'm not familiary with all the JVMy stuff ;) 11:50
11:50 bjz_ joined, bjz left
toolforger I don't know how asm injects classes into the parent class loader, I think it should do that 11:50
:-)
I know a bit more, but I try to avoid ClassLoader trickery 11:51
because I know enough to know it's dangerous, but not enough to deal with the dangers
RabidGravy [ptc], so yeah at a guess I'd say they renamed their github account and forgot to update the modules list and the meta files 11:52
I'll fix the meta list and PR the changes to the META files 11:53
toolforger Hmm... IgnoreNameClassLoader looks whacky to me 11:54
each newly created class has its own IgnoreNameClassLoader
dalek osystem: 1a11045 | RabidGravy++ | META.list:
Change 'carbin' to 'nbrown'
 11:56
toolforger Issue is that each Class has a reference to the ClassLoader that creates it, so the IgnoreNameClassLoader objects will stick around until the class goes away, which it doesn't until you unload the ClassLoader
RabidGravy right that's part of the problem fixed
toolforger I'll have to look up how this kind of stuff is supposed to be done, proliferating ClassLoaders just to get access to a defineClass() function without having to bother with names is going to create problems, especially once you start to seriously do interop with Java code 11:58
[ptc] RabidGravy: dunno if crabin is nbrown though... 11:59
I think nbrown just happened to have a fork of the repo; they have many forks of perl6-related repos 12:00
toolforger anyway - thx, gtg, rl interrupt, will be back after reading up things on the ASM site - the ASM guys should be pretty knowledgeable about how the bytecode their lib generates should be injected into a JVM
12:00 toolforger left
RabidGravy No they're not forks 12:00
and if our assumption is wrong at least the modules will be installable again :) 12:02
12:03 canopus left 12:08 canopus joined
[ptc] true :-) 12:10
RabidGravy I've left a note on the PRs to get on here and discuss if that isn't the right thing, but all done now :) 12:12
12:18 rurban joined
[ptc] RabidGravy: did a bit more digging and found that nbrown was active on this channel in 2009, and that he joined GitHub in the same year. 12:26
other than that, he'd forked almost all perl6-related repos, and that's my guess how the "carbin" repos got there...
RabidGravy and if the "carbin" did delete their account then it might explain why they don't appear to be forks 12:28
[ptc] oh well, the modules are installable again :-) 12:29
RabidGravy yeah and if nbrown doesn't want to have anything to do with them we can arrange to have them moved elsewhere in an orderly fashion 12:31
[ptc] sounds good 12:41
RabidGravy now let's see about this "modules being in a weird order" thing 12:44
12:54 vendethiel left, rurban left, vendethiel joined 12:59 rurban joined
RabidGravy which I instantly regret due to the immense dependencies of the modules.perl.org code 13:04
ho hum
13:11 TimToady left 13:13 TimToady joined
masak read it as "modules being the World Order" 13:16
13:22 lizmat left
pmurias really dislikes how modules.perl6.org looks 13:26
13:32 lizmat joined
RabidGravy I've got a thing somewhere that loads up the modules list in CouchDB for a bit of munging goodness 13:45
13:46 rurban1 joined
masak pmurias: I don't disagree, but could you be more speicifc? 13:46
13:48 rurban left 13:50 rurban1 left
RabidGravy The projects.json is there to consume for all, it's make something that pleases you :) 13:56
pmurias masak: I don't have a good idea how it should look as I would have fixed it already 13:57
RabidGravy maybe I should have limited the number of modules the build project list retrieved this could take all afternoon 13:58
13:58 bjz_ left
moritz RabidGravy: do you want a dump of module.perl6.org sqlite3 database? 14:00
RabidGravy it'll be fine, it's about half way through now 14:01
masak pmurias: well, it has to have *some* visual appearance... :P I was mostly asking in case you were sitting on a concrete suggestion for improvement, but seems not 14:05
maybe one way to ask is "should it look more like metacpan.org/ ?"
