481 lines
35 KiB
Markdown
481 lines
35 KiB
Markdown
# Decisions
|
||
|
||
## Phase 1 evidence contract — wave 1
|
||
|
||
Session/receipt schema1 records original/copy/package/media identities separately.
|
||
Imported commands are inert data; duplicate IDs are idempotent only for identical
|
||
bytes. Retained cases bind departure/return and copy-state evidence; fresh cases
|
||
have independent lifetime IDs. Eight dialogue/lifetime/input cells each require
|
||
continuous plus dialogue-return, left/right bank and landing-idle scope on original
|
||
Windows/DxWnd, native Linux port and native Windows port (120 obligations).
|
||
Synthetic/component/Wine/cross-publish/historical evidence cannot substitute.
|
||
The GSD worktree fork-base guard selected sequential native executors because
|
||
origin/HEAD is absent in this Gitea/main repository. Existing user authorization
|
||
for main commits is encoded in git.allow_default_branch_commits.
|
||
|
||
## Phase 1 execution structure — 2026-11-12
|
||
|
||
Build the agreed evidence workflow in seven plans/five waves: preservation and
|
||
session/scenario contracts; parallel local viewer, full-game capability ledger
|
||
and passive port observer; local Windows/DxWnd bundle; native Linux capture; then
|
||
one combined hands-on session and strict acceptance audit. Materials readiness
|
||
and native/original acceptance have separate verdicts. Missing actual captures
|
||
leave EVID-02/EVID-04 open. Five research uncertainties now have explicit planning
|
||
policies while their actual runtime outcomes remain execution dependencies.
|
||
|
||
## Phase 1 comparison decisions — 2026-11-12
|
||
|
||
Use full lab → VR04 → lab → outside journeys plus focused normal/edge-input
|
||
replays, with equal natural/skipped dialogue and separate fresh/retained coverage.
|
||
Review original/port playback together with optional overlays, actual timing drift
|
||
and switchable audio. Browse evidence by capability with current gaps first and
|
||
expandable revision history; order discrepancies by player impact. Prepare one
|
||
local bundle and combined Windows session per checkpoint. The original runs from
|
||
a generated copy through DxWnd, with version/settings recorded; port-native
|
||
receipts remain separate. Full decisions are in
|
||
`../.planning/phases/01-reproducible-fidelity-baseline/01-CONTEXT.md`.
|
||
|
||
## Full reconstruction planning — 2026-11-12
|
||
|
||
The user scoped GSD planning to the remaining full reconstruction and explicitly
|
||
required accuracy as high as possible to the original. Organize work as playable
|
||
slices, retaining the movement-first stabilization gate and all original-required
|
||
content. Use autonomous mode, standard phases, parallel independent plans,
|
||
tracked planning docs, research/plan checking/verification, full instructions
|
||
and inherited session models; optional PR body sections are disabled.
|
||
See `../.planning/PROJECT.md`, `config.json`, `REQUIREMENTS.md` and `ROADMAP.md`.
|
||
These decisions introduce no gameplay changes or new compatibility claims.
|
||
|
||
User workflow update (2026-11-05): completed, verified milestones are now pushed
|
||
to the existing `Gitea` remote at `https://gitea.unixnight.com/UnixNight/Neutron.git`
|
||
on `main`. This explicit user instruction supersedes earlier local-only Git
|
||
restrictions in prompt.MD and AGENTS.md. Builds/tests remain local; original
|
||
assets and generated packages stay excluded.
|
||
|
||
1. **Preserve input directory.** Keep Original in place, baseline all file hashes,
|
||
generate outputs elsewhere. No legacy binary execution in the source directory.
|
||
2. **Nested local repository.** Workspace `.git` is reserved and not a functioning
|
||
repository. Initialize GamePort/.git, commit code/docs/tools only. The original
|
||
no-remote rule is superseded by the user workflow update below.
|
||
3. **Evidence-first data recovery.** Native x86 executable does not imply little
|
||
endian formats: GAM and OMT use big endian metadata. Require bounded readers,
|
||
full consumption and real original-file regressions.
|
||
4. **Initial scene.** NewGame.tsk starts LV1B and player GAM references STARTEXP.
|
||
Prioritize that scene; use smaller VR01 for data/render verification. The LV1B
|
||
LABEXP start trigger means simple camera spawning is insufficient for fidelity.
