Verify every claim on the Nil Python page by compiling it

• Type: docs (Track D — prose only, no compiler or lib/** changes)
• Status: done found false in a single pass.
• Owner: frank2-D

Why this page specifically

Nil Python moved faster than its documentation, and the page has measurably drifted. Found and fixed on 2026-08-09, both by testing rather than reading:

• "Nil Python is limited to four parameters per function" — false. Five- and eight-parameter functions build and run against the pinned compiler.
• "there is no operator-overload protocol yet (__eq__, __len__, __iter__, __getitem__, __call__, __enter__/__exit__ are not hooked in)" — mostly false. All of those dispatch except __iter__, which is a real gap; with works on user classes.

Two stale claims in one sitting on one page is a pattern, not bad luck. The rest of the page was written at the same time and has not been checked.

The one already known to be unverified

"It requires explicit type annotations on function parameters and return types, while local variables are automatically inferred."

Plausible, matches the design intent, never tested. Start here.

Everything else worth a compile

Walk the page top to bottom and run each behavioural claim:

• Type inference — numeric widening to a float slot; retroactive variant promotion on incompatible rebinds; static rejection for records/classes/dyn arrays. Three claims, three programs.
• C interop — return-lifting of trailing T** out-params, automatic string marshalling both directions, #define integer macros as constants. (The sqlite CRUD example on the page was verified working on 2026-08-09.)
• Classes — @property/@x.setter, module-level @dataclass, super().__init__ as a statement-only form, single inheritance refused with a diagnostic. (Single inheritance and the diagnostic were verified.)
• Syntax rules — tabs/spaces mixing is a compile error; indentation suspended inside brackets.

Rule

Track D's gate applies: verify by compiling, never by reading the frontend source — the source shows intent, the binary shows behaviour, and the two already disagreed twice here. Any gap found is filed for Track N, not fixed under D.

Acceptance

Every behavioural claim on docs/targets/nil-python.md either verified against the pinned compiler or corrected, with the gaps that turn up filed as N tickets.

Log

• 2026-08-09 — filed. Two false claims corrected the same day (docs(D): Nil Python targets CPython behaviour; correct the dunder claims); this covers the rest of the page.
• 2026-08-19 — resolved, commit 66bebdce5.