blin-rakuast-triage.md

RakuAST Blin Module Board — 2026-08-20

Source: https://github.com/coke/raku-ast-blin-results at 5b64b48.

Blin ran with RakuAST as the default frontend (rakudo 19a125e7e9 "Change default to AST"; two modules bisected one commit earlier to 86dc6eac "test moarvm fix", which is the same flip). Every failure in this run is therefore RakuAST-attributable.

Two things this run does not show.

1. It was run with the proposed MoarVM serialization fix applied, so the whole "Object does not exist in serialization context" bucket (29 modules in the 2026-07-21 triage) is absent. That fix has not been turned into a PR.
2. It predates rakudo main as of 2026-08-20 by a week, and its ecosystem snapshot is older still. 34 of the 53 were reinstalled and retested against a rakudo built from main today; the results are in Re-verification below. Three rows are already resolved: SQL::Abstract and Data::Generators pass on main, and Net::IP::Parse was fixed and released as 0.0.7 after blin's snapshot was taken.

Counts. 1422 OK, 53 fails, 2 flappers (Net::DNS, Selkie::UI — load-dependent, not triaged here), 1 bogus ecosystem entry (Foo:: Foo), 400 AlwaysFail and 554 Unknown (fail on both frontends or never ran; out of scope).

---

How to read this

Verdict

• RAKUDO — genuine RakuAST bug. Fix belongs in rakudo.
• MODULE — RakuAST is correctly stricter, or the module reaches into compiler internals it was never promised. Fix belongs in the module.
• MODULE (fragile) — the module's test asserts an implementation detail (a backtrace index, a coverage line list, a mixin's exact .WHAT). Both frontends are defensible; the test expectation has to become frontend-agnostic.
• DEP — fails only because a dependency in another row fails.
• DEFERRED — known, decided not to chase.

PR status

• PR delivered — a pull request exists upstream (linked).
• fix ready, not delivered — a finished commit sits on a local branch in this workspace and was deliberately not sent upstream.
• rakudo PR open/merged — the rakudo-side change that clears it.
• none — nothing written yet.

