agentic-workflow-manager 9.16.0 → 9.16.1

package/dist/src/core/journal/process.js

@@ -470,7 +470,10 @@ function descendantActivity(psTable, rootPid) {
     const children = new Map();
     let rows = 0;
     for (const line of psTable.split('\n')) {
-        const match = /^\s*(\d+)\s+(\d+)\s+(\S+)\s*$/.exec(line);
+        // time may be empty on some Darwin samples for a just-spawned process;
+        // treat blank as "0" so the pid still joins the descendant tree (otherwise
+        // a healthy detached grandchild is invisible and activity reads as frozen).
+        const match = /^\s*(\d+)\s+(\d+)(?:\s+(\S+))?\s*$/.exec(line);
         if (match === null)
             continue;
         if (++rows > MAX_PS_ROWS)
@@ -478,7 +481,7 @@ function descendantActivity(psTable, rootPid) {
         const pid = Number(match[1]);
         const ppid = Number(match[2]);
         const siblings = children.get(ppid) ?? [];
-        siblings.push({ pid, time: match[3] });
+        siblings.push({ pid, time: match[3] && match[3].length > 0 ? match[3] : '0' });
         children.set(ppid, siblings);
     }
     const seen = new Set([rootPid]);
@@ -499,13 +502,24 @@ function descendantActivity(psTable, rootPid) {
     return `${found.length}:${crypto_1.default.createHash('sha256').update(found.join('\n')).digest('hex')}`;
 }
 function observeDescendants(rootPid) {
-    try {
-        const table = (0, child_process_1.execFileSync)('ps', ['-A', '-o', 'pid=,ppid=,time='], { encoding: 'utf8', stdio: exports.EXEC_STