Repository navigation
Description
Two landing figures are read out of the interpreter repository by hand
conformancefigure already reads three facts out of go-embedded-ruby/ruby and checks them against a tree's claims — the passing total, the Ruby version those guards resolve at, and which mspec helpers the shim stubs. It is wired in both consumers: this repo's docs.yml, and the landing's deploy-pages.yml:147 (pinned @v0.8.0).
Two more figures on the landing have the same shape and no gate at all.
1. 178 modules across 176 orgs — source: the ruby repo's go.mod
hugo.toml's Phase 8 blurb states:
178 modules across 176 orgs are now bound into rbgo as native modules
Measured against go.mod on main today: 178 modules, 176 orgs, with go-ruby-widgets contributing three (widgets, tui, mvvm), so modules = orgs + 2. Correct right now. It is a hand-copy in another repository of a number that moves whenever a module is bound or dropped.
⛔ The ruby repo's README carried the same pair and documented how to get it as a shell pipeline:
(count: `grep -oE 'github\.com/go-ruby-[a-z0-9-]*/[a-z0-9-]*' go.mod | sort -u | wc -l`; `go-ruby-widgets` contributes three)
That pipeline cannot tell a changed count from a broken pattern — and it was structurally blind to the defect that was actually there: six names (commonmark, kramdown, liquid, mustache, nokogiri, rouge) sat in the README's go-ruby-* family list claiming github.com/go-ruby-<name>/<name> addresses that do not exist, while go.mod has them as github.com/go-<name>/<name>. The counts were right the whole time; only the list was wrong, so anyone checking the figure was reassured. go-embedded-ruby/ruby#954 fixes the list and gates both counts and the list membership, both directions, inside that repo.
The landing repeats the counts and no gate here reads go.mod.
2. 2,202 of 2,206 spec files produce a result, the other 4 recorded as FILEFAIL — source: the ruby repo's BASELINE
Also in the Phase 8 blurb. Measured against scripts/conformance/rubyspec/BASELINE on main today: OK=2202 FILEFAIL=4 total=2206. Correct right now.
These are a different claim kind from the one conformancefigure checks: it sums the passing examples, and the ruby repo's own TestReadmeConformanceCountMatchesTheBaseline skips FILEFAIL rows on purpose — "a FILEFAIL row carries no count". Measured, not assumed: setting the ruby README's table to 2,203 of 2,206 leaves that test green. Two numbers on one line, one guarded and one not. go-embedded-ruby/ruby#955 gates the file counts inside that repo.
⛔ And a counting trap worth carrying into any implementation here: grep -c FILEFAIL over BASELINE answers 5 where there are 4 records. The fifth hit is the header line that documents the format:
# One record per spec file: OK<TAB>path<TAB>passing, or FILEFAIL<TAB>path.
Key on the record shape — a tab-separated first field on a non-comment line — never on the word.
Suggested shape
Add both as further facts to conformancefigure, rather than a new tool: it already fetches from the interpreter repository, already walks a tree for claims, and is already wired in the two places that would need it. Then bump the pin in deploy-pages.yml and in docs.yml.
The existing lane's own comment states the rule that makes the pin bump part of the work, not a follow-up:
Keep this pin moving forward — a gate pinned to a version that cannot see a line is indistinguishable from a page with nothing wrong.
⛔ Worth checking as part of this: the landing pins @v0.8.0 while docs.yml runs the tool from the local tree. If the tool has gained a claim kind since v0.8.0, the landing is already behind by exactly the amount that comment warns about. I did not verify which versions those two are at — only that one is a pin and the other is not.
Why both are filed and not fixed here
Refs go-embedded-ruby/ruby#954, go-embedded-ruby/ruby#955 — both stay open on their own merit; they gate these two claim kinds inside the interpreter repository, which is what makes the landing's hand-copies the remaining gap rather than the whole problem.
Both figures are correct today, verified rather than assumed, so nothing on the page is wrong at the moment. What is missing is the derivation, and adding it spans three repositories — the tool, its release, and two pin bumps — which is more than belongs in a drive-by. Filed so the next corpus sweep or bound-module change is caught by something other than a reader who happens to count.
🤖 Generated with Claude Code
Prior art: priorart "conformancefigure derives the bound module count and the spec-file counts from the interpreter repo" at 2026-11-16 04:10 CET reported prior art; checked: The 62 hits are conformancefigure's own history, one PR per fact it learned to read: #41 the version, #47 the stubbed helpers, #40 .toml prose, #44 the 4.0.9 move. None reads go.mod, and none reads BASELINE's RECORD counts -- the tool sums passing examples, a different claim from how many spec FILES there are. #26 is the scientific-stack pages, unrelated. This asks for a fourth and fifth fact in the shape those PRs established.
Activity