Modern configurable software systems are routinely built and tested across thousands of configuration variants to ensure correctness under diverse settings. Reusing build artifacts across configurations is therefore crucial to reduce redundant effort, yet remains challenging. Existing approaches estimate rebuild cost using syntactic option differences and apply greedy scheduling, which often misestimate true costs and yield suboptimal build orders. We present SIBS, the first semantics-aware scheduling framework for incremental builds across configurations, addressing both inaccurate cost estimation and the limitations of greedy scheduling. The key idea is a semantic distance metric that models the impact of configuration changes on rebuild scope by capturing option-level effects on code generation, accounting for global configuration headers, and propagating changes through dependency graphs. This enables cost-aware global scheduling formulated as an Asymmetric Traveling Salesman Problem (ATSP). Across 21 real-world systems and five independently sampled configuration batches, SIBS reduces observed build time relative to the state-of-the-art build scheduler in 17 systems. On average, it saves 196s (21%), with a median saving of 40s (3%), and achieves a 27.7% geometric-mean reduction in build time across projects. Batch-level reductions range from 25.7% to 29.3%, and the scheduling overhead remains modest (median 3.1s; mean 12.9s). These results demonstrate that semantics-aware cost modeling and global scheduling provide substantial and robust efficiency gains across diverse configurable systems.

@inproceedings{paperbot12483,
  title = {Beyond Syntactic Differencing: Semantic-Aware Scheduling for Cross-Configuration Incremental Builds},
  author = {Xiaolong Dai and Shanshan Li and Jun Lyu and Xintao Niu},
  booktitle = {ASE},
  year = {2026},
  doi = {10.1145/3832783.3837533}
}