Authors:
Renáta Hodován
and
Ákos Kiss
Affiliation:
University of Szeged, Hungary
Keyword(s):
Test Case Minimization, Delta Debugging, Parallelization.
Related
Ontology
Subjects/Areas/Topics:
Software Engineering
;
Software Engineering Methods and Techniques
;
Software Testing and Maintenance
Abstract:
The first step in dealing with an input that triggers a fault in a software is simplifying it as much and as
automatically as possible, since a minimal test case can ensure the efficient use of human developer resources.
The well-known minimizing Delta Debugging algorithm is widely used for automated test case simplification
but even its last published revision is more than a decade old. Thus, in this paper, we investigate how it
performs nowadays, especially (but not exclusively) focusing on its parallelization potential. We present new
improvement ideas, give algorithm variants formally and in pseudo-code, and evaluate them in a large-scale
experiment of more than 1000 test minimization sessions featuring real test cases. Our results show that with
the help of the proposed improvements, Delta Debugging needs only one-fourth to one-fifth of the original
execution time to reach 1-minimal results.