14:06 kurahaupo joined
masak I mean, it doesn't actually have to be a list 14:06
in fact, it can only remain a single list for as long as we see ourselves as not having all that many modules
moritz as long as a browser can comfortable keep the whole list, I'm all for having it right there
14:07 rurban joined
RabidGravy If more people used "tags" it would be nice to have a tag cloud kind of thing 14:08
moritz the problem with tags is always: what do you use them for?
and can you make usage somewhat consistent?
RabidGravy well that's for individual module authors and/or end-users to decide :) 14:10
the downside of tags is exactly that people will ask the questions you did :) 14:11
moritz and I'm hesitant to implement a system where I'm not confident I can tell people how to use or, or where I'm confident they'll use it in a way that makes sense
masak what moritz said -- it's a rare thing that "tag cloud" comes in and magically makes things better or more usable in some way 14:12
usually it takes quite a bit of work, tweaking, and experimenting. I'm not sure we should give module authors that extra work without seeing clear benefits from it. 14:13
pmurias masak: using a fork of metacpan.org could be the right solution
stmuk what problem are we trying to solve here? 14:14
masak * pmurias really dislikes how modules.perl6.org looks 14:15
stmuk: ^
stmuk what does he mean by "looks"? web design or something else?
pmurias stmuk: ugliness 14:16
it used to be uglier before I remove the kwalitee-or-whatever-it-was-called 14:17
RabidGravy well regarding the "fork metacpan" isn't there a move afoot to integrate with the actual metacpan?
stmuk the ecosystem doesn't remotely resemble cpan
pmurias RabidGravy: there is some work in a branch of fork on github 14:19
stmuk forking metacpan.org wouldn't buy you much
you could just stick a search box up using elastic search or whatever 14:20
pmurias there are plans to allow perl 6 releases on cpan 14:23
stmuk there are experimental versions already on cpan 14:24
pmurias stmuk: just stealing the metacpan design might be one solution 14:25
14:26 rurban1 joined, rurban left
pmurias hopefully I'll talk it over with a friend who does both programming and art and he'll tell me how modules.perl6.org should look like ;) 14:26
stmuk actually Isn't there already a forked metacpan running on hack? 14:27
RabidGravy I'm relaxed about the appearance but nice features metacpan has include the in-site browsing, dependency look-up, testing info etc 14:30
stmuk hack.p6c.org:5001 14:31
14:34 rurban1 left
moritz Hosting generously provided by: The Perl 6 Community! 14:35
14:35 rurban joined
stmuk the CPAN-ecosystem experiment seemed to show quite a mismatch last time I looked 14:36
14:37 nwc10 joined
nwc10 I'm reading docs.perl6.org/language/phasers#INIT and I'm confused by what "When phasers are in different modules, the INIT and END phasers are treated as if declared at use time in the using module." means 14:37
or is it as simple as I think - it's the same as Perl 5 14:38
use Foo; where Foo.pm has `INIT {warn} END {warn}`
is just the same as
stmuk stealing design and ideas is good .. but direct code reuse probably isn't possible
moritz nwc10: it means that they are run at the compile time of the outer module
nwc10 `BEGIN {INIT {warn} END {warn}}` in the outer module?
which I think is what you just typed 14:39
14:39 wamba left
moritz yes 14:39
nwc10 good thanks.
stmuk how would metacpan deal with modules with the same name Foo::Bar which are different? :)
moritz the nesting of compile and run time are dificult enough to understand and explain; mixing in more phases doesn't make it easier :/ 14:40
nwc10 yes. agree.
