code wiki / _hdl_build / nx_wiredclaim_lib.nx
nx_wiredclaim_lib.nx
buildroot/runtime/_hdl_build/nx_wiredclaim_lib.nx
about
nx_wiredclaim_lib.nx -- IS AN ORGAN ACTUALLY CALLED, OR ONLY CLAIMED TO BE? NO main() ON PURPOSE.
THE CLASS THIS EXISTS FOR is the ecosystem's most-repeated defect: a capability that is BUILT, GATED,
and DESCRIBED AS WIRED, but has no production call site -- so it is, in effect, the baseline it was
meant to beat. Banked instances: public registration open while gate-proven ls_signup_allowed sat
unwired (1785445704); the pre-deploy gate computed, published and structurally unreachable from the
act it guards (1785453118); and F881, where nx_magicratchet is asserted "wired into /api/build" in
FOUR comments while grep finds ZERO call sites in the source AND ZERO in the deployed mgmt binary,
and a two-build experiment shows the enforcement does not fire (1785530277).
METHOD, AND WHY IT IS DELIBERATELY NOT AN LLM: the 2026 literature on code/comment inconsistency
reports that plain LLM judging yields unacceptably high false positives, because a model reads the
normal gap between high-level prose and detailed code as a defect (DocPrism). The sound technique is
CASCADE-style DOCUMENT TESTING: turn the claim into something executable and run it. This organ is the
STATIC half of that -- it decides, over a corpus grep, whether every mention of an organ is PROSE.
A claim with no call site is not proof of absence, it is the trigger to run the behavioural test.
INPUT is the output of `nx_shelltool grep <organ> <dir> nx out=<f>.out` -- lines of `path:lineno:text`.
Composing the grep rather than walking keeps this organ free of a corpus cap it would have to be
honest about, and reuses the ONE canonical search path.
u26a0DECLARED ENVELOPE (L011): line-shape heuristics, not a parser. A call written across a line break, or
dispatched through a runtime-assembled path (the way a builder may exec `_offc/<name>.elf`), reads as
NO CALL SITE. That is why the verdict for zero-calls-plus-claims is CLAIMED-UNWIRED -- a REVIEW
TRIGGER naming the behavioural test to run -- and never "proven absent".
license_tier: ORIGINAL No hw writes (Rule 26).
dependencies 1 imports · 1 importers
imports: nx_gatedry_lib.nx
imported by: nx_wiredclaim.nx
structs
| none |
consts
| 29 | const WC_WIRED: i64 = 0 // at least one non-comment, non-self mention: a real call site candidate |
| 30 | const WC_CLAIMED: i64 = 1 // zero call sites, but prose mentions it -- the F881 shape |
| 31 | const WC_NOEVIDENCE: i64 = 2 // no mentions at all |
| 32 | const WC_UNREADABLE: i64 = 3 |
| 34 | const WC_SLASH: i64 = 47 |
| 35 | const WC_SPACE: i64 = 32 |
| 36 | const WC_TAB: i64 = 9 |
| 37 | const WC_NL: i64 = 10 |
functions
| 40 | func wc_is_comment(buf: *u8, s: i64, e: i64) -> i64 called by 1: wc_scan |
| 55 | func wc_is_self(buf: *u8, s: i64, e: i64, organ: *u8) -> i64 |
| 73 | func wc_scan(buf: *u8, n: i64, organ: *u8, cnt: *i64) -> i64 |
| 105 | func wc_verdict(cnt: *i64) -> i64 called by 1: main |
| 112 | func wc_name(v: i64) -> *u8 called by 1: main |