[00:28] *** topnep left [00:29] *** topnep joined [00:38] *** Sgeo joined [01:47] @Coke The Windows GitHub actions / tests of "LLM::Functions" have been "always" failing. And they show the error you reported. There might be a "deeper" problem, so, I will spend some time trying to see what are the reasons for failure. [01:59] *** hulk joined [02:00] *** kylese left [02:02] <[Coke]> Let me know if you need me to test something. [02:06] <[Coke]> FWIW, it means I can't use App::Crag on windows [02:15] *** hulk left [02:15] *** kylese joined [02:25] <[Coke]> oh. this is a familiar error. [02:25] <[Coke]> you can't do .all ~~ Str:D [02:26] <[Coke]> it doesn't work on windows for some reason, and hasn't for ages. [02:26] <[Coke]> I forget what other module we tripped over that was doing that. They ended up changing the test to not use a junction. [02:27] <[Coke]> m: say .all ~~ Str:D # False on windows [02:27] [Coke], rakudo-moar f1c184644: OUTPUT: «True␤» [02:28] *** atcroft joined [02:29] <[Coke]> thought we had a rakudo ticket for this, don't see one, added https://github.com/rakudo/rakudo/issues/6109 [02:54] *** atcroft left [03:12] *** atcroft joined [03:34] *** vasko4535586 left [03:37] *** vasko4535586 joined [05:32] *** Aedil joined [05:36] *** soverysour joined [05:36] *** soverysour left [05:36] *** soverysour joined [05:41] *** soverysour left [06:07] *** soverysour joined [06:07] *** soverysour left [06:07] *** soverysour joined [06:12] *** soverysour left [06:33] *** atcroft left [06:56] *** constxd left [06:57] *** constxd joined [07:11] *** soverysour joined [07:16] *** soverysour left [07:25] *** wayland joined [07:43] *** johnjay left [07:44] *** abraxxa joined [07:46] *** johnjay joined [08:02] *** Sgeo left [08:03] Ok, good. I will make changes, test, and upload a new version within a 4-5 hours. [08:16] *** abraxxa left [08:18] *** abraxxa joined [08:53] *** topnep left [08:55] *** topnep joined [08:57] I thought this looked like an interesting project (perl LSP written in Rust) as a template for what could be done to make a Raku LSP https://github.com/EffortlessMetrics/perl-lsp/ [09:03] fwiw `say all() ~~ Str:D` does work on windows [09:03] and `say all() ~~ Str:D` doesn't [09:03] erm [09:03] and `say all(@ = ) ~~ Str:D` doesn't [09:03] quite puzzling [09:17] *** librasteve_ joined [09:21] *** wayland76 joined [09:22] *** wayland left [10:16] *** tobs left [10:17] *** tobs joined [10:27] *** johnjay left [10:32] *** johnjay joined [10:44] *** sjn left [10:58] *** abraxxa left [10:59] *** topnep left [11:00] *** topnep joined [11:46] weekly: https://www.reddit.com/r/rakulang/s/5oO7noad1V [11:46] librasteve, Noted! (weekly) [12:46] *** nine left [12:48] *** nine joined [13:18] I can only deal with Windows installations via GitHub's Windows actions. I usually have enough patience for one or two attempts fiddling with it. [13:28] *** kylese left [13:28] *** annamalai left [13:28] *** tobs left [13:28] *** kylese joined [13:28] *** tobs joined [13:29] *** simcop2387 left [13:35] *** tobs left [13:36] *** annamalai joined [13:37] *** tobs joined [13:37] *** simcop2387 joined [13:38] Unfortunately, is made with Rust... [13:41] @Coke The version 0.5.6 of "LLM::Functions" should not have the installation failing you reported. (Just uploaded that version.) But, otherwise, the Windows GitHub actions of that package are failing. [13:44] *** soverysour joined [13:44] @librasteve Why implement another LSP? Raku has LSP implementations for IntelliJ and VS Code. Do you have other IDEs in mind? [13:48] *** soverysour left [14:07] @antononcube I am just repeating what I saw here previoously eg << @librasteve this is just my humble opinion, but I think the most crucial thing for engaging the next 1000 Raku programmers would be having stuff like a static analyzer and LSP and treesitter grammar>> [14:07] quite likely I am conflating a treesitter LSP