Original Barrage 1.0.5: aimed target and score investigation

The target/score goal remains unproved. Four manually aimed large grenades were released, but the reviewed HUDs stayed at 000000 and the native result screen reported 0 points. Native Quit succeeded, with actual exit zero after 240.568 seconds. No further original or input was authorized or attempted.

This task sought a visible target destruction or positive score consequence of an ordinary manually aimed grenade, released settled evidence, and native Quit. The accepted 2026-10-18 run establishes crosshair movement, 36-to-30 ammunition consumption and visible impact only. Those facts are not new target/score proof.

Evidence: scratch/runs/20261114-barrage-target-score/.

Exact original recovery and controls

No local matching executable or ZIP was found among 818,844 examined files in /home/user, /tmp, /opt and /var/tmp; 23 private directory exclusions are recorded. The single directly authorized HTTPS GET chain succeeded, following one official redirect to master.dl.sourceforge.net, with inherited proxy settings and TLS certificate verification enabled. There was no retry or manual mirror/provider selection. The request had 15-second connection and 60-second total limits and a 1,455,807-byte size ceiling. Its actual process exited zero with both output streams at EOF.

ZIP path: test/binaries/media/barrage-original-20261114/barrage-1.0.5-win32.zip. It is exactly 1,455,807 bytes, SHA-256 bc4f43d1bc67c7e0db8168083d5b901a6123bbc5ccc9c9346852e68c3dcbd080. The original tree is alongside the ZIP, outside runs. Containment, member path and symlink checks, CRC and extracted-byte comparisons passed for all 45 files, 4,256,910 bytes. The original verifier confirmed inventory 25615b915a23144e4874ed96b670397f29eb0bd341288b79460b79bc91141db7 and authentic CRT 887eb5ce93edb7192ca3e9220f07f9ca0f94db02af5862ebcbdfcb852db99fd1. The verifier prepared a fresh ignored test/binaries/candidates/barrage-105 fixture and verified its exact 44 companion mappings. No config was supplied.

The bundled README, GPLv2 COPYING and windowed BAT were read before input. README controls are mouse movement to aim, left button for large grenades, middle button or Ctrl-left for small grenades, and right button or Space to reload. Escape quits the game without highscore checking. README lists target scores of 5 for a soldier, 20 for a jeep and 50 for a tank, with vehicle escape penalties and a 36-point reload cost. These are documented rules; they do not establish a new observed hit. The registered route retains barrage_105, C:\BARRAGE\barrage.exe, cwd C:\BARRAGE, and original -w.

Manual boundary and runtime status

The task owns tools/barrage-controller.py, tools/barrage-guard.py and their focused tests. The guard derives from the released Quake II transport, with Airstrike's pre-DOWN 1,750ms fallback margin and a 1,500ms pulse ceiling. Positive all-unsent recovery requires identity-matched cancelled statuses, independent release and a new reviewed scene/intent. Mixed, missing or UNKNOWN receipts close input. Actual process waits are distinct from physical UP ACKs. The adapter requires the submitted full scene to be reviewed before permanent binding, precision scenes under five seconds, menu scenes under thirty seconds, and an irreversible native Quit phase with a mechanical 180-second reserve. It contains no target recognition, tracking, aiming algorithm or score reader.

Preparation passed eight policy tests and twelve HTTP/terminal tests. Heavy work began only after recovered LTris's actual native/watchdog/outer-child/outer waits, both EOFs and heavy release. The first production run passed seven of nine checks: its two failures correctly rejected a fixture replacing mouse handlers after binding and a fixture continuing after Solitaire Escape changed menu ownership. Only these test setups changed: instrumentation moved before binding, and independent key fixtures use separate native identities. The next run passed those cases but failed its existing 250ms timing expectation with a 331ms physical hold for a requested 120ms pulse. The release child fired on schedule; the renderer received UP later. All failure evidence is retained. An unchanged full rerun passed all nine checks, including pointer/left/Ctrl-left, 1,500ms pulse, slow real fsync, late/dropped ACK, actual EOF waits, and parent death. The parent-death fallback physically released after 1,737ms.

flock /tmp/wine-build.lock bash tools/build.sh then passed with actual exit zero/both EOFs. Original/fixture verification and the tested closure passed again immediately before launch. No emulator, SDL, guard library, rendering or game-state fix was made. Shared helpers and foreign WIP were preserved.

