Files

36 lines
1.6 KiB
Markdown

# SpeedTreeRT static analysis
The scripts analyze the original x86 Windows DLL without executing or modifying it.
Generated pseudocode and Ghidra databases belong under `build/verification/` and are ignored by Git.
See `../../docs/SPEEDTREE-REVERSE-ENGINEERING.md` for the completed analysis and entry map.
Use an official Ghidra distribution with its native decompiler installed for your host,
and a compatible JDK (the completed run used Ghidra 12.1.4 and OpenJDK 21):
```sh
python3 tools/reverse_speedtree/run.py \
--ghidra-home /path/to/ghidra \
--java-bin /path/to/jdk/bin
```
The runner imports/replaces `SpeedTreeRT.dll` in the generated Ghidra project, runs
automatic analysis and calls `ExportSpeedTree.java`. It writes `decompiled.c`,
per-function files, function/call indexes, PE headers, disassembly, hashes and logs.
The completed run also has an `exports.tsv` cross-reference and `validation.json`
from a separate output completeness check.
On macOS arm64, if the distribution lacks `Decompiler/os/mac_arm_64/decompile`,
build the bundled C++ decompiler in Ghidra's own directory, outside the client repo:
```sh
cd /path/to/ghidra/Ghidra/Features/Decompiler/src/decompile/cpp
make -j 8 ghidra_opt ARCH_TYPE='-arch arm64' \
ADDITIONAL_FLAGS='-mmacosx-version-min=11.0 -w' OSDIR=mac_arm_64
mkdir -p ../../../os/mac_arm_64
cp ghidra_opt ../../../os/mac_arm_64/decompile
```
Decompilation completion is not proof of recovered source correctness. Runtime
library helpers are included, indirect calls may be missed, and inferred types
and local layouts require review and comparison with real DLL outputs.