Reconstruct lab story, interactions, patrol and jet pack practice flow

This commit is contained in:
UnixNight committed 2026-10-24 20:33:03 -04:00
1 parent 58424120bb
commit c929f942e1
84 files changed
+7957 -305

No files matched your search

+89 -120
View File
@@ -1,134 +1,103 @@
# Progress — current checkpoint
# Progress — 2026-10-24
**Full reconstruction is incomplete.** The current milestone is an evidence-backed
asset recovery toolchain and a controllable original lab with local Linux/Windows
packages. This is a movement checkpoint, not a completely reconstructed scene.
**The full reconstruction is incomplete.** The current implementation extends the
user-verified Windows rendering checkpoint into an interactive original lab and
jet pack tutorial flow. This is not a verified faithful complete scene or release.
## Controllable lab checkpoint
## Current integrated behavior
The user confirmed that the previous Windows build loads and renders and asked
to advance. The default now runs walking/backpedaling/turning, ballistic no-gadget
jumping, source-triangle collision, a chase camera, and original idle/run/backpedal/
jump/fall clips. Arrows move, Shift/Insert/numpad0/right mouse jump, R resets, C
pauses for inspection/resumes, F11 fullscreen, Escape quits. The intro is skipped.
- Default play runs LV1B STARTEXP and its16-cue CAM/MCA/AI camera/speech sequence.
Escape/Space skips actions and restores control; `--inspect` retains inspection.
- Original walking/backpedal/turn/jump arithmetic, source clips, chase camera and
fixed60Hz port simulation. Powered jumping now translates the recovered state
table,8-second effect timer,60/100 contact radii and energy/VR immunity rules.
- Original Carl/Goddard morphs, sprite quads and9 lab model actors, including the
OMT phones, source buttons and moving bars. Goddard's unregistered STOP/WALK/TALK
requests retain the current source clip rather than invented aliases.
- Shared patrol states1/3/7, state9 base motion and Carl's WALK/STOP profile.
Body support uses recovered radii, factors, quick magnitude and one-sided trace;
full wall response and Carl sensing remain incomplete.
- Lab explanation -> SCENE35 -> jet pack pickup -> SCENE50 -> Carl contact/talk
-> SCENE60 -> outside LOAD dispatch. Pickup adds50, Carl callback adds100 before
dialogue. Source task gates/objective selector and inventory slot0/3 persist.
- The pickup's internal tag `shrinkray` names the lab jet pack, proven by original
GET JET PACK art and powered-jump inventory checks. It is not a shrink-ray weapon.
- Original objective pad, three lab objective images, digits and PracticeYesNo
are rendered at source640x480 coordinates. O changes stationary panel offset.
Practice pauses the simulation; Y enters VR04/PHONEBOOTH and N resumes.
Return context is current GAM/latest STRT (LV1B/STARTEXP for NewGame), separate
from the still-unrecovered runtime checkpoint-root restoration.
- Button contact/cooldown/down/up/next-trigger and vertical-door duration,
reversal/hold/next-trigger logic. Source dynamic-box broadphase/triangle ordering
is translated; narrowphase/response still uses a modern approximation.
- Ordinary stars/trophy, score and acquired flags; all12 TSK files round-trip.
CLI checkpoint load/write preserves unknown fields, inventory, tasks, energy,
objective selector and RETURN context. Complete runtime/menu saving is unfinished.
- Music and speech use independent SDL streams. Positional effects/banks and
physical audio verification remain unfinished. Original default-bank pickup/button/
moving-door one-shots and owner loop stop are now connected; attenuation/volume
policy and jetpack sound callbacks remain unfinished.
Recovered arithmetic and addresses are in `analysis/movement-checkpoint.md`.
Notable corrections: Left precedence; current-yaw target offsets rather than
accumulating destination yaw; Up halves keyboard turn; turning alone decays speed;
no ordinary midair arrow steering; disabled-gadget long hold can repeat on landing;
normal camera eye `(0,180,-350)` and target `(0,99,0)`. Exact original frame order
is unresolved; the transient disabled-gadget state3 transition is collapsed.
## Source availability and evidence
Explicit port approximations: fixed60Hz cadence, capsule radius24 and standing
height209.4361, floor-derived model/root separation, static sweep/slide solver and
all-source-triangle mask, camera sphere obstruction, animation state/loop timing.
Collision retains all2869 lab triangles including325 invisible proxy triangles.
Original solver/masks, actor pivot/scale, stairs/dynamic objects and state activation
remain unresolved. At walls the clipped camera can move close to Jimmy.
All539 original hashes remain the baseline. All35 GAM files are supplied and33
have matching OMT geometry. Level3B/E GAMs exist, each1777 bytes, and are identical
after their four-byte level identifiers. Their classes name absent OMT archives,
but no serialized incoming GAM transition or supplied save selects them; loader
null handling does not establish a fatal failure. Normal gameplay use is unverified,
so these are not an established blocker and no extra files are currently requested.