Sole original and observations

The exact command is in original/recipe.json and original/launch.json: registered barrage_105, --no-build, default batch 1000, real ticks/calendar, time scale 1, UTC, 1024x768 host scene, no threads, and a fixed 900-second native guard. An independently armed observable watchdog had an absolute 920-second deadline. Original cwd/-w/configuration remained intact. There was no box, browser, deployment, alternate executable or second launch.

The tested source base was 6108273f95fcffc2a5d86956a3856114c5aec116. Executing module: 2,200,987 bytes, SHA-256 c94ccf1b179f18bd4b47867445afed8cef2f1393b19aabfd259c7f3b94a7be09. tested-relevant.patch records relevant dirty tracked hunks; separate patches record the four new controller/test files. Git plus patches is the source archive. No source/media/WASM bundle was copied into this run.

The initial submitted full scene was personally reviewed before the sole permanent owner binding. A reviewed left click selected Enter Shooting Range. range-ready.png shows score 000000, timer 2:55 and all 36 rounds. The first separate aim intent and a later shot intent expired before DOWN at approximately 5.984s and 6.587s; both remained locally NOT_SENT. A third stale shot intent was also refused. A small run-specific CLI invocation wrapper shortened command submission; it imports the unchanged tested adapter and accepts only explicitly supplied manual coordinates. It contains no image processing or move selection.

Four manually chosen large-grenade clicks were consumed against fresh reviewed scenes at desktop (229,231), (333,247), (212,359) and (412,323). Scene ages at consumption were 4.573s, 4.913s, 4.829s and 4.225s respectively. These images showed a tank or soldiers at the chosen points; no target prediction or tracking was used. The native crosshair changed to those points and back from red to green, and ammunition decreased 36 → 30 → 24 → 18 → 12. Separate released views show new impact craters and continuing native targets, but no unambiguous destroyed target and no positive score. They do not establish a collision defect or prove that no transient hit ever occurred between observations. Ammo depletion and craters are expressly excluded from success criteria.

The native round ended with “The time is up! Your result is 0 points.” shot-quit-menu.png preserves that result; the subsequent Topgunners values are the native score list, not a new earned score. No new round was started.

Native Quit, cleanup and evidence limits

The irreversible Quit phase began about 207.459 seconds after launch, before the fixed 720-second gameplay cutoff. Each further action had a fresh reviewed native scene: documented Escape at range/end, Escape at the result screen, Escape at Topgunners, then a left click on the personally visible main-menu Quit at (342,450). The final click was consumed and independently released; the original logged [Exit] code=0, unloaded its DLLs, and exited normally. There was no host-quit command, watchdog kill, rebound owner or replay.

All nine consumed gestures have independently prearmed UP, ACK, actual release child wait and both EOFs. Maximum UP ACK latency from consumption was 518.292ms. All precision inputs were under five seconds; all menu/Quit scenes were under thirty. Three stale intents sent no DOWN. No original UNKNOWN receipt occurred. Actual native, watchdog, supervisor-child and outer waits all exited zero with both EOFs. The watchdog retired on actual native wait without firing, the heavy lock was released, all recorded runtime PIDs disappeared and port 8587 closed.

Postflight verified zero drift across 323 executing source/module/helper pins, 93 media/fixture/CRT pins and 12 read-only input pins. Twenty selected manually reviewed PNGs have SHA-256, PNG chunk CRC and inflate verification and are named in result.json; unused/redundant captures were removed with hash records. Logs, actual waits, failed and successful fixture checks, patches and postflight remain under the run ID. No frame series, source archives or repeated media copies are retained. Free /home remained above 1.5GB.

Box ID: none; state NOT_CREATED; expiry and stop receipt: not applicable. There is no full-27, FPS, audio, browser or complete-game compatibility claim. Next: the aimed target-destruction/positive-score proof remains missing. A further original attempt requires a new assignment. This single task is stopped.