code wiki / _hdl_build / nx_memvel_gate.nx
nx_memvel_gate.nx
buildroot/runtime/_hdl_build/nx_memvel_gate.nx
about
nx_memvel_gate.nx -- referee for the SUSTAINED-GROWER predicate (rm_sustained_count), 2026-08-14.
WHY THIS GATE EXISTS: nx_memvel had no referee at all, and the column its own header calls decisive
(grew_in = k/N) was printed PER ROW and never totalled, so no caller could branch on it. Every
consumer therefore fell back to nx_resmon's leak census, which is a SNAPSHOT relation
(VmSize==VmPeak) -- a LEVEL, and a level cannot express a leak. Measured on this box the same day:
17 leak suspects against a red threshold of 6, so nx_resmon was RED PERMANENTLY, while the true
sustained-grower count was 0 in back-to-back runs.
THE TOOTH THAT MATTERS IS T7. It asserts the two predicates DISAGREE on the shape that caused this
work: the screen calls nx_hub_gw a candidate, the velocity predicate clears it. A change that reduced
rm_sustained_count to echoing the screen would pass every other tooth here and fail only T7 -- and
that echo is precisely the defect this work removed.
Every input is synthetic: no /proc, no fixture on disk, no added load, deterministic by construction.
license_tier: ORIGINAL expect_exit: 0
dependencies 4 imports · 0 importers
imports: nx_syscalls.nxnx_gate_verdict.nxnx_resmon_lib.nxnx_ctxtop_lib.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
| 22 | func main() -> i64 |