|
||
5. **Local dependencies.** Missing SDK/native tools installed under `.tools` or
|
||
`/tmp` after approved official downloads, without assuming root or altering host.
|
||
6. **Useful parallelism.** Three independent agents own binary/rendering, asset
|
||
recovery and infrastructure, while primary owns gameplay data and integration.
|
||
Shared runtime/format changes must be coordinated. No external coordination.
|
||
7. **Honest compatibility.** Build, headless Vulkan frame, SDL presentation,
|
||
gameplay fidelity and native Windows execution are separate evidence categories.
|
||
8. **Preview ground contact and idle clarity.** User testing exposed floor overlap
|
||
and a misleading animated-default description. Original idle vertices are
|
||
identical; do not invent idle motion. Expose moving source clips explicitly.
|
||
Until original actor transforms/collision settling are recovered, apply one
|
||
model-only floor correction derived from standing soles and visible geometry,
|
||
with explicit preview labeling and regression tests. Preserve GAM/ASE data and
|
||
per-frame morph displacement; this is not the original collision algorithm.
|
||
9. **Controllable lab before complete scene.** Integrate evidence-backed no-gadget
|
||
walking/jumping/camera now, with separate stateful simulation and explicit
|
||
provisional source-triangle capsule collision. Rendering does not set the number
|
||
of movement calls: use60Hz until original cadence is recovered. Preserve source
|
||
GAM/ASE positions; keep physical feet, inferred actor-root height and visual
|
||
pivot separate. Include invisible geometry rather than inventing visibility
|
||
masks. Document the skipped intro, missing interactions, transient jump-state
|
||
simplification, animation timing and close-wall camera limitation. Test against
|
||
source geometry and build both packages without calling this a faithful full scene.
|
||
|
||
10. **Source-driven scene interpreter.** Preserve synchronous AI action dispatch,
|
||
original camera formulas and audio-duration advancement rather than adding a
|
||
fixed intro timer or completingSCENE on intro end. Separate task activation,
|
||
sprite playback, draw visibility and physical collision. Report unrecovered
|
||
behavior explicitly. Preserve unknown TSK fields and original checkpoint resume;
|
||
no claim of arbitrary-position/full-state save support.
|
||
11. **Separate supplied records from required content.** All35 GAM entity files
|
||
exist;33 have matching OMT geometry archives. Level3B/Level3E GAMs contain the
|
||
same four entities apart from their level code, with no incoming serialized
|
||
GAM transition or supplied save selecting them. Their original classes name
|
||
absent OMT archives, but registration alone does not prove normal gameplay use.
|
||
Record unavailable geometry without substitution; do not classify it as a
|
||
reconstruction blocker or request more files unless progression requires it.
|
||
|
||
12. **Preserve source physics boundaries.** Recovered quick magnitude, support trace,
|
||
radii/factors and3DUD broadphase/triangle ordering stay separate from provisional
|
||
wall/capsule response. Check isolated original DLL results without executing or
|
||
modifying originals in place. Keep visual soles distinct from logical root75.9.
|
||
13. **Correct prototype names from original evidence.** The pickup tagged SHRINKRAY
|
||
is the jet pack in original objective art and inventory slot0/3 code. Practice0
|
||
entersVR04/PHONEBOOTH with current GAM/STRT return context. Do not invent a
|
||
shrink-ray weapon or VR01 mapping. Later recovery identifies TSK StoredVector
|
||
as the separate transition departure root; see decision17.
|
||
14. **Compose original UI and effect assets.** Draw source640x480 UI in a separate
|
||
depth-disabled overlay; retain clipped original coordinates. Use exact default
|
||
SoundEffects.omt identities for lab effects, preserve encoded RIFF resources
|
||
and document unsupported decoders. Do not substitute editor fonts/audio banks.
|
||
15. **Default recovered scene play.** Start the integrated lab story/interactions by
|
||
default; retain --inspect and deterministic smoke modes. Passing direct-contact
|
||
tests does not establish navigation, original fidelity or full-game completion.
|
||
16. **Use recovered control and UI lifetimes.** MCA PlayerControlled restores its
|
||
target at completion; active valid shots clear control. O changes the objective
|
||
animation timer rather than gadget layout. Reset air-turn state at the recovered
|
||
grounded/released branch. Use per-level geometry-layer far limits; the current
|
||
merged renderer uses their maximum until separate layer masks are implemented.
