pub fn commit_barrier_ready(
canonical: &ReplicaState,
editors: &[&ReplicaState],
) -> Result<bool>Expand description
State-vector commit barrier (#crdtauth3): is canonical a consistent
cut — has it integrated every live editor’s ops up to that editor’s current
state vector?
This replaces the fragile finalize patch-ack (“did a queued patch
round-trip?”) with a state-vector ack (“is the canonical replica synced
through SV=N for every live editor?”), the structural root fix for the
no_ack / ipc_proof_insufficient / post-commit-worktree-corruption class:
a commit can only snapshot a state that provably holds every editor’s last
keystrokes, so un-propagated editor ops can never be lost at the commit
instant. With no live editors the barrier is trivially satisfied (the headless
/ git-authoritative path).