fix(supervisor): treat incomplete marker/kind signals as fallthrough to schtasks probe

samson1357924 · samson1357924 · commit d7ba263f846a · 2026-06-21T09:59:37.000+08:00
Refine marker/kind detection logic in detectRespawnSupervisor: only
return null (skipping probe) when BOTH OPENCLAW_SERVICE_MARKER and
OPENCLAW_SERVICE_KIND are present and kind is not 'gateway'.
Previously, setting only marker without kind would short-circuit the
schtasks probe, producing false negatives on Windows scheduled tasks.

test coverage:
- probe via spawnSync with status=null (killed by signal)
- probe via spawnSync timeout exception (ETIMEDOUT)
- unrelated env vars (no known markers) still triggers probe
- whitespace-only marker after trim treated as absent
- marker set but service kind missing (incomplete signal)
- safe-integer boundary (Number.MAX_SAFE_INTEGER) for content-length
diff --git a/src/infra/supervisor-markers.test.ts b/src/infra/supervisor-markers.test.ts
@@ -183,4 +183,66 @@ describe("win32 schtasks probe fallback", () => {
     ).toBeNull();
     expect(mockSpawnSync).not.toHaveBeenCalled();
   });
+
+  it("returns null when probe status is null (process killed by signal)", () => {
+    mockSpawnSync.mockReturnValue({
+      status: null,
+      pid: 0,
+      output: ["", "", ""],
+      stdout: "",
+      stderr: "",
+      signal: "SIGTERM",
+    });
+    expect(detectRespawnSupervisor({}, "win32")).toBeNull();
+  });
+
+  it("returns null when probe throws on timeout", () => {
+    mockSpawnSync.mockImplementation(() => {
+      throw new Error("ETIMEDOUT");
+    });
+    expect(detectRespawnSupervisor({}, "win32")).toBeNull();
+  });
+
+  it("probes when env has unrelated keys but no known markers", () => {
+    mockSpawnSync.mockReturnValue({
+      status: 0,
+      pid: 0,
+      output: ["", "", ""],
+      stdout: "",
+      stderr: "",
+      signal: null,
+    });
+    expect(detectRespawnSupervisor({ PATH: "/usr/bin", HOME: "/home/user" }, "win32")).toBe(
+      "schtasks",
+    );
+    expect(mockSpawnSync).toHaveBeenCalled();
+  });
+
+  it("probes when marker is whitespace-only after trim", () => {
+    mockSpawnSync.mockReturnValue({
+      status: 0,
+      pid: 0,
+      output: ["", "", ""],
+      stdout: "",
+      stderr: "",
+      signal: null,
+    });
+    expect(detectRespawnSupervisor({ OPENCLAW_SERVICE_MARKER: "   " }, "win32")).toBe("schtasks");
+    expect(mockSpawnSync).toHaveBeenCalled();
+  });
+
+  it("probes when marker is set but service kind is missing (incomplete signal)", () => {
+    mockSpawnSync.mockReturnValue({
+      status: 0,
+      pid: 0,
+      output: ["", "", ""],
+      stdout: "",
+      stderr: "",
+      signal: null,
+    });
+    expect(detectRespawnSupervisor({ OPENCLAW_SERVICE_MARKER: "openclaw" }, "win32")).toBe(
+      "schtasks",
+    );
+    expect(mockSpawnSync).toHaveBeenCalled();
+  });
 });
diff --git a/src/infra/supervisor-markers.ts b/src/infra/supervisor-markers.ts
@@ -1,6 +1,10 @@
 // Defines process supervisor marker labels for gateway diagnostics.
 import { spawnSync } from "node:child_process";
-import { GATEWAY_LAUNCH_AGENT_LABEL, GATEWAY_WINDOWS_TASK_NAME, resolveGatewayLaunchAgentLabel } from "../daemon/constants.js";
+import {
+  GATEWAY_LAUNCH_AGENT_LABEL,
+  GATEWAY_WINDOWS_TASK_NAME,
+  resolveGatewayLaunchAgentLabel,
+} from "../daemon/constants.js";
 
 const SUPERVISOR_HINTS = {
   launchd: ["OPENCLAW_LAUNCHD_LABEL"],
@@ -100,9 +104,11 @@ export function detectRespawnSupervisor(
     if (marker && serviceKind === "gateway") {
       return "schtasks";
     }
-    // If service markers are present but don't match, respect the explicit
-    // signal (e.g. marker=worker means "not a gateway").
-    if (marker || serviceKind) {
+    // If both service markers are explicitly set and don't match gateway,
+    // respect the explicit signal (e.g. marker=worker means "not a gateway").
+    // If only one is set (incomplete signal), we can't be certain — fall
+    // through to the schtasks probe below.
+    if (marker && serviceKind && serviceKind !== "gateway") {
       return null;
     }
     // Fallback: probe schtasks directly when no env vars are set at all (e.g.
diff --git a/test/scripts/bounded-response.test.ts b/test/scripts/bounded-response.test.ts
@@ -99,6 +99,28 @@ describe("scripts bounded response reader", () => {
     },
   );
 
+  it.each(helpers)("accepts safe integer %s content-length values", async (_name, read) => {
+    const response = {
+      headers: new Headers({
+        "content-length": String(Number.MAX_SAFE_INTEGER),
+      }),
+      body: {
+        async cancel() {},
+        getReader() {
+          return {
+            async read() {
+              return { done: true, value: undefined };
+            },
+            async cancel() {},
+            releaseLock() {},
+          };
+        },
+      },
+    } as unknown as Response;
+
+    await expect(read(response, "probe", Number.MAX_SAFE_INTEGER)).resolves.toBe("");
+  });
+
   it.each(helpers)(
     "rejects unsafe decimal %s content-length values before reading",
     async (_name, read) => {
