vrurg jnthn: I just would like to merge it after the release due to many other things fixed. Don't want the ops to impeed the merge. 00:03
AlexDaniel please no more operators, especially whole family of operators, especially ones that won't be used that often
jnthn AlexDaniel: They're not really operators, at least in how they're parsed; more just modifiers on `.` 00:04
But yeah, I'm not yet convinced whether they're worth a short syntax. 00:05
AlexDaniel jnthn: yeah but excuses won't make the code using them more readable :S
vrurg AlexDaniel: besides, I beeter have $obj.?+foo(...) than $obj.^batch-call('foo', \(...))
I doubt the second is more readable than the first. 00:06
AlexDaniel it kinda is 00:07
jnthn AlexDaniel: Well, "different things should look different" (and perhaps "odd things should look odd") are pretty much design principles...
But yeah, I'm far from sure about the new . modifiers at this point.
The problem with the second one isn't just readability, but also efficiency. 00:08
vrurg I'm not a big fan of new ops either. But found no better solution.
jnthn But again, I'm not sure how relevant that is
I mean, we have .* and .+ but in reality BUILDALL is custom-compiled to incldue calls to them 00:09
vrurg That's why I proposed experimental status.
jnthn Rather than using the operators
(To be clear, to include calls to the BUILD submethods)
Plus BUILD and TWEAK have to be interleaved
I can't remember actually using .+ and .*. I can imagine situations when they'd be useful, I just haven't ended up factoring things that way. 00:10
Maybe that's 'cus I tend to be a light user of inheritance anyway... :)
vrurg Interleaving is the biggest reason ops are not applicable to them. But current design makes submethods pretty much useless for a developer.
AlexDaniel jnthn: does Comma show docs for operators when you hover on them? 00:11
vrurg The PR changes the situation and provides more opportunities for using submethods.
jnthn AlexDaniel: It will. We didn't get to it in the current release (the current one is just the start of the integrated docs effort). 00:12
It's only this one when we started showing docs from doc comments in current source and from used modules, plus from perl6/doc for the built-ins. 00:13
And that's *nice*, but there's still a load to polish there 00:14
AlexDaniel I see 00:19
jnthn That's why I added a bunch of doc comments to the current Cro release :) 00:20
japhb jnthn: The problem I had earlier with die and react whenever turned out to be a braino. Rereading rt.perl.org/Public/Bug/Display.html?id=130716 carefully yet again yielded new insights. That's one seriously info-dense bug. 03:08
04:08 squashable6 left 04:11 squashable6 joined 05:16 squashable6 left 05:19 squashable6 joined 07:22 AlexDani` joined 07:23 AlexDaniel left 08:28 sena_kun joined
lizmat Files=1277, Tests=109554, 210 wallclock secs (28.37 usr 7.63 sys + 2933.11 cusr 260.04 csys = 3229.15 CPU) 08:28
08:29 sena_kun left
tyil AlexDaniel`: new output is here gitlab.com/tyil/rakudo-star/-/jobs/339846981 08:54
08:55 sena_kun joined
tyil github.com/rakudo/rakudo/issues/3274 09:04
lizmat afk for most of the day& 09:12
tyil looks like there's an Any in between each signal which I guess shouldn't be the case 09:31
I could apply a grep and filter out all Any values, but then everything would be shifted one to the left, which could be fixed by inserting an Any at the front again 09:32
but that sounds very dirty ;~;
09:37 ExtraCrispy left 09:43 Altai-man_ joined 09:45 sena_kun left
MasterDuke pmurias: how should decont be done for truffle? 10:29
huh, why is the jvm spurt implementation so complicated? 10:37
10:49 timo1 joined
pmurias MasterDuke: re decont I think a node with @Specialization for various things it gets 10:53
MasterDuke: most of the time decont is a noop
MasterDuke: what do you mean by the jvm suprt implementation? 10:54
MasterDuke: if you are looking at nqp::spurtasync it doesn't seem to be used anywhere so I don't think there is any need to implement that 10:56
11:44 sena_kun joined 11:45 Altai-man_ left 12:00 ggoebel joined 12:24 donaldh joined 12:31 [Tux] left 12:36 donaldh left 12:39 [Tux] joined 12:46 lucasb joined 13:03 ggoebel left
MasterDuke pmurias: spurt, slurp, and lines from github.com/perl6/nqp/blob/truffle/...andle.java are what i was surprised by. nqp::open on the jvm creates a AsyncFileHandle, which is why i'm doing it 13:09
pmurias MasterDuke: the low level spurt doesn't seem to be used for anything 13:11
MasterDuke pmurias: as i understand it, those are for the convenience subs nqp provides 13:12
so yeah, probably not a high priority 13:13
pmurias MasterDuke: spurt seems to be only used by the nqp::spurtasync op 13:14
MasterDuke: which I assume is some no longer used relic 13:15