|
||
Source evidence and regression limits are recorded in the focused analysis docs.
|
||
17. **Keep transition restoration separate from pickup checkpoints.** Source
|
||
RETURN flags preserve the departure root at+9A8, restore player/companion
|
||
position and suppress STRT once. Serialize only the proven root in StoredVector;
|
||
carry pending RETURN in memory. Retain authored STRT yaw and ordinary fresh-save
|
||
spawning. Pickup42AF00's+1E38/+1E48 is a different, still-unconsumed capture.
|
||
18. **Apply terrain eligibility at shape registration.** Source level loaders skip
|
||
NC-prefixed shape names. Use prepared shape metadata for player and actor
|
||
terrain queries; retain rendered geometry and null-material proxies. Do not
|
||
infer a generic water-material rule or full water gameplay from this finding.
|
||
19. **Keep user-requested debug tools explicit.** F1 opens a modern diagnostic menu
|
||
with code-generated glyphs, separate from recovered UI assets. Pause gameplay
|
||
and audio together, capture held input until release and stage SCENE edits until
|
||
Enter. Invoke existing source-state consumers for actions; offer only prepared
|
||
geometry and source startpoints. No debug behavior is attributed to the original.
|
||
Force Control is a persistent debug override over authored ownership, with
|
||
FALSE restoring the current scripted state. Reset/new scene clears it, and
|
||
source music handoff continues to use authored ownership.
|
||
20. **Keep live diagnostics passive and session-owned.** F2 and the F1 Live Overlay
|
||
action change visibility without pausing gameplay/audio. Preserve this setting
|
||
across scene rebuilds in memory, independently of TSK state. Reuse the existing
|
||
font/buffers, give F1 display priority and measure wall-frame intervals separately
|
||
from simulation ticks. Report the rendered camera/projection, not a guessed mode.
|
||
21. **Own presentation at application lifetime.** Replace scene GPU resources
|
||
transactionally, retaining the SDL window, Vulkan core and user window state.
|
||
Size changes may recreate targets/swapchain. Per-level audio/gameplay remain
|
||
separate; source fades are still open.
|
||
22. **Uniform UI with anchors.** Preserve4:3 source positions, center modal/F1
|
||
canvases and anchor objectives bottom/right and F2 top-left. Map logical input
|
||
through actual drawable dimensions with double precision for strict bounds.
|
||
This is authorized modern widescreen policy, not original algorithm recovery.
|
||
23. **Pause feature expansion for architecture recovery.** Preserve clean69d0187
|
||
as a local tag and independent package copies. Use the lab/tutorial/return/
|
||
outside route to map state owners, callers, phases and persistence before more
|
||
gameplay. `checkpoint/` supersedes older architecture/priority summaries.
|
||
Evidence probes and directly evidenced corrections are permitted. A documented,
|
||
gated implementation plan ends the investigation checkpoint; continue through
|
||
it without waiting to understand every level.
|
||
24. **Separate task objective and entity phases.** Original45FB5B selects the
|
||
objective before task record mutation45FB7F and entity dispatch45FBD9. Expose
|
||
pre-assignment notification rather than relying on subscriber registration
|
||
order. Preserve repeated/missing assignment dispatch. Source loader's additional
|
||
task callbacks/PostLoad are a distinct lifecycle question, not implied by this fix.
|
||
25. **Consume prior player motion before callbacks.** Split controlled player
|
||
physics, input, controller and fire effect phases. Fresh input precedes Jimmy;
|
||
retained Jimmy/fire precede the appended destination game/input. Track this
|
||
relation explicitly across scene continuations. Mode-based input and the
|
||
player's grounded-release reset are separate boundaries. Keep fixed60Hz,
|
||
capsule/support and other grouped scene callbacks explicit approximations.
|
||
26. **Use the source static camera predicate on proven route scenes.** Build
|
||
sparse grids in source shape/polygon order, prefer the first wall-grid hit,
|
||
then the first all-grid hit. Use the individual source segment kernel without
|
||
nearest reduction; apply source desired-eye weighting and gain before follow
|
||
interpolation. Only LV1B/VR04/LEV1 filters are proven. Shared camera handoff,
|
||
special follow/input state and exact x87 boundaries remain separate work.