vs. the IntelliJ / VSCode LSPs we already have [14:08] Oh, yeah, now I remember that discussion. [14:09] BTW, Geany 2.1.0 (July 06, 2025) has officially included Raku as a recognized / supported filetype. [14:11] Sweet, I used to use Geany. [14:12] I tried to use in general and with Raku, but so far have not. Trying out its newest version, though... [14:14] *** Sgeo joined [14:17] that's cool - my personal opinion is that we have "good enough" tooling (always room for improvement ofc), but that we are lacking a reason for folks to take raku to their boss and say "I wanna build X and raku is the best tool because of Y" [14:20] Now I remember -- geany has fairly unnatural for macOS keybindings. [14:22] It seems that they can be changed programmatically, though. [14:23] There's a chicken & egg problem really. The reason to take it to the boss and say let's use this fundamentally has to include 'you can hire someone when I leave'. I think building awareness and developer base is how you fix that problem. [14:24] Having had to OK or deny tech choices, a huge part of that is just risk management. If Bob's the only guy who knows Raku, I can't OK it for anything substantial. [14:25] On the other hand there's a great reason to use it in this day & age: token efficiency, Raku is very terse and when you're spending thousands on Claude Enterprise, that matters. [14:25] I agree on chicken and egg ... the best way to solve this is set out in the book "crossing the chasm" imo [14:26] I think ecosystem expansion helps too. Cro exists, Cro is good. We need Cro for ML, Cro for TUI, Cro for GUI apps etc. [14:26] Something at that level with that level of documentation. [14:27] https://github.com/Raku/problem-solving/issues/507 [14:28] I just did full Notcurses bindings & Jinja2 template with 99% of the Python impl [14:28] Working on TUI framework atm [14:29] counter-intuitively to make a general purpose tool (like Raku) successful in general, you first need to focus on a small number of use cases where Raku has a compelling advantage over the alternatives (to overcome the chicken and egg) [14:30] I think the obvious standout is data transformations, which suits LLMs very well. Grammars are the killer feature imo. [14:31] I have also worked on web tooling (Air) with a very small hope that folks would see that for a new Rails type revolution ... my current feeling is that the world is not ready for a new web framework [14:31] lol - did you read the issue already? [14:32] I am reading it [14:32] https://github.com/m-doughty/Template-Jinja2/blob/master/lib/Template/Jinja2/Lexer.rakumod You just couldn't do this as simply in any other lang. [14:34] And thats probably suboptimal, I'm a noob with grammars. [14:35] so in a nutshell - I agree that Raku Grammars are the killer feature and that DSLs are the best focus [14:37] "The risk is that this may not turn out to be a horse." - I think this is my worry with DSL focus, but I agree it's the standout area. [14:38] I think we can drive more uptake by working on DSLs and sharing our stories on that vs. the same amount of effort on (say) another LSP [14:38] I would like a treesitter impl that isnt really slow in nvim :P [14:38] I also agree on horse risk [14:40] My M4 Max should not chug in nvim lol [14:40] but we have to make some bets or we will never drink any Champagne [14:42] [it may seem to many that I have forgotten that issue - not so, it is next on my list after Air stubs] [14:45] I think there's 2 problems that both need solving and imo the issue covers problem 1: getting interest. [14:46] Problem 2 is people bouncing off when something significant that they want isn't there in zef, or looks unmaintained. [14:47] there are 40 million active devs in the world - lets say we can interest 100 more of those in Raku, then problem 2 will get some help, eh? [14:48] Depends which 100 :P [14:48] A lot of those devs are