#3013's merge of same-named declarations fixed a false **negative** and
introduced a false **positive**.
`ModelSelectorTab` declares its own two-parameter
`resolveEffectiveExecutor(task, settings)` — a pass-through with nothing
lane-related — while an unrelated exported function of the same name in
`effective-model-resolution.ts` takes `columnFlags`. Both local calls
were reported unwired against a signature they have never had.
Only **exported** declarations enter the accepting map, so the rule is
exact: if the calling file declares the name itself and the map entry
came from a different file, the call resolves to the local declaration
and is not a lane call here.
## The version I did not ship
My first attempt re-ran the detector over the single calling file and
used that result. It scored *better* on this tree — **19 → 16** instead
of 19 → 17, also clearing `bucket-mapping.ts` — and I threw it away.
A single-file pass cannot resolve an **imported** options interface. A
locally-declared function with an imported context type would quietly
stop being lane-accepting, and every call to it would stop being
checked. That is a false negative, which is the one failure a ratchet
must not have; the better-looking number came from the gate seeing less.
The global pass still does all type resolution here — only the *choice*
of declaration is local.
## Measured
| | |
|---|---|
| new tests | 3 |
| against the old census | **1 of 3 fails** — the positive |
| baseline | **19 → 17**, exactly the two `ModelSelectorTab` sites |
Both negatives pass either way and they are the ones that matter: a file
declaring its **own** exported lane function is not shadowed by itself,
and a file declaring nothing is judged normally. Shadowing must not
become a way to disappear a genuine unwired call.
## Still flagged, honestly
`bucket-mapping.ts:75` stays in the baseline. `bucketForTask(task:
TaskItem)` is only lane-accepting because `TaskItem` *declares*
`columnFlags` — the lane data rides on the domain object, so passing
`task` forwards it inherently. That is a different limitation
(options-bag vs domain-entity parameters) and I have not tried to fix it
here; it accounts for 2 of the remaining 17 along with
`otherBucketSecondaryLabel`.
## Verification
`node --test scripts/__tests__/check-lane-wiring.test.mjs` **19 passed**
· `pnpm test:gate` 13 + 161 + 487 + 71 · lint · lifecycle census
`--strict` · lane-wiring · fnxc-dates (TZ=UTC) · changesets — green.
5897d87e95