nx_neverbrick_gate.nx
buildroot/runtime/nx_neverbrick_gate.nx
about
nx_neverbrick_gate.nx -- proves the Rule-26 reversibility classifier actually discriminates, and that
it fails CLOSED. Rule 26 says never-brick must be proven mechanically, never asserted; this is the
proof, and it calls the SHIPPING functions in nx_mcu_brick.nx so it cannot drift from what a real
write chokepoint runs.
NON-VACUITY IS THE POINT. Two failure modes would make this gate a liar, so both have controls:
- a classifier that returns RED for EVERYTHING passes every safety test and blocks the product.
T4 and T8 are the positive controls: the real esp32-ai flash workload MUST come back GREEN.
- a classifier that returns GREEN for everything would sail through the positive tests. T1/T2 are
the negative controls: undeclared and null namespaces MUST come back RED.
- T7 is the neg-control on T6's READER: if nb_efuse_kills_recovery returned 1 unconditionally, T6
would pass while measuring nothing. A check and its own neg-control failing together means the
instrument is broken.
T5 is the keystone assertion: the same flash write that is GREEN with the ROM download path intact
MUST be RED once DIS_DOWNLOAD_MODE is burned. Flash reversibility is conditional, not intrinsic.
license_tier: ORIGINAL No hw writes (Rule 26). expect_exit: 0
dependencies 2 imports · 0 importers
imports: nx_syscalls.nxnx_mcu_brick.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
| none |
functions
| 21 | func nbg_len(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } return n } called by 1: nbg_p |
| 22 | func nbg_p(s: *u8) -> i64 { sys_write(1, s, nbg_len(s)); return 0 } |
| 23 | func nbg_pn(v: i64) -> i64 |
| 36 | func nbg_ck(name: *u8, c: i64) -> i64 |
| 42 | func main() -> i64 |