---

Board

Module	Repo	Group	Verdict	Status
App::APOTD	deoac/apotd	G4 call analysis	MODULE	none
Array::Circular	lizmat/Array-Circular	G8 containers	RAKUDO	none
BioInfo	MattOates/BioInfo	G2 slang	RAKUDO	none
BioPerl6	cjfields/bioperl6	G6 RakuDoc	MODULE	none
CSS::Module	css-raku/CSS-Module-raku	G11 versions	DEP	none
CSS::Specification	css-raku/CSS-Specification-raku	G11 versions	MODULE	PR delivered (#6, merged, unreleased)
Code::Coverage	lizmat/Code-Coverage	G10 attribution	MODULE (fragile)	none
Control::Bail	skids/perl6-Control-Bail	G3 type strictness	MODULE	none
Data::Generators	antononcube/Raku-Data-Generators	—	flake	passes on main today
Dawa	~bduggan/raku-dawa	G1 $*W	MODULE	none
Env	lizmat/Env	G1 $*W	MODULE	none
FindBin	lembark/raku-findbin	G3 type strictness	MODULE	none
FindBin::libs	lembark/raku-findbin-libs	G3 type strictness	MODULE	none
French	raku-multilingual/raku-french	G2 slang	MODULE	none
GD::Raw	raku-community-modules/GD-Raw	G6 RakuDoc	MODULE	PR delivered (#8, open)
Grammar::BNF	raku-community-modules/Grammar-BNF	G2 slang	MODULE	none
IRC::Log::Colabti	lizmat/IRC-Log-Colabti	G3 type strictness	MODULE	none
IRC::Log::Textual	lizmat/IRC-Log-Textual	G3 type strictness	MODULE	none
Image::Resize	raku-community-modules/Image-Resize	G6 RakuDoc	DEP (GD::Raw)	none
Inline::Perl5	niner/Inline-Perl5	G1 $*W	MODULE	fix ready, not delivered
JSON::Fast	timo/json_fast	G3 binding strictness	MODULE	none
JobQueue	m-doughty/JobQueue	G12 unclassified	unknown	none
LIVR	koorchik/perl6-livr	G5 P5 regex	MODULE	none
List::Allmax	thundergnat/List-Allmax	G3 type strictness	MODULE	none
MergeOrderedSeqs	FCO/MergeOrderedSeqs	G3 type strictness	MODULE	PR delivered (#1, merged, unreleased)
Net::IP::Parse	bradclawsie/Net-IP-Parse	G3 type strictness	MODULE	PR delivered (#8, merged and released in 0.0.7)
OpenMPT::Bindings	sasha/OpenMPT-Bindings	G8 NativeCall at CHECK	RAKUDO	none
P5__DATA__	lizmat/P5__DATA__	G2 slang	MODULE	none
P5tie	lizmat/P5tie	G7 dynamic-scope pragma	MODULE	none
Pod::Perl5	dnmfarrell/Pod-Perl5	G9 regex/Match	unknown	none
Polyglot::Regexen	alabamenhu/PolyglotRegexen	G1 $*W	MODULE	none
Portuguese	raku-multilingual/raku-portuguese	G2 slang	MODULE	none
Proc::Easier	sdondley/Proc-Easier	G10 attribution	MODULE (fragile)	none
Regex::FuzzyToken	alabamenhu/RegexFuzzyToken	G9 regex/Match	RAKUDO	none
RegexUtils	grizzlysmit/RegexUtils	G5 P5 regex	MODULE	none
Router::Right	pheix/raku-router-right	G5 P5 regex	MODULE	none
SION	dankogai/raku-sion	G9 regex/Match	RAKUDO	none
SQL::Abstract	Leont/sql-abstract	G4 call analysis	—	fixed since the run
SQL::Lexer	dmaestro/SQL-Lexer	G9 parse strictness	MODULE	none
ScaleVec	samgwise/p6-ScaleVec	G12 unclassified	unknown	none
SelectiveImporting	wukgdu/raku-SelectiveImporting	G2 slang	MODULE	none
Selkie	m-doughty/Selkie	G6 RakuDoc	MODULE	none — retest 0.15.1 first
Slang::Forgiven	mustafaaydn/Slang-Forgiven	G2 slang	MODULE	none
Slang::Mosdef	bduggan/mosdef	G2 slang	MODULE	none
Slang::Predicate	samgwise/p6-Slang-Predicate	G2 slang	MODULE	none
Sustenance	repo unavailable (github:atweiden)	G12 unclassified	unknown	none
T	CIAvash/T	G2 slang	MODULE	none
Test::Stream	raku-community-modules/Test-Stream	G10 attribution	MODULE (fragile)	PR delivered (#5, open)
Tie::Array	lizmat/Tie-Array	G7 dynamic-scope pragma	DEP (P5tie)	none
Tie::StdArray	lizmat/Tie-StdArray	G7 dynamic-scope pragma	DEP (P5tie)	none
Web::Cache	scmorrison/perl6-Web-Cache	G10 attribution	MODULE (fragile)	none
XHTML::Writer	gfldex/perl6-xhtml-writer	G3 type strictness	MODULE	none
auto-dynamic	lizmat/auto-dynamic	G1 $*W	DEFERRED	none

---

G1. Reaching for $*W (4 + 1 deferred)

RakuAST has no $*W. Reading it throws Dynamic variable $*W not found, usually from a sub EXPORT running at BEGIN time.

• Dawa, Env, Polyglot::Regexen — direct $*W in sub EXPORT.
• Inline::Perl5 — same, plus the v5 slang, which is a separate porting gap. 35 of 42 test files pass with the $*W half worked around; the use v5 slang half needs its own port. A rakuast branch in scratch/Inline-Perl5 carries the $*W half and was never delivered.
• auto-dynamic — DEFERRED by earlier decision. Its exported &DYNAMIC hooks an implementation-detail sub; a one-line fix is known but the module is asking for something the frontend does not promise.

