Idea
Checking distributed systems by exhaustive simulation
Two servers read a shared counter, both add 1, both write back, and one increment is lost, but only when both reads land before either write. A test run sees one order of events, and a reviewer tracing the orders by hand can miss the bad one.
yichus dist tries every delivery order and declared fault up to a bound, and each
property comes back holds, violated with the schedule that broke it, or
inconclusive when the bound cut the search short.
The system is a Bosatsu model whose handlers return
their next state and messages as data.
See a lost update caught →See the reviewer benchmark →
Task 002 worked example: prevent a lost update with compare-and-swap
The checker’s schedule for a server that reads and writes back blindly:
1. deliver app1 -> db [Get] db(v=0,ver=0) app1(waiting) app2(waiting) 2. deliver app2 -> db [Get] db(v=0,ver=0) app1(waiting) app2(waiting) 3. deliver db -> app1 [GotValue(0,0)] db(v=0,ver=0) app1(done) app2(waiting) 4. deliver db -> app2 [GotValue(0,0)] db(v=0,ver=0) app1(done) app2(done) 5. deliver app1 -> db [Put(1)] db(v=1,ver=1) app1(done) app2(done) 6. deliver app2 -> db [Put(1)] db(v=1,ver=2) app1(done) app2(done) quiescent; both_counted requires v == 2, found v == 1: violated
The compare-and-swap version holds over all 106 schedules, and the search is complete.
Benchmark: reviewer accuracy with and without the checker
Fresh reviewers judged three defective servers that read as correct:
| Reviewer | Without the checker (reading only) | With the checker |
|---|---|---|
| Default-tier reviewer | 3/3 caught, correct schedules derived by hand | not run; reading already suffices at this size |
| Haiku-tier reviewer | 2/3; certified one defective variant as correct, high confidence | 3/3 caught, root causes taken from the checker's schedules |
Each cell is one trial, so this shows the case exists, not a rate. The task 010 run artifact records a second case, an attacker reflecting a server’s own nonce.
Ask your agent about anything here; it has the detail.