cid e943a1ec739b33ee8c9492080f38548d58ff0a16eba461d33fdeea820595670e · original · open access · via openalex
Title: Argos: Detecting Collaboration Antipatterns in Perforce Server Logs Authors: Hanifah Alpha-Boda; Yannick Francillette; Sylvain Hallé Venue: IEEE Gaming, Entertainment, and Media Conference (GEM) Year: 2026 Canonical: doi:10.1109/gem72276.2026.11747047 URL: https://doi.org/10.1109/gem72276.2026.11747047 Collector: openalex OA: true ABSTRACT Version control servers silently record every action performed by a development team, yet these logs are rarely exploited to detect counterproductive collaborative behaviors. In the video game industry, Perforce logs particularly dense activity on non-mergeable binary assets, rendering textual diff-based analysis inoperative and requiring an approach grounded in event sequences. We present Argos, a streaming tool designed to automatically detect five collaboration antipatterns from Perforce server logs. Among the 168 antipatterns catalogued in ReliSA, five have been operationalized: Collective Procrastination, Warm Bodies, Lone Wolf, Bystander Apathy, and a proxy for Cascading Branches. Argos’s architecture relies on two complementary phases. The first reproduces, offline, the four OpenMetrics instruments popularized by Prometheus (Counter, Gauge, Histogram, and Summary) to characterize the overall activity of the depot. The second integrates complex event processing via BeepBeep 3.13, enabling the detection of order-sensitive patterns, such as a file modified but never submitted. Order-sensitive detection relies on per-file finite-state automata that track edit, submit, and revert transitions without requiring the entire log to be loaded into memory. Evaluated on a 1.7 GB log from a multi-team game development, Argos extracts 16,262 unique events in about 14 seconds. It identifies 217 Edit-then-Revert sequences and 250 abandoned files. The agreement between the terminal value of the open-files gauge and the number of apathy alerts (250 in both cases) constitutes an internal cross-validation of the two phases. Argos’s source code is publicly available on GitHub to foster reproducibility and adoption by the scientific community.