pure 9-5ers with no interest in contributing upstream. OSS contributors is a much smaller proportion. [14:49] anyway my bet == my tuits and I hope I can convince others to help with the DSL story [14:49] I might join when I'm done with my ridiculous todo list :D [14:49] =b [14:50] Just bought 8 A100s so I can have agents running all night now. [14:52] Tired of spending $2000/mo on cloud AI :D [14:53] yikes [14:55] <[Coke]> .seen arkiuat [14:55] [Coke], I saw arkiuat 2026-03-27T21:03:23Z in #raku: nice [14:55] .seen Xliff [14:55] apogee_ntv, I saw Xliff 2025-07-05T21:33:48Z in #raku: apogee_ntv: I'm not against that approach. [14:55] :( [14:55] Yeah, in Raku the distance to Cro matters. I do not see as that good though. I think you still have to install with —/test. [14:55] He was working on notcurses a while back and I managed to get that working. [14:57] [Coke]: are you able to test again with most recent? i believe the issue to be fixed et the warnings gone [15:06] <[Coke]> tonyo: all tests pass, test and install both silent of warnings [15:07] <[Coke]> thanks! [15:07] alright i'll send up 0.18 [15:09] <[Coke]> tonyo++\ [15:10] <[Coke]> tonyo++ [15:10] *** topnep left [15:12] *** topnep joined [15:15] *** merp left [15:17] *** merp joined [15:33] @Coke No idea how to fix the rest of the Windows "bugs" of "LLM::Functions". I assume you have to install that package on Window separately with --force-test or --test. [15:58] <[Coke]> note: just running RAKULIB=. raku t/03-cloning.rakutest does not give the huge warning that 'zef test .' does - it just says that test 5 failed. [16:00] Is zef linked to a different raku installation that the raku in PATH? [16:01] <[Coke]> get-command shows them both in rakubrew under moar-2026.03 [16:01] <[Coke]> antononcube - if I put in a dd before the 5th test in that file, the stop-tokens shows as just [] [16:03] <[Coke]> ... which is the same as on linux. [16:04] <[Coke]> so your "all" isn't really doing anything anyway, is it? the list is empty [16:04] <[Coke]> m: say all() ~~ Str:D [16:04] [Coke], rakudo-moar f1c184644: OUTPUT: «True␤» [16:06] <[Coke]> (why is the test is , True, and not ok $conditional?) [16:06] alright 0.0.18 is published [16:06] <[Coke]> tonyo++ [16:08] <[Coke]> ugexe: ah, it's a future test int he list that has all the warning output, nevermind. [16:08] @Coke Hmm.. this seems like a rakudo Windows bug, then. [16:10] I am not sure is that going to be clearer or not. [16:14] <[Coke]> That's just the failure on 3. 6 and 7 also die, for different reasons. [16:14] Right. [16:15] <[Coke]> I'll see if I can track those down. [16:16] *** soverysour joined [16:16] *** soverysour left [16:16] *** soverysour joined [16:17] <[Coke]> ... ah. 5, which complains about info, is going through a code path that .all ~~ Map:D; in lib/ [16:17] <[Coke]> so I'm guessing it's all the same root issue. [16:20] <[Coke]> and the can't find multi for llm-function... one of the multis has a where clause of .all ~~ Str:D [16:20] <[Coke]> so you're off the hook. :) [16:20] @antononcube I prefer to recommend zef install Cro --/test because the test takes a long time ... otherwise it works fine - to fix that we probably can move a bunch of Cro test out from /t to /xt [16:21] @librasteve Ok, good to know. [16:21] (with the exception that if your machine does not have 0.0.0.0 and port open fpor loopback test) [16:28] *** sibl joined [16:33] *** sibl left [17:08] *** soverysour left [17:22] *** soverysour joined [17:22] *** soverysour left [17:22] *** soverysour joined [17:26] *** soverysour left [17:33] *** soverysour joined [17:33] *** soverysour left [17:33] *** soverysour joined [17:40] *** soverysour left [17:46] i install cro with --/test too [17:47] for exactly that reason ^ [17:59] ¦ raku.org: librasteve++ created pull request #312: Stubs [17:59] ¦ raku.org: review: https://github.com/Raku/raku.org/pull/312 [18:03] on