Zengo Furukawa

A prototype of a concurrent behavior monitoring tool for testing of concurrent programs

Testing of concurrent programs is much more difficult than that of sequential programs. A concurrent program behaves nondeterministically, that is, the program may produce different results with the same input data according to execution timings of the program. In testing of concurrent programs, test data must specify not only input data but also sequences of statements. Ordered Sequence Testing Criterion for length k (OSC/sub k/), which was proposed by the authors, requires execution of all sequences of length k of concurrency statements which cause concurrent actions in a concurrent program. A monitoring tool has been developed for applying the testing criterion OSC/sub k/ to the testing of C concurrent programs on UNIX system. The tool measures coverage with regard to k-tuples of concurrency statements (OSC/sub k/) in source codes of a C concurrent program using a probe insertion method. The analysis of the tools output for a practical C concurrent program shows not only applicability of the tool for testing of concurrent program but also the necessity of a supporting tool for forcing execution of concurrency statements.

Coverage Based Testing for Concurrent Software

Concurrent programming is playing key roles for improving software performance since multi-core processors are used on various computing system. However, there may be not standard methodology for testing concurrent program. Especially, it would be difficult to know the quality of concurrent software. In this research, one of quality indexes, which would be defined by our research, is concurrent coverage criteria and is used for concurrent software testing. The concurrent coverage criteria would intend to find concurrent software specific defects, such as race conditions.

Design and implementation of test-case generation for concurrent programs

Test-cases play an important role for high quality of software testing. Inadequate test-cases may cause bugs remaining after testing. Overlapped ones lead to the increases in testing costs. This paper proposes the Event InterActions Graph (EIAG) representing behavior of concurrent programs including any task-type and the cooperated paths (copaths) on the EIAG as test-cases, and describes the test-case generation tool (TCgen) for concurrent programs written in Ada programming language. The EIAG consists of Event Graphs and Interactions. An Event Graph is a control flow graph of a program unit in a concurrent program. The interactions represent interactions such as synchronizations between the program units. TCgen generates test-cases as copaths from an Ada concurrent program. The generated copaths satisfy given testing criteria. They can find some communication errors in testing and detecting unreachable statements which concern interactions. It is, however, necessary to validate feasibility of the generated copaths.

A method for structural testing of Ada concurrent programs using the Event Interactions Graph

Software testing generally proceeds as follows: generating test-cases, selecting test-data, executing a test target program, inspecting execution result and evaluating whether testing has already been sufficient or not yet. As for methods for structural testing of programs, the way using a coverage, where the coverage means what extent given testing criteria are satisfied, is noted. At the evaluating step, whether or not we finish the testing is determined in view of the coverage. This paper proposes a method for structural testing of concurrent programs written in Ada programming language, especially, test-case generation and execution of the programs. The Event InterActions Graph (EIAG) is used as a model for concurrent programs. The EIAG consists of Event Graphs and Interactions. An Event Graph is a control flow graph of a program unit in a concurrent program. The Interactions represent interactions between the program units. Program units are such as procedures, functions and task-types. After generating test-cases on the EIAG, a method for selecting test-data is described and measures to cope with infeasible test-cases with which are generated in this step is clarified. And a forced execution of a test target concurrent program in order to solve the nondeterministic execution is investigated. The nondeterministic execution is characteristic of concurrent programs.

Mutation Testing of "Go-Back" Functions Based on Pushdown Automata

A go-back (GB) function for canceling recent user or system operations and going back to and resuming of previous state(s) is very often used regardless of the application domain. Therefore, faulty handling of them can cause severe damages in those applications. This paper proposes a mutation-based approach to testing GB functions modeled by pushdown automata. Novel mutation operators, recent coverage criteria, and a new algorithm for test case generation are introduced. A case study validates the approach and discusses its characteristics.

Concurrent N-Switch Coverage Criterion for Generating Test Cases from Place/Transition Nets

A PN (place/transition net) is known as an algebraic model that is useful for describing concurrent, distributed and asynchronous properties of software, and it is also recognized as a model of MBT (model-based testing) that is an important technique for developing high quality software. We propose a concurrent N-switch coverage criterion to generate test cases of high coverage levels from a PN representing test specifications of software. All sequences of successive fireable transitions of length N+1 on the PN can be covered by executing test cases that satisfy the concurrent N-switch coverage criterion. This paper shows its processes, work products, algorithms, and examples.

Constructing a usage model for statistical testing with source code generation methods

In statistical testing, testcases are generated using a usage model which defines usage of software as probability distribution on a state machine diagram. The aim of this study is to automatically construct a usage distribution from operational data, which assists construction of a reasonable usage model at low cost. This paper shows that applying source code generation methods based on a state machine diagram makes it possible. The prototype system has been developed to support the presented method.

Construction Method of a High-Order Markov Chain Usage Model

Statistical testing is a method for evaluating software reliability. Test cases are generated according to a usage model that represents the behavior of users. Accuracy of a usage model is an important factor since effectiveness of statistical testing (i.e., accuracy of software reliability) depends on it. This paper shows the usage model construction method using high-order Markov chains in order to construct accurate usage models. It is illustrated that the usage models constructed by this method are more accurate than traditional ones based on a simple or time-homogeneous Markov chain and can be constructed automatically. Some problems for more effective statistical testing are then found.

Development of virtual learning environment for discovery learning in school education

This paper focuses on learning which emphasizes experience among discovery learning schemes. Knowledge gained by experience is clearly stable compared with learning acquired by simple symbol manipulations. Thus, realizing a learning environment supporting discovery learning using virtual reality (VR) technology is attempted. The VR technology has many characteristic functions and can provide an interactive learning environment with a high degree of freedom which is required in supporting discovery learning which relies on experience. In this paper, considering applications to curriculum education, in which discovery learning is performed frequently, a framework of a virtual learning environment (VLE), which is a learning environment, integrating the functions of a VR environment and an ITS environment is proposed.

Test Strategies Using Operational Profiles Based on Decision Tables

In this paper, we propose a DTOP (Decision Table-based Operational Profile) that is a novel operational profile constructed by giving probability distributions to a decision table, and then discuss test strategies using the DTOP in order to effectively apply OPBT (Operational Profile-Based Testing) to software that includes complex logical relationship among inputs and outputs.