maddo/Crusader_Decomp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

more work done

Browse source
This commit is contained in:
MaddoScientisto 2026-04-09 00:32:12 +02:00
commit d323bb28fc
68 changed files with 714 additions and 19 deletions
Download patch file Download diff file Expand all files Collapse all files

20
docs/presentation-callback-broker-layout.md
View file

@ -200,6 +200,26 @@ Why not first:
3. decide whether `0x45a6` is an owned buffer, fallback object, or adapter scratch lane
4. tighten the constructor/installation provenance in the live NE program, not only the raw notes
## Live Ghidra Authoring Status
Verified first live `PresentationCallbackBroker` batch landed on 2026-04-08.
- Created class owner `Remorse::PresentationCallbackBroker` in the active `CRUSADER.EXE` database.
- Re-anchored the global install/teardown pair into live `12d0:` helpers by direct access to the `0x4588/0x458c/0x4590/0x4594/0x4595/0x45a6` broker globals:
- `12d0:0513` -> `InitOnce`
- `12d0:0656` -> `TeardownOnce`
- Added short decompiler comments to both methods so the current `0x4588` lifecycle remains visible in-session:
- `InitOnce` now records the `0x4594` guard, state snapshot into `0x458c/0x4590`, broker install at `0x4588`, and fallback-buffer allocation at `0x45a6`.
- `TeardownOnce` now records the `0x4595` guard, broker clear, conditional slot `+0x0c` emit when `0x4590 != 0x458c`, slot `+0x04` release, and final cleanup path.
- Re-anchored the finalize-phase caller as well: `1288:0fc3` is now `allocator_phase_finalize_pass` in the live database with a comment recording that it exercises broker slot `+0x08` twice before sweeping allocator heads.
- Preserved two verified live slot `+0x0c` callers with decompiler comments instead of forcing premature renames:
- `1278:0616 Vport_1278_0616` uses the installed broker callback when the local vport path takes its fallback branch with `param_2 == 0`.
- `1320:1588 Dispatch_ModalGump` emits the broker callback before and after modal dispatch when the requested state pair differs from the current mode snapshot.
- This remains intentionally conservative live authoring:
- no forced `this` typing yet for the broker pointer parameter on `InitOnce`
- no live promotion yet of `allocator_phase_finalize_pass` under the class owner itself, because it is a broker caller rather than a broker method
- no attempt yet to rename the wider subsystem beyond the existing `PresentationCallbackBroker` placeholder
## Bottom Line
The `0x4588` family is now documented well enough to be treated as a real object candidate.