nx_virtio_spec_oracle_gate.nx
buildroot/runtime/nx_virtio_spec_oracle_gate.nx
about
nx_virtio_spec_oracle_gate.nx -- THE ORACLE ROW FOR `smallos`: corroborate OUR declared virtio-MMIO
register map against the PUBLISHED OASIS SPECIFICATION -- a document WE DID NOT WRITE.
WHY THIS EXISTS (and why it is not just another self-test). nx_evidence_verdict.nx clause 5
TRIANGULATION requires >=2 distinct method classes with >=1 MECHANISTIC, and defines the ORACLE class
as agreement with an INDEPENDENT implementation or external ground truth -- explicitly "something WE
DID NOT WRITE". Every other gate in this lane is MECH and SELF-AUTHORED: I wrote the questions AND the
answers. MEASURED 2026-08-15: smallos therefore carried exactly ONE method class, min_classes=2 binds,
and the domain COULD NOT REACH PROVEN no matter how many of its own gates went green.
******A DOMAIN THAT CAN ONLY EVER CITE ITSELF HAS A CEILING NO AMOUNT OF GREEN CAN RAISE.
WHAT I REFUSED, recorded so nobody quietly does it: this estate has TWO rv64 simulators
(rv64im_min_sim, nx_rv64_fast) and a differential between them LOOKS like clause 5's
"second-implementation cross-check". It does not qualify -- BOTH ARE OURS. **A CROSS-CHECK BETWEEN TWO
THINGS YOU WROTE MEASURES YOUR CONSISTENCY, NOT YOUR CORRECTNESS.**
THE SUBJECT: knowledge/specs/EXTERNAL-virtio-v1.2-oasis.html, fetched 2026-08-15 over the sovereign
HTTPS stack from docs.oasis-open.org (2,678,562 B, byte-verified on copy). It is the ONLY external
virtio ground truth in the corpus -- before the fetch there was NONE (`virtio-v1` across knowledge/ =
0 matches, corpus_complete=1 over 176,785 files).
THE CHECK, and its CEILING STATED PLAINLY: for every register offset and identity constant OUR spec
DECLARES, assert the literal appears in the OASIS document. This is PRESENCE-IN-THE-NORMATIVE-TEXT,
which is WEAKER than parsing the register table: a literal could in principle occur in unrelated prose.
It is corroboration, NOT derivation, and it must never be quoted as more. What makes it worth having
anyway is the NEG-CONTROL below: a FABRICATED offset must be ABSENT, which proves the check can
discriminate rather than saying yes to everything. *A PRESENCE TEST WITHOUT AN ABSENCE CONTROL IS A
FUNCTION THAT RETURNS TRUE.
DATA-DRIVEN (rule 11): the constants are READ OUT OF OUR SPEC at runtime, never retyped here, so the
gate cannot drift from the thing it certifies. Adding a register to the spec extends this gate for free.
exit: gv_verdict (0 GREEN / 1 RED / 3 SKIP) license_tier: ORIGINAL. Read-only. No hw writes (Rule 26).
dependencies 2 imports · 0 importers
imports: nx_syscalls.nxnx_gate_verdict.nx
imported by: nobody (leaf or entry point)
call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown
structs
| none |
consts
| 36 | const VO_SPEC_EXT: *u8 = "knowledge/specs/EXTERNAL-virtio-v1.2-oasis.html" |
| 37 | const VO_SPEC_OURS: *u8 = "knowledge/specs/virtio_blk_sector_virt.spec" |
| 40 | const VO_BOGUS: *u8 = "0x0AB7" |
| 41 | const VO_NL: i64 = 10 |
| 42 | const VO_SP: i64 = 32 |
functions
| 44 | func vo_p(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } sys_write(1, s, n); return 0 } |
| 45 | func vo_n(v: i64) -> i64 |
| 58 | func vo_len(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } return n } |
| 63 | func vo_lc(c: i64) -> i64 { if c >= 65 { if c <= 90 { return c + 32 } } return c } called by 1: vo_has |
| 65 | func vo_has(buf: *u8, n: i64, lit: *u8) -> i64 |
| 87 | func vo_is_sovereign(buf: *u8, ls: i64, le: i64) -> i64 called by 1: main |
| 96 | func vo_starts(buf: *u8, ls: i64, le: i64, pfx: *u8) -> i64 |
| 106 | func vo_lastfield_tail(buf: *u8, ls: i64, le: i64, out: *u8) -> i64 called by 1: main |
| 121 | func main() -> i64 |