Preliminary defect data from the iterative development of a large C++ program (experience report)

Abstract

Preliminary results from the development of Rational Rose, a large (> 100 KLOC) program written in C++, suggest that high product quality can be achieved during a telescoped development schedule through the use of an iterative-development methodology. Data on defect density and discovery rate gathered on one phase of Rose development implies a low error rate in code delivered for functional test and supports 80-20 defect clustering, 80-20 distribution of noncritical versus critical defects, an association of defects detected during functional test with depth in the subsystem hierarchy, and the utility of data on rates of defect discovery for C++ programs.

Improving quality, one process change at a time

“…Rational followed the Booch iterative-development methodology for the third major iteration of the Rose product and found that high code quality was achievable, thus supporting the 80-20 defect clustering and bug criticality (80% of all bugs are found in 20% of the code). The author focused largely on bug distribution and established that more abstract subsystems had higher rates of defects, so showing that subsystem hierarchy plays an important part on the bug distribution [13]. Berger discusses her longitudinal study findings based on field observations and interviews dealing with the issues of trust, the lack of collaboration, and the censure between business managers when an Agile process is introduced into a bureaucratic government agency [9].…”