Assess supplied jn-engine research and capture evidence
This commit is contained in:
1 parent
c0acf58d9c
commit
c6eb1861cd
8 files changed
+347
-1
No files matched your search
@@ -0,0 +1,118 @@
|
||||
# Supplied jn-engine assessment
|
||||
|
||||
2026-11-03. User requested review of `../../../SURVEY-for-Rockman.md` and
|
||||
`../../../jn-engine-master/`. The survey was read in full; its generated claims
|
||||
were checked against supplied source/artifacts and our existing evidence.
|
||||
Read-only review of that tree; no friend code, legacy executables, deployment,
|
||||
network services or Ghidra scripts executed. GamePort source baseline: d6ca417;
|
||||
documentation HEAD on entry: c0acf58. No gameplay changes or new runtime claim.
|
||||
|
||||
## Conclusion
|
||||
|
||||
**Useful research and validation material, especially the saved decompiler bodies
|
||||
and original capture tooling. It is not a more complete faithful gameplay
|
||||
implementation to substitute for this port.** Keep C#/.NET10, SDL3 and Vulkan.
|
||||
Use their class/address index to accelerate REA investigations; independently
|
||||
check each imported interpretation. Their C/SDL2/OpenGL gameplay intentionally
|
||||
contains approximations which our recent recovered gadget slices already exceed.
|
||||
|
||||
Detailed independent reviews:
|
||||
[gameplay](jn-engine-gameplay.md), [capture/tooling](jn-engine-capture.md).
|
||||
A ready-to-share [evidence request](jn-engine-capture-request.md) records the
|
||||
most valuable additional material; nothing has been sent externally.
|
||||
|
||||
## Identity and corpus checks
|
||||
|
||||
Local SHA-256 recomputation versus `J:docs/binaries.sha256`:
|
||||
|
||||
| Binary | Our original | Friend's declared original | Consequence |
|
||||
|---|---|---|---|
|
||||
| Neutron.exe | 1,019,904 bytes; `f5281076e1a2492f2d61f87aae3f4544b6d0f535ebe33f9fbb6f78710fe1aea8` | Identical declared size/hash | EXE addresses can be cross-indexed |
|
||||
| OMT2.dll | 1,343,488 bytes; `3ec41ead011adf56387a6dbd4580e5b5a9721b5f8020c8a56420f42824354943` | 1,339,392 bytes; `a94f13e148bff790aab556d0381055df57f2ff4f9abc5947c25847c7370b0fea` | Reconcile DLL build before reusing DLL addresses or treating captures as our exact engine version |
|
||||
|
||||
Friend originals are not included, so the second column is a direct measurement
|
||||
and the third a supplied manifest declaration. A DLL difference does not make
|
||||
their measurements invalid; it limits their automatic applicability.
|
||||
|
||||
Case-insensitive filename pairing and SHA-256 comparison establish **all 254
|
||||
Original/ASE, 126 Original/png and 35 Original/gam files are byte-identical to
|
||||
their supplied counterparts**. Friend directories contain two additional ASEs
|
||||
(`bigbotstop.ase`, `sporkystop.ase`) and thirty additional PNGs. Do not merge these
|
||||
extras into our authoritative asset set. The survey's asset-count explanation is
|
||||
stale; our earlier 128 PNG count includes two root font PNGs outside Original/png.
|
||||
Our full original manifest verifies all539 files unchanged after this review.
|
||||
|
||||
## Evidence value and reuse order
|
||||
|
||||
Here `J:` means `../../../jn-engine-master/`, relative to this report.
|
||||
|
||||
| Material | Verified scope | Recommended use |
|
||||
|---|---|---|
|
||||
| Class ledger and reports | 208 ledger rows, all `spec`: 181 Medium/27 Low; reports present | Naming/address/dependency index, not a completion percentage |
|
||||
| Jimmy target6 decompiler output | Saved helper bodies including 428D50, 427FF0, 429D00, 426A70; generic signatures have artifacts | First gameplay reuse: compare Grappler/Rocketship recovery with REA and guarded EXE bytes |
|
||||
| C3DAnimated report/dispatch | Completion and selection timing rules, including repeated end callbacks | Candidate for replacing limited animation ownership after independent verification |
|
||||
| GAM schema, class IDs, SCENE/KITTY transition table | Source/data research over shared level corpus; several effects deferred | Expand campaign coverage checklist and identify missing owners/callers |
|
||||
| Linkage oracles | 16 substantive Python scripts plus template, building their C dumpers | Adapt selected expected-value functions/corpus cases to a GamePort adapter; not drop-in original-runtime tests |
|
||||
| DDraw proxy and OMTC protocol | Real draw/state/texture capture implementation; derived Level1 fixtures supplied | Calibrate selected Vulkan states after matching binary/version and matrix conventions |
|
||||
| Ghidra scripts and offered database | Scripts included; analysed project absent | Apply reviewed naming/function-boundary ideas to a copied project through our REA workflow |
|
||||
| Gameplay state tap / deterministic original replay | Proposed in survey; not implemented/included | Request a minimal phase-aware trace before building a replay consumer |
|
||||
|
||||
**Immediate recovery lead:** `J:docs/decomp/evidence/c3djimmy_target6.md:2556`
|
||||
contains Jimmy429D00, previously unidentified by our default analysis. Its
|
||||
header and `J:tools/ghidra/CreateFunctions.java` explain how explicit boundaries
|
||||
were created. This can unblock investigation, but forced `__thiscall`,
|
||||
`unaff_retaddr` and uninitialized aliases prevent treating the output as a
|
||||
finished algorithm or ABI. Maintain REA Evidence and byte checks when continuing.
