JSVT4: A review of software testing by means of statistical methods

A methodology for software validation (correctness checking from a statistical point of view) was presented in [1]. The methodology was implemented for Java software validation, by means of JSVT3, an Eclipse plugin. However, JSVT3 had some usability problems making it impractical. Particularly, the assertions checker used has several implementation problems. Furthermore, the methodology relies on the developer to define probabilistic distributions for the test data. However, an unconscious developer that does not know about the related data statistical distributions could take bad decisions (e.g., always choosing classical distributions) that could lead to select non-representative test data. An initial solution for these problems is presented, by means of JSVT4, a new implementation of the methodology introduced in [1]. The asserttions are checked by means of Junit+, an extension of Junit that allows validation of quantified first order logic predicates. Junit+ is integral part of JSVT4, so that this does not depend on external software for its execution.