What has to change. Port each $*W use to $*R (the RakuAST resolver) plus the $*CU/$*LANG surface. Four modules in this family have already been ported and merged this way and no longer appear in the run — CompUnit::Util (PR #1), Benchy (PR #3), overload::constant (PR #1), and FINALIZER (PR #4). Those are the templates to copy.

G2. Slangs and grammar augmentation (10)

Modules that mix a role into the MAIN grammar, augment Raku::Grammar, add package_declarator rules, or hook World.do_import.

• French, Portuguese — declare custom infix operators; RakuAST reports Lexical symbol '&infix:<...>' is already bound to an outer symbol.
• Slang::Forgiven, Slang::Mosdef — new statement/term syntax; RakuAST reports Unexpected block in infix position.
• Slang::Predicate — custom operators; output carries a stray α\t prefix.
• T — mixes a T::Grammar role into $*LANG.slang_grammar('MAIN') to add a t term. Under RakuAST the term never registers, so raku -M T -e 't 2 == 2' reports Undeclared routine: t.
• P5__DATA__ — augments the grammar; Method Raku::Grammar+{Data}.pod_newline not found.
• SelectiveImporting — uses the World do_import hook for :select/:except import tags; RakuAST reports no such tag 'select'.
• Grammar::BNF — the serialization half was fixed; two slang test files remain.
• BioInfo — the odd one out and the only RAKUDO verdict in this group: Cannot find method 'to-begin-time' on object of type QAST::Op. A slang path hands RakuAST a raw QAST node where it expects an AST node.

What has to change. For everything but BioInfo this is a language-surface question, not a bug: RakuAST needs a supported slang story, or each module needs rewriting against $*LANG.slangs and the RakuAST grammar's own extension points. Test::Async was ported this way successfully (PR #3, merged), so custom declarators are portable; custom operators and terms are the harder half. BioInfo needs a rakudo-side fix at whichever slang entry point accepts a QAST::Op.

G3. Binding and type-check strictness — RakuAST is correctly stricter (10)

Legacy skipped a nominal or return type check; RakuAST performs it and the module's own signature turns out not to describe what it is passed.

• List::Allmax (expected Callable[Callable] but got Block), MergeOrderedSeqs (expected Order but got Block), XHTML::Writer (expected Positional but got Str) — a *@slurpy ahead of a typed parameter made legacy skip the typed parameter's nominal check.

• IRC::Log::Colabti, IRC::Log::Textual — expected Positional[Str] but got Array (["tyil"]). The value has to be an explicitly typed array.

• Net::IP::Parse — Type check failed for return value; expected Positional[...UInt8]:D. Fix is to build and return a typed array.

• Control::Bail — This type (Failure) does not support associative operations.

• FindBin, FindBin::libs — This type cannot unbox to a native string: P6opaque, Junction. A Junction reaches a native-string slot.

• JSON::Fast — Parameter '$pos' expects a writable container (variable) as an argument, but got '5' (Int) as a value without a container. The module passes ++$pos to another int $pos is rw parameter. prefix:<++> yields a value, not a container, so this is not supposed to bind — and both frontends agree in three of the four shapes:

  shape	legacy	RakuAST
  Int $n is rw, ++ on a lexical	rejects	rejects
  int $n is rw, ++ on a lexical	rejects	rejects
  Int $n is rw, ++ on a forwarded is rw param	rejects	rejects
  int $n is rw, ++ on a forwarded is rw param	accepts	rejects

  Legacy accepting the fourth row is the outlier, inconsistent with its own answer in the other three. RakuAST is the consistent one. Module fix: use the $pos = nqp::add_i($pos, 1) form the same file already uses elsewhere.

What has to change. Module side, one signature or one construction each. Do not relax rakudo for these. Two are already done: MergeOrderedSeqs (PR #1) is merged but unreleased, so blin still installs the broken 0.0.1; Net::IP::Parse (PR #8) is merged and released — 0.0.7 passes on main today, and blin only failed because its snapshot still tested 0.0.6.

