How javapathfinder/jpf-core works
Project type and purpose
• Java PathFinder (JPF) is an extensible model-checking framework for Java bytecode. It is not a web app or database-backed service, but a standalone CLI tool and library that loads Java class files, interprets or simulates their execution in a custom VM, and systematically explores thread interleavings and state spaces.
File structure
• Top-level scripts (bin/) and build configs (build.gradle, settings.gradle, jpf.properties) form the CLI and bootstrap.
• src/main/gov/nasa/jpf: core engine (Config, JPF, JPFClassLoader, Shell, Property, State, VM).
• src/main/gov/nasa/jpf/jvm: classfile reader, bytecode interpreter, InstructionFactory.
• src/main/gov/nasa/jpf/search: state‐space search strategies (DFSearch, RandomSearch, heuristics).
• src/main/gov/nasa/jpf/listener & report: observer plugins and output publishers (ConsolePublisher, XMLPublisher).
• src/peers: native‐method “peers” that replace real JVM native methods.
• src/annotations: Java annotations to mark JPF options, filters, perturbations.
• src/main/fault: fault/injection examples and utilities.
• src/classes: generated annotation proxy classes.
• src/tests & src/examples: test suite and usage examples.
README
• Confirms that JPF is a CLI tool that runs on OpenJDK 11, highly plugin-oriented, with extensive wiki and community documentation.
Final advice
• Major modules and how they interact without overwhelming with every single class.
• Extension points explicitly.
• Document configuration loading, execution loop, and plugin invocation.
• Remain detailed: break the VM into ClassLoader, Interpreter, InstructionFactory, but group low-level instructions under “Bytecode Interpreter.”