[tannevaled]

conformance.md told readers the shim stubs "a few MSpec matchers (complain, output, ruby_exe, …)". Read against spec_helper.rb:

	helpers
stubbed (raise SpecSkip)	ruby_exe (:882), ruby_cmd (:883), output_to_fd (:280)
implemented	complain (:249), output (:279), be_computed_by (:199)

Two of the three named are implemented, and ruby_cmd — stubbed one line below the ruby_exe that was named — appeared nowhere on the site.

The error runs in the direction that hides work. An example using a genuinely stubbed helper skips; one using complain runs, and when the behaviour is missing it fails. A reader sent to look for it among the skips does not find it there — ruby#909 was nearly missed exactly so. ruby#911 corrected the same sentence in the ruby repository and this page kept the wrong one, which is why this PR adds a check rather than a third corrected paragraph.

The check

tools/conformancefigure already fetches spec_helper.rb for CUR_VERSION. It now also reads every def … raise SpecSkip out of it and judges the pages against that set in both directions — a helper called stubbed that is not, and a stubbed helper the site names nowhere.

Three decisions worth the review:

• It is a vocabulary with its own self-test, like the total's and the version's. "Stub" is an ordinary word on this site: webassembly.md has a graceful stub row listing sqlite3, sequel, activerecord, and the paragraph recording this very correction says "an example using a stubbed helper skips; one using complain runs". Matched on the word alone, both read as claims — measured: the first run of this check failed conformance.md on its own explanation. It is now three phrasings, with those two lines as negative cases.
• The unreadable-shim refusal is held until the findings are printed. The file's own ⛔ note says a structural complaint must never arrive before a concrete disagreement. Written the obvious way (refuse early, like the baseline and the version), a missing stub list swallowed every stale-total finding in the suite — eight tests went from naming a stale figure to naming an unreadable shim.
• skip is deliberately outside the set, and the comment says why: it raises SkippedSpecError because the spec asked to be skipped. That is mspec behaviour rbgo implements, not a hole in the shim. A census that does not state its population gets read as a total.

One hole the walk cannot close by itself: the completeness check only applies where the list is made at all, so deleting the enumeration outright would leave nothing to be incomplete about. TestTheSiteNamesEveryStubbedHelper walks the real content/en and requires every stubbed helper to be named, independently of any figure on the page, so a figure refresh never makes it noisy.

Also

The unswept ruby/spec trees are library/, optional/, command_line/ and security/. command_line/ — 32 spec files — was missing from that list here, on the landing page, and in the ruby README. Corrected on this page; the landing follows once this tool is tagged.

Verification

$ go test ./tools/conformancefigure/
ok

$ go run ./tools/conformancefigure --root content/en --baseline … --runtime … --shim …
5 total claim(s), 2 version claim(s) and 6 stubbed-helper mention(s) checked
against the gate's 23672 at Ruby 4.0.9 (derived from RUBY_VERSION);
the shim stubs output_to_fd, ruby_cmd, ruby_exe

run against the real BASELINE, kernel_introspect.go and spec_helper.rb from a current rbgo checkout.

Refs ruby#911.
Refs go-embedded-ruby/ruby#909 — it stays open; this page only stops sending its
reader to look for that failure among the skips.

🤖 Generated with Claude Code