G4. Compile-time call analysis reaches calls legacy skipped (2)

• App::APOTD — Calling mail-die(Str, Str, Str, Str) will never work with declared signature (Str $msg). The module really does pass four arguments to a one-argument sub. Legacy skipped the check because one of the arguments is an interpolated string; RakuAST checks it anyway. Minimal difference:

  sub main() {
      my $r = 1;
      mail-die "a", "b {$r}", "c", "d";        # legacy: silent. RakuAST: SORRY
      sub mail-die (Str $msg) { die $msg }
  }

  Module fix: concatenate the four pieces into one string.

• SQL::Abstract — Calling split-source(SQL::Abstract::Source:D) will never work with any of these multi signatures. No longer reproduces: the released 0.0.9 tarball compiles and tests clean on rakudo main as of 2026-08-20. Nothing to do.

G5. Perl 5 regexes (3)

• LIVR — a rx:P5 literal backs the url rule, so the validator rejects valid URLs.
• Router::Right — s:Perl5:g/^\s+|\s+$//; strings keep their surrounding whitespace, so " DELETE " never matches and %h<controller> comes back undefined.
• RegexUtils — EVAL "rx:Perl5$flags/$rg/" in CreatePerlRegex.

What has to change. Module side: use Raku regexes. Note that the decision has already been made to make :P5 a hard compile-time error (rakudo PR #6534, open), so these three will change from wrong answers to a clean X::Syntax::Regex::P5 once it lands. That is an improvement, not a regression, but it does mean these modules break loudly rather than quietly.

G6. RakuDoc strictness (4)

RakuAST's RakuDoc parser rejects markup that the legacy parser silently repaired.

• GD::Raw — one L<...> link is missing its closing >; the legacy parser consumed the next link's > to close it. Fix is one character, delivered as PR #8, still open.
• Selkie — RakuDoc markup code C missing endtag '>>' in lib/Selkie/App.rakumod. Blin tested 0.14.1; 0.15.1 has since been released and its App.rakumod no longer contains a C<< at all, so retest before touching this one.
• BioPerl6 — =begin end in lib/Bio/Role/Range.pm6; =begin may not be followed by the directive end.
• Image::Resize — DEP: it fails only on use GD::Raw.

What has to change. Fix the pod. Worth a decision on whether rakudo should be lenient here instead; the current behaviour is defensible and the modules are wrong, but four modules is a real cost for a documentation typo.

G7. The P5 dynamic-scope pragma (3)

P5tie, and Tie::Array / Tie::StdArray which depend on it. FETCH/STORE never fire, so tied variables read back as Any and pushes return the wrong thing.

Legacy only worked because World.nqp hardcodes a P5-name-prefix hack that applies the dynamic-scope pragma to importing files. With that hack removed, legacy fails 15 of 17 exactly like RakuAST — so this is module-side, not a RakuAST regression. The fix is for P5tie to request the pragma itself from sub EXPORT via $*LANG.set_pragma, before the unit declaration; that gets 17/17 on both frontends with a workaround-free compiler. Nothing has been delivered upstream.

G8. Rakudo bugs the module cannot work around (2) — RAKUDO

• Array::Circular — a trait whose argument is a declaration is silently not applied:

  use Array::Circular;
  my @b is circular(my $);   # legacy: SORRY, requires a compile-time constant
                             # RakuAST: trait never runs, @b is a plain Array

  Silently dropping a trait is worse than the error the module wants to raise.

• OpenMPT::Bindings — a NativeCall sub cannot be called from a CHECK phaser:

  use NativeCall;
  sub getpid() returns int32 is native { * }
  CHECK { say "CHECK: ", (try getpid()) // "FAILED: $!" }
  say "RUN:   ", getpid();
  # legacy   CHECK: 53854            RUN: 53854
  # RakuAST  CHECK: FAILED: lang-call cannot invoke object of type 'VMNull' ...   RUN: 53857

  The module probes openmpt_get_library_version() from a CHECK block to fail early with a friendly message, so under RakuAST every consumer sees the VMNull text instead. The same shape also masks a genuinely missing library: with the .dylib absent, legacy reports Cannot locate native library and RakuAST reports the VMNull invoke, which is how this looked in the blin log.

