Diagnosing Software Faults Using Multiverse Analysis

Authors: Prantik Chatterjee, Abhijit Chatterjee, Jose Campos, Rui Abreu, Subhajit Roy

IJCAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments show that the Multiverse Analysis not just improves the efficiency of fault localization but also achieves better coverage and generates smaller test-suites over DDU, the current state-of-the-art technique. On average, our approach reduces the developer effort over DDU by over 16% for more than 92% of the instances. Further, the improvements over DDU are indeed statistically significant on the paired Wilcoxon Signed-rank test.
Researcher Affiliation Academia Prantik Chatterjee1 , Abhijit Chatterjee1 , Jos e Campos2 , Rui Abreu3 and Subhajit Roy1 1Indian Institute of Technology, Kanpur, India 2LASIGE, Faculdade de Ciˆencias, University of Lisbon, Portugal 3INESC-ID and IST, University of Lisbon, Portugal
Pseudocode Yes Algorithm 1: Ulysis
Open Source Code Yes Ulysis is available in EVOSUITE as part of pull request #293, https://github.com/Evo Suite/evosuite/pull/293.
Open Datasets Yes We have performed our experiments on DEFECTS4J version 1.4.0 [Just et al., 2014] which is a benchmark suite consisting of six diverse Java project repositories.
Dataset Splits No The paper describes evaluating performance on '111 valid instances' from the DEFECTS4J benchmark and generating multiple test-suites, but it does not specify explicit training, validation, or test dataset splits in terms of percentages or counts for data partitioning.
Hardware Specification Yes We have done these experiments on a 16 core virtual machine with Intel Xeon processors having 2.1 GHz core frequency and 32 gigabytes of RAM.
Software Dependencies No The paper mentions using 'EVOSUITE' and 'GZOLTAR tool' but does not provide specific version numbers for these software dependencies in their experimental setup.
Experiment Setup Yes To take into account the randomization within EVOSUITE, for each fault, we have generated 5 test-suites using a time limit of 600 seconds on each fitness function. ... All experiments are performed at branch granularity, i.e., the program components are branches.