raku.org and stubs ... coke raised concerns that the default Air strategy for routes was an issue for the raku.org site some time back https://github.com/Raku/raku.org/issues/240 [18:05] going forward you just need to go Page.new: :stub and all will work (alongside the other route method still in play for Air::Components) and a sitemap and robots.txt is generated for you automagically, of course [18:07] sorry it took a while, but quite a deep change so needed me to clear some quality time [18:09] full example (from Air::Examples) https://www.irccloud.com/pastebin/YVJu70Of [18:09] *** soverysour joined [18:09] *** soverysour left [18:09] *** soverysour joined [18:15] *** annamalai left [18:16] *** annamalai joined [18:18] *** sjn joined [19:09] .hug [19:09] * huggable6 hugs simon_sibl [19:09] Ah only in #raku I guess not in beginner channel [19:10] I regularly forget about the existence of the -begginner channel, I should put that in my autojoin list [19:23] *** mehbark joined [19:43] *** soverysour left [20:04] Voldenet: do you know when say all(@ = ) ~~ Str:D stopped working on windows? i tried on 2023.04 (I cant try on anything else) and it worked. that is arm though so no jit [20:12] <[Coke]> oh, I can bisect it. (I'm not sure it ever worked) [20:12] <[Coke]> Thanks for the reminder, I should have thought of that. [20:14] the first thing to try would be running with spesh and jit disabled [20:14] <[Coke]> confirm it also worked for me in 2023.04 [20:15] <[Coke]> I tried that per timo's suggestion earlier, no change. [20:15] ah nice [20:19] <[Coke]> 2024.01 is the first release it fails in [20:21] <[Coke]> I'll do a bisect on the commits between those releases. [20:29] [Coke]++ [20:29] ¦ raku.org/main: f45ecfe2b0 | librasteve++ | 21 files [20:29] ¦ raku.org/main: stubs ok [20:29] ¦ raku.org/main: review: https://github.com/Raku/raku.org/commit/f45ecfe2b0 [20:29] ¦ raku.org/main: 2587aebf5f | librasteve++ | META6.json [20:29] ¦ raku.org/main: pin to Air:ver<0.1.16> [20:29] ¦ raku.org/main: review: https://github.com/Raku/raku.org/commit/2587aebf5f [20:29] ¦ raku.org/main: c7f98471cd | librasteve++ (committed using GitHub Web editor) | 22 files [20:29] ¦ raku.org/main: Merge pull request #312 from librasteve/stubs [20:29] ¦ raku.org/main: [20:29] ¦ raku.org/main: Stubs [20:29] ¦ raku.org/main: review: https://github.com/Raku/raku.org/commit/c7f98471cd [20:32] <[Coke]> as soon as I reinstall strawberry perl. As soon as I figure out how to run the MSI as an admin. as soon as... [20:37] winget install -e --id StrawberryPerl.StrawberryPerl [20:38] i dunno if winget helps with the permissions issue [20:39] what if you run with RAKUDO_OPTIMIZE_NOSM=1 [20:42] <[Coke]> yup, that fixes it [20:42] <[Coke]> (on 2026.03) [20:42] i got the suggestion from claude in this response where it was asking for more information https://gist.github.com/ugexe/4e2d80629fdedb8d9305e437510ea58a [20:44] <[Coke]> Added notes to the ticket. [20:47] <[Coke]> Weirdly, there were no commits to the optimizer file between those release tags [20:47] <[Coke]> so the per-revision bisect probably still needed. [20:47] <[Coke]> (finally working on the first commit there.) [21:00] <[Coke]> ok, the first three commits don't build for me [21:02] <[Coke]> so, giving up for now [21:03] i assume it works without the :D? [21:06] <[Coke]> correct. [21:40] <[Coke]> looks like "rakudo-m-early-build" is dying trying to build gen\moar\CORE.c.setting on this old version. Any thoughts? [21:40] <[Coke]> return code is the not very helpful 0xff [21:45] is everything clean, in terms of files and such? [21:46] <[Coke]> fresh clone [21:46] <[Coke]> 2023.12 era [21:48] try turning JIT off [21:48] you probably have a newer windows that requires the setjmp fix in moarvm [21:52] <[Coke]> Thanks, that helped. [21:52] "newer windows" really means newer MSVC [21:54] *** japhb left [21:56] *** japhb joined