|
||
27. **Apply gadget availability at level entry.** Source PostLoad enables eight
|
||
group0 gadgets4→1, then disables owned1→4 in VR01..08. Preserve unknown states,
|
||
item8 and group1. This is availability, not acquisition or save normalization;
|
||
snapshots/constructors retain raw wire data. VR04's source jetpack bypass remains.
|
||
|
||
28. **Expose implemented systems through modern debug tools.** The user explicitly
|
||
requests expansion and noclip during the gameplay feature pause. Preserve
|
||
General-page controls; add Player/Systems pages and F3 flight. Keep authored
|
||
control ownership separate from flight and bypass player contact triggers.
|
||
Clear pending motion/held input at toggles, retain an entry escape position,
|
||
and reset transient overrides on scene rebuild. Mute by stream gain so clocks
|
||
continue. Debug checkpoint uses the existing TSK/start progression contract,
|
||
never pretends to save live physics/cutscene state. These are investigation
|
||
features with port verification, not recovered original behavior.
|
||
|
||
29. **Persist bounded retained scalars independently of resets.** Ordinary route
|
||
loads keep movement mode, jump-hold and owned-fire timer in a separate memory
|
||
container for proven no-gadget/jetpack states. Reset jump/body/presentation
|
||
through their own boundaries; do not serialize these fields in TSK or carry
|
||
them through debug checkpoint load. Unsupported levels/gadget cleanup remain
|
||
the existing explicit approximation until traced. First destination ticks
|
||
can change retained values; boundary and post-tick tests are different.
|
||
30. **Read full PostLoad tails before asserting final availability.** The prior
|
||
common-prefix interpretation was incomplete. Each VR01..08 re-enables its
|
||
practice gadget after the shared disable pass; LV1B disables1/7. Correct the
|
||
implementation and expectations from the later dispatch/call evidence and
|
||
retain the earlier report as a scoped historical audit with a correction.
|
||
|
||
31. **Control identity gates movement, not the whole player.** Run eligible body
|
||
physics and owned fire while raw control is absent; skip input/controller
|
||
tail. Preserve queued work and distinct resets. Apply source-specific ordered
|
||
preparations at CAM/MCA/AI activation; initial MCA target-none returns without
|
||
completion, and later shots do not repeat preparation. STOP notification follows
|
||
body clear. Authored transforms update the real player body. Keep modern
|
||
Force/Noclip and frame pauses separate from raw ownership. Other wrapper/class
|
||
callbacks and collision remain evidence gates rather than empty-input guesses.
|
||
32. **Use local REA as an additional static evidence source.** Keep Ghidra/JDK
|
||
and process environment local, analyze generated copies, retain complete
|
||
Evidence/profile/hash/snapshots and close owned sessions. CLI avoids changing
|
||
the existing MCP registration. Unknown procedure definitions remain unknown;
|
||
they do not invalidate byte/vtable evidence or imply missing game code.
|
||
|
||
33. **Integrate collision in verified boundaries.** Preserve source sphere kernel
|
||
nearest state/face/edge semantics as an isolated component with original-code
|
||
probe references. Keep live capsule replacement gated on traversal, callbacks,
|
||
support/friction and dynamic response. Independently correct the proven Jimmy
|
||
overhead wrapper using its root ray and mode-specific cancellation, retaining
|
||
fresh/retained input order. Modern camera debug controls affect camera queries
|
||
only. Do not reinterpret this partial correction as complete original physics.
|
||
34. **Compose body response only inside recovered contact coverage.** Preserve
|
||
signed material database IDs in collision metadata and reject legacy prepared
|
||
data lacking identities. An ordinary-only adapter may start with VR04, whose
|
||
static materials have no special Player374 policies; lab/outside contain
|
||
special contacts. The tested grid dispatcher is an isolated component until
|
||
body/landing/friction order and root visuals are compared; dynamic reclip
|
||
must explicitly own shared nearest/last-polygon state. Keep source published
|
||
speed160 separate from controller608. Geometry-off gravity cancellation must
|
||
not be reinterpreted as an ordinary prior-grounded jump rule.
|
||
35. **Enable the tested static body subset in VR04.** Compose verified traversal,
|
||
support and friction using original body-reference outputs. Reject special
|
||
scenes/provisional dynamic resolver and keep lab/outside on the baseline.
|
||
Preserve source60/100 radius changes through both sphere response and support;
|
||
restore60 in the later Jimmy wrapper. Published speed160 survives control
|
||
preparation separately from controller608. No dynamic-door geometry does not
|
||
establish absence of other source dynamic actor classes. Modern feet/sole and
|
||
noclip admission remain explicit policies; this is not whole-tutorial fidelity.
