Tut’s Tomb: reviewed original deal, 2026-11-14

Evidence directory: scratch/runs/20261114-tutstomb-reviewed-outcome/. One registered original default deal produced thirteen visibly verified removal events: ten pairs and three single Kings. It ended through native Game → Exit with fifteen pyramid cards remaining. This is a partial result, not a native win, loss, or exhausted-deal claim. Final review identified a missed legal 10D + waste 3H pairing; the original was not restarted.

The route uses registered wep16_tutstomb, the existing local test/binaries/wep16/WEP2/TUTSTOMB.EXE, its help file, imported CARDS, and runtime WEPUTIL. No downloaded or substituted executable is involved. The original’s module-reference table names win87em, KERNEL, GDI, USER, SOUND and CARDS. CARDS and WEPUTIL each import KERNEL, GDI and USER. The runner handles the system modules and win87em internally; CARDS is a real local NE DLL. A registered manifest alone is insufficient to establish the media closure. Before-launch pins and actual module logs must agree; an unexpected unpinned dynamic module remains unknown. The earlier TicTactics MSVCRT before-hash omission remains a gap in that separate run.

The native notes identify the important input behavior: WM_LBUTTONDOWN acts on the card remembered by WM_MOUSEMOVE, not the DOWN coordinates. The existing coordinate-press path in lib/renderer-input.js, enabled by test/run.js, supplies pointer arrival and a separate 30ms hover pump before DOWN. The controller chooses one coordinate press, never a hidden second click or raw native-message bypass. Transport acknowledgements establish neither selection highlighting nor actual card removal.

tools/tutstomb-controller.py requires a personally reviewed complete default pyramid before permanent PID/start/window binding and admission. Gameplay reviews expire after five seconds; menu reviews after thirty. Each first pair member requires a personally chosen visible partner. A second member requires a separate current scene showing the first selection, its earlier action, and visible ranks totaling thirteen. A King is single. Stock actions require a visible legal target and personal justification; recycling additionally requires the visible draw-three rule. There is no Start, Deal, New session, undo, solver, card reader or move generator route.

Every gesture independently prearms its 1750ms fallback UP and has a finite 1500ms limit, unique nonce and no replay. UNKNOWN, mixed identity, cancelled input, late acknowledgement and release-child failure close gameplay. The fixed runtime reserves an irreversible final 180 seconds for native Quit. Native UI Exit and host teardown are recorded separately. A normal native wait can justify OWNER_EXITED when final physical UP has no acknowledgement; it cannot turn that absence into a physical UP claim.

Rules and controls reviewed before launch

The built WAT help parser decoded all eleven topics of the existing local TUTSTOMB.HLP. The native rules define a free card as one with no card above it, pairs totaling thirteen, Ace one, Jack eleven, Queen twelve and King thirteen. Kings are removed singly. Only the top turned-over stock card can be played. Clicking the face-down stock turns one or three cards according to the untouched options; the green-circle recycle is allowed only for draw-three. Winning means removing the whole pyramid. The help explicitly says the first selected card is highlighted, then a successful pair moves to the upper-right corner. The Casino score is +13 per match, -28 per deal, -1 per flipped stock card, -1 per ten seconds and -5 per recycle. No Options control is used to change the original default.

Twenty-five policy tests, nine HTTP controls and eleven actual CLI controls passed. The latter use a separate Solitaire fixture, establishing transport behavior without claiming Tut’s Tomb gameplay. Actual renderer dequeues and native DispatchMessage traces show hover at 1794699467810ms and DOWN at 1794699467840ms: separate pumps thirty milliseconds apart. One native menu control opened the fixture's first menu. The full build passed under both the heavy-runtime lock and /tmp/wine-build.lock in 49.081 seconds. The controls cover UNKNOWN after a consumed click and dropped reply, window replacement, cancellation, late ACK, release-child death before DOWN, independent fallback, parent death and capture after UP. All eleven native, watchdog, supervisor and outer fixture waits have actual EOF evidence; orphan release-child waits after parent death remain unknown.

The initial light policy test caught a missing diagnostic; the initial HTTP parent-death check raced the release child's terminal-audit write. Both were corrected and their failed logs retained. The older JavaScript HLP helpers did not decode this help format correctly; the built native parser above provides the complete readable rules instead.

Actual original and visible evidence

The published controller/test route was commit 69a3e19980fbc6fdbfe7bf7307a378a1620d5a65 before launch. The original ran once, with real ticks, default 1000 batch size, 800×600 screen, fixed 1200-second limit and independently prearmed 1220-second watchdog. Its native runtime was 809.740 seconds. No configuration, clock, private card memory, seed, solver, undo, Deal or New session was used. The untouched original showed draw-three and Casino scoring. The complete actual board was reviewed and permanently bound to PID 4159762/start token 274682277/main HWND 0x10001 before the first card action. Each actual pair first selection was separately reviewed as visibly inverted before its second member was chosen. Freshness refusals happened before sending input and required a new personal scene.

