A curious facet of software development is the poor uptake of many of the principles of software testing in industrial environments.In some industries, notably the avionics and nuclear sectors, software producers have no choice but to demonstrate thorough testing of their software.Interestingly, parts of the telecommunications industry, although answerable to no certification body, have also always had a tradition of rigorously testing their software, probably as a result of pioneering research at Bell Labs in New Jersey back in the 1960s.From an industry viewpoint, there is often a fear that bad news, brought to light by a rigorous testing process, will extend the time-to-market and result in a loss of market share.However, part of the problem of poor uptake must be borne by academic researchers because industry still lacks answers to key motivating factors such as:• What is the quality gain per unit cost?• What is the benefit in terms of customer satisfaction?• What is the long-term financial benefit?Answers to these and other questions would help testers make a strong case to managers for adequate resources (time, tools and people) to perform testing to significantly higher standards.It is pleasing to see that the two papers included in this issue of the journal do address real-world issues of industrial relevance.Although the techniques discussed in the papers are very different (one is concerned with software inspections, the other with testing embedded systems using modelling, simulation and design-of-experiments), they do, in essence, share a common concern: the need for an effective oracle.An oracle is an arbiter of software behaviour that is able to judge what is correct or, at least, what is deemed to be acceptable.In the case of inspections the oracle is human of course; in the other case discussed here, the oracle is a simulation model of the desired system and is therefore amenable to automation.In the first paper, 'Eight maxims for software inspectors', Diane Kelly and Terry Shepard describe how they collected the comments of inspectors over a period of three years and organized these comments into themes, with the net result being the synthesis of the eight maxims, i.e. suggested rules for effective inspections.Software inspection is one of the few verification and validation (V&V) techniques widely used in industry.It could be argued that it should always be used because there are things that humans can find which no automated tool will detect.On the other hand, the expense of deploying human resources is considerable and a balance needs to be struck between human resource usage and automation.The paper does not consider the relationship with automation but it does address important aspects such as maximizing effectiveness by measurement, standards and control.For example, given the expense of such a people-centric activity, it is important to match people's expertise with the task at hand, otherwise effort could be wasted or, at best, it might be less fruitful

Abstract The paper considers some of the reasons advanced by commercial and industrial institutions in order to avoid using analysis and test tools. These insights serve to explain why so few tools are currently in use in such organisations. There is clearly a challenge for the testing community to meet, and this paper is a faltering step towards formulating the goals.

This paper reports results obtained from a static analysis of the NAG Mark 4 Fortran numerical algorithms library.

There are a number of practical difficulties in performing a path testing strategy for computer programs. One problem is in deciding which paths, out of a possible infinity, to use as test cases. A hierarchy of structural test metrics is suggested to direct the choide and to monitor the coverge of test paths. Another problem is that many of the chosen paths may be infeasible in the sense that no test data can ever execute them. Experience with the use of "allegations" to circumvent this problem and prevent the static generation of many infeasible paths is reported.

In this paper we describe the extension of an existing FORTRAN IV numerical software testbed (Hennell, 1978) to enable ALGOL 68 programs to be investigated. The extensions necessary were two-fold; firstly, a complete rewrite of the first phase, a static analysis in which the source text is reformatted (for reasons stated within this paper), analysed for all possible control jumps and statistics on language constructs are collected. The second major extension was to incorporate into the second phase an existing ALGOL 68 compiler which after some modification enables dynamic execution histories to be collected in a data base. These modifications to the compiler represent extensions to the language definition which enable user programs to trace themselves. The utilisation of this compiler restricts source code programs to be written in ALGOL 68s, an official ALGOL 68 subset (Hibbard, 1974). The third, analysis, phase is essentially identical in both the FORTRAN and ALGOL 68 systems.

This paper discusses the need for measures of complexity and unstructuredness of programs. A simple language independent concept is put forward as a measure of control flow complexity in program text and is then developed for use as a measure of unstructuredness. The proposed metric is compared with other metrics, the most notable of which is the cyclomatic complexity measure. Some experience with automatic tools for obtaining these metrics is reported.

This paper describes an experimental testbed facility designed to examine some of the problems which arise in the implementation of high quality numerical software libraries. The testbed is used to measure the effectiveness of test programs. Effectiveness here is used in the sense that these test programs should ensure that the routine implementation is error free rather than to examine the numerical properties of the algorithm. The testbed has been used in extensive investigations of the stringent test programs of the NAG numerical algorithms library (Ford and Hague, 1974) and continuation of this work is seen as a major application for the testbed.