melezhik. . 06:32
Voldenet grondilu: I think you're right - posted code should work 07:14
Voldenet m: class A { method !foo { "bar" }; class B { method x { A.new!foo } } }; A::B.new.x 07:17
camelia ===SORRY!=== Error while compiling <tmp>
No such private method '!foo' for invocant of type 'A::B'
at <tmp>:1
------> o { "bar" }; class B { method x { A.new!<HERE>foo } } }; A::B.new.x
Voldenet very similar code in C# works, so that is definitely sane behavior dotnetfiddle.net/GQiPxf 07:25
and also works in java
librasteve Voldenet: (and grondilu), the docs are quite expansive on how raku differs from other languages … Unlike in some languages where private is an accessibility modifier on a method, in Raku "private methods" and "methods" are quite different things - that is to say, it's better to read "private method" as a compound noun rather than an adjective describing a noun. 08:00
that is not to say that you are wrong, since I do not claim to understand the raku design well enough, or to know the reason why the docs say that 08:01
I suspect it may boil down to “we need a way to do fully private things in raku since “my” variables cannot be hidden from outer scopes” or maybe “a module is a class, so we want to suppress lexical scoping across module boundaries” … as I say I don’t know and this is wild speculation… I’m not that deep into using private methods so I could live with changing/fixing this providing it is clear 08:07
and simple either way :-)
lizmat you probably should think of private methods as lexically scoped subs that have an implicit invocant "self" specified as the first positional argument 08:47
Voldenet librasteve: Well, hence "I think", I'm not very fixated on this - syntactically the class being inside of the other should have access to things on the same level IMO, but those details are not the same even in java/kotlin/C# 08:49
I feel like there's no single idea on what's the point of subclasses in the first place 08:50
Voldenet but it might be difficult to say (a'la C++): I want class B and C to be friends of A and have access to its private methods, but other parts of code should not have this access 08:53
grondilu The thing is, even if I define the nested class inside the definition of a method of the outer class, the private call still fails. This is odd to me. The private call works, unless it's inside a local class? 09:01
m: class A { method !foo {}; method bar { class B { method x { A.new!foo } } } }; 09:02
camelia ===SORRY!=== Error while compiling <tmp>
No such private method '!foo' for invocant of type 'A::B'
at <tmp>:1
------> method bar { class B { method x { A.new!<HERE>foo } } } };
grondilu ^see?
besides there is no reason raku should think the method is called by an invocant of type A::B and not A, is there? 09:03
golfing it even more: 09:04
m: class { method !foo {}; method bar { class{ method x { A.new!foo } } } };
camelia ===SORRY!=== Error while compiling <tmp>
Whitespace required after keyword 'class'
at <tmp>:1
------> ass { method !foo {}; method bar { class<HERE>{ method x { A.new!foo } } } };
grondilu oops
m: class { method !foo {}; method bar { class { method x { A.new!foo } } } };
camelia ===SORRY!=== Error while compiling <tmp>
Undeclared name:
A used at line 1
grondilu nevermind 09:05
m: class A { method !foo {}; method bar { class { method x { A.new!foo } } } };
camelia ===SORRY!=== Error while compiling <tmp>
No such private method '!foo' for invocant of type '<anon|1>'
at <tmp>:1
------> ; method bar { class { method x { A.new!<HERE>foo } } } };
grondilu m: class A { method !foo {}; method bar { my &foo = { A.new!foo }; class { method x { foo } } } }; 09:07
camelia ( no output )
grondilu ^that would be a workaround, I guess.
m: lass A { class B {...}; trusts B; method !foo {}; method bar { class B { method x { A.new!A::foo } } } } 09:11
camelia ===SORRY!=== Error while compiling <tmp>
No such method 'add_trustee' for invocant of type
'Perl6::Metamodel::PackageHOW'
at <tmp>:1
grondilu m: class A { class B {...}; trusts B; method !foo {}; method bar { class B { method x { A.new!A::foo } } } }
camelia ( no output )
grondilu ^or this
Voldenet oh, raku has "trusts", huh 09:12
patrickb I think I'm facing a Native all bug. paste.sr.ht/~patrickb/1de35011b308...634f1739b1
On Linux both the C and the Raku version work fine. On an ARM MacOS the C version works, but in the MacOS Version the ioctl call succeeds, but does not update the $size struct. 09:14
melezhik .tell patrickb: you might be interested in this - sparky.sparrowhub.io/report/~patri...-API/24326 HTH 09:15
tellable6 melezhik, I'll pass your message to patrickb
patrickb Any clues on how to best debug this? Any best guesses as to which component is the most potential offender?
