What the Result column means, written down before the next rule asks

[wshallwshall]

The owner settled a question the census fix left implicit: the Result column says what happens when
a rule fires, never how wide the rule is.

So a row reading hard fail counts as enforced however narrow its scope, and only a qualified result
like hard fail, on those two counts as half-covered. PD-12 is the case that forced it — its gate
hard-fails over exactly one count on one page — and under this reading it is a test row, which is why
the census reads seven.

Nothing changes in the numbers

The page and the test already encode this reading. What was missing was the reason. Read the other
way, PD-12 would join PD-8 as half-covered and the census would read six — defensible, and not what
this page does.

Why it earns a paragraph rather than a commit message

The next narrowly scoped gate lands on the same fork, and the numbers alone don't say which way the
last one went. A reader recounting the table would have to guess whether PD-12 was classified
deliberately or by accident.

That is precisely the defect shape this page keeps finding in itself: a figure that is right, with
the derivation that makes it checkable left in somebody's head. So it's recorded in both places a
person could be standing when the question arises — beside the table on the rule sheet, and in the
docstring of the function that does the counting.

146 tests.

🤖 Generated with Claude Code