|
||
36. **Keep gadget display, equipped mode and owned effects distinct.** The user
|
||
parked collision work and requested gameplay gadgets. Integrate original
|
||
group0 resources/state and default item3 first: its display index is3 but raw
|
||
mode remains−1. Do not treat the default shrink-ray/jetpack as special mode3
|
||
or call Bubble mode0 Scooter. Reuse original heading, strict multi-target
|
||
gates and nonloop soldier morph; ordinary hits award10 through the score/save
|
||
owner. Expiry restores authored/reset AIState, not pre-hit dynamic mode.
|
||
Keep modern full-panel pause, upright effect transform, additive branch and
|
||
blended-depth policy explicit. Other gadgets, full AI and warning-audio
|
||
lifecycle remain independent evidence gates, not guessed feature completion.
|
||
37. **Preserve timed gadget and cleanup ownership.** Invisible's seven-second
|
||
recharge/active clocks run after movement with any nonnull controlled pointer;
|
||
raw-null pause is distinct from cleanup. CAM/MCA default DeactivateInv to1
|
||
and explicitly release named Jimmy before target validation, independently
|
||
of raw-owner438. Keep failed display2/raw−1 selection, source blend flashing,
|
||
original meter art and companion body flags separate from AI activation.
|
||
Energy scalar and warning latch share one session owner; raw saved/refill
|
||
assignment is not a helper-driven warning rearm. Full enemy sensing,
|
||
suppression globals, graphics options and canvas alignment remain explicit.
|
||
38. **Integrate Bubble's controller and body together.** Mode2 is persistent and
|
||
has fire throttle, its own turn cap and direct300/900 supported bounce; it
|
||
must not enter the ordinary550/jetpack table. Pass radius160/factor.415 to the
|
||
tested static body, preserve queued work/root and restore source650 maximum
|
||
on cleanup. VR01's floor111 has a retained-material callback even in air, so
|
||
implement that bounded policy with original bypass/debounce/material ordering.
|
||
Original Canv1 remains an independently updated sprite with full RGBA fade,
|
||
not invented sphere geometry. Keep the render bridge, generic camera/Euler,
|
||
other damage states and retained helper lifetime explicit. Bubble practice
|
||
precedes live equip; menu gating reads raw mode, not display selection.
|
||
39. **Keep Baseball ownership beyond selection.** Preserve the original20-slot
|
||
ring, pending release after cleanup, prior-capture label and source release
|
||
cost/sound order. Landing queues a full impulse before support clears Y;
|
||
use a ball-specific profile rather than Jimmy callbacks. VR07 target state
|
||
belongs to source instances despite duplicate tags, with strict parent hit
|
||
threshold and conditional collision activation. Exact TIME glyph binding
|
||
replaces the temporary font policy. Keep current sphere-pair discovery,
|
||
angular reflection, mouse/analog aim and helper retention across loads explicit.
|
||
40. **Save Scooter recovery before implementing it.** Its separate C3DJEEP actor,
|
||
mount/control handoff and wheel-dependent drive/support cannot be represented
|
||
by a Jimmy mode or an invented constant wheel count. Preserve the source
|
||
report/exporter and close wheel bodies before live Scooter integration.
|
||
41. **Use supplied jn-engine as research, not a replacement runtime.** Same EXE
|
||
declaration and415 byte-identical shared assets support cross-indexing its
|
||
class specs/decompiler bodies and adapting bounded tests. Different declared
|
||
OMT2.dll, mixed pointer bases, forced prototypes and explicit native gameplay
|
||
tunings require independent checks. Retain C#/.NET10/SDL3/Vulkan and REA byte
|
||
evidence. Capture tooling/derived scenes are present; raw OMTC, analysed
|
||
Ghidra database and proposed gameplay state traces are not. Record offered
|
||
permission before source import. Assessment: reviews/jn-engine-survey.md.
|
||
42. **Verify vehicle arithmetic before connecting live ownership.** RocketFlight
|
||
preserves per-callback input acceleration, support/brake ordering and retained
|
||
vertical-command publication, compared with28 isolated copied-original cases.