melezhik sorry you are talking about this? "[Terminal-API] get-config failed" ? 09:16
patrickb No. This is about get-win-size and it only affects MacOS. 09:17
wayland Hi all! I asked a question about a month ago about operators that take multiple parameters. I've written a short gist about what I'm trying to achieve at gist.github.com/wayland/2876a7c8ef...24ad9ed8c7 . If anyone has ideas about how I could achieve these things, I'd be interested to hear about it :)
Voldenet patrickb: I'd build C library that'd set the struct's property without ioctl and see if that works 09:21
melezhik sure, np 09:23
Voldenet if not, then it's good test case, if it does, then it's ioctl-related somehow
patrickb I'll have to verify, but I'm pretty sure that's covered by the native all tests. I suspect it's related to ioctl being a vararg function.
but yes, creating an isolated test case is a good idea. 09:24
Voldenet wayland: you don't need an operator to take multiple parameters, `html⪪body⪪h1{.id eq "red"}` can be written as ((html⪪body)⪪h1).CALL-ME({ .id eq "red" }) 09:31
lizmat m: class A { method CALL-ME(|) { say "hi" } }; say A() 09:43
camelia (A(Any))
lizmat m: class A { method CALL-ME(|) { say "hi" } }; say A.new()
camelia A.new
lizmat m: class A { method CALL-ME(|) { say "hi" } }; say A.new()()
camelia hi
True
lizmat Voldenet: is the .CALL-ME needed there ? 09:44
librasteve_ . 09:46
Voldenet hm 09:49
m: class h1 { method CALL-ME(&n) { n() } }; h1({ .say }) # I was expecting this to work without () 09:51
camelia (Any)
Voldenet m: class h1 { method CALL-ME(&n) { n() } }; h1 { .say } # but it doesn't
camelia ===SORRY!=== Error while compiling <tmp>
Unexpected block in infix position (missing statement control word before the expression?)
at <tmp>:1
------> ss h1 { method CALL-ME(&n) { n() } }; h1<HERE> { .say } # but it doesn't
expecting any o…
Voldenet h1 would work as a sub 09:54
librasteve_ www.irccloud.com/pastebin/lv83BHqX 09:57
librasteve ^^ I'm fixated on the grondilu thing and thus made a little snippet to try and get in my head what lizmat said in that context ... ultimately I think the thing that is confusing me is that grondilu original has a subclass that makes a new outer ... not sure that that would ever be something I would do (but then I guess there may be a reason like nested class trees) 09:59
anyway must to $day-job 10:00
Voldenet m: multi infix:<A>(&a, &b) {}; multi sub html() { -> {}; }; sub h1(&n) { -> { }; }; html A h1 { .id eq 'red' }
camelia ===SORRY!=== Error while compiling <tmp>
Undeclared name:
A used at line 1
Voldenet this doesn't quite work
m: multi infix:<A>(&a, &b) {}; multi sub html() { -> {}; }; sub h1(&n) { -> { }; }; (html) A h1 { .id eq 'red' }
camelia ( no output ) 10:01
wayland Thanks Voldenet and lizmat. It seems I've been unclear. I've amended the gist to say "The elements such as body and h1 are supposed to be undeclared labels; you could equally well substitute in filenames if the tree in question happened to be a filesystem instead of a document." 10:03
That's why I said I could quote them if necessary. HTH,
Voldenet m: role N {}; multi infix:<A>(Str $a, Str $b) { N }; multi infix:<A>(N $a, Str $b) { N }; multi infix:<A>(N $a, Block $b) { N }; "html" A "head" A "h1" A { .id eq "red" } 10:09
camelia ( no output )
wayland Voldenet: Thanks! That looks like my least bad option at the moment. 10:13
Voldenet I'm not sure if for "xpath-like" operator it wouldn't be better to define some other operator 10:14
m: role N {}; multi infix:<A>(Str $a, Str $b) { N }; multi infix:<A>(N $a, Str $b) { N }; multi infix:<B>(N $a, Block $b) { N }; "html" A "head" A "h1" B { .id eq "red" } 10:15
camelia ( no output )
Voldenet m: role N {}; multi infix:<A>(Str $a, Str $b) { N }; multi infix:<A>(N $a, Str $b) { N }; multi infix:<B>(N $a, Block $b) { N }; "html" A "head" A "h1" B { .id eq "red" } A "h2" B { .className eq "x1" } 10:16
camelia ( no output )
Voldenet m: role N {}; multi infix:<A>(Str $a, Str $b) { N }; multi infix:<A>(N $a, Str $b) { N }; multi infix:<A>(N $a, Block $b) { N }; "html" A "head" A "h1" A { .id eq "red" } A "h2" A { .className eq "x1" } # it works either way, but with B i feel that it's more readable
