fix(computer): report idempotent Stop state accurately

supervisor/src/docker-stop-state.test.ts

@@ -0,0 +1,21 @@
+import { describe, expect, test } from "bun:test";
+import { wasRunningBeforeStop } from "./stop-state";
+
+describe("Computer Stop state classification", () => {
+  test("reports only a running container as wasRunning", () => {
+    expect(wasRunningBeforeStop("running")).toBe(true);
+    expect(wasRunningBeforeStop("RUNNING")).toBe(true);
+
+    for (const status of [
+      "created",
+      "restarting",
+      "paused",
+      "removing",
+      "exited",
+      "dead",
+      "unknown",
+    ]) {
+      expect(wasRunningBeforeStop(status)).toBe(false);
+    }
+  });
+});

supervisor/src/docker.ts

@@ -1,4 +1,5 @@
 import Docker from "dockerode";
+import { wasRunningBeforeStop } from "./stop-state";
 import {
   BOT_LABEL,
   type ComputerNames,
@@ -1166,7 +1167,9 @@ export async function ensure(
 
 /** Stop this Bot's computer. Its storage is untouched, so its logins survive. */
 export async function stop(names: ComputerNames): Promise<boolean> {
-  if (!(await inspectOwned(names))) return false;
+  const existing = await inspectOwned(names);
+  if (!existing) return false;
+  const wasRunning = wasRunningBeforeStop(existing.status);
   try {
     // Long enough for Chromium to flush its profile, matching the compose grace period.
     await docker.getContainer(names.container).stop({ t: 30 });
@@ -1176,7 +1179,9 @@ export async function stop(names: ComputerNames): Promise<boolean> {
       throw new DockerUnavailableError(String(error));
     }
   }
-  return true;
+  // The gateway/audit contract is "was it running before this Stop?", not "did a container exist?".
+  // Docker keeps stopped containers so existence alone would report a second idempotent Stop as work.
+  return wasRunning;
 }
 
 /**

supervisor/src/stop-state.ts

@@ -0,0 +1,4 @@
+/** The gateway's `wasRunning` flag describes active state before Stop, not container existence. */
+export function wasRunningBeforeStop(status: string): boolean {
+  return status.toLowerCase() === "running";
+}

supervisor/tests/docker.integration.test.ts

@@ -16,7 +16,7 @@ async function available() {
 }
 
 test.skipIf(!(await available()))(
-  "supervisor Docker lifecycle in an isolated namespace (nine cases)",
+  "supervisor Docker lifecycle in an isolated namespace (ten cases)",
   async () => {
     const namespace = `supervisor-test-${crypto.randomUUID()}`;
     const child = Bun.spawn(
@@ -49,7 +49,7 @@ test.skipIf(!(await available()))(
     if (!summaryLine)
       throw new Error(`Missing fixture result: ${stdout} ${stderr}`);
     const summary = JSON.parse(summaryLine.slice(prefix.length));
-    expect(summary.completedCases).toBe(9);
+    expect(summary.completedCases).toBe(10);
     expect(summary.cleanup).toBe("complete");
     // Do not echo nested Bun summaries: scripts/test-ci.ts counts the outer suite's summary.
     console.log(summaryLine);

supervisor/tests/fixtures/docker-lifecycle.ts

@@ -295,6 +295,23 @@ describe("fleet lifecycle timestamps", () => {
   }, 90_000);
 });
 
+describe("idempotent Stop result", () => {
+  test("reports true only when the Computer was running before Stop", async () => {
+    await withDocker().supervisor.ensure(names, {
+      image: IMAGE,
+      environment: [],
+    });
+
+    expect(await withDocker().supervisor.stop(names)).toBe(true);
+    expect(await withDocker().supervisor.stop(names)).toBe(false);
+
+    const inspected = await withDocker()
+      .docker.getContainer(names.container)
+      .inspect();
+    expect(inspected.State?.Running).toBe(false);
+  }, 90_000);
+});
+
 describe("a computer that never answers", () => {
   test("fails instead of being handed out as ready", async () => {
     // A wait that cannot fail is a sleep: every computer that never came up was reported ready, and