MasterDuke it's not used in rakudo at all
huh, it's only in nqp-jvm 13:16
pmurias and it's not implemented on other backends so we don't need that
AsyncFileHandle is created by nqp::openasync 13:17
not nqp::open
MasterDuke argh, i was looking at the wrong thing! 13:18
jnthn Pretty much all string I/O is coordinated at Rakudo level these days, only demanding binary I/O of the runtime level (and decoder stuff) 13:19
13:21 [Tux] left, [Tux] joined
MasterDuke jnthn: how recent were any nqp changes that were done for that? maybe we should do the work to rebase up to head if it means we don't have to implement a bunch of stuff that would then just get thrown away 13:22
jnthn MasterDuke: Long ago I don't remember when :) At least a year, surely. 13:24
Possibly 2+ years
MasterDuke hm, i think the truffle branch is pretty old. do you know an easy way to check if it's after that change? 13:25
jnthn I'd just git log on NQP's src/core/IO.nqp or whatever that file is actually called :)
MasterDuke i see things like `Implement readchars in NQP IO handle.`, `open/close/spew/spurt use NQPFileHandle.`, `Implement set-encoding in NQP file handle.` by you in may 2017 13:28
jnthn That sounds like it.
So, 2+ years. :D
MasterDuke ok, sounds like we're good then 13:29
pmurias MasterDuke: I don't think there have been gigantic changes to nqp-jvm backend since the start of the branch 13:30
MasterDuke: there are no longer used ops in HEAD nqp
jnthn I'm pretty sure I didn't do any cleanup of the JVM stuff after doing the I/O changes, just the additions needed to get things working. So there's surely dead code. 13:32
13:43 Altai-man_ joined
MasterDuke pmurias: truffle doesn't have a SixModelObject, right? github.com/perl6/nqp/blob/truffle/....java#L389 13:44
13:46 sena_kun left
El_Che AlexDaniel`: �� No regressions found, dot file not saved 13:48
will save the output dir
pmurias MasterDuke: we don't, we use Object instead (will look at the code after food) 13:50
MasterDuke: I assume that "types" put in hllConfig are TypeObjects 14:06
MasterDuke pmurias: i think i have nqp::open done, but it's obviously not easy to test until i also have nqp::(read|write)fh 14:11
Geth_ nqp/truffle: d49239bb03 | (Daniel Green)++ | 8 files
[truffle] Add nqp::open

However, it's hard to test if it actually works without nqp::(read|write)fh also implemented, so caveat emptor.
MasterDuke pmurias: ^^^ anything look off? 14:14
El_Che AlexDaniel`: for you information: # du -sh Bl 14:21
AlexDaniel`: 12G
so it looks like you need around 14gb space to run Blin on docker 14:22
AlexDaniel`: output is here: claudio.ulyssis.be/var/tmp/output.tar.gz 14:24
AlexDaniel`: output is here: claudio.ulyssis.be/var/tmp/output.tar.gz
AlexDaniel`: I will run it again, as it seems the newer image was not there yet when it started 14:35
AlexDaniel`: I wonder if ssh isn't a problem on itself. I fixed the unknonw host problem, but it does not mean you can clone those repos: gist.github.com/nxadm/d51c57553835...67c13bb282 14:39
AlexDaniel`: so the unknow host thing may be a red herring and those modules should really point to https 14:40
pmurias MasterDuke: seems correct 14:48
MasterDuke good 14:49
14:52 timo1 is now known as timotimo, AlexDani` is now known as AlexDaniel 14:53 AlexDaniel left, AlexDaniel joined
El_Che AlexDaniel`: ^-- 14:54
AlexDaniel El_Che: actually… 14:55
El_Che: your output says that 1100 modules are OK
El_Che: previous run was 1055 OK modules
so whatever the difference is, your setup is better
El_Che now running with the ssh fix, so curious about that, but see the remark above about ssh clonin 14:56
AlexDaniel El_Che: we should just PR all modules with that issue
so… hm 14:57
interestingly, “Unicode::GCB – Flapper” 14:58
this one should've been bisected, actually 14:59
El_Che so missing dep? 15:00
I changed to deps from the indructions because they were outdated
libqt-something-3 to 4 and jre-default to default-jre 15:01
AlexDaniel El_Che: no, Flapper means it tested it on HEAD and it failed, then it tested it on 2019.07.1 and it succeeded. Then to make sure it's not random it did a few extra runs and got different results
El_Che Oh 15:02
AlexDaniel this could be because it just crashed for random reason on 2019.07.1
but in my experience rakudo is not that unstable… 15:03
so hmm
El_Che: you're doing another run, right?
either way, previous run was clean, and this time we got more OK results, so it can't possibly be bad
El_Che Yes, running one with the ssh changes as we speak 15:04
AlexDaniel samcv: green light from me. We'll be doing some extra checks because I'm paranoid, but there are no issues whatsoever
El_Che mm, I didn't clean the output dir, is that bad? 15:05
restart with a new one?
