ShimmerFairy [Coke]: To answer your earlier question, I checked the JSON spec and it's still very much a valid JSON file, it just can't be parsed as a sequence of graphemes. (It should've caused problems a while ago, but IIRC MoarVM never handled Prepends exactly right.) 02:17
Assuming my quick glance at the JSON spec is right, in order to correctly parse all possible JSON files, you should be using something like NFC (possibly Uni if you can manage to leave the text unnormalized?) to do the parsing with. Since JSON::Fast isn't using Grammars, that's hopefully not too bad a change to make. 02:18
ab5tract Hmmm, this is also maybe another motivation for me to finish the work on adding a third-party JSON parser to MoarVM. Besides the speed, we can avoid some pitfalls like this\ 08:55
ShimmerFairy: Nice to see you around these days btw! I went lookking for your blog a while ago but couldn't recollect the domain 08:56
ShimmerFairy Thanks! Don't worry about that, I've not blogged in many many years. 08:57
About the JSON thing, the issue isn't really with the JSON parser itself, rather a bigger issue with Raku that would be nice to fix someday. Namely, while NFG strings are really nice for a lot of work, they flat out aren't the right tool for every job, and right now Raku's support for codepoint-based strings is pretty terrible (about as bad as C and C++, I'd say). 08:58
A related issue is that, in retrospect, we didn't really think NFG through as much as it perhaps needed to be. For example, right now Raku accepts "ൎ"" as a single double-quoted string (as it parses source code as NFG), but should it? Or is it technically OK but a security issue? 09:03
lizmat from what I can see, that roundtrips... not sure I see a security issue here? 09:27
ab5tract lizmat: Fixed my patch! :D 09:47
ShimmerFairy I was thinking along the lines of "confusables", like you could possibly trick someone reading source code into misinterpreting a " if it's actually part of a grapheme that gets swallowed up into a string literal.
ab5tract it needed to guard against $checkee being a role group itself 09:48
lizmat ab5tract++
ab5tract still needs a roast run, but it's working fine with Hash::Ordered now
lizmat cool! 09:49
ab5tract sometimes we can have nice things ;)
lizmat almost Xmas :-)
ab5tract I tend to get too excited and announce news to soon 09:52
which is to say, this still needs more work :) 09:53
Geth rakudo/main: 6b47541e27 | (Elizabeth Mattijsen)++ | src/Raku/Grammar.nqp
RakuAST: make sure "constant" is localizable

Spotted by habere-et-dispertire++ in
   github.com/Raku-L10N/L10N/issues/16
09:56
lizmat ab5tract: understood :-)
ShimmerFairy One thing that's on my mind now, after going through the Unicode upgrade, is idly wondering what parts of Raku are based on old, obsolete ideas from Unicode (along the lines of Windows's 16-bit wchar_t), or what would be different in a new language today. 09:57
ab5tract Those are excellent questions. Have there ever been attempts to re-implement the NFG plumbing in a separate library? 10:28
It feels like NFG is mostly unknown in the outside world, which is a bit sad considering what an achievement it is 10:29
Just thinking out loud, but maybe implementing a libnfg in say, Zig, would surface some of those spots you’re pondering about 10:30
ShimmerFairy Writing a library (or multiple) to let me work with strings in C++ like I can in Raku has long been an idea in the back of my head. The funny thing about NFG is that it's not that complicated an idea really, it's just a way of working with grapheme clusters without having to think about it too much. 10:44
10:53 melezhik joined
melezhik [Coke]: will do. I am working on improving brownie reports , once it’s settled I will probably fill few tickets 10:54
Meanwhile , once you have a time please upgrade agents and run it - cc ab5tract [Coke] patrickb - thanks 🙏 I would like to run new rounds with improved reports data 10:55
ab5tract melezhik: "wonder-thunder has entered the chat" 11:21
I wonder if running under qemu would solve my issues with capping ressources 11:22
melezhik ab5tract: ++
Ok, tests reports are more informative now. I use prove6 t/ — verbose -I lib 12:06
brw.sparrowhub.io/file_view/job.1.r...Traced.log
To get details on failing modules
Ok. This clear some details “ZMQ error: Address already in use (code 98)” 12:15
brw.sparrowhub.io/report/job.1.report/784
Why some tests fail
[Coke] melezhik: upgraded 12:16
melezhik [Coke]: ++ 12:17
[Coke]: could please pull the latest ? 12:18
0.0.47
use prove6 verbose to get full report for failed modules
[Coke] so not just the "update inside the container" but start over?
melezhik Ah, interesting, this should pulled it , let me see 12:19
Maybe it’s just old job
[Coke] send termination request as /root/.sparky/work/agent/.states/1765023402.terminate
start new agent job
melezhik Ok . Looks good now 12:20
Yep . This is what upgrade does
heartbeat from cokebot-53585969, version: 0.0.47, max threads: 4, jobs run cnt: 0
Now up update 👍
ugexe: looks like zef install . --deps-only does not install test dependencies, how can install all modules dependencies without installing a module? 12:34
Or maybe it does … 🤔 still not sure 12:37
Ok . Looks like I found yet another bug in brownie, it uses prove6 from default Raku, no good , stopping 010r 12:40
Sigh
13:02 melezhik_ joined
melezhik_ [Coke]: ab5tract: I've fixed one important bug, could you please upgrade one more time? thanks 13:16
13:21 melezhik_ left
[Coke] done 13:36
14:43 melezhik left 15:17 melezhik joined
melezhik Thanks 15:17
16:21 melezhik2 joined
melezhik2 [Coke]: fixed yet another bug, please upgrade again , thanks ) 16:21
working version should be 0.0.49, 16:22
finally I think reports format is good enough , there is enough details in failing tests
16:25 melezhik2 left
ugexe if you want to run tests without installing test dependencies then use --/test-depends --test 16:50
16:58 Pixi left 17:03 Pixi joined
ab5tract melezhik: my box is going to be indisposed for a bit, so I've deactivated the agents for now 17:19
melezhik ab5tract: no worries and thanks for all your help 17:20
[Coke] laptop crashed, coming back online 17:22
melezhik “7:50 PM <ugexe> if you want to run tests without installing test dependencies then use --/test-depends --test” ugexe thanks . Finally I figured out that I had to run ‘zef —deps—only install . && zef —verbose test .’ 17:27
[Coke]: ++ 17:28
18:04 Pixi` joined 18:05 Pixi left 18:10 Pixi` left 18:19 Pixi joined 18:21 finanalyst joined 19:37 melezhik left
ab5tract [Coke]: I’ve had some crashes while running brownie too. Are you running with podman-compose? 21:14
Because I haven’t managed to actually get any resource constraint options to work reliably 21:15
22:01 finanalyst left
[Coke] ab5tract: using docker. 22:29
(host OS is mac) 22:30