camelia ( no output )
Voldenet and after all, no CALL-ME was needed 10:17
wayland Voldenet: Yes, agreed about the separate operator; was thinking the same. 10:18
Voldenet wayland: btw, I read the second part after left join - RED was doing something similar and it leads to emitting proxies and making a lot of custom operators 10:31
however RakuAST might improve that dev.to/fco/the-future-of-red-orm-for-raku-44gf 10:32
SmokeMachine It still is doing that… the RakuAST implementation of Red is still in very early stages
Voldenet my overly positive news reporting style treats early stages as finishing touches ;P 10:33
Voldenet hm, thinking about how this could be passed this gives me an interesting idea, probably making xpath a lot simpler 10:36
m: sub xpath(*@n) {}; xpath "html", "head", "h1", { .id eq "red" }, "h2", { .className eq "x1" }
camelia ( no output )
Voldenet wayland: ^ the style is easier and requires no custom operators at all 10:38
I have no idea which one is better, but second one has hardly any magic in it, which probably is good 10:39
bad part is that it has no magic :P 10:41
patrickb I have it reproduced without ioctl. It's exactly as I guessed. It fails on MacOS when trying to modify a struct passed in via varargs. 11:02
That specific thing is not covered by tests.
This makes me think that the issue is probably with dyncall /libffi 11:03
SmokeMachine m: class Node { has $.parent; has $.child; has @.tests; method AT-KEY(&code) { @!tests.push: &code; self }; method add($child) { without $!parent { $!parent = $!child; return self }; self.new: :parent(self), :$child }}; sub term:<html> { Node.new: :child<html> }; multi infix:<< <- >>(Node $parent, $child) is looser(&postcircumfix:<{ }>) { Node.new: :$parent, :$child }; say html <- "body" <- "h1"{ *.id eq "red" } 11:06
camelia Node.new(parent => Node.new(parent => Node.new(parent => Any, child => "html", tests => []), child => "body", tests => []), child => Failure.new(exception => X::AdHoc.new(payload => "Type Str does not support associative indexing."), backtrace => Back…
SmokeMachine Odd… locally that works…
Voldenet there's no postcircumfix<{ }> 11:07
SmokeMachine No, I got it wrong…. It does not work locally either… :( 11:12
But it works with ()…
glot.io/snippets/h5t1sg55xd 11:14
lizmat and yet another Rakudo Weekly News hits the Net: rakudoweekly.blog/2025/03/25/2025-...n-rakuast/ 11:34
leont Hmmm, Air looks pretty cool 12:30
Xliff \o 13:20
tellable6 2025-03-22T23:17:32Z #raku <lizmat> Xliff: if you're not using those routines, it's all good
Xliff Was reading FCO 13:27
D'oh
Was reading FCO's post on Red's future and was wondering how that would work. Are there plans on exposing a mechanism that will generate the AST for a classes methods (or at least selectively preserving the AST in stage parse? 13:28
if not, it could be as simple as having a trait...let's say "ast-preserved" which will add a .AST method to a Callable which will emit the AST as presered during the parsing phase. 13:29
I was working on something like that a while ago, but got sidetracked 13:30
Would the core devs be interested in me reviving that work for a future PR?
lizmat the idea is that a CHECK phaser would have access to the complete RakuAST tree 13:31
so at that time, you would be able to do whatever you want to that tree, before it gets turned into QAST / bytecode 13:32
SmokeMachine My plan is, on CHECK time, replace all map/grep/first/etc to add a named parameter with the block ast… 13:41
lizmat Xliff: so such a trait would only make sense if you'd like to be able to access the original AST **at runtime** 13:42
librasteve_ I was wondering how the AST would get applied … CHECK sounds cool 14:02
leont: appreciate the kind words … it’s still early days for Air so now is a great time to come in a shape things the way you want ;-) 14:03
leont Getting form handling right is the main hard thing really. I think pretty much all the other things are obvious 14:04
All the other things are just output, form handling is also handling input if done right.