|
||
Radius50 is collision size, not render scale; rates650/−650 are acceleration,
|
||
not assigned velocity. Do not substitute Jimmy's body/camera for aircraft
|
||
callbacks or claim the scalar component is a live Rocketship. Resolve owned
|
||
versus placed properties, passenger/effect lifetime and control/camera phase
|
||
before VR02 integration. Grappler's geometry/material/swing contract likewise
|
||
forbids nearest-hook/direct-pull replacement. See gameplay/rocketship-component.md.
|
||
43. **Compose the owned Rocketship in boundedVR02.** Preserve source registration
|
||
order, rider static-body work, separate raw control/menu permission and
|
||
named-Jimmy cleanup.1E4 is a dynamic-pair flag, not an integration gate.
|
||
Dedicated source callbacks, Euler seam, maneuvers, LOOK-to-eye camera and
|
||
passenger/flame pivots replace provisional movement/render assumptions.
|
||
Require source/probe admission for other materials/levels; no nearest-ship,
|
||
invented boost or direct vertical-velocity substitute.96 isolated component
|
||
observations support contracts; live SDL/Vulkan replay verifies the port,
|
||
not whole-original or native Windows fidelity. See gameplay/rocketship-live.md.
|
||
44. **Identify field consumers before assigning render semantics.** Rocketship's
|
||
four impact writes are first-material diffuse RGBA, not shape scale. The
|
||
copied suffix probe validated numeric values but its synthetic buffer and
|
||
historical `scale` label did not establish their meaning. Constructor and
|
||
actual DLL renderer consumers now guard the contract. Use a ship-only red
|
||
material flash with unit geometry; regress vertex positions as well as values.
|
||
See gameplay/rocketship-impact-render.md.
|
||
45. **Keep inspection in the gameplay coordinate convention.** A right-handed
|
||
view/perspective pair can render a valid frame while mirroring a positive-Z
|
||
source scene relative to gameplay. OrbitCamera now owns the matching
|
||
left-handed view/projection; apply the shared Vulkan Y correction only once.
|
||
Regress screen right/up and depth across camera poses, not frame presence
|
||
alone. The modern inspector is separate from recovered original camera logic.
|
||
46. **Keep Grappler cursor/attachment and input/normal gates distinct.** The leaf
|
||
attachment helper lacks cursor filters, but aim's earlier true return ends
|
||
active input before that helper. Do not interpret it as a rejection bypass.
|
||
Attached input may return early while ordinary Jimmy frame publication still
|
||
runs. Compare isolated sqrt/first-scale/reel/damping with original snippets
|
||
before composing body/camera/helper ownership; do not invent250 clamps or a
|
||
detach impulse. GrapplerMotion remains isolated until those gates close.
|
||
47. **Keep Grappler query metadata and component boundaries explicit.** Capture
|
||
the actual first selected polygon/material with its hit. Preserve root versus
|
||
swing-pivot origins, quick-magnitude constraint order and strict>60 contact
|
||
detach. Native aim/constraint/post-base suffix probes do not establish the
|
||
full body context, base contact, camera or helper lifecycle; close those
|
||
contracts before liveVR08. See gameplay/grappler-query.md.
|
||
|
||
48. **Bound Grappler's shared collision and persistent helper state.** Reuse the
|
||
preceding source-body sphere context in attached47B4B0, including stale last
|
||
polygon callbacks; support/ray metadata is separate. Keep scalar memory at
|
||
session scope and helper/audio state at scene scope. Explicit RETURN cleanup
|
||
is recovered; ordinary reconstructed-scene disposal is a modern safety
|
||
boundary. Preserve first geometry scale across reattach/reequip/load. Camera
|
||
offsets are locals consumed by separate ground/air branches, not world poses.
|
||
Exact SWING uses its own source-pivot draw. See gameplay/grappler-live.md.
|
||
|
||
46. **Scooter owns a vehicle body and real wheel supports.** Normal item4 selection
|
||
summons at Goddard; strict later contact mounts. Keep scalar/live velocity/
|
||
pending impulse and body/input/frame/wheel phases distinct. Enable bounded
|
||
VR05 and outsideLEV1 from the normal owned tile. Use source body150/.2,
|
||
four narrow wheel offsets, explicit prevent-lowering/bounce overrides,
|
||
face-only camera clip and exact unscaled art. Preserve parked Jimmy collision.
|
||
Native component comparison, managed integration and port rendering establish
|
||
separate coverage; no complete original/Windows parity is inferred.
