Internal view
 |
|
 |
| PROGRAMMABLE AUTOMATIC CODE INSPECTION | ||
|---|---|---|
| Tuesday 3 November 1998 | 17.00 | Brian Spratt Room |
| Professor Keith Paton Paton System Design, Quebec, Canada |
||
Work by Boehm and others suggests that in much software development, the majority of the effort is spent in a cycle of test-diagnose-fix operations. Fagan showed that manual inspection of the code for breaches of coding standards could detect code faults early in the process. The effort devoted to inspection and fixing faults early was much less than the effort that would have been spent fixing faults later in the process. Despite this, manual code inspection has not really caught on. Its costs are clear and have to be paid now; its benefits will appear only in the future and many developers believe they will not appear at all. Many of the tests traditionally carried out in manual code inspection can be carried out automatically by suitably programmed computer systems. Closed systems for automatic code inspection, eg LINT and GOLD-C, provide a very wide range of pre-defined tests that detect many of the faults that have been found to occur in practice. The tools also provide advice on stylistic issues, such as code indentation, that many users will want to ignore. In closed systems, therefore, the user is allowed to disable the tests individually but there is no mechanism for adding new tests. Open (or programmable) systems for automatic code inspection, eg Code Check and YACC, allow the user to define new tests to cover situations not covered in existing closed systems. Tools based on Code Check have been devised to allow a major auto maker to detect breaches of its company-specific coding standards, which include a fixed point mathematics library, an authorized subset of ISO Standard C and a standard convention for naming ISO Standard C and a standard convention for naming variables.
The methods used in programmable automatic code inspection can also be used to transform code from one style to another. This has advantages when humans prefer one style but a machine insists on another.
For further information please contact Professor Peter Brown on ext 3636 or Kim Holborn on ext 7695.
