EN FR
EN FR


Section: New Results

Software Quality: History and Changes

Representing Code History with Development Environment Events. Modern development environments handle information about the intent of the programmer: for example, they use abstract syntax trees for providing high-level code manipulation such as refactorings; nevertheless, they do not keep track of this information in a way that would simplify code sharing and change understanding. In most Smalltalk systems, source code modifications are immediately registered in a transaction log often called a ChangeSet. Such mechanism has proven reliability, but it has several limitations. We analyse such limitations and describe scenarios and requirements for tracking fine-grained code history with a semantic representation. We want to enrich code sharing with extra information from the IDE, which will help understanding the intention of the changes and let a new generation of tools act in consequence. [24]

Mining System Specific Rules from Change Patterns A significant percentage of warnings reported by tools to detect coding standard violations are false positives. Thus, there are some works dedicated to provide better rules by mining them from source code history, analyzing bug-fixes or changes between system releases. However, software evolves over time, and during development not only bugs are fixed, but also features are added, and code is refactored. In such cases, changes must be consistently applied in source code to avoid maintenance problems. We propose to extract system specific rules by mining systematic changes over source code history, i.e., not just from bug-fixes or system releases, to ensure that changes are consistently applied over source code. We focus on structural changes done to support API modification or evolution with the goal of providing better rules to developers. Also, rules are mined from predefined rule patterns that ensure their quality. In order to assess the precision of such specific rules to detect real violations, we compare them with generic rules provided by tools to detect coding standard violations on four real world systems covering two programming languages. The results show that specific rules are more precise in identifying real violations in source code than generic ones, and thus can complement them. [25]