librasteve_ I have done all the HTMX examples in raku crotmp already … github.com/librasteve/raku-HTMX-Examples 14:07
and I think that quite a lit of these are form-relavant and could/should be included in forms
(which I did not really do any examples / any Components or Tags for in Air yet) 14:08
one general question I have is whether there is “core Cro::WebApp” forms support currently used by Cro Template and if so how easy that would be to carve out and re-use… so I guess next step is to rootle around in there 14:10
leont: do you have a role model form implementation that you like (not necessarily in raku) … or maybe some words about do & donts? 14:17
s/do/dos/
leont I've been trying to stay away from anything front-end for at least 15 years so I wouldn't really know what the good solutions are. 14:19
librasteve_ very wise 14:47
liztormato Librasteve do you think it would be feasible / make sense to rework irclogs.raku.org in hArc? 14:50
librasteve feasible - yes. make sense - not really since quick glance at the code shows many lines of cro templates …. so I am not sure it makes sense to just rewrite something that works unless there is a need for big changes 15:27
what I have in mind is a rework of raku.org … since tbh that is a bit in need of a more modern design and some of the content (raku example, and so on) looks very stale and also imo, there is about 10 times too much info than is needed to appeal to a potential raku wannabe to get started 15:30
what do you think of that idea?
[Coke] Folks - had to reload the MoarVM tarball into the release, and *create* the nqp tarball. both should be in place now.
the git repos were fine, but any tools that depended on the release would have had an issue (e.g. rakudo/star) 15:31
Xliff Oh! You can access the AST at CHECK time? How would that work? The .ast method? 15:41
Xliff librasteve: Do you think a rework of raku.org to use Air might be more flexible and reliable than Cro templates? 15:44
liztormato librasteve feels like we can use a restyling to coincide with a 6.e release 15:45
Xliff I love Cro templates, but they have limitations that I suspect Air does not.
Xliff As long as coding styles enforce separation of display logic and module logic it could work well. 15:46
librasteve_ I, also, love cro templates (so much that I added fragment support to them back in December) … they are the gold standard in terms of usability and syntax for templates (and there is even good syntax highlighting in the IntelliJ raku plugin v2) then I got into HTMX and I implemented the HTMX examples in Cro templates github.com/librasteve/raku-HTMX-Examples … then I got kinda frustrated since I had forgotten how mind 15:54
numbing and restrictive coding a website in a template language can be and I missed all the cool stuff I could do with a full language and since I had been listening to some podcasts about elmlang, I realsed that the functional ness of raku is a fantastic way to bring HTML tags right into raku (my favourite coding medium) … this is the way
liztormato Xliff: re CHECK time access. Probably a dynvar such as $*CU
SmokeMachine Would that make sense to use Cromponents once it already uses Cro templates? 15:55
librasteve_ liztormato … good idea … I will make a prototype of a new design and share around for feedback … 6.e is due to release at Easter, right? 15:57
SmokeMachine Xliff: what kind of limitations you see on Cro templates? 15:58
librasteve_ Xliff: the great thing about HTMX is that all the coding can be done in the back end (yes even for dynamic UX) … I have taken a lot of inspiration from the HTMX essays - I recommend htmx sucks first, then locality of behaviour and complexity budget are my favourites htmx.org/essays/#complexity-very-very-bad 16:03
Xliff SmokeMachine: Ability to set constants. This is deliberate according to jnthn. Minor shortcomings to expression logic. LTA handing of objects that are not defined (particularly, if a template tries to evaluate Nil, it will crash the route 16:04
Xliff librasteve: And yes, this is the way! :) 16:05
liztomato: OK, so $*CU hasn't been implemented, yet?
SmokeMachine Xliff: it does not fix that, but have you tried Cromponent?
Xliff SmokeMachine: How could Air not support that? It's Raku!? 16:06
SmokeMachine: No. I have not tried that yet. Link?
SmokeMachine $*CU is implemented (I’m searching for my example using it) but it only exists índice a CHECK block
Xliff Ah. OK. 16:07
Can you modify $*CU at CHECK time?
SmokeMachine Xliff: github.com/FCO/Cromponent 16:08
Yes, you can… I’m still searching my test/example
Xliff Dynamic variable $*CU not found 16:09
Xliff SmokeMachine: Huh! Looks interesting. I'll give it a spin and see if I can use it. Thanks! 16:10
SmokeMachine Xliff: you need to set RAKUDO_RAKUAST=1 if I remember it well...
librasteve_ forgive me - what is it that Air does not support? 16:11
Xliff Ah, THAT'S what I'm missing.
