Empirical Study Reveals Unsoundness in Git's Default Merge Algorithm

rickasaurus · x · 2026-08-12

Marc Brooster released an empirical study on the soundness of Git's three-way merge. It reveals that under the default ort strategy, Git can silently produce an "unsound" merge — yielding incorrect results without flagging any conflicts.

Motivated by a 2007 paper on Diff3, the project provides a minimal reproducible counterexample. It also includes a fuzzer to hunt for such flawed merges and a minimizer tool to reduce counterexamples. The issue has been successfully reproduced in Git version 2.50.1.

Original post →

More from coding & agent

coding & agent channel →