Results of the research project
First step of the project was discovering technics suitable to measure source code quality. Based on the results of this research we determined various source code quality indexes. These indexes can give good hints on the places containing or possibly containing bugs. We published an article on the proposed source code quality indexes in IEEE Transactions on Software Engineering.
Next step of the project was analysing and documenting the OpenOffice.org source code architecture. We compared various versions of the code, to get a good picture also on the source code evolution process (OpenOffice quality assesment report).
We developed a description language of the code pieces possibly containing errors ("bad smells") and, using this language, we prepared descriptions of patterns to be searched for in the following steps. We also developed a program to discover bugs in the code using these patterns. Based on the results we got 55 patches in 38 files. These patches have been shared with the OpenOffice developer community (Patches I in html | in pdf).
The database of the detected "bad smells" was analyzed and corrected, then handed over to a computer learning algorithm to enhance the search algorithm. The enhanced algorithm found 129 bugs in 58 files. These patches were also sent to OpenOffice. (Patches II in html | in pdf).
We also analyzed the issue database of OpenOffice to enhance our algorithm. (Issue database in html | in pdf).
The project results will also be presented on a press conferece on February 7, 2007.