Not Air. Cro templates.
librasteve_ phew
SmokeMachine Xliff: I finally found it: github.com/FCO/ACME-Overreact 16:24
Xliff Thanks!
SmokeMachine github.com/FCO/Acme-Overreact/blob...ct.rakumod 16:25
SmokeMachine That was one of my first tests to see if it would be possible to use RakuAST on Red… 16:31
grondilu wasn't there a syntax for read-only variables? 16:37
m: my $x is ro = pi 16:38
camelia ===SORRY!=== Error while compiling <tmp>
Can't use unknown trait 'is' -> 'ro' in variable declaration.
at <tmp>:1
------> my $x is ro<HERE> = pi
expecting any of:
TypeObject
default
dynamic
export
grondilu sigilless variables won't do, as they'd bind to the variable, right?
m: my $x = pi; my \var = $x; $x = 3; var
camelia WARNINGS for <tmp>:
Useless use of var symbol in sink context (line 1)
grondilu m: my $x = pi; my \var = $x; $x = 3; say var 16:39
camelia 3
SmokeMachine also, I was thinking... I don't know how macros are going... but I think that another way of solving the Red AST replacing would be if we could not only define macros that would be called like `macro($par1, $par2)` but also `$par1.macro($par2)` that way, Red could define a `map` macro for example that would do: `$par1.map: &par2, :raku-ast(&par2.ast)` 17:00
m: my $x = pi; my \var = $x<>; $x = 3; say var 17:02
camelia 3.141592653589793
SmokeMachine grondilu: ^^
when you do `my \var = $x` you are setting the container to `var`, if you remove it from the container, it will work the way I think you want... 17:04
SmokeMachine m: my $x = pi; given $x -> $var is copy { say $var; $x = 3; say $var } 17:09
camelia 3.141592653589793
3.141592653589793
lizmat Xliff: try: RAKUDO_RAKUAST=1 raku -e 'CHECK say $*CU; say "hello world"' 17:13
Xliff lizmat: Got it. Now I want to see if I can make it easier to obtain just the AST for methods rather than having to introspect the entire compunit. 17:16
lizmat grep is your friend: CHECK .say for $*CU.grep(RakuAST::StrLiteral); say "hello world" 17:17
grondilu SmokeMachine: what does $x<> mean? 17:18
lizmat $x deconted
Xliff lizmat: Yeah? Try doing that to a full compunit with a few classes that contain a dozen methods apiece!
lizmat so how would you select them then? 17:19
SmokeMachine Xliff: Maybe this could help: github.com/FCO/ASTQuery
lizmat or that :-) 17:20
Xliff Hmmm... $*CU.grep( RakuAST::Method ) actually works a treat.
lizmat SmokeMachine++ slipped my mind
Xliff FCO: Are you finished with that project or is there a future roadmap?
grondilu sees the so-called decont operator in the docs. Assumes it's fairly new because he doesn't remember it at all. 17:21
Xliff nqp::decont has been around for a while
SmokeMachine Xliff: not finished, still working on that… when possible…
lizmat grondilu: actually, <> is the zen slice. and the zen slice will return its invocant decontainerized 17:22
as all code blocks do, unless they have an "is raw" :-)
grondilu Oh, I see. I was wondering if that notation would interfere with postfix:<> for hashes. 17:24
lizmat it doesn't: it's the same thing :-)
grondilu I wonder if $$x would do the same
m: my $x = pi; my \var = $$x; $x = 3; say var
camelia 3
grondilu nope
lizmat itemization makes sure that it only itemizes deconted value 17:25
s
m: my $x = 42; say $x =:= $$$$$x 17:26
camelia True 17:26
lizmat it actually doesn't re-containerize 17:26
grondilu FYI I have just finished translating chess.js into raku. It's seems to work, it's much faster than my previous code, but still much slower than javascript. 17:27
lizmat grondilu: cool! 17:27
grondilu lots of room for optimization, though. native types and so on. 17:27
lizmat blog post ?
grondilu is not really a blog guy 17:27
anyone can feel free to check out my repo, though : github.com/grondilu/chess 17:28
I think all programming language should have a chess engine for performance tests. IMHO. 17:30
[Coke] grondilu: I think the example is slightly better with -I. instead of -Ilib 17:30
[Coke] or, if someone is installing it, without the -I 17:31
[Coke] grondilu++ 17:31
[Coke] didn't realize it had a gui. :) 17:31
grondilu it's not a GUI lol 17:32
[Coke] er, ui. :) 17:39
grondilu in chess.js there is a perft method, with depth as argument. I translated it to raku too. 17:42
in javascript perft(4) takes about 100ms. Guess how long it takes in raku.