G9. Regex and Match behaviour (4)

• SION — RAKUDO, and the widest-reaching find in this run. m:p($pos) ignores its argument:

  my $b = 'abc123';
  say ($b ~~ m:p(2)/\w+/).?from;   # legacy 2, RakuAST Nil
  say ($b ~~ m:p(4)/\w+/).?from;   # legacy 4, RakuAST 0
  say ($b ~~ m:p(1)/\w+/).?from;   # legacy 1, RakuAST Nil

  RakuAST appears to fall back to the string's stored continuation position, which is why results alternate as each match leaves a new .to behind. Any hand-written parser that walks a string with m:p($pos) is affected; SION is just the one that showed up here.

• Regex::FuzzyToken — RAKUDO. Importing the same sub EXPORT module twice in one compilation unit breaks the second import's wrap:

  grammar A { use Regex::FuzzyToken; token TOP { $<a> = <fuzzy: <aaa bbb ccc>, /<.alpha>+/> <.digit>+ } }
  grammar B { use Regex::FuzzyToken; token TOP { $<b> = <fuzzy: <111 222 333>, /<.digit>+/> <.alpha>+ } }
  say A.parse("aab323892")<a>.Str;   # legacy "aaa", RakuAST Any

  With only grammar A present both frontends agree. Adding the second use breaks both. The module's &fuzzy.wrap(...) mixes a role into the match using two dynamic variables set inside a <?{ }> assertion; under RakuAST that communication is lost once the module is imported twice.

• Pod::Perl5 — $match.caps yields entries whose .value.made is Any, so the HTML and Markdown serializers emit empty strings. Not root-caused; same neighbourhood as the two above.

• SQL::Lexer — MODULE. grammar Basic:ver<0.2.2> is Lexer:ver<0.2.1..*> — legacy ignored a :ver adverb in is position, RakuAST reads it and reports Invalid type smiley ':ver'. The module has to drop the adverb from the is clause.

G10. Line, file, and frame attribution (4) — MODULE (fragile)

None of these are wrong answers; they are tests asserting a compiler-internal shape.

• Proc::Easier — walks Backtrace.new.list with hardcoded +1/+2 offsets after filtering SETTING frames. RakuAST's frame layout differs by one, so it reports the module's own file and a line one short.

• Test::Stream — expects a specific reported line (186) and gets another (76). Already addressed by PR #5 (open), which makes the expectations frontend-agnostic.

• Code::Coverage — the coverage line list drops line 9. Line 9 sits in an if branch that never executes on that run, so RakuAST's answer is arguably the correct one; legacy emitted a coverage annotation for it anyway. Real consequence: coverage numbers shift for everyone, so this is worth a conscious decision rather than only a test edit.

• Web::Cache — expected: (Block) / got: (Block+{Callable[Str]}). RakuAST attaches the Callable[T] mixin to a pointy block with a return constraint; legacy attached it to sub but not to ->. RakuAST is the consistent one:

  my $c = -> --> Str { "y" }; say $c.WHAT;   # legacy (Block), RakuAST (Block+{Callable[Str]})
  my $b = sub (--> Str) { "z" }; say $b.WHAT;  # both (Sub+{Callable[Str]})

G11. Version resolution (2)

• CSS::Specification — Could not find CSS::Grammar:ver<0.3.0..*>. Legacy ignored :ver, RakuAST honours it, and a Range is not a valid :ver argument. Fixed upstream by PR #6 (merged), which has not been released, so blin still installs 0.5.3 and still fails.
• CSS::Module — DEP: its only failing assertion is the CSS::Specification version check.

G12. Not yet root-caused (3)

Worth a look but nothing verified beyond the observable.