The catalog preserves absent geometry without substitution.
Verification: 11 collision and13 movement groups pass, including real lab source
geometry, long/high-speed sweeps, held/re-pressed jumping, ten-minute stability,
deterministic replay and60/120/144Hz presentation yielding the same simulation.
A240-step GPU walk/jump replay travels1250.12 units, jumps once, lands, and ends
grounded; maximum rise137.62. This is port behavior, not original fidelity proof.
The Linux SDL window run receives injected movement/jump keys, resizes800x600,
and reports zero Vulkan validation errors on GTX1070. Headless llvmpipe also
reports zero errors. User-confirmed native Windows loading/rendering applies to
the previous build; native Windows movement validation remains outstanding.
Both self-contained packages were rebuilt. The Linux package passes the same
GTX1070 window replay; the Windows package passes240 windowed frames under isolated
Wine64 with SDL keys/jump/resize and identical movement statistics (validation
disabled there). Wine is preliminary compatibility evidence, not native Windows.
One additional inspector-window regression was not executed because automatic
approval review was unavailable; the full headless inspector regressions passed.
Detailed recovery: analysis/lab-progression.md, scene-entities.md, scene-models.md,
patrol.md, original-collision.md, dynamic-collision.md, buttons-doors.md, ui.md,
jetpack.md, level-resource-mapping.md, and save-format.md. Addresses, source offsets,
confidence and approximation boundaries are retained there.
## Latest user-reported preview fix
## Verification
The user ran the Windows build and observed a static Jimmy partially inside the
lab floor. This verifies initial Windows startup/rendering by user report, with
OS/GPU/version/input/audio details unspecified. The user subsequently confirmed that the corrected build loads and renders.
Managed checks cover source data, director/actions/cameras, task/interactions,
patrol, recovered support, moving-box collision, jetpack states and HUD composition.
Python currently has35 top-level regression tests plus2 targeted UI extraction tests.
Existing11 collision and13 movement groups pass;14 jetpack groups pass. Eight
quick-magnitude float results are checked against an isolated original DLL probe.
Original mode2 standing/jump animation probes confirm zero dynamic-point offsets
for the tested clips. This is DLL evidence, not observed original game behavior.
Both issues were reproduced from actual assets: jimstop has two byte-identical
vertex poses, and its minimum local Y=-64.6165 plus GAM Y=26.6619 places the soles
at -37.9546 while the original visible floor is near Y=0.5486. Calling the default
an animated idle was inaccurate. The idle stays faithful to its still source pose;
1/2/3/4 now explicitly select idle/run/jump/scratch previews and the window title
states the mode/controls. A source-data-based preview correction of +38.503 Y
places the standing model on visible floor triangles, leaving GAM data unchanged.
This is an explicit preview policy, not the recovered game collision solver.
Headless llvmpipe runs render the story, model actors, practice prompt, lab exit
and VR04 entry with zero Vulkan validation errors. The300-step VR04 powered replay
jumps once, rises418.98 units and lands; VR immunity retains20 energy. The ordinary
240-step replay travels1250.12 and rises137.62; port landing-count telemetry changed
with the explicit state table. These measurements verify port behavior only.
See artifacts/full-integration-test.log and practice/jetpack logs for current checks.
Tests verify real lab grounding, ceiling/invisible-proxy rejection, still idle and
moving clips, automatic animation advance in actual GPU readbacks, and SDL keyboard
switching to a run cycle. Khronos validation reports zero errors on llvmpipe and
GTX1070. Source assets remain unchanged. See rendering.md for evidence/scope.
The rebuilt Windows package additionally passed an isolated Wine64 window test:
six frames, animation key2 selected moving source samples, resize800x600, exit0.
Validation was disabled in that Wine test; corrected native Windows retest is pending.
The full scripts/test.sh integration passed:35 Python regressions,2 targeted UI
checks, all managed groups, GPU story/route/practice/VR04 replays, SDL dummy streams
and539 unchanged original hashes. Both self-contained packages rebuilt successfully.
Linux package runs the12s story without SDK/source/Original mounted, zero Vulkan
validation errors. Its GTX1070/Xvfb practice window pauses, receives N, resumes11
steps and resizes800x600 with zero validation errors and original pickup audio.
The Windows package passes the same12-frame practice/input/resize/stream check
under isolated Wine64, matching simulation statistics (validation disabled).
It also renders the12s story under Wine. Generated package logs live under
artifacts/package-integration.log, linux-package-integration/, package-practice-window.log
and windows-practice-package/. Initial native Windows loading/rendering remains
user-confirmed for the earlier checkpoint; current native Windows story/input/audio
is still unverified. Wine is preliminary compatibility evidence only.
## Verified
## Next work and limits
- Full `../prompt.MD` specification read; originals preserved and local Git initialized.
- All 539 original files (224,601,234 bytes) still match recorded SHA-256 hashes.