pmurias stmuk: the have to have distinct fully qualified named 14:42
nwc10 I guess the reason this mattered enough here to need that note in the docs, is because INIT and END have (er) FIFO and LIFO semantics, so the ordering (relative to other INITs and ENDs) needs to be clear)
pmurias stmuk: names
nwc10 er, free ')'
stmuk pmurias: yep! 14:43
14:44 nwc10 left
stmuk yay nom seems to have enough windows fixes over 2016.10 for an MSI 14:47
so there will probably be a R* 2016.11 so the windows users can install a binary
14:48 domidumont joined 14:51 kurahaupo__ joined 15:05 rurban left
lizmat stmuk++ 15:13
15:16 rurban joined 15:18 raiph joined 15:20 canopus left
RabidGravy I really wish I understood this jquery DataTable plugin to figure out why the weird ordering 15:22
ah, okay it's the dataType thing 15:24
timotimo recently the page has been very unresponsive to searching ... at least once 15:25
oh 15:26
perhaps it was because i searched for "http" and everything matched by way of their urls ;_;
15:27 canopus joined 15:30 wamba joined 15:31 rurban left 15:42 AlexDaniel joined 15:45 qub joined, qub left 15:46 cdg joined
RabidGravy timotimo, I think this is entirely related to what has gone wrong with the sorting 15:55
15:55 rurban joined 15:57 rurban1 joined
timotimo mhm 15:59
16:00 rurban left, araujo joined 16:01 lizmat left
RabidGravy yeah it appears that it is taking the whole of the first TD into consideration for searching and sorting - including the tags, href and everything 16:04
16:06 cdg left
timotimo so if you search for "a" or "href", that'll go in, too? 16:06
[ptc] just tried typing "href" into the search field; didn't seem to drop the number of items at all 16:17
doing a search for e.g. 'wig' turned up the entry one would expect
timotimo i wonder how difficult it'd be to nativecall into fortran libraries? 16:21
RabidGravy [ptc], I'm just trying to get my head round what it thinks it is doing 16:23
DrForr Doen't FORTRAN use column-major ordering, among other things? 16:24
timotimo yes
16:24 rurban1 left
DrForr Not saying it's impossible by any stretch of the imagination, just thinking about what would have to be changed. 16:26
RabidGravy right, I've got the search right but I don't appear to have cracked the ordering 16:28
16:40 rurban joined
pmurias what's the difference between throwpayloadlex and throwpayloadlexcaller? 16:41
timotimo one skips a stack frame 16:42
so pretends to be in the caller's frame
at least that's what i think it does
the kind of thing you'd implement "die" with
RabidGravy there, that's fixed up
though rather fugly but I'm no client side developer 16:43
16:45 ale1ster joined 16:50 Khisanth left 16:54 rurban1 joined
dalek href="https://modules.perl6.org:">modules.perl6.org: c288107 | RabidGravy++ | assets/js/dataTables.filter.perlModule.js:
Strip the html from the text, this improves search
 16:56
href="https://modules.perl6.org:">modules.perl6.org: 09d7f7a | RabidGravy++ | assets/js/main.js:
Strip to text for order data type

Fixes #65
16:56 rurban left, zakharyas joined
RabidGravy of course if anyone else better understands the datatable plugin please intervene 16:57
16:59 rurban joined, rurban1 left 17:01 rurban1 joined 17:03 Khisanth joined, rurban left 17:04 rurban joined 17:06 rurban1 left 17:08 rurban1 joined 17:09 rurban left 17:11 rurban joined 17:12 ale1ster left, rurban1 left 17:13 Guest11976 left 17:14 rurban1 joined 17:16 rurban left 17:17 rurban joined 17:19 rurban1 left, mxco86 left 17:21 rurban1 joined 17:22 rurban left
RabidGravy if twingling the modules.perl6.org is necessary to pick up the changes to the javascript then someone might need to do that :) 17:23
stmuk RabidGravy++ # full stack 17:30
RabidGravy :-\ well in the case of client side stuff it's more tinker with stuff until it works 17:31
timotimo i know that feeling 17:32
RabidGravy there's something not quite right about that datatable plugin though 17:33
stmuk hows that really different to any programming? :)
RabidGravy if you set a type on a column, even though it's implementation only effects the searching it impacts the sorting, so you have to set the orderTableType as well 17:34
17:37 skids joined 17:44 rurban joined, rurban left 17:47 rurban1 left 17:48 rurban joined
RabidGravy timotimo, the searching and sorting should be fine on the modules.perl6.org/ now it appears to have reloaded 17:50
timotimo cool :)
thank you for your work
[ptc] RabidGravy++ 17:51
just tried it and it works well
RabidGravy: you're a full-stack developer now :-P
timotimo stacks some more stuff on top of RabidGravy
moritz now RabidGravy is a stack overflow developer :-) 17:52
[ptc] aren't we all? 17:55
stmuk ttp://www.commitstrip.com/en/2016/11/07/which-full-stack-developer-are-you/ 17:57
+h
I'm the first sort :P
TimToady .tell garu if you want to return multiple values from a function, you should really use binding, which will work argument by argument, rather than list assignment, which will flatten like Perl 5's list assignment 17:58
yoleaux 01:21Z <AlexDaniel> TimToady: I was expecting this to work: ｢my ($x, [@y]) = 24, [40, 50, 60]; say @y.perl｣, but right now it is completely bugged. What would be the right result in this case? See #130079
TimToady: I'll pass your message to garu.
