Whole Program Inference (WPI) Repository

This repository contains all the necessary scripts, files, and data for running Whole Program Inference (WPI).

Repository Link

GitHub Repository

Execution Steps

The specific steps to run WPI are as follows:

cd wpi
python3 run_wpi.py

This script manages the WPI outer layer and logs the results produced by the tool. 3. After execution, the resulting .ajava files will be stored for each benchmark in the directory root-dir/dataset/benchmark-directory/src/wpi-iterations. 4. The results of running the checker before and after WPI inference can be found in root-dir/wpi/checkerframework-3.34.0-results-WPI.

In case there is a problem:

Configuration

Ensure to configure the parameters used in root-dir/wpi/run_wpi.py and root-dir/wpi/wpi/wpi.sh as needed. The setup has been tested and run on macOS.

Benchmark Structure

To run WPI on all benchmarks, the benchmarks must be structured in a standard format as proposed by the developers of the NJR dataset. This structure should include directories for classes, source files (src), information (info), JAR files (jarfiles), and libraries (lib). The most critical directories are src, which contains all the .java files, and lib, which contains all necessary .class files.

Compiler Compatibility

Before execution, ensure that the benchmarks can be compiled using javac, as required by the checker framework. Download and prepare the latest release of the checker framework from GitHub Releases.

Checker Framework Setup

Copy the checker directory from the downloaded framework to reside within the root directory at scripts/tools/checker-framework-3.34.0.