StarCraft: manually completed supply depot and native Quit

The current registered starcraft_shareware completed a newly commanded Terran Supply Depot in one live mission. A personally selected SCV received native construction-menu and placement clicks. The finished building was then selected: its panel showed Terran Supply Depot, 500/500, Supplies Provided 8, Total Supplies 50. The initial cap was 42; minerals changed from 250 to 150. The scaffold, finished structure, and finished building's identity panel were each personally reviewed in complete scenes. This establishes completed construction; the mission's “Destroy the rebel base” objective was not completed and no enemy damage or mission victory is claimed.

Evidence: scratch/runs/20261114-starcraft-unit-mission-progress/result.json. The candidate is the existing starcraft-shareware; this is a distinct RTS gameplay result, not a new title or expanded 27-app/3D coverage claim.

Current run and manual decisions

The heavy turn followed actual Minesweeper and fresh GL-startup terminal waits, both EOFs, absent PIDs, and heavy-lock release. Their records are preserved in predecessor-release.json; the earlier closed GL run was never reopened.

One original used test/run.js --app=starcraft_shareware, cooperative execution, default 1000-block batches, real ticks, an 800×600 complete canvas, and fixed 1200/1220-second runtime/watchdog deadlines. The existing registered ophelia terran1 arguments, startup registry and game media were unchanged. The frozen scripted gameplay runner was neither invoked nor adopted. No new cheat, map, AI, speed, guest-state write, hidden game-state inspection, replay, restart, right click, or scripted action list was used.

The first reviewed complete scene was the native STRONGARM mission title card. It established one permanent PID/start-token/window binding, truthfully marked bootstrap. A subsequent scene had naturally advanced to the briefing; its visible Start button was separately reviewed and clicked. The Tips modal was separately dismissed. An unobstructed mission scene was then admitted.

Selection, command-menu choice, placement-command state, placement target, and results were reviewed separately. The SCV's selected portrait/green ring and 60/60 health established actual selection. The native “Select Location” text and Cancel-only command card established the active placement command. A left click on personally chosen clear dirt commanded construction. Natural game time continued through travel, scaffold construction and completion.

Selected personally reviewed complete scenes, relative to the run directory:

Evidence What it establishes
original/bootstrap-observation.png Native title before permanent binding
original/result-selected-scv.png Actual selected SCV and status panel
original/placement-review.png Native placement command active
original/result-construction-started.png SCV constructing new scaffold
original/result-construction-complete.png Completed structure; cap now 50
original/result-depot-identity-complete.png Completed Supply Depot, 500/500, +8 supplies
original/quit-confirmation-review.png Native Exit Program confirmation, with completed depot still present

personal-scene-reviews.json records the observations and image hashes. input-audit.json records eleven consumed, independently released clicks. The five gameplay clicks used scenes aged 3.720–4.330 seconds at intent creation; the controller also rechecked freshness immediately before DOWN. Menu scenes had a 30-second bound. All pulses were 400 ms with an independently prearmed 1750 ms fallback and unique nonces. Two admission reviews and three command reviews expired before input/arming; their scenes were never reserved or sent. Fresh scenes and new choices were used, with no UNKNOWN or rebind.

Native exit and scope limits

At 292.791 seconds the irreversible Quit phase began, well before the 1020-second gameplay cutoff. The separately reviewed native route was MENU → End Mission → Exit Program → “Are you sure you want to exit Starcraft?” → Exit Program. Restart Mission and Quit Mission were not chosen. The final physical UP was acknowledged. No host quit, cancellation request or watchdog signal ended the original.

The log records guest [Exit] code=0; CLI duration was 339.356 seconds, and the actual native wait completed at 339.607 seconds. Native, watchdog, supervisor child and outer wrapper all returned 0 with actual waits and both EOFs. The watchdog exited unfired, the heavy lock was released, runtime PIDs were absent and port 8605 was closed. These are actual records, not inferred exit codes.

Shutdown also logged a secondary-thread EIP=0 diagnostic and fourteen abandoned parked critical-section enters; the final thread summary retained another active thread. These diagnostics remain in the log. The demonstrated claim is native Exit Program followed by guest/process exit 0, not a guarantee of clean shutdown for every emulated thread. No runtime patch was attempted.

No FPS, audio, browser parity, enemy destruction, full mission completion or full-suite compatibility claim follows from this run. Box ID/expiry are null; state is NOT_CREATED, and a box-stop receipt is not applicable.

Preservation and validation

Tested source base: f6931d34b7ada7386d335624fe77f41803a2ea4f, with the relevant uncommitted runtime and owned new-file binary patches in the run directory. The source tree was not archived or copied. The tested WASM was 2,200,987 bytes, SHA-256 c94ccf1b179f18bd4b47867445afed8cef2f1393b19aabfd259c7f3b94a7be09; it was referenced and hashed, not copied into evidence.

launch-pins.json and postflight.json establish 2,945 pins, zero drift. registry-before.json/registry-after.json preserve the full app entry, paths, resolved dependency graph, imports, missing dependencies and twelve existing media/system-data file hashes. The five owner-specified EXE/DLL/MPQ/installer sizes and SHA-256 values matched exactly. No media was downloaded or copied. The existing non-seed shell32.dll absence and builtin-handler fallback were identical before/after and were also reported by the actual loader.

Validation passed: 20 own policy tests, 12 actual endpoint tests, 10 actual CLI tests, and flock /tmp/wine-build.lock bash tools/build.sh under the heavy lock. Controls cover phase/selection/command/freshness policy, permanent identity, sticky UNKNOWN/no replay, late/dropped ACK, actual child waits/EOF and independent release, including parent death. Instrumentation and synthetic scene metadata belong only to isolated Solitaire fixtures; none was loaded by StarCraft. All 31 complete-scene PNGs passed CRC/inflate/800×600 validation. Fourteen selected scenes have personal review notes; this run's evidence checker found zero errors, missing artifacts or invalid metadata.

Changes are confined to the StarCraft controller, its tests and this report. lib/apps.js, the frozen runner, historical StarCraft notes and peer runtime sources remained read-only. The original is closed; no further mission is required for this bounded task.