|
||
|
||
47. **Fix Scooter handoff consumers and sound channel processing.** Preserve
|
||
copied dismount Euler, then run source Goddard X/Z leveling with.3 factors
|
||
in the supported follow bridge. Preserve the nominal engine pitch command,
|
||
then apply recovered47D9D0 positional frequency/history and DX bounds.
|
||
Use shared listener state for mounted/parked/script cameras; global audio
|
||
cadence and attenuation remain separate. See the focused gameplay reports.
|
||
|
||
48. **Stabilize Jimmy from authoritative components while preserving the playable baseline.**
|
||
Use JN's original decompiler index as leads only. Integrate source ground/air
|
||
input, velocity, full-axis base facing and precise jump hold/gravity after
|
||
copied-original comparison. Preserve retained targets and keyboard-turn latch;
|
||
facing runs without raw input ownership. Require real support polygon338 for
|
||
downward ground pressure; do not translate capsule Grounded into that state.
|
||
The failed stair route and successful compatibility rerun are retained evidence.
|
||
Keep ordinary air pitch/actor-root visual bridge together as the next chunk,
|
||
and special/root-changing/dynamic body behavior behind independent gates.
|
||
See gameplay/jimmy-movement-stabilization.md.
|
||
|
||
49. **Keep air input, facing and the logical-root visual bridge together.**
|
||
Pitch seeds are retained apply-owned fields, not per-input mode constants.
|
||
Init clears destination Euler/dirty independently of current full STRT Euler.
|
||
Publish full authored/current XYZ and enable ordinary pitch/turn lean with
|
||
source hierarchical rotation/degree caching. Retain upright foot alignment
|
||
as an explicit local model correction around the logical root; do not call it
|
||
an original pivot offset or mutate source geometry/body support. Defer OMT's
|
||
final approximate-angle model decomposition with its measured seam mismatch.
|
||
Broader special/dynamic collision retains separate gates. See jimmy-air-bridge.md.
|
||
|
||
50. **Preserve contact-owned root and body modes instead of treating world-off as noclip.**
|
||
Keep current/previous full Vec4, anchor and source integration/support/
|
||
collision/world/alternate gates independently. Only fresh XYZW equality permits
|
||
publication of the original candidate; callback roots do not inherit stale
|
||
falling displacement/clamp. Landing callback root/anchor/acceleration/impulse/
|
||
modes persist, while cached velocity/support publication follows afterward.
|
||
World-off still integrates and may support; collision-off has its own hold/
|
||
translation branch. Preserve inactive-path context. Retain the lab/outside
|
||
capsule until real Jimmy callbacks and multi-frame/dynamic ownership are proven.
|
||
See checkpoint/body-contact-implementation.md.
|
||
|
||
51. **Camera orientation preparation replaces the facing destination too.**
|
||
Source CAM/MCA flatten current and game+274 destination to current yaw with
|
||
pitch/roll0; direct writes preserve facing dirty. Reset both on that existing
|
||
preparation boundary, not on every input lock/STOP/DEFAULT.20 isolated native
|
||
cases verify this state ownership. Neutral grounded input instead retains
|
||
residual destination bank in seven source trajectories; no unproven universal
|
||
landing snap is added. Another whole-frame writer or original quirk remains
|
||
open. See gameplay/jimmy-rotation-handoff.md and its two source reports.
|
||
|
||
52. **Recover full cutscene lifecycle before further symptom fixes.**
|
||
User redirects work to decompilation first after confirming a distinct
|
||
post-skip idle bank. Preserve the current runtime/packages; diagnose and
|
||
document original entry/contact/skip/control/input ordering before resetting
|
||
movement fields or changing the scheduler. Current source/component findings
|
||
do not yet explain the whole-original upright return. See
|
||
gameplay/jimmy-cutscene-decompilation.md for the recovery gate.
|
||
|
||
53. **Observe the actual original handoff when composed source matches the port.**
|
||
292 new bounded original-code cases do not establish another movement reset;
|
||
fresh input-before-player composition itself restores bank. Retain the user's
|
||
upright-original requirement without inventing an owner/DEFAULT/animation reset.
|
||
Use the copied-original read-only observer to measure actual route state and
|
||
registry membership. Its successful Wine32 attach is tool validation, while
|
||
MAIN-only black frames are not gameplay evidence. Keep runtime/packages fixed
|
||
until the source/differential gate closes. See gameplay/jimmy-cutscene-lifecycle.md.