AlexDaniel El_Che: doesn't matter
El_Che ah ok, I'll keep it running then
AlexDaniel: in the future we can run the image with podman instead of the docker daemon because it doesn't require root 15:06
MasterDuke pmurias: ugh, readfh is going to be more annoying that i thought. and the existing writefh will likely need some work too 15:07
AlexDaniel there are now release branches in nqp and rakudo 15:08
discord6 <kelly> what are you guys testing?is there a release coming? 15:29
pmurias MasterDuke: doing readfh and writefh properly requires implementing decoding/encoding (t/nqp/082-decode.t) 15:31
not the most fun stuff
MasterDuke yeah, i'm seeing that 15:32
AlexDaniel kelly: yes :)
15:45 Altai-man_ left
El_Che AlexDaniel: ⏳ 403 out of 1343 modules processed 16:00
(off to the supermarket)
16:05 ggoebel joined
pmurias MasterDuke: something at least a little more interesting would be to implement 'if ... -> $cond {}' 16:10
MasterDuke shouldn't github.com/perl6/nqp/blob/master/s...#L287-L288 just be `my str $litconst_lc := nqp::fc($node[0]);`? 17:33
(with appropriate changes after) 17:34
17:58 SmokeMachine left 17:59 SmokeMachine joined
Kaeipi should Metamodel::Primitives have methods for nqp's can and type parameterization ops? 18:12
those are the main reason a library i'm writing is stuck using nqp
jnthn Yes for the type parameterization ones; for nqp::can you can just do .^can, no? I mean, it's already something you ask the MOP, not really a primitive. 18:23
I thought the type parameterization ones were already in Metamodel::Primitives.
Kaeipi ik .^can exists, but it isn't guaranteed to exist on a HOW 18:28
afaik at least 18:40
jnthn Right, but .^find_method *is*, because nqp::can is defined in terms of nqp::findmethod which at least falls back to .^find_method 18:41
Possibly it's worth it for avoiding going via the MOP, though 18:42
Well, to hit the cache
pmurias MasterDuke: that seems to be incorrect 19:02
MasterDuke: ß will match sS using nqp::eqatic
MasterDuke: so the NFA matching will have differen semantics
19:27 pmurias left 20:09 Kaeipi left 20:10 Kaeipi joined 20:20 MasterDukeMobile joined
MasterDukeMobile pmurias: to confirm, you also think the *existing* code is incorrect? 20:21
tellable6 MasterDukeMobile, I'll pass your message to pmurias
20:25 pmurias joined
pmurias MasterDuke: yes, the NFA and regular case insensitive matching *currently* works differently when matching ß and sS 20:26
tellable6 2019-11-02T20:21:29Z #raku-dev <MasterDukeMobile> pmurias: to confirm, you also think the *existing* code is incorrect?
MasterDukeMobile Should be a relatively easy fix. Not exactly sure how to craft a test case, but I think something with interpolating a literal value should do it 20:27
20:28 Kaiepi joined, Kaeipi left
MasterDukeMobile E.g., match ß against an interpolated sS 20:29
20:30 MasterDukeMobile left 20:32 MasterDukeMobile joined, MasterDukeMobile left
pmurias MasterDuke: match something using the NFA 20:52
MasterDuke: and match something without using the NFA
MasterDuke: using say | for the NFA case and having for example an action that stops NFA for the other case 21:01
discord6 <kelly> will the new release contain jnthn's EA and other optimization that he talked about in conf 21:11
AlexDaniel it has a bunch of stability and performance improvements 21:48
don't know about EA, but regardless of that it's a much better release 21:49
21:53 ggoebel left
jnthn There's some EA stuff that ain't ready for merge yet. I'll get back to it soon; I've been distracted with a mixture of AFK stuff, the rename stuff, etc. 21:57
I really don't want to merge it until I'm feeling really confident about it.
22:16 lucasb left
discord6 <kelly> cool 22:30
23:31 pmurias left