00:14 MasterDuke joined
MasterDuke ugexe: i meant how he and jnthn got to 90% done (being very vague here, done ~~ (100% roast && building rakudo) ) pretty quickly, but the last bits seemed to be a terrible slog, pretty much burning nine out and turning him off raku 00:17
i think nine himself has suggested going back through the logs to see what he mentioned as being particularly bad pain points, and things he would prefer be removed from the language 00:19
away now to build a metal box (...so many boxes, i feel like i'm in a video game...) 00:29
00:34 MasterDuke left 01:49 Pixi` joined 01:52 Pixi left 01:54 Pixi` is now known as Pixi
Geth__ rakudo: ugexe++ created pull request #6615:
RakuAST: make a regex's slurpy hash behave like a method's
02:20
02:41 kurahaupo left
Geth__ rakudo/main: 4 commits pushed by (Nick Logan)++ 03:43
04:32 sjn_ is now known as sjn
Geth__ rakudo: ugexe++ created pull request #6616:
Share unit lexicals between BEGIN code and the mainline
04:44
04:47 Pixi` joined 04:50 Pixi left 05:05 kurahaupo joined 06:35 kurahaupo left, kurahaupo joined 06:38 kjp left 06:42 kjp joined 07:46 jjatria left, jjatria joined 08:02 vrurg_ joined 08:03 vrurg left 08:57 Geth__ left, Geth joined 10:11 finanalyst joined 10:44 finanalyst_ joined, finanalyst left 11:02 finanalyst_ left
lizmat FYI: dev.to/lizmat/mainstreaming-rakuast-49j8 12:48
Geth rakudo/main: c68f3e170d | (Nick Logan)++ (committed by Elizabeth Mattijsen) | 3 files
Take the Slip raw where SlippyIterator recognizes Empty

slip-all decides whether it has been handed Empty with nqp::eqaddr. Reached the way an iterator reaches it, with the result of a map or grep block, that comparison was testing a container rather than the Slip and never matched, so slip-all built an iterator over Empty and pushed from it to push nothing. Every element such a block rejected cost that ... (13 more lines)
13:19
rakudo/main: de9d8b5548 | (Nick Logan)++ (committed by Elizabeth Mattijsen) | 2 files
Only sum the split adverbs once one has been given a value

Every Str.split checks that at most one of :v, :k, :kv and :p was passed, by adding the four boolifications together. That is four method calls and a chain of Int additions boxing a result per step, on a call that almost always passes no adverb at all.
... (10 more lines)
[Coke] I don't think we can default to RAKUAST until we make that moarvm change. 13:25
The cleanish results from that blin run included a moarvm patch that is not applied. 13:26
lizmat that is understood :-)
[Coke] OK. Just making sure we're all on the same page there. 13:28
lizmat yeah, we are... :-)
[Coke] If you're thinking september: 1) get the moarvm patch in ASAP. 2) start doing blin runs ASAP
That can't wait until the week-to-go run.
lizmat understood
Geth rakudo/main: 242b33ac92 | (Elizabeth Mattijsen)++ | src/core.c/ThreadPoolScheduler.rakumod
Make setting up $*SCHEDULER *not* vivify %*ENV

The requirement for setting up $*SCHEDULER do not actually need val() processing at all, so separate them from normal setup, thus reducing startup time.
14:00
ugexe i feel like we would want to do github.com/rakudo/rakudo/pull/6534 but because its a compile time error i'm not sure how we should work around the roast tests for p5 regex since compile time error will occur before e.g. skip-rest can handle things. i guess we could wrap everything in an eval or something 15:05
ab5tract did we leave dir(:recurse) as a user exercise for similar reasons as to why we didn't implement factorial? 15:17
because it bugs me that it's missing :) 15:18
related: trying to make an LLM use Raku instead of Python in its processing/glue scripts is made nearly impossible due to our leniency on named arguments. it's not a good mix for the hallucination machine 15:19
[Coke] ab5tract: i use File::Find 15:51
Geth rakudo/main: 4 commits pushed by (Nick Logan)++, (Elizabeth Mattijsen)++ 15:52
16:04 Pixi` is now known as Pixi
lizmat notable6: weekly 16:33
notable6 lizmat, 1 note: 2026-08-31T12:47:58Z <lizmat>: dev.to/lizmat/mainstreaming-rakuast-49j8
lizmat notable6: weekly reset
notable6 lizmat, Moved existing notes to “weekly_2026-08-31T16:33:27Z”
17:18 ShimmerFairy left 17:20 ShimmerFairy joined
lizmat and yet another Rakudo Weekly hits the Net: rakudoweekly.blog/2026/08/31/2026-...revisited/ 17:44
ugexe huh, it looks like this thing got Inline::Perl5 passing all of its tests... including the v5-inline stuff 18:16
patrickb "this thing" being mutsu? 18:40
ugexe no, claude code 18:50
the missing context being Inline::Perl5 wasnt working under rakuast yet
20:19 finanalyst joined
ab5tract File::Find is great, but I'm still unclear why the functionality isn't included in core. I'm expecting there to be a good reason, it's my lack of imagination that I don't see it 20:46
the "we leave factorial as an exercise to the user because it is so easy to implement" makes more sense to me than leaving it to users to write a DFS/BFS
for processing the full contents of a directory 20:47
so what am I missing? :) 20:48
ugexe probably because what should happen in various situations isn't well defined. how to treat symlinks, what order to return them in. further, dir is lazy so an argument that makes it not lazy would make such an api a bit inconsistent 20:50
ab5tract fair :) 20:51
I figured the answer would probably be related to how FS stuffs is rarely as simple as one expects it could be 20:52
ugexe i can't really think of a good reason for dir to be lazy honestly 20:53
ab5tract maybe we can keep that on a backburner for 6.f
it would be interesting to see if anything in the ecosystem would choke on a non-lazy dir 20:54
some mechanism for making un-consumed named args fatal would also be useful 20:55
if perhaps impossible? 20:58
21:21 finanalyst left
Geth rakudo: ugexe++ created pull request #6617:
RakuAST: fix package naming and version parsing for Inline::Perl5
21:57
rakudo/main: 578952e6ae | (Nick Logan)++ | 3 files
RakuAST: don't rename a package with a raw VM string

Previously a package's meta-object was created with the name it was declared with and then renamed with its fully qualified name, always handing `set_name` a raw VM string. The legacy frontend passes the fully qualified name to
  `new_type` once and never renames.
... (10 more lines)
22:48
rakudo/main: 45d527bdf4 | (Nick Logan)++ | 3 files
RakuAST: don't parse a module name like v5-inline as a version

Previously the optional `+` or `-` that may follow a version's numbers would take the hyphen out of a module name such as `v5-inline`. `vnum` is `\w+` and stops at the hyphen, so `v5-inline` matched as version `5-`, and the
  `<!before '-'>` guard after it cannot help because the hyphen has already been
consumed by then. ... (9 more lines)
rakudo/main: a35b0af00f | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | 6 files
Merge pull request #6617 from ugexe/ugexe/rakuast-package-name-and-version-adverb

RakuAST: fix package naming and version parsing for Inline::Perl5