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=130079
yoleaux 01:50Z <MasterDuke> TimToady: do you mind looking at irclog.perlgeek.de/perl6/2016-11-13#i_13557942? is there something wrong or are we just expecting incorrect behavior?
03:29Z <AlexDaniel> TimToady: ↑ #130081
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=130081
TimToady AlexDaniel: same to you, use binding there, not list assignment 17:59
18:00 autark left
AlexDaniel TimToady: example? 18:01
yoleaux 07:20Z <lizmat> AlexDaniel: see github.com/rakudo/rakudo/commit/37d0e46614
18:01 girafe joined
AlexDaniel ah 18:02
got it
TimToady .tell MasterDuke TOP is not going to magically start backtracking merely because .parse fails, especially since backtracking probably means you have a smelly grammar, which is why tokens and rules don't backtrack by default in the first place; so use regex and explicit $ if you really need to backtrack 18:03
yoleaux TimToady: I'll pass your message to MasterDuke.
AlexDaniel TimToady: so what should be the result if you attempt to use list assignment like this?
should it be a syntax error, a run time error of some sort, or what is it?
TimToady it should be like P5's list assignment, to the extent possible, but assigning to [@y] should probably fail since that's not sensical for either assigning or binding a flat array 18:05
that would only bind an array contained as a single elem of another array
and using [@y] makes no sense under a flattening regime like p5 list assignment 18:06
just bare @y would slurp the rest of the args
AlexDaniel TimToady: well, I can barely follow any of this, so feel free to leave a comment on #130079
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=130079
TimToady alas, don't really have good internet here, except a few minutes at a time... 18:07
this hotel is not quite so enlightened as our previous one 18:08
18:08 araujo left
TimToady and this coming week we may not have much internet at all, on a cruise 18:08
18:20 domidumont left 18:24 domidumont joined
grondilu geekosaur: would you have a few minutes to help me make the wrapper you were talking about earlier? I wrote it but and it compiles but I can't link it properly. 18:31
geekosaur sure
grondilu hang on
the wrapper I wrote is: paste.debian.net/895284/
I compiled it with gcc -lbinaryen -I../binaryen/src -c -Wall -Werror -fpic binaryen-constants.c 18:32
I made the shared library like this: gcc -shared -o libbinaryen-constants.so binaryen-constants.o 18:33
but then when trying to call it with NativeCall I get:
Cannot locate native library '/usr/local/lib/libbinaryen-constants.so': /usr/local/lib/libbinaryen-constants.so: undefined symbol: BinaryenConst
yet BinaryenConst is defined in binaryen-c 18:34
(FYI binaryen is at github.com/WebAssembly/binaryen)
geekosaur putting -lbinaryen at the beginning instead of the end is at minimum a smell and potentially the problem 18:36
grondilu ok
geekosaur linkers tend to search libraries only when they have undefined symbols, so a library at the beginning may be ignored because it doesnt have any symbols to look for yet 18:37
grondilu same error
geekosaur -l is not really an option like -c or -o, it's a macro of sorts that expands to the path to a libraru
grondilu also frankly I thought I was supposed to put it in the linking line, not the compilation one 18:38
but I've seen it done as such somewhere so *
geekosaur the linking line is usually better. shared object handling is somewhat odd, it may well work as part of the compile line, then again it might not 18:40
you could try moving it
oh, creating a .o, probably not going to be used because linker not invoked, so yes
grondilu full error message: paste.debian.net/895287/
geekosaur you can also use ldd on the .so to see if it has the reference to libbinaryen.so.whatever 18:41
grondilu it doesn't apparently 18:42
18:43 labster joined, labster left 18:44 labster joined
grondilu I've tried adding ../binaryen/CMakeFiles/binaryen.dir/src/binaryen-c.cpp.o on the linking line but I still don't see it in ldd's output 18:44
geekosaur that is static, it won't show in ldd
grondilu ok 18:45
geekosaur again, try moving that -lbinaryen to the second gcc instead of the first
[ptc] anyone got an idea what this error message is trying to tell me? Cannot find method 'export-package' on object of type CompUnit
18:45 labster left
geekosaur (also it would likely fail because other objects would be needed) 18:45
[ptc] a `git grep` on the source doesn't turn up any `export-package` routine...
