Research On Approach To Software Fault Analysis And Quality Evaluation

The quality assurance of software produ~t acts a key role in software engineering. It is very difficult to evaluate software quality systematically and objectively. Although it is important to user of software product and the software development organization, it is a pity that the quality assurance technology is immature both in theory and practice presently.Firstly this dissertation studies the problems of evaluating software quality systematically and objectively. It puts forward a model and a standard draft of evaluating software quality. Then a suite of tools of measurement, testing and defect analyses is developed to support the model. With these tools, data of software characteristic can be obtained and accumulated.The primary content of this dissertation is as follows:1)The dissertation analyses software quality characteristics and some famous software quality models, puts forward a model and a standard draft of software quality evaluation. The model is supported by a suite of tools which includes the four tools we have developed. The software quality evaluation model defines and classifies the factors of software quality characteristics. By using these tools, the factors can be mapped to quality profiles, so the relationship between factors and quality characteristic can be set up to evaluate software quality. Based on lots of experiment data, by using neural network algorithm, the model has the ability of knowledge discovering and evolution.2)The dissertation studies the theory and technology of software complexity measurement, analyses the shortcoming of the C&K metrics suite for object-oriented design and provides a lot of improved metrics suite. Then the dissertation introduces a practical C++ measurement tool, SMTCPP, implemented with the improved metrics. SMTCPP parses to Ci--- programs by LL(l) method, extracts a lot of program抯 information, such as classes, members and objects; counts the indications, such as the number of method per class, the biggest complexity among methods, depth of inheritance tree, the number of children, coupling between object classes, response for class, and relative lack of cohesion in methods. The measure values are very useful to guide software process. The tool may also put the values into a database to collect sufficient data for building software quality evaluation model. The dissertation analyses the experiments for three practical programs. The result shows SMTCPP is useful.3)The dissertation studies the theory and technology of testing, analyses the characteristic of program errors and combines these technology into 00 technology to develop the tool of testing, which named ErrSeed, based on error injection arithmetic. By using this tool to a lot of C++ projects, a mass of testing data is obtained. The experiments show that the testing tool is valid and useful for software developer to calculate the number of error remained in program and the rate between the error injected?and the error checked out, and evaluate the ampleness of testing. The tool can also be used to collect the data into a database for building software quality evaluation model.4)The dissertation puts forward the concept of software defect measurement for the first time, studies the theory and technology of software fault analysis, the classification of software defect and automatic analyses method. In this part, a knowledge-based analysis tool, which extracts the components from source code of software by the syntax analysis, was developed. According to the defect抯 knowledge stored in the knowledge base and reasoning rules, it can reason out the software defect. Experiment results show that the approach is effective.5)The dissertation analyses the shortcoming of traditional software reliability theory and points out the limitation of Nelson model. Then it introduces a software reliability model based on software input domain.