#243: match the pair, and let the corpus decide when the scan is finished

[Polichinel]

Closes #243. Gate 2 of epic #241.

The parse is gone

The scan no longer reads a line, captures what it thinks the value is, and compares. It knows both halves before it starts — the declared names and the declared values — so it matches NAME=VALUE directly, with findall rather than one match per line.

Every blindness C-57 records came from that parse: the swallowed comment, the missing terminator, the six-character lookbehind, the second assignment on a table row. None is expressible now.

The halves need not correspond — APPWRITE_X=<value declared for APPWRITE_Y> is the ordinary copy-paste slip, and it names the coordinate that declares the value.

The stopping rule is a test, not a paragraph

It enumerates every line in this repository's own tracked markdown that assigns a declared coordinate, and fails if the matcher cannot read one. A form no document here uses is not a gap. A new form is taught in the same change as the document introducing it.

That is what stops a fifth widening, which is what this story was for.

The four mutations routed from #242 are closed

By one richer fixture rather than four tests — two copies, two sections, one short, one in markdown. Each element defeats a specific narrowing:

mutation	before	after
scanned_sections narrowed to target	survives	caught
length floor reinstated	survives	caught
SyntaxError swallowed	survives	caught
break after the first finding	survives	caught

Also: the secret exemption is gone — an inline name literal sitting under a comment saying the scope came from the declared partition and not from an inline list.

An exclusion list built, then deleted

The plan called for excluding values spelled like this repo's own code. Measured against what the AST branch actually sees — non-docstring string constants — the collision does not exist: the only registry values appearing that way are the two contract rows, and contract is MIRRORED, never in the scanned set.

I had built a narrowing of a security scan for a problem this branch doesn't have. Deleted. The latent version is C-97's.

The line budget, honestly

tests/*.py: 9,200 → 9,308. Still the wrong direction. Detail and what I propose in the story comment — I am not restating the criterion to make it pass.

ruff check .    # clean
pytest -q       # 415 passed, 1 skipped, 40 xfailed

🤖 Generated with Claude Code