grondilu I got /usr/bin/ld: cannot find -lbinaryen 18:46
collect2: error: ld returned 1 exit status
geekosaur [ptc], something is trying to use `is export` on a CompUnit for some reason
[ptc] geekosaur: oh, ok. Thanks :-) Now I can dig further
MasterDuke src/core/CompUnit/Handle.pm:36: method export-package() returns Stash {
yoleaux 18:03Z <TimToady> MasterDuke: TOP is not going to magically start backtracking merely because .parse fails, especially since backtracking probably means you have a smelly grammar, which is why tokens and rules don't backtrack by default in the first place; so use regex and explicit $ if you really need to backtrack
geekosaur grondilu, you didn;t actually install binaryen? you will need -L and -R options in the link line to point to wherever the libbinaryen.so.* is
18:47 espadrine left
grondilu geekosaur: I did put the libbinaryen.so file in /usr/local/lib and it works fine with NativeCall 18:47
geekosaur then ld is not using -L/usr/local/lib -R/usr/local/lib by default
(you could check /etc/ld.so.conf to see if /usr/local/lib is in there; if so then you don't need the -R, but you might still need the -L) 18:48
grondilu gcc: error: unrecognized command line option '-R'; did you mean '-R'?
geekosaur ??? 18:49
oh
grondilu IKR
geekosaur wait, is my client changing that?
grondilu nope, it was sic
18:50 labster joined
grondilu $ grep local /etc/ld.so.conf.d/* 18:50
/etc/ld.so.conf.d/libc.conf:/usr/local/lib
geekosaur ok, you don't need the -R then
(and I was wondering if my client had changed a normal hyphen to a Unicode smart hyphen of some kind, but I just checked that and it didn't) 18:51
(also come to think of it that would not cause *that* error, it would somehow have to be the 'R' not the '-')
grondilu it still can find the lib
s/can/can not/
geekosaur which part can't find it now? 18:52
MasterDuke .tell TimToady i think i figured out that i needed a regex, but i'm having some problem implementing it with some further discussion here, does this look like a bug to you? irclog.perlgeek.de/perl6-dev/2016-...i_13559207
yoleaux MasterDuke: I'll pass your message to TimToady.
grondilu $ LANG=C gcc -L/usr/local/lib -shared -l binaryen -o libbinaryen-constants.so binaryen-constants.o 18:53
/usr/bin/ld: cannot find -lbinaryen
collect2: error: ld returned 1 exit status
oh wait
geekosaur aside from the fact that "-l binaryen" should not have a space and should be at the very end --- the error suggests it figured that out (which surprises me) but still can't find it. unless -l with a space does something weird 18:54
grondilu I'm so dumb
geekosaur [ptc], MasterDuke's thing suggests you may in fact have an internal error where something that expects a CompUnit::Handle got a CompUnit instead 18:55
grondilu geekosaur: I'm so sorry. I really thought I add put it in /usr/local/lib. Turns out I didn't.
geekosaur you may also have to put the one with the version suffix there, if it has one (and if it doesn't, it's a bad package and will bite someone someday... but that's not your problem) 18:56
grondilu well, I still get the error anyway
geekosaur ls -lL /usr/local/bin/libbinaryen.so
er
ls -lL /usr/local/lib/libbinaryen.so
wait no, sigh
ls -l /usr/local/lib/libbinaryen.so
grondilu -rwxr-xr-x 2 grondilu staff 2349192 nov. 12 20:00 /usr/local/lib/libbinaryen.so
geekosaur ok, then I don't know why it's still not finding it 18:57
grondilu the error I was talking about was the NativeCall one, not the gcc.
