In software testing, many troublesome faults are caused by interaction of input parameters. If automatic efficient test-case generator（AETG）, in parameter order（IPO） or orthogonal Latin square is used in the softw...In software testing, many troublesome faults are caused by interaction of input parameters. If automatic efficient test-case generator（AETG）, in parameter order（IPO） or orthogonal Latin square is used in the software system under test, the whole test set cannot be run completely due to time or budget constraints. In this paper, according to the number of parameter k and their values n, a software system under test can be divided into four types. As for k-2 system, an algorithm was proposed to construct test rases, based on the longitudinal binary string set and method of controlling variables. As for k-n valued parameters whose n was a prime or power of prime, a method of covering array of test sets was designed to construct test sets by taking orthogonal array and derive arrays from orthogonal array and two useful conditions. As for k parameters whose n-value was not all equal, an experience algorithm was presented in this paper. The experimental results show that the size of test sets from the proposed methods is better than that from AETG, IPO, and orthogonal Latin square.展开更多
基金National Natural Science Foundation of China ( No. 61073163)Project of Science and Technology Commission of Shanghai Municipality,China ( No. 09220503000)
文摘In software testing, many troublesome faults are caused by interaction of input parameters. If automatic efficient test-case generator（AETG）, in parameter order（IPO） or orthogonal Latin square is used in the software system under test, the whole test set cannot be run completely due to time or budget constraints. In this paper, according to the number of parameter k and their values n, a software system under test can be divided into four types. As for k-2 system, an algorithm was proposed to construct test rases, based on the longitudinal binary string set and method of controlling variables. As for k-n valued parameters whose n was a prime or power of prime, a method of covering array of test sets was designed to construct test sets by taking orthogonal array and derive arrays from orthogonal array and two useful conditions. As for k parameters whose n-value was not all equal, an experience algorithm was presented in this paper. The experimental results show that the size of test sets from the proposed methods is better than that from AETG, IPO, and orthogonal Latin square.
