RTC: Reject stale persisted CRDT documents#77890
Conversation
|
Size Change: +475 B (+0.01%) Total Size: 7.87 MB 📦 View Changed
ℹ️ View Unchanged
|
|
Following up on this because we've been chatting about it, I wouldn't take this too seriously as is, but here's an AI audit of the code: https://github.com/danluu/gutenberg/blob/codex/audit-pr-77889-77890/pr-77889-vs-77890-audit.md. From that doc, the lack of atomicity / CAS in this PR seems like an issue. Some of the other issues called out seem like straightforward things to fix, like using a 32-bit FNV hash as a token. For many of the other issues, I think I'd need a better understanding of WordPress and Gutenberg to really comment. If this gets into somewhat more solid shape, I think we can run the #77716 fuzzer on this. We could also do it now, but I think it would make sense to get this cleaned up first (although, as a future piece of work, automatically triggering the fuzzer against PRs and letting contributors choose to run "extra fuzzing" for some amount of time for a PR so that this is trivial to do would be nice). |
See also #77716
See also #77889
Vibe coded with GPT-5.5 in Codex