|
||||
|
||||
## Corrections before adoption
|
||||
|
||||
- Movement is explicitly tuned: instant tank movement, jump650 and gravity1600.
|
||||
This is not evidence to replace our recovered550/800 behavior.
|
||||
- Their native gadget list includes Jetpack and omits Baseball; Shrink Ray uses
|
||||
raw6, conflicting with our recovered item6/raw6 Baseball. Invisible activation,
|
||||
Bubble placement, Grappler pull, Scooter mounting and Baseball lifetime also
|
||||
contain native design/tunings. Use class evidence, not these replacement rules.
|
||||
- Jimmy field maps mix complete-object and game-interface bases. Our established
|
||||
interface is complete+C0. Movement mode is complete+7C4/interface+704; jump state
|
||||
complete+7E8/interface+728; shadow distance complete+86C/interface+7AC.
|
||||
The survey mixes these bases in its proposed tap list. Its corrected shadow
|
||||
interpretation is also not propagated to `J:docs/decomp/C3DPlayer.md`.
|
||||
- Capture fixtures are derived evidence, not an included raw stream: 390 files,
|
||||
13,512,573 bytes; anchor8881 has3523 draws/342 textures at640x480. No `.omtc`
|
||||
or named original reference screenshot is present. The “oracle diff” checks
|
||||
fixture structure/counts/configuration, not pixels or gameplay trajectories.
|
||||
- No-fog/vertex-alpha/projection observations are capture-specific. Verify state
|
||||
and matrix storage before changing global rendering policies.
|
||||
- The survey compared older builds; its “other seven gadgets” and “no vehicles”
|
||||
coverage statements do not describe current GamePort recovery/integration.
|
||||
- No licence file was found in the supplied tree. Record the offered written
|
||||
permission or permissive licence before directly importing friend code/scripts.
|
||||
|
||||
Field/loop cross-checks: [movement checkpoint](../analysis/movement-checkpoint.md),
|
||||
[Bubble pointer/slot contracts](../gameplay/bubble-motion.md),
|
||||
[loop ownership](../checkpoint/loop-motion.md). Detailed citations for the C
|
||||
runtime differences and capture limits are in the two independent reviews.
|
||||
|
||||
## Concrete next steps and remaining questions
|
||||
|
||||
1. For the next gadget slice, use the saved429D00/426A70 bodies and Rocketship
|
||||
class report to guide a bounded REA recovery. Resolve pointer bases, callers,
|
||||
ownership and body dependencies before implementation; do not port prototype
|
||||
gadget movement.
|
||||
2. Request the analysed Ghidra database, raw OMTC with matching screenshots and
|
||||
exact capture DLL hashes. Their existing Level1 fixtures are useful now, but
|
||||
raw data is needed to reproduce the claimed renderer observations independently.
|
||||
3. Start the proposed original-state instrumentation with VR04 movement/jetpack,
|
||||
then lab→tutorial→lab→outside including Carl/RETURN. Require game-update
|
||||
phases, initial state and stable actor identity before planning replay tooling.
|
||||
4. Later adapt one camera oracle and animation-completion cases, explicitly
|
||||
separating recovered-model consistency from original-runtime comparison.
|
||||
|
||||
Open: capture DLL version/availability; raw evidence and analysed project;
|
||||
corrected/typed field maps and function signatures; original simulation cadence;
|
||||
renderer parity; actual state-tap feasibility; final-level/credits behavior;
|
||||
code reuse permission. None is silently resolved by this source intake.
|
||||
|
||||
Validation this turn: source/artifact review, exact binary/corpus hashes, ledger
|
||||
counts and original-manifest verification. No new build, port regression run,
|
||||
original execution or native Windows validation was needed or claimed.
|
||||
Reference in new issue
Block a user