SmokeMachine Xliff: did you look at ASTQuery?
librasteve 100s? 17:46
grondilu less 17:47
librasteve btw if its faster than raku, than alone is worthy of a blog
less than 100s or 100ms?
nahita3882 π seconds
grondilu less than 100s 17:48
librasteve m: say e ** (i * π), 's'; 17:49
evalable6 -1+1.2246467991473532e-16is
Raku eval -1+1.2246467991473532e-16is
grondilu anyway it's 37s against 122ms. It's not too bad for a first draft, and a straightforward code translation with zero optimizations.
it's much better than my own implementation anyway, which was ridiculously slow. 17:50
Xliff SmokeMachine: I remember when you started working on it! 17:52
SmokeMachine: Are you finished with it or is it still under development? 17:53
SmokeMachine Still under development… but it can already be useful… 17:57
grondilu it's not finished, of course. There are so many things to implement.
grondilu I should add entries in the TODO, though. 17:59
one of the most ambitious goal would be a fully-fledge, uci-capable chess engine. 18:00
(www.chessprogramming.org/UCI)
maybe a GUI one day?
though I think if I could display the board and animate the moves in the terminal with Kitty's protocol, that'd be already pretty cool. 18:02
Xliff Kitty? 18:03
Is there a working Kitty for Linux?
grondilu sw.kovidgoyal.net/kitty/ 18:04
holmdunc github.com/vushu/raylib-raku would be good for making real graphical chess 18:56
antononcube Hm... "Graphviz::DOT::Chessboard" is (most likely) much more convenient to use. 19:43
It produces SVG images. 19:44
melezhik. .tell librasteve: I look at github.com/librasteve/Air/tree/main/t and just for fun I am thinking to implement those one by using sparrow task checks . It’ll be interesting to compare . In the end of the day air generates a lot of html code which is essential to test … just a though 🤔 19:45
tellable6 melezhik., I'll pass your message to librasteve
antononcube Also, "JavaScripdt::D3" can produce D3.js code of chess positions.
The ability to make a chess-GUI with Raku is demonstrated here: www.youtube.com/watch?v=fwQrQyWC7R0 19:47
librasteve melezhik: please do - but i am worried about the ease of installing the Air deps since I’m struggling on macOS with DBIish on pq with 2025.02 20:26
tellable6 librasteve, I'll pass your message to melezhik
Xliff How can I get raku to run a command in the background and exit? 20:31
(effectively daemonizing said program)
librasteve Proc::Async maybe? 20:33
tellable6 hey librasteve, you have a message: gist.github.com/3624428b1f78b4fa66...8ce77c1685
librasteve docs.raku.org/type/Proc::Async 20:34
[Coke] I don't think you can daemonize it, but you could use proc::async, then have your code wait until it completes then exit 20:56
[Coke] googles and finds raku-knowledge-base.podlite.org/mo.../README.md - wtf is this? 20:57
ah, it's github.com/hipek8/p6-UNIX-Daemonize 20:58
ugexe lol 21:03
it is using fork()?
yeah no one should be using github.com/hipek8/p6-UNIX-Daemonize 21:04
ugexe we do not support forking in raku (even if there is a nqp::fork()), since moarvm itself cannot support forking past a certain point in the startup 21:07
[Coke] Yah, it's only there, not published. 21:10
jdv iirc brrt put real work into that 21:27
Xliff Could I use nohup and background that? 21:42
antononcube @.xliff 21:44
@.xliff See here: 21:45
cdn.discordapp.com/attachments/633...6dc3d&
Xliff antoncube: What model is that? 21:52
antononcube OpenAI's "gpt-4o" with the prompt "CodeWriterX". 22:10
See that prompt here : github.com/antononcube/Raku-LLM-Pr...son#L10658 22:11
Actually, it might not be clear that the screenshot is of Raku chatbook, not the web interface of a certain LLM provider service. 22:13
ugexe jdv: into nqp::fork yes, and the way it was used was within the early startup contain i mentioned earlier. early startup being before libuv initialization or some such 22:23
ugexe s/contain/constraint/ 22:23

Powered by Cro and the Raku® Programming Language.

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

Thank you!