Quake II required-motion upper route, 0754
Onward-door/guard progress was not achieved. The first runtime qualification attempt failed in the run-scoped fixture preload before Chrome or the original entrypoint started. I treated that attempt as terminal: no runtime retry, original launch, game input or capture followed. This is an invocation blocker, not evidence of a defect in the published motion repair or a failed game route.
Evidence: scratch/runs/20261109-quake2-required-motion-upper-route-0754/.
Repair integration and offline checks
The prepared production closure integrates reader 5bc5f86f9 plus published
terminal repair 4ba9f3a1b08c11219c50023f4abffb3bf6c6f9cf, retaining caller
1473c5952a52d5d2b4546b93a0d462b64fe5c6cd and custody
4341b87723a7d56bc98223215cca4cdc878fcae2. CPU remains 2185596 bytes,
SHA-256 a320d4c85156644fd12b6134d66b11784f3100b9b135756206c087cdacd43295.
The preceding 722-provider recipe is preserved, adding one run-specific native
fixture generator (723 pins). Historical host/Worker providers and the shared
Q3 caller are read from their exact Git revisions without editing peer files.
The task's input wrapper requires requireMotion: true on every gameplay key
or combo. The actual repaired caller checks held
attackInput.playerMotion.status before dependent held screenshots, waits,
automatic capture and successful ACK. The fresh offline composition exercised
initial SNAPSHOT to held UNKNOWN, UNKNOWN with UP failure, owner retirement,
optional motion and valid required motion. All five cases passed, including
joined independent release and sticky first failure. These are authored
protocol/native-memory controls, not original input-consumption evidence.
Fresh passive-reader controls and both prepared persistent-PIPE controller cases passed. Compiled production and fixture preflights passed. Source-matched prior accepted tests, the published 21-case composition and locked-build proofs are referenced and hash-checked; they were not rerun merely to duplicate unchanged evidence. No production code changed in this task, so no build of different shared sources is presented as qualification of this frozen closure.
First runtime blocker
The native fixture was intended to exercise the actual production caller, reader, hold/capture/ACK and cleanup path in Chrome before any original game. Its generator derives an image in memory from the canonical EXE: preserved reader windows, a separate authored PE entry, native initialization/publication and an ordinary imported Sleep loop. Its initialization is fixture code only; it is not original movement, consumption or a host write into a live game. The generated image was not archived or substituted for original game media.
Two local prelaunch invocation errors were corrected before a remote supervisor was spawned: the fixture phase initially used the wrong transport-import parent, then launch composition attempted to make a local path relative to the remote root. Their failures and actual waits remain retained. One intervening barrier publication refused an existing receipt; that earlier receipt was preserved before publishing the corrected prelaunch generation. These are superseded invocation preparations, not native passes.
The first actual runtime attempt reached the retained kernel supervisor. The
fixture preload used process.cwd() as its repository root, but Boat's launcher
working directory was /home/user, whereas the authenticated staged tree was
/home/user/q2-required-0754. Node therefore tried to open
/home/user/test/binaries/candidates/quake-2-demo-installer/installed-extracted/Install/Data/quake2.exe
and exited with ENOENT. The command log retains the complete path and stack.
The actual caller latched owned launch transport stdio EOF: 1 at
2026-11-09T08:01:00.627Z and joined cleanup. No Chrome spawn, native keyboard
preflight, owning Worker, renderer, original entry, game command or PNG was
reached. Native qualification did not pass; its later identity/flag/CPU/DLL gates
were not observed at runtime. No caps or gates were weakened to proceed.
The concrete next change would be an explicit staged-root derivation in the fixture invocation, with a working-directory-independent offline preflight. That change and any new qualification/runtime require a successor task; this attempt was not retried.
Evidence and joined shutdown
One box: bx_dwvy9e4p, immutable UUID 366ee613-c19a-4a6a-8121-7fe27773ef6d. The independent watchdog was observable before create, with 25-minute deadline 08:23:35.511Z. Create returned TTL1491 and initial expiry/archiveAfter 2026-11-09T08:23:35.616Z. The exact prior Redline0724 ARCHIVED PHYSICAL_DRAINED release was checked before create.
Custody retained 7 owned / 7 actual reaps, complete true, command-log EOF true, and no remaining/unreaped/unresolved entries. The remote supervisor and caller parent actually waited status1 with separate stdout/stderr EOF. Local caller children actually waited statuses1 and0, with both EOFs; the retained controller actually waited status1. Physical absence is additional census evidence, not a substitute for those waits.
All 14 remote artifact rows, 12 unique contents, were retrieved and independently SHA/size verified before archive. Both remote gzip streams passed GNU gzip CRC and independent decompression/hash. The browser profile was absent, profile writers were empty, runtime lock was free and input writer lock absent. No browser profile ports existed; no port-close or guest key-release event is invented for an uncreated browser. Kernel all-writers/nonleader coverage remains UNKNOWN independently of this census.
Final state ARCHIVED, stopped 2026-11-09T08:02:32.302Z. Stop receipt stop_618fb446736111cd, completed 08:02:32.319Z, forced false, error null. Final archiveAfter is null; expiresAt was not returned. The independent watchdog and lifecycle supervisor actually waited0 with separate EOF and released the local runtime lock. PHYSICAL_DRAINED was posted promptly at 08:02:48Z.
Own outer logs are gzip drains from first byte. Retained caller/fixture-internal formats remain explicit, including the existing plain supervisor streams. The initial local board watcher exceeded its separate 600-second test watchdog: its controller was actually waited -9, but the tail-child wait receipt is missing and both watcher-internal gzip files lack final trailers. Those two streams are retained as truncated evidence; tail reaping and their complete EOF are UNKNOWN. This does not alter the separately verified runtime/caller/watchdog joins above. Recovery uses exact Git revisions, a small relevant-source patch and a small run-input patch, plus existing media/module path and SHA references. No source tree, media or module archive was created. Free /home stayed above 1.5 GB.
Last original remains HP94 then HP0 LOAD GAME, damage cause UNKNOWN. Pane traversal, ladder and move-up consumption remain unestablished. No blind descent, crate-side turn or assumed ladder input was attempted. Received-frame motion SNAPSHOT does not prove contact, consumption, freshness between samples or damage cause; raw flag names are not inferred. Native lifetime remains REVOKED; all-writers and lifetime edict ABA remain UNKNOWN. No peer Q2/Redline/NFS/shared host/native-custody edit, agent, worker or public deployment occurred; Heroes II and Claude were untouched. Single-task STOP.