[tannevaled]

What

CRuby's pathname.rb writes alias / +. rbgo's prelude wrote a forwarding def
instead:

def /(other)
  self + other
end

because the parser refused the alias: after alias, a / opened a regexp
rather than naming the division method. Fixed in go-ruby-parser v0.11.1
(go-ruby-parser/parser#47), which this bumps to.

Measured — and it was not only a reflection difference

	ruby 4.0.5	before	after
Pathname.instance_method(:/).original_name	:+	:/	:+
instance_method(:/) == instance_method(:+)	true	false	true
with #+ redefined, Pathname.new("/a") / "b"	#<Pathname:/a/b>	"HIJACKED"	#<Pathname:/a/b>

The third row is behaviour: an alias keeps the body it was made from, while
a forwarding def follows whatever #+ later becomes. That is the row that makes
this more than a tidy-up, and it is in the test.

Conformance

library/pathname/divide_spec.rb goes 0 → 1, which is MRI's own ceiling for
it. The file holds one example, and its title is "is an alias of Pathname#+" —
the corpus naming exactly this. The pathname library specs go 38 → 39 with
no file down.

That suite lives under library/, which the ratchet does not sweep (only
language/ and core/), so the gain is real but ungated — and nothing in
language/ or core/ moves.

Scope

A scan for the workaround's shape — an operator-named def whose body just
forwards to another operator — found one site in the prelude, this one.

How it was found

A test written to pin the parser README's claims swept CI's stdlib. CI's runner
carries Ruby 3.2, not the 4.0.5 the project targets, so it swept 66 files
instead of 47 and refused pathname.rb on alias / + — a construct valid in
every Ruby that no 4.0.5 top-level file happens to use. Fixing the lexer
upstream is what made this prelude change possible.

Green

go test ./... -count=1 -timeout=90m exit 0 (10 packages), gofmt clean,
go vet clean, GOOS=wasip1 and GOOS=js clean. The frozen prelude was
regenerated, as its drift test requires.

🤖 Generated with Claude Code