• JobQueue — JobQueue::EventSink.new.dispatch('x') should die with "must be implemented" and instead dies with "Stub code executed". Reduced versions of the role, the unit role form, precompilation, and composing-then-punning all behave identically on both frontends, so something about the installed dist is doing it.
• ScaleVec — Cannot pop from an empty Array[ScaleVec::Chord::Graph::PV] in the progression planner.
• Sustenance — Use of uninitialized value of type Rat in numeric context inside a WhateverCode. The upstream repo appears to be gone, so this may be unfixable module-side.

---

Re-verification against rakudo main, 2026-08-20

34 of the 53 were reinstalled and retested with zef against a rakudo built from main today (ac2ce1c1ce, which is origin/main plus two unrelated local commits), with RAKUDO_RAKUAST=1. This is what changed in the week since the blin run.

Module	Still fails	Note
Array::Circular	yes
Code::Coverage	yes
Env	yes
French	yes
GD::Raw	yes
Grammar::BNF	yes
IRC::Log::Colabti	yes
IRC::Log::Textual	yes
JSON::Fast	yes
JobQueue	yes
List::Allmax	yes
MergeOrderedSeqs	yes
Net::IP::Parse	0.0.6 yes, 0.0.7 no	the released fix postdates blin's snapshot
OpenMPT::Bindings	yes	root-caused today, see G8
Polyglot::Regexen	yes
Portuguese	yes
Proc::Easier	yes
RegexUtils	yes
Router::Right	yes
SION	yes	root-caused today, see G9
Selkie	yes (0.14.1)	0.15.1 released since; its App.rakumod no longer has the bad markup
SelectiveImporting	yes
Slang::Forgiven	yes
Slang::Mosdef	yes
T	yes
Test::Stream	yes	PR #5 not merged yet
Tie::Array	yes
Tie::StdArray	yes
auto-dynamic	yes
App::APOTD	yes
Data::Generators	no	0 failing test files; the 0.4169-vs-0.4 assertion was a flake
SQL::Abstract	no	released 0.0.9 compiles and tests clean
P5tie	not retested	already installed as a Tie::Array dependency, so zef skipped its tests
CSS::Specification	inconclusive	local run hit a dependency-resolution artifact, not the :ver failure

The 19 not retested here are the ones whose cause was already settled by the 2026-07-21 triage or by inspection: BioInfo, BioPerl6, CSS::Module, Control::Bail, Dawa, FindBin, FindBin::libs, Image::Resize, Inline::Perl5, LIVR, P5__DATA__, Pod::Perl5, Regex::FuzzyToken (reproduced separately, see G9), SQL::Lexer, ScaleVec, Slang::Predicate, Sustenance, Web::Cache, XHTML::Writer.

---

Suggested order of work

1. G9 SION's m:p($pos) — a silently wrong regex match position is the worst kind of bug in this list, and nothing in the ecosystem can work around it.
2. G8 Array::Circular's silently dropped trait — small, self-contained, one-line repro, and silence is the worst possible failure mode for a trait.
3. G9 Regex::FuzzyToken's double import — "importing a module twice breaks it" is a broad hazard that happens to have surfaced in one module.
4. G8 NativeCall in a CHECK phaser — one repro, and it degrades every module that probes its native library early.
5. G2 BioInfo — the only rakudo bug hiding in the slang group.
6. Chase the module PRs that are already sitting upstream — GD::Raw #8 needs a review, MergeOrderedSeqs #1 needs a release, Test::Stream #5 needs a review, and CSS::Specification #6 needs a release. Four rows clear with no new code.
7. G3 and G5 — straightforward module PRs, one signature or one regex each.
8. G6 — decide leniency-or-fix once, then act on all four.
9. G1 and G2 — the $*W and slang batch needs a policy decision before any more module PRs, since the port pattern is only settled for the $*W half.