Selected before/result PNGs are named by action in selected-captures.json; the second-member before images show the actual first selection. Removal claims come from the visible cards disappearing and the native score, never from a transport ACK. Every pair below totals thirteen, and each King is single. A net +12 reflects the native timer crossing a ten-second boundary.

Event Removed Native score before → after Newly free pyramid cards
king1 KC -38 → -25 —
pair1 10C + 3S -44 → -32 4H
king2 KS, top waste -33 → -21 —
pair2 AH + QD -29 → -17 JD
king3 KH, top waste -17 → -4 —
pair3 JD + 2H -7 → 6 JC
pair4 JC + 2S -3 → 10 —
pair5 8D + 5H 8 → 21 —
pair6 4H + 9C 12 → 24 10H, 6C
pair7 6C + 7H 4 → 17 —
pair8 10H + 3C -4 → 9 3D
pair9 3D + 10S 8 → 20 7C, 10D
pair10 7C + 6S 8 → 21 JH

The sixth removal event was pair3. Total removed cards were 23; thirteen came from the pyramid, leaving fifteen there. Nineteen ordinary stock-draw clicks and three visible green-circle recycles occurred. Native help permits reuse in draw-three mode. After the third recycle the stock was face-down and the waste empty; no further draw occurred before Quit. The last visible native score was -23 at time 804 in native-exit-visible.png.

The help documents -5 per recycle. The actual selected recycle images show -10, apart from timer deductions: 15 → 5 on the first recycle, 19 → 8 on the second (one timer point), and -9 → -19 on the third. This discrepancy is unresolved; the observed UI values take precedence over an assumed formula.

This was not exhaustive play. At scene s54 the top waste was 3H and 10D was already exposed by pair9. I overlooked that legal pair and drew again, incorrectly recording that no ten was free. The transient s54 PNG was pruned before this final review; its original capture metadata/hash remains, and selected pair9/final-board PNGs show the exposed 10D. The third recycle also occurred before deciding to stop. No native outcome dialog appeared. The result therefore preserves the missed opportunity and partial status rather than claiming an unwinnable or completed deal.

Closure, provenance and shutdown

Imports and runner auto-resolution were inspected before launch. The actual native module log loaded CARDS and registered WEPUTIL for LoadLibrary; no unexpected external dynamic module appeared. System modules and win87em are implemented internally. Exact complete known media sizes/SHA-256 values below were equal before and after; the HLP is included as a registered dependency.

Existing local WEP2 media Bytes SHA-256 before = after
TUTSTOMB.EXE 101472 d02dd9ce2e6bbead9e9464691b6db08c20a4158b09b93be5a414606f9d1476e8
TUTSTOMB.HLP 15527 7616f28c70beb61d94128eb8284e8154b44b00e5b0a5837440bdb54b173f0b0b
WEPUTIL.DLL 19200 7bd426e4d1ca0afea88e5ad22d9add44d4df79432cdc4b5cb165c10c6023e6bd
CARDS.DLL 148528 bcc968e7c8b376eae9dc91dad9dc43a40502fd91df1663c6dc82fb6c50b10a0f

launch-pins.json records 361 exact source/runtime/built-artifact and local WEP2 candidate-media hashes. release-pins.json verified all 361 unchanged before releasing the runtime slot. Git commit plus the 58821-byte relevant-uncommitted.patch preserves relevant uncommitted source hunks, including new files. No source tree, WASM, game media or frame bundle was copied into evidence. Thirty-five selected decision/result PNGs remain; redundant transient PNGs were removed with their hashes recorded.

Q2 then JezzBall predecessor actual native/watchdog/supervisor waits, both EOF, heavy release and absent PIDs were personally checked before this run; their ancillary outer UNKNOWN fields remain unchanged. On this run, the personally reviewed native Game menu showed Exit. Selecting it produced [Exit] code=0, native actual wait 0 with both EOF, watchdog actual wait 0, and supervisor actual wait 0 with both EOF and heavyReleased=true. All four original lifecycle PIDs were absent and port 8633 closed. There was no host-close gesture. Final transport completion itself claims neither removal nor native Quit; actual UI choice plus native termination establishes Quit. The ancillary caller tool ended 143 before writing its own wrapper receipt, so that wrapper's actual wait/EOF remain UNKNOWN. No restart or rebind was made. Runtime release was announced immediately; all later work was report and evidence cleanup. No box was created: ID/expiry null, state NOT_CREATED, stop receipt not applicable. Free disk was 12 GB, above the 1.5 GB floor.

Machine-readable removal/exposure/score/stock/partial/nativeQuit evidence is scratch/runs/20261114-tutstomb-reviewed-outcome/result.json. No browser, FPS, audio, deployment, full-27 or new-title-win claim is made. The earlier TicTactics MSVCRT before-hash gap is not repaired by this run.