now -lbinaryen works
MasterDuke does anyone here have any experience parsing the json output of rakudo profiles? i profiled building rakudo, but it creates a 830mb file, which the two usual viewers can't handle 18:58
[ptc] geekosaur: could be. I was digging in the `v5` repository and trying to work out if the test suite worked (it doesn't)
geekosaur ok, use ldd on your .so to make sure it has the link to libbinaryen.so and that it can resolve it
[ptc] geekosaur: and once one gets the PERL6LIB right, that's the error I was getting
geekosaur MasterDuke, I think people have been slamming into that one for over a year now >.> 18:59
every so often timotimo grumps about it 19:00
grondilu no link to libbinaryen.so shown by ldd
$ ldd /usr/local/lib/libbinaryen-constants.so
linux-gate.so.1 (0xb7754000)
libc.so.6 => /lib/i386-linux-gnu/libc.so.6 (0xb757f000)
/lib/ld-linux.so.2 (0x800ae000)
geekosaur then it's still not actually linking it 19:01
what does that link command currently look like?
grondilu $ LANG=C gcc -shared -o libbina
ryen-constants.so $(find ../binaryen -name 'binaryen-c.cpp.o') binaryen-constant
s.o -lbinaryen
geekosaur you broke it...
grondilu wyeah
geekosaur please do not link against the .o
grondilu bad paste
MasterDuke geekosaur: yeah, i looked into modifying the qt viewer, but don't know (or like) c++ enough to change its parsing library 19:02
geekosaur you would need to link EVERY .o and not use -lbinaryen
do one or the other, not both
grondilu ok
finally it shows up! 19:04
19:04 darutoko left
grondilu but as "libbinaryen.so => not found" 19:04
19:05 zakharyas left, labster left, labster joined
geekosaur but it's in /usr/local/lib and you showed it being in the ld.so.conf stuff. siiiigh 19:06
ok, guess you get to try to force it
grondilu I get a different error from NativeCall now. Progress I guess. 19:07
geekosaur since gcc was being weird about -R, try passing it through as a linker option: -Wl,-R,/usr/local/lib
19:09 Guest22943 left, Guest22943 joined, Guest22943 left, Guest22943 joined
grondilu yes, that works. I see it correctly in ldd now 19:10
now I get a segmentation error 19:11
geekosaur sigh 19:12
grondilu nevermind. I'll go back to it later. 19:13
geekosaur wait, that wrapper makes not a lot of sense unless you are wrapping things that take refs already, and if you aren't then it should be throwing C errors unless they're doing something heinous in C
grondilu I guess I will have to look in the make file how they did compile and link binaryen. I'll just cargo cult it 19:14
geekosaur (which...)
because if something is struct foo f(struct foo a) then the wrapper is struct foo *f_(struct foo a_) {static struct foo evil = f(*a_); return &evil } # except this is not re-entrant nor thread safe, you need to be more clever in that case 19:15
grondilu I do more than turning struct value call into struct ref call if that's what you're saying. I think I can hide the struct all together. 19:16
19:17 Guest22943 is now known as giraffe
grondilu because the function that returns a struct is used by only one function anyway. 19:17
and that function does return a reference. 19:18
rather a BinaryenConst
which is a (void *)
I supposed the binaryen team had reasons to make the struct visible but it does not seem obviously necessary. 19:19
(sorry I meant BinaryenExpressionRef earlier, not BinaryenConst) 19:20
geekosaur you might want to make sure that does what you think it does. sadly, with C it's not always obvious 19:22
and if you get it wrong, your diagnostic is indeed a core dump :/
grondilu I guess to be safe I could do a wrapper using struct ref as you suggested. 19:23
but the struct has an anonymous union in it, that makes it painful to do with NativeCall 19:24
19:25 cdg joined, cdg left, cdg joined
grondilu oh man, I was silly again 19:27
I forgot the module argument
it works!