Deliberately not on this board

• The MoarVM serialization fix this run was built with is still a local patch. It is the single largest lever in the ecosystem (29 modules in the 2026-07-21 run) and needs to become a PR.
• Net::DNS and Selkie::UI came back Flapper. Blin flappers on this box have historically been load-dependent rather than real.
• Foo:: Foo is a malformed ecosystem entry, not a module.

The P5tie, Tie::Array, Tie::Hash, Tie::StdArray and Tie::StdHash distros have been reduced to modules that pass testing, but can no longer be installed.

New release of Image::Resize with updated dependency on GD::Raw uploaded.

The Slang::Predicate distro is not a slang at all: it just exports a number of infixes.

The testing issue seen is caused by the fact that ⇒ (aka 21D2 RIGHTWARDS DOUBLE ARROW ⇒) is a Unicode version for => in the RakuAST grammar. And somehow the test just creates a Pair (instead of honoring the imported infix), and so the test stringifies that which causes "output carries a stray α\t prefix".

Haven't been able to golf the issue further.

I'm surprised P5__DATA__ is showing up here. Apparently I released a version once to CPAN, and honestly just forgot about it.

It's a slang indeed, so would need adaptation. Not sure yet whether I want to put the effort in, as it currently is not even listed on raku.land, so its visibility has been minimal anyway.

RakuAST needs a supported slang story, or each module needs rewriting against $*LANG.slangs and the RakuAST grammar's own extension points.

Slangify provides both legacy and RakuAST support, but of course slang modules need to find the right hooks into the Raku grammar.

Re Sustenance: the author has given up on it, I've looked at the code from the dist and decided I'm not going to spend any more time on it.

I've given up on porting P5__DATA__: it goes off into some weird situation where $/ becomes VMNull, and I don't want to spend more time on that.

About the SION case: the slightly adapted golf using .match shows that it is not an issue with match itself, but with the codegenning of m:p(2)/foo/:

my $b = 'abc123';
say ($b ~~ m:p(2)/\w+/).?from;   # legacy 2, RakuAST Nil
say ($b ~~ m:p(4)/\w+/).?from;   # legacy 4, RakuAST 0
say ($b ~~ m:p(1)/\w+/).?from;   # legacy 1, RakuAST Nil

say ($b.match(/\w+/, :2p)).?from;   # legacy 2, RakuAST 2
say ($b.match(/\w+/, :4p)).?from;   # legacy 4, RakuAST 4
say ($b.match(/\w+/, :1p)).?from;   # legacy 1, RakuAST 1

Indeed, for m:p(2)/o/ legacy codegens:

 - QAST::Stmts+{QAST::SpecialArg}(:named<p>) <wanted> 2
    │   │             - QAST::Want <wanted> :statement_id<2> 2
    │   │               - QAST::WVal(Int) 
    │   │               - Ii
    │   │               - QAST::IVal(2)  2

and RakuAST codegens:

- QAST::Op+{QAST::SpecialArg}(if :named<p>) 
    │               - QAST::Var(lexical $/) 
    │               - QAST::Op(callmethod to) 
    │                 - QAST::Var(lexical $/) 
    │               - QAST::IVal(0) 

???

re SION: proposal for fix in rakudo/rakudo#6585

Re Array::Circular, golfed the issue to:

$ raku -e 'multi sub trait_mod:<is>(Variable:D \v, :$foo! is raw) { dd; die "foo" }; my @a is foo'
sub trait_mod:<is>(Variable:D \v, :$foo! is raw)
===SORRY!===
foo
$ raku -e 'multi sub trait_mod:<is>(Variable:D \v, :$foo! is raw) { dd; die "foo" }; my @a is foo(my $)'
sub trait_mod:<is>(Variable:D \v, :$foo! is raw)

As far as I can see, passing any value to the trait_mod is ok, EXCEPT if a container is passed.

