Details

Autor(en) / Beteiligte

• Brykczynski, Bill R

Titel

Improving defect detection ability with adaptive software inspection checklists

Ort / Verlag

ProQuest Dissertations Publishing

Erscheinungsjahr

1999

Link zum Volltext

• ProQuest

Quelle

ProQuest Dissertations & Theses A&I

Beschreibungen/Notizen

• Defects are widely acknowledged to be a significant factor contributing to the cost of developing software. One way of helping to control cost is through software inspection, a group-based method for finding defects during software development. The inspection process is widely recognized as an effective method for detecting defects. In practice, however, an inspection generally does not detect all defects that exist in the reviewed workproduct, one reason being that the reviewers had not been prompted to check for the full range of defect possibilities. This dissertation addresses the problem of these latent post-inspection defects by describing a new process for creating inspection checklists that change in response to analyses of two defect distributions: (1) the distribution of defects previously found as a result of inspecting the developer's code, and (2) the distribution of recent defects likely being injected into the code by the entire development project. We call checklists created by this process adaptive checklists because they are updated after every inspection and are tailored to specific individual developers and development environments. These adaptive checklists are designed to improve a reviewer's defect detection ability during the preparation phase of an inspection. We conducted an industry experiment using adaptive checklists for code workproducts. The results of the experiment showed that most reviewers found more defects during the inspection preparation phase when they used an adaptive checklist as opposed to a static checklist. Also, inspections that used an adaptive checklist exhibited a higher detection rate of frequently occurring defects (both in terms of counts of defects and defects per thousand lines of code) than inspections using a static checklist. The results of this empirical investigation have led to the identification of a defined and repeatable process for creating inspection checklists that are regularly updated based on feedback from defect analysis. We also describe prescriptive guidelines for constructing adaptive checklists.