|
||
|
||
54. **Classify successful TSK story entry as retained-player order.**
|
||
Source application startup loads backgroundLevel1D/Jimmy; successful TSK
|
||
destination requests use an existing controlled C3DJIMMY and the ordinary
|
||
selective teardown/reuse path. The new level game follows that retained node,
|
||
including first playable New Game. Program's carriedSave-only test was too
|
||
narrow. Include TSK-backed story entry, keep fresh direct-scene simulations,
|
||
and preserve rotation/movement algorithms. Baseline first-neutral input
|
||
republishes−10° bank; corrected startup order stays upright through300 idle
|
||
ticks. Original runtime/native Windows comparison remains distinct. See
|
||
gameplay/main-menu-player-lifetime.md.
|
||
|
||
55. **Retain source pair ownership before admitting a replacement scene loop.**
|
||
47A430 uses physical registration order, live callbacks and eligibility gates
|
||
that preserve old contacts. Sphere47CA90 checks only the earlier actor's
|
||
exclusions; successful responses precede independent membership notifications.
|
||
The28-case original/C# component implements that bounded contract with an
|
||
explicit box solver. Keep it outside live gameplay until full registration,
|
||
real actor callbacks and shared reclip dependencies close. Render visibility
|
||
and raw-control flags are not substitutes for physical eligibility. See
|
||
checkpoint/pair-lifecycle.md.
|
||
56. **Keep physical membership,49E and active collision as separate state.**
|
||
Common4623D0/464210 only use49E to gate the base hook; they can append the
|
||
same pointer again. Jimmy42AB50 guards the whole bridge. Registration runs
|
||
Init before physical append, while loading runs Init then PostLoad per live
|
||
registry node and visits new tails. Twelve original component cases confirm
|
||
these boundaries. Inventory actual constructor/owner-driven helper edges;
|
||
neither visibility nor an exposed InitPhysics method establishes membership.
|
||
Do not synthesize a full live body count or retained re-entry from the72-record
|
||
serialized partition. See checkpoint/physical-registration.md.
|
||
57. **Preserve shared context and validate kernel arguments before live admission.**
|
||
Full47A5E0 execution retains old hit/last polygon and can issue its static
|
||
callback after a dynamic hit. Final local-root publication overwrites callback
|
||
changes unless inhibited. Seventeen complete comparisons pass in a test-only
|
||
composition. Kernel A,C,B/retained construction normal is confirmed by all12
|
||
captured arguments; a same-position match with wrong nearest state was rejected.
|
||
Keep source topology and kernel boundary distinct. Empty-grid/synthetic actor
|
||
results do not admit live scene collision. Guard default Jeep deletion while
|
||
leaving cleanup-mutated active lifetime open. See checkpoint/dynamic-context.md.
|
||
58. **Follow evidenced float-store boundaries instead of accepting nearby roots.**
|
||
Populated reclip exposed an edge projection mismatch at triangle6, then a ray
|
||
hit mismatch at7. Delayed X projection/edge-root products, retained quick-
|
||
magnitude normalization and delayed hit-coordinate stores close14 complete
|
||
float32 comparisons and12 kernel steps without epsilon or near-zero clamps.
|
||
Keep public float magnitude/caller nearest state and source per-coordinate
|
||
stores distinct; double intermediates do not establish general x87 identity.
|
||
Count static node visits before the geometry gate, separately from actual
|
||
primitive tests. Real actor callbacks/lifetime still gate live integration.
|
||
See checkpoint/dynamic-reclip.md.
|
||
59. **Bind normal response by real class and preserve late scalar reads.**
|
||
Jimmy ordinary/null/jetpack374 does not enter inherited reflection378.
|
||
The inherited handler samples live Y again, preserves unclamped damping and
|
||
has a separate angular-heading tail. Implement that as an isolated component,
|
||
confirmed by30 complete original/C# snapshots. Soldier mode10 additionally
|
||
invokes complete174 with500; its RNG/action policy remains outside this
|
||
boundary. Sample inhibition/radius/42C/scene geometry at their source points,
|
||
not once before callbacks. Explicitly reviewed relocation operands replace
|
||
the probe's rejected cross-instruction numeric scan. No new live actor is
|
||
admitted from these finite fixtures. See checkpoint/actor-normal-contacts.md.
|