Automated Fault Localization for Verification-Aware Languages: A Study on Dafny

Replication package for "Automated Fault Localization for Verification-Aware Languages: A Study on Dafny".

This archive contains a Docker image with all dependencies, datasets, and cached results pre-installed.

## Quick Start

1. Load the image:
   docker load -i dafny_research.tar.gz

2. Enter the container:
   docker run --rm -it dafny_research:latest bash

3. Follow REVIEWER_README.md inside the container (at /app/REVIEWER_README.md) for full reproduction instructions.

All research questions can be reproduced from cached results in under 5 minutes.
Individual techniques can be re-run from scratch to verify correctness.

## Contents
- Pre-built Docker image with Dafny v4.11.0, Z3 4.12.1, Daikon, .NET 8, Python 3
- 500-mutant dataset (sample_original_can_run) from DafnyBench
- Cached predictions for all techniques (RAND, CNTB, CNTS, CNTM, SNAP, LLM)
- Source code for all fault localization strategies
- Scripts to regenerate the dataset from scratch

## Requirements
- Docker (tested on Docker 24+)
- ~8 GB disk space (3.5 GB compressed, ~7 GB loaded)
- No internet connection needed (everything is self-contained)