Supporting the grow-and-prune model for evolving software product lines

Zusammenfassung

Software Product Lines (SPLs) aim at supporting the development of a whole family of software products through a systematic reuse of shared assets. To this end, SPL development is separated into two interrelated processes: (1) domain engineering (DE), where the scope and variability of the system is defined and reusable core-assets are developed; and (2) application engineering (AE), where products are derived by selecting core assets and resolving variability. Evolution in SPLs is considered to be more challenging than in traditional systems, as both core-assets and products need to co-evolve. The so-called grow-and-prune model has proven great flexibility to incrementally evolve an SPL by letting the products grow, and later prune the product functionalities deemed useful by refactoring and merging them back to the reusable SPL core-asset base. This Thesis aims at supporting the grow-and-prune model as for initiating and enacting the pruning. Initiating the pruning requires SPL engineers to conduct customization analysis, i.e. analyzing how products have changed the core-assets. Customization analysis aims at identifying interesting product customizations to be ported to the core-asset base. However, existing tools do not fulfill engineers needs to conduct this practice. To address this issue, this Thesis elaborates on the SPL engineers' needs when conducting customization analysis, and proposes a data-warehouse approach to help SPL engineers on the analysis. Once the interesting customizations have been identified, the pruning needs to be enacted. This means that product code needs to be ported to the core-asset realm, while products are upgraded with newer functionalities and bug-fixes available in newer core-asset releases. Herein, synchronizing both parties through sync paths is required. However, the state of-the-art tools are not tailored to SPL sync paths, and this hinders synchronizing core-assets and products. To address this issue, this Thesis proposes to leverage existing Version Control Systems (i.e. git/Github) to provide sync operations as first-class constructs.