Context: Defect reporting is an important part of software development in-vivo, but previous work from open source context suggests that defect reports often have insufficient information for defect fixing. Objective: Our goal was to reproduce and partially replicate one of those open source studies in industrial context to see how well the results could be generalized. Method: We surveyed developers from six industrial software development organizations about the defect report information, from three viewpoints: concerning quality, usefulness and automation possibilities of the information. Seventy-four developers out of 142 completed our survey. Results: Our reproduction confirms the results of the prior study in that "steps to reproduce" and "observed behaviour" are highly important defect information. Our results extend the results of the prior study as we found that "part of the application", "configuration of the application", and "operating data" are also highly important, but they were not surveyed in the prior study. Finally, we classified defect information as "critical problems", "solutions", "boosters", and "essentials" based on the survey answers. Conclusion: The quality of defect reports is a problem in the software industry as well as in the open source community. Thus, we suggest that a part of the defect reporting should be automated since many of the defect reporters lack technical knowledge or interest to produce high-quality defect reports.
|Conference||INTERNATIONAL SYMPOSIUM ON EMPIRICAL SOFTWARE ENGINEERING AND MEASUREMENT|
|Period||01/01/1900 → …|