一句话：提出 SplitCA，可在资源限制内为绝大多数超千选项的大规模可配置系统生成 4-wise 覆盖数组，显著提升可扩展性。
高度可配置软件系统的配置空间随选项数指数增长，难以穷尽测试。组合交互测试（CIT）通过生成 t-wise 覆盖数组（CA）来缓解该问题，其中 4-wise CIT 能发现低强度测试遗漏的重要故障。然而，现有约束覆盖数组生成（CCAG）工具难以为选项数超过一千的大规模系统生成 4-wise CA。本文提出 SplitCA，一个面向大规模 4-wise CA 生成的有效工具。论文在 118 个公开真实世界大规模高度可配置软件实例上进行评估，每个实例均超过一千个选项。在合理时间与内存限制下，现有 4-wise CCAG 工具未能为任何实例生成 4-wise CA；SplitCA 则为其中 117 个实例成功生成，大幅提升了 4-wise CCAG 的可扩展性。
Highly configurable software systems are difficult to test exhaustively because their configuration spaces grow exponentially with the number of options. Combinatorial interaction testing (CIT) addresses this problem by generating t-wise covering arrays (CAs), and prior studies show that 4-wise CIT can detect important faults missed by lower-strength testing. However, existing constrained covering array generation (CCAG) tools still struggle to generate 4-wise CAs for large-scale systems with more than one thousand options. To address this limitation, we propose , an effective tool for large-scale 4-wise CA generation. We evaluate on 118 public real-world large-scale highly configurable software instances, each with more than one thousand options. Under reasonable time and memory limits, existing 4-wise CCAG tools fail to generate a 4-wise CA for any of these instances. In contrast, successfully generates 4-wise CAs for 117 instances, substantially improving the scalability of 4-wise CCAG. Video: https://youtu.be/9uM1Vl2ph-o Repository: https://github.com/chuan-luo-group/SplitCA or https://doi.org/10.5281/zenodo.21360896
@inproceedings{paperbot12338,
title = {SplitCA: An Effective Tool for Generating 4-wise Covering Arrays for Large-Scale Highly Configurable Systems},
author = {Shuangyu Lyu and Ruizhi Shi and Yuechen Duan and Chunming Hu and Chuan Luo},
booktitle = {ASE},
year = {2026},
doi = {10.1145/3832783.3834623}
}