scripts for building func-io datasets on oss-fuzz supported open-source projects.
- python3.8 or above
- docker
- clang+llvm14, our init script will get this for you.
source ./scripts/env.sh
./scripts/init.sh
python3 scripts/projects.py -d <project> -p build_w_pass
python3 scripts/projects.py -d <project> -p fuzz_w_pass -ft 1s
where <project>
can be chose from build_w_pass.md.
We are working on C/C++ projects first since thet have the most fuzzer targets. The next step is to extend this project to Python and Java since these two languages are also widely used in LLM code-understanding tasks.