| In this research, we focused on the bugs (error fixes and non-error changes) in the Eclipse Project, which has 220 modules with 33,904 files, and 3,780,201 lines of code (LOC). We developed a tool to collect the bug data from the Eclipse CVS repository and Bugzilla. We analyzed the data by testing 10 hypotheses. Our research showed that total changes (error and non-error together), error changes, and non-error changes over the decade-long evolution had the Power-Law distribution at the module and file level. When we classified error changes into more detailed categories, we found that all but one bug category had the Power-Law distribution at the module level. At the file level, about two thirds of the categories did not have the Power-Law distribution. Our research showed that there was no correlation between LOC, eLOC, 1LOC, Comment, and Lines and the number of bugs at module and file level. However, a small percentage of modules having a large percentage of bugs also contained most of the code size (LOC). Nevertheless, a small percentage of files having a large percentage of bugs did not contain most of the code size. |
