MO-HO Replication: Crash Reproduction Data for 124 Hard-to-Reproduce Crashes

Pouria Derakhshanfar from Delft University of Technology provides the replication package for the ASE 2020 study 'Good Things Come In Threes: Improving Search-based Crash Reproduction With Helper Objectives'. The dataset supports the evaluation of the MO-HO multi-objectivization approach, which was tested on 124 hard-to-reproduce crashes from open-source projects. Results indicate MO-HO reproduced five crashes unreproducible by prior methods and improved effectiveness by up to +10% in reproduction ratio.

Use Cases

• Benchmarking multi-objective evolutionary algorithms (e.g., NSGA-II, SPEA2) for crash reproduction based on the described helper objectives.
• Evaluating the impact of test length and method sequence diversity objectives on search-based test generation.
• Comparing the MO-HO approach against single-objective search and decomposition-based multi-objectivization (De-MO) for software debugging.
• Analyzing the performance of evolutionary algorithms on hard-to-reproduce crashes from open-source projects.

Strengths

• Dataset is associated with a peer-reviewed study accepted at the ASE 2020 conference.
• Evaluation is based on 124 hard-to-reproduce crashes from open-source projects.
• Results include specific performance improvements, such as a +10% reproduction ratio increase over a single-objective baseline.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Delft University of Technology

Collection Method

Replication package for an academic study; likely contains experimental results, code, and data used for evaluation.