Consider me boggled: if the value that is being passed is a container, the subsequent die is ignored.

rakudo/rakudo#6586 fixes the Array::Circular issue

FWIW, I'm working on Grammar::BNF

rakudo/rakudo#6588 fixes JobQueue

Ok, the Grammar::BNF code is now up-to-date (as in using modern Raku): will look at what needs to be taken from the Raku grammar to make it work with RakuAST

@ugexe: the action part of porting Slang::BNF (https://github.com/raku-community-modules/Grammar-BNF/blob/main/lib/Slang/BNF.rakumod#L26-L46) is looking to me it's going to take me a long while to figure out. Any comments, suggestions, place of inspiration I could take a look at?

Hmmm... starting to think maybe Slang::BNF should be rewritten using EXPORTHOWlogic?

Pushed my work on Grammar::BNF (https://github.com/raku-community-modules/Grammar-BNF) in the hope for some (external) inspiration.

Created a PR for JSON::Fast

Created a PR for List::Allmax

I think we can consider Bioperl6 as abandoned: never got beyond a 0.0.1, author did not respond to the notification that it wouldn't get updated, documentation has "So, keep in mind that nothing is set in stone yet and things may change under your feet. Hic sunt dracones."

Looked at LIVR, decided to not spend time on trying to convert rx:P5/^(?:(?:https?)://(?:(?:(?:(?:(?:(?:[a-zA-Z0-9][-a-zA-Z0-9]*)?[a-zA-Z0-9])[.])*(?:[a-zA-Z][-a-zA-Z0-9]*[a-zA-Z0-9]|[a-zA-Z])[.]?)|(?:[0-9]+[.][0-9]+[.][0-9]+[.][0-9]+)))(?::(?:(?:[0-9]*)))?(?:/(?:(?:(?:(?:(?:(?:[a-zA-Z0-9\-_.!~*'():@&=+$,]+|(?:%[a-fA-F0-9][a-fA-F0-9]))*)(?:;(?:(?:[a-zA-Z0-9\-_.!~*'():@&=+$,]+|(?:%[a-fA-F0-9][a-fA-F0-9]))*))*)(?:/(?:(?:(?:[a-zA-Z0-9\-_.!~*'():@&=+$,]+|(?:%[a-fA-F0-9][a-fA-F0-9]))*)(?:;(?:(?:[a-zA-Z0-9\-_.!~*'():@&=+$,]+|(?:%[a-fA-F0-9][a-fA-F0-9]))*))*))*))(?:[?](?:(?:(?:[;/?:@&=+$,a-zA-Z0-9\-_.!~*'()]+|(?:%[a-fA-F0-9][a-fA-F0-9]))*)))?))?)$/ to a Raku regex, unless we haz an automatic tool for that. Do we?

Also: LIVR hasn't really been touched in 9 years, the author didn't bother to convert the regex to Raku (probably for the same reason I don't want to spend time on it), so I'd say we consider this module abandoned.

Created a PR for Router::Right. It fixes all of the :Perl5 related issues, but is still failing one test that doesn't appear to be related to using :Perl5

The FindBin issue is NOT a module issue: made an issue for it.

The RegexUtils distribution is not easy to fix, as it appears to specifically supports Perl5 regexen of some sort. Left an issue.

I think Control::Bail should be marked abandoned. The author did not react to the issue made about it no longer getting updated, and the last commit is 9 years go.

Created a PR for SQL::Lexer

The Slang::Predicate issue is due to the unicode variant that was added to RakuAST doesn't emulate how all other unicode alternatives were added. The unicode variant of := has a similar issue due to be implemented similarly. Either more thought needs to be put into a mechanism for providing unicode alternatives to things generally, or they should be implemented the same way the others are i.e. as overridable subs (although I don't think the binding one can work that way at all)

Regarding Grammar::BNF it looks like we needed some raku level machinery implemented still. I'll also link to the changes I have made to Grammar::BNF to get it to pass its tests, although I haven't reviewed the code.

rakudo/rakudo#6591
https://github.com/raku-community-modules/Grammar-BNF/commits/ugexe/rakuast-support/

Re: Grammar::BNF. Merged the branch, streamlined it (abstracted a lot into a local Slang module, which a lot of the code might find its way to a DSL module at some point in the future). Version 1.2 on its way to the ecosystem now.