Research On GUI Defects And Location Methods

The bugs in graphic user interface(GUI) may be derived from the code smells in code source, giving rise to refactoring problems, and leading to GUI bugs. In the assistance of bug reports, it is easier to locate the defect code and fix the bugs in GUI. However, there are increasing bug reports, and programmers cannot find the bug reports effectively.For the purpose of studying:(a) relationship between code smells and the GUI bugs,(b) Locating GUI bugs through bug reports. This paper mainly focus on the following three aspects—code smells detection, bug reports extraction approach and GUI testing.For the code smells detection, this paper made a research on detection features, such as differentiation, antigen and signal processing. Essential concepts and signals are migrated to software engineering. The paper presents a detection paradigm, where the algorithm is based on dendritic cell algorithm in danger theory, which regards code smells as antigen. Software metrics values convert to the danger signal and the safe signal for processing, in which mature signal and semi-mature signal is calculated by weight equation. Code smells can be confirmed in comparison of relative values. Variety of code smells’ priority is determined by mature context antigen value. There are lower false positive rates in the paradigm. The experiments proves that this approach is competitive effectiveness in F-score(0.784) as well as Kappa analysis(0.756)and outperformance compared to other detection technique.For bug reports extraction approach, the paper proposes a novel extraction approach, which synonyms were merged into one specific word firstly in the approach. Then, it sets up a vector space model, and some text mining methods, such as TF-IDF and information gain, are presented to collect word features for bug reports specifically, but also there is a word number based algorithm for determining sentence complexity, so as to choose the sentence in long length. This work introduced Bayes classifier into bug report extraction. TPR is increased and FPR is decreased in this approach. The experiments proves that bug report extraction by using data mining and Bayes classifier is competitive through the evaluation of AUC(0.71), F-score( 0.80) and Kappa value(0.75).For GUI testing, the paper brought in the Petri net theory in the discrete and parallel system, defining concepts of event, events sequence, and events decomposition in graphic user interface. The paper introduced some significant properties of Petri net, such as reachability, roundedness, liveness and strong connectedness to this field, so as to improve the coverage and efficiency of graphic user interface testing. In addition, an attempt to solve six categories bug in graphic user interface, such as non-reachability, not strong connected, dead-lock, unbounded, not suitable to the original model and error jumping is conducted. The experiments proves that graphic user interface testing based on Petri net is more effective than traditional simple random test in coverage of events, code lines as well as the number of fault detection.After the previous three basic research, there are two main conclusions in this research through experiments:(a) based on mutual information correlation analysis, code smells will bring in bugs in GUI. Particularly, if the software updates frequently, the reliably of GUI is hampered seriously, and(b) bug reports extraction can help the bug locating significantly.