- Native PE32 x86 MSVC C++ game, Open Media Toolkit 2.0 DR4, Direct3D7/DirectDraw7,
DirectInput8 and DirectSound identified with addresses in analysis reports.
- 35 GAM files: 3,299 entities, 93 class codes, 84,044 properties, byte-for-byte
round-trip in Python and full corpus decoding in native C#.
- 100 OMT files / 7,027 indexed resources audited: 2,155 meshes, 2,183 materials,
1,614 textures, 875 bounded RIFF resources; zero decode failures.
- ASE/OMT coordinate and UV conversion proven with duplicate assets. Jimmy morph
sample vectors retain original base UV/topology; normal texture corrected to
jimycarl.png via actual InitObject code, with shock atlas kept separate.
- .NET SDK10.0.402/runtime10.0.13, Silk.NET2.23.0, SDL3.4.2 provisioned locally.
- C# builds without warnings/errors; 20 Python tests plus native C# corpus,
malformed-data, arithmetic, visibility, case lookup, morph and merge tests pass.
- Native Vulkan lab rendering: 2,869 triangles / 33 textures; combined Jimmy
preview: 3,683 triangles / 34 textures. Real GPU readback, zero validation errors
under llvmpipe; scene/player screenshots visually reviewed.
- Fixed original null-material skipping removes false white collision-proxy
surfaces. Original startpoint/camera and recovered projection are available.
- SDL3 hardware window/presentation on GTX1070 under Xvfb: six frames, injected
keyboard received, resize to800x600, zero validation errors.
- SDL3 recovered WAV stream consumes bytes on dummy device. Speaker output untested.
- Self-contained linux-x64/win-x64 packages built under dist/. Linux package also
renders inside a separate filesystem namespace with no SDK/project/original
directory mounted. Initial Windows rendering is now user-reported as noted above.
- Windows x64 package also renders two headless frames under isolated Wine64,
exit0, without source/SDK/originals mounted. Its fixed-time PPM is byte-identical
to the Linux package on the same lavapipe driver (checkpoint670adb4). Broader native Windows remains untested;
Wine test did not enable validation or exercise SDL window/audio/input.
- Official Wine11 original-copy runs attempted in isolated filesystem/network
namespaces. Both GL and Vulkan backends produce unusable frames; details below.
No usable original gameplay reference has been captured: isolated Wine GL/Vulkan
attempts fail rendering. Continue static recovery and isolated DLL probes locally;
comparison on a working original installation remains required for fidelity.
## Build and run
Next priorities are lab navigation/door collision and checkpoint return-position
fidelity; original collision47BD80/479D20 and phase/cadence; audio-bank/SFX mapping;
UI slide/highlights/energy gauges; actor sensing/animation completion. Then expand
remaining gadgets, damage/combat/enemies, menus/configuration, full saves and all
level progression. Lighting/fog/blend/culling and sprite/pivot fidelity are open.
Do not claim complete gameplay based on direct-contact route or GPU smoke tests.
From GamePort: `source scripts/env.sh`, `dotnet build`, then `dotnet run -- --with-player`.
Asset preparation: `python3 tools/extractors/prepare.py --audio sounds musicsecretlab`.
Tests: `scripts/test.sh`; packages: `scripts/package.sh`. See README for exact options.
Generated outputs and dependencies are ignored, and everything remains local.
## Remaining and current constraints
No original gameplay reference has been captured. Wine GL cannot obtain a suitable
pixel format/3D surface; Wine Vulkan logs unsupported fixed-function state and a
swapchain image count mismatch. Native Windows/VM access has been requested for
reference/fidelity and native port validation. These failures do not establish
original behavior. Static reverse engineering can continue independently.
Original collision resolution, introductory LABEXP cutscene, entity activation/
task/trigger execution, interactions, combat, AI, progression, menus, mouse aiming
and saves remain incomplete. The no-gadget controller checkpoint requires original
runtime comparison. Animation looping/speed and collider/pivot remain approximations.
Rendering material/lighting/fog/blend state mapping also remains incomplete.
Two investigators contributed movement evidence/tests and collision implementation/
tests for this checkpoint. Primary integrated the runtime, input, animation, packages
and verification. Reports distinguish static recovery from runtime checks.
## Continuation
Read AGENTS, tasks, architecture, compatibility, reverse-engineering index, and
analysis reports. Check Git status before editing. Avoid repeating solved inventory,
GAM decoding or OMT extraction. LV1B/STARTEXP is initial scene evidence; its LABEXP
trigger still needs reconstruction. VR01 provides a smaller, 16-entity test scene.
Next actionable addresses/tasks are in tasks.md; original execution procedure and
errors are in reference-testing.md. Commit history preserves local milestones.
All previous agent file ownership has returned to primary; investigators that
reached usage limits are not dependencies. Generated output stays separate from
Original and the nested local repository has no remotes. Read tasks.md and Git
status before continuing; preserve source evidence and update these notes.