Variability-Aware Event Sequence Graphs for Scalable Software Product Line Testing

=== Version 1.0.1 — Fault Detection Re-Run ===This version supersedes version 1.0.0. The RQ3 fault detection results have been re-computed after fixing an issue in the FaultDetector component used for the original RQ3 measurements. The original detector relied on a name-based vertex lookup that, when two events in a product-level model shared the same cleaned name, could resolve to the wrong vertex during mutant execution. This caused some mutants to be reported as killed by sequences that did not in fact traverse the omitted edge or visit the omitted vertex, which inflated the mutation scores reported in version 1.0.0 for several SPLs. The fix replaces the name-based lookup with a direct reference to the original product-level model held in memory during mutant execution, so each mutant is now compared against the exact graph that produced the test suite. The fix affects only RQ3; the RQ1 and RQ2 measurements (test generation time, throughput, edge coverage, peak memory) are not affected and are reproduced unchanged in this version. The corresponding manuscript revision uses the RQ3 numbers from this version. === Original description ===This archive contains the raw experimental data produced by the distributed evaluation in the manuscript "Variability-Aware Event Sequence Graphs for Scalable Software Product Line Testing." It accompanies the source-code repository at https://github.com/esg4aspl/esg-with-feature-expressions, which holds the implementation, the orchestration scripts, the feature-model inputs, and the analysis pipeline. A 10-node, 80-core DigitalOcean cluster generated the data over approximately 2,440 droplet-hours, evaluating the proposed Model Once, Generate Any approach against two baselines — a Structural Baseline (Event Flow Graph testing via the GUITAR framework) and a Stochastic Baseline (Random Walk traversal) — across eight Software Product Lines ranging from 12 valid products to 124 billion valid products: Soda Vending Machine, eMail, Elevator, Bank Account, Student Attendance System, Tesla, syngo.via, and Hockerty Shirts. For each Software Product Line, the archive contains per-product DOT graph files, per-product Event Flow Graphs, the GUITAR test sequences for the Structural Baseline, and the per-shard measurement CSV files for the three research questions on comparative efficiency, extreme scalability, and fault detection. The per-product test sequences produced by ESG-Fx and Random Walk are intentionally excluded — the manuscript's quantitative claims rely on the aggregated CSV measurements, and these sequences are regenerable by re-running RQ1 on the cluster. The tarball mirrors the directory layout of files/Cases/ in the source repository and is intended to be extracted in place over that directory. After extraction, the Python analysis pipeline in the source repository runs without modification and reproduces the manuscript's tables and figures. Detailed reproduction instructions are in REPRODUCIBILITY.md in the source repository at https://github.com/esg4aspl/esg-with-feature-expressions/blob/main/REPRODUCIBILITY.md