mcp-watcher: safety-poll fallback for dropped inotify events

Adds a periodic timer (default 30s) that calls drain() unconditionally,
covering the case where chokidar/inotify silently drops an IN_MODIFY
event. Observed twice in production: ping appended to inbox, file mtime
updated, but no event delivered to the watcher; a sibling-file touch
unblocked it. Root cause is Linux inotify under brief idle gaps + atomic
writes — not consistently reliable on its own.

drain() is already idempotent (HWM comparison short-circuits when
nothing's new), so the steady-state overhead is one stat + JSON parse
per poll cycle. Event-driven path remains the primary; the poll just
masks the rare miss within the cycle interval.

- safetyPollMs option: default 30_000, set to 0 to disable
- stop() clears the interval before closing chokidar
- Two new tests: safety-poll delivers when fs-event never fires;
  safetyPollMs:0 truly disables the timer

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
bob-boat 2026-05-23 12:54:26 -04:00
parent 39a95ad49f
commit 95ff60ce94
2 changed files with 87 additions and 0 deletions

View file

@ -130,6 +130,67 @@ describe("InboxWatcher initial drain", () => {
await w.stop();
});
it("safety-poll delivers pings appended while no fs-event fires", async () => {
// Simulates the dropped-inotify-event case: chokidar mock never fires
// onChange. A ping is appended after start(); only the safety-poll
// timer can catch it. Asserts the watcher recovers within one poll cycle.
const paths = makePaths(dir);
mkdirSync(paths.pingsDir, { recursive: true });
writeInbox(paths.inbox("bob"), [ev({ id: "before-start", ts: "2026-05-06T10:00:00Z" })]);
const delivered: PingEvent[] = [];
const w = new InboxWatcher({
paths,
agent: "bob",
notify: async (e) => { delivered.push(e); },
// Chokidar mock that NEVER fires onChange — simulates a dropped event.
startWatcher: () => ({ async close() {} }),
safetyPollMs: 20,
});
await w.start();
expect(delivered.map((d) => d.id)).toEqual(["before-start"]);
// Append a new ping post-start, without notifying chokidar.
appendFileSync(
paths.inbox("bob"),
JSON.stringify(ev({ id: "after-start", ts: "2026-05-06T12:00:00Z" })) + "\n",
);
// Wait long enough for at least one safety-poll cycle.
await new Promise((r) => setTimeout(r, 80));
await w.stop();
expect(delivered.map((d) => d.id)).toEqual(["before-start", "after-start"]);
const hwm = JSON.parse(readFileSync(paths.hwm("bob"), "utf8"));
expect(hwm.last_delivered_ts).toBe("2026-05-06T12:00:00Z");
});
it("safetyPollMs: 0 disables the safety-poll timer", async () => {
const paths = makePaths(dir);
mkdirSync(paths.pingsDir, { recursive: true });
writeInbox(paths.inbox("bob"), [ev({ id: "a", ts: "2026-05-06T10:00:00Z" })]);
const delivered: PingEvent[] = [];
const w = new InboxWatcher({
paths,
agent: "bob",
notify: async (e) => { delivered.push(e); },
startWatcher: () => ({ async close() {} }),
safetyPollMs: 0,
});
await w.start();
// Append a ping; with no fs-event AND no safety-poll, it must NOT be delivered.
appendFileSync(
paths.inbox("bob"),
JSON.stringify(ev({ id: "missed", ts: "2026-05-06T11:00:00Z" })) + "\n",
);
await new Promise((r) => setTimeout(r, 50));
await w.stop();
expect(delivered.map((d) => d.id)).toEqual(["a"]); // only the initial-drain ping
});
it("skips pings already covered by HWM on restart", async () => {
const paths = makePaths(dir);
mkdirSync(paths.pingsDir, { recursive: true });