code wiki / _hdl_build / nx_gk_fixture_gate.nx
nx_gk_fixture_gate.nx
buildroot/runtime/_hdl_build/nx_gk_fixture_gate.nx
about
nx_gk_fixture_gate.nx -- THE REFEREE FOR gk_fixture_dir, the per-RUN gate fixture directory (2026-09-03).
THE DEFECT IT GUARDS, measured the day it shipped. Two runs of nx_bck_elide_gate launched in the same
second against two DIFFERENT compilers BOTH returned verdict=SKIP with "the named compiler built none of
the fixtures". Neither run was faulty; the pass under test was working (its own log showed
BCK-ELIDE sites=4 elided=2 throughout). They had overwritten each other's fixtures in the single
directory both had hardcoded. Run sequentially, the same gate on the same box is 13/13 GREEN.
The estate already enforced the PER-GATE half of this law (gate scratch in /tmp/<gate>/, never shared
with a production beat). The PER-RUN half did not exist, and its absence fails in the ABSTAINING
direction -- a false SKIP acquits nothing and alarms nobody, it just deletes the evidence.
THE ANTI-VACUITY PROBLEM THIS GATE HAD TO SOLVE. A tooth saying "gk_fixture_dir returned a path" passes
for a function that returns a CONSTANT, which is precisely the broken shape. So the load-bearing tooth is
that TWO calls differ, and it is paired with a NEGATIVE CONTROL that builds the OLD constant-path shape by
hand and asserts it COLLIDES. Without that control a reader cannot tell a working discriminator from a
test that could never fail.
exit: 0 GREEN . 1 RED license_tier: ORIGINAL No hw writes.
dependencies 3 imports · 0 importers
imports: nx_syscalls.nxnx_gate_verdict.nxnx_gatekit_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
| 24 | const GF_PATH: i64 = 1024 |
| 25 | const GF_SUBJECT: *u8 = "nx_gk_fixture_gate_subject" |
functions
| 27 | func gf_streq(a: *u8, b: *u8) -> i64 called by 1: main |
| 34 | func main(argc: i64, argv: *i64) -> i64 |