geekosaur: thanks a lot for your help :) 19:28
geekosaur yay
grondilu github.com/grondilu/binaryen-bindings 19:33
it will be cleaner to do the kind of wrapper you were talking about though, I'll work on it. 19:35
timotimo geekosaur: yeah, i grump that we don't use ldconfig to find .so files to load with nativecall 19:46
that's what that's about, right?
geekosaur no, that was MasterDuke complaining about tools choking on profile output 19:47
timotimo oh, you mean the profile output being far too big 19:49
i see
geekosaur yep
timotimo dinner time over here now :) 19:50
moritz bisectable6: my %h; %h{class { has str $.s }.new.s} = 1
bisectable6 moritz, On both starting points (old=2015.12 new=59bb1b1) the exit code is 0, exit signal is 11 (SIGSEGV) and the output is identical as well
moritz, Output on both points:
geekosaur grondilu, it's simpler to do the wrapper I suggested, but it's too limited --- note my comment about reentrancy. if yours works, go with it 19:51
timotimo moritz: requires an moar+nqp bump to work
grondilu I now understand why they made the struct visible. My wrapper can not re-use the same literal for different modules, for one. 19:52
moritz timotimo: oh, tehre's a fix for that already?
19:53 setty1 joined
AlexDaniel this, I guess: github.com/MoarVM/MoarVM/commit/28...193ab31a53 20:01
20:02 firstdayonthejob left
AlexDaniel MasterDuke: should it even report exit code if there's a SEGV? 20:02
20:03 rindolf left 20:06 firstdayonthejob joined 20:07 rindolf joined
AlexDaniel m: say 8.9 gcd 2 20:09
camelia rakudo-moar 59bb1b: OUTPUT«2␤»
20:11 bjz joined 20:13 devmikey joined
moritz has been working so much with gocd lately that he parses gcd as a misspelling of gocd :-) (go.cd/ for those wondering) 20:13
AlexDaniel I just threw it in as an interesting edge case, there's nothing wrong and nothing new with it. 20:14
… otherwise Zoffix wonders why is it so quiet here :) 20:16
geekosaur quiet? 20:17
yeek-yeek >.>
MasterDuke AlexDaniel: i guess there's no reason to
20:26 rurban left 20:31 nowan left
RabidGravy coo a live issue with the futuremusic.fm, haven't had one of those for a while 20:33
20:36 setty1 left 20:37 setty1 joined 20:38 BenGoldberg joined 20:39 domidumont left, leego left, leego joined 20:41 leego left 20:42 leego joined 20:49 bjz left 20:54 lizmat joined
RabidGravy I've so got to finish the radio software 21:03
good show on the stream.futuremusic.fm:8000/mp3 btw :) 21:05
21:08 devmikey left 21:24 Vynce joined
El_Che De gustibus non est disputandum :) 21:24
21:24 FROGGS_ left
timotimo moritz: yup, AlexDaniel found the right commit 21:37
21:39 bjz joined 21:41 seatek joined 21:42 kyclark_ joined 21:43 cdg_ joined 21:46 bjz left, cdg left 21:47 cdg_ left, Vynce left 21:48 bjz joined, nowan joined 21:51 Vynce joined 21:55 Vynce left, bjz left 22:00 kyclark_ left 22:06 bjz_ joined 22:11 bjz_ left 22:15 pmurias left 22:24 Ben_Goldberg joined, BenGoldberg left, Ben_Goldberg is now known as BenGoldberg 22:29 setty1 left, cdg joined, kurahaupo__ left 22:31 RabidGravy left 22:34 lostinfog joined, lostinfog left
nightfrog 0 22:38
seatek I just want to thank you all for exceptions. They work so nicely and so flexibly. 22:41
rindolf Hi all! make install paste.debian.net/895338/ in rakudo fails here - 22:48
22:52 danlei joined
danlei how can I redefine classes/roles/... at the repl without getting Redeclaration of ... errors? 22:53
22:54 kurahaupo left
timotimo JSON::Fast: 63.87user 0.48system 1:04.40elapsed 99%CPU (0avgtext+0avgdata 874728maxresident)k 22:56
JSON::Tiny: 338.42user 1.88system 5:40.57elapsed 99%CPU (0avgtext+0avgdata 4634628maxresident)k 22:57
about 19% as much ram usage, more than 5x as fast
for a 50 MB file 22:58
seatek timotimo: so if i were to, say, right now, change use JSON::Tiny to use JSON::Fast, no bad things would happen and my world would be much better? 23:00
timotimo correct
23:00 cdg_ joined 23:02 cdg left
seatek timotimo: and you also are correct! :) 23:02
rindolf Seems like the travis-CI build fails as well - for similar reasons - travis-ci.org/rakudo/rakudo/jobs/175494053
seatek no failed tests. that felt nice
danlei (I found MONKEY-TYPING/augment and that kinda works, but there has to be some easier way for interactive use, no?) 23:03
gfldex danlei: no, Perl 6 is a lot less dynamic that it looks 23:08
danlei gfldex: is that the way it's supposed to be and going to stay, or is it a deficiency of the implementation (rakudo)? that seems to be some basic stuff for interactive repl use 23:10
gfldex danlei: it will stay that way
danlei gfldex: hm, I'm a little -- shocked :) 23:11
gfldex: but maybe that's just because I'm used to the lisp (and other modern dynamic languages) way to do things 23:12
anyway, thanks
gfldex danlei: if you don't need type checks, you can do the dynamic stuff by hand
23:13 cpage left, cpage joined
danlei gfldex: you mean without using the class/rule/... sugar? 23:13
gfldex: or what do you have in mind? 23:14
gfldex m: my $C = class { method m { say 'oi?' } }; $C.new.m; $C = class { method m { say 'oi!' } }; $C.new.m
camelia rakudo-moar 59bb1b: OUTPUT«oi?␤oi!␤»
danlei gfldex: ah, I see
gfldex: that seems ok for playing around at the repl, but it's not ideal when, say, loading files in the repl 23:15
but it's at least something
gfldex: could you maybe point me to some document explaining the rationale behind this (perl6 being more static than I expected)? some synopsis maybe? 23:19
timotimo you can't really load a file "into" the repl at the moment 23:21
the repl is quite underpowered and also buggy in general
23:22 cpage left
danlei timotimo: I've written a little Emacs comint repl for perl6, I use EVALFILE for whole files, but often, I'm just sending the region to the process 23:22
23:22 Upasaka left, cpage joined 23:23 rindolf left
danlei timotimo: what gfldex showed is better than nothing, but of course I don't want to write my code this way :) 23:23
timotimo mhm 23:25
i got a major speed boost in JSON::Fast by throwing out the CATCH that translate the internal error from ordat going over the end of the string into a nicer error text 23:27
danlei I thought I just missed some pragma or command line switch -- I didn't expect that to be by design. it's still a great language, as far as I can tell from a few weeks of dabbling. but it's a little sobering for me. 23:28
anyway, thanks for the help guys
timotimo good luck and have fun with perl6 :) 23:29
danlei thanks :)
23:29 cpage left
timotimo your problem is just not being allowed to redefine classes in the repl? 23:30
23:30 cpage joined 23:31 cdg_ left 23:32 cyphase left
danlei timotimo: well, whatever yields the Redeclaration errors is suboptimal for interactive use, imho. (I was playing with classes and roles today and that's where I became aware of it. no problems with variables, subs etc. until now) of course, you could always just restart the repl, but that's a little awkward if you're used to languages where redefining is no problem. 23:35
timotimo have you tried spelling it "my class Blah { .. }"?
it won't affect things that have already been declared to use your class 23:36
because the name gets overwritten for future stuff, but the old thing is still there
23:36 kyclark_ joined, cyphase joined
danlei timotimo: that's even better than gfldex' suggestion, thanks (but I'd still have to write the code in my files this way) 23:37
thanks
23:39 kyclark_ left
danlei (actually, it's good to know about both :) 23:41
23:45 girafe left 23:52 kyclark_ joined 23:53 kyclark_ left
danlei thanks again and happy hacking -- I'm off 23:54
23:54 danlei left 23:59 pecastro left

Powered by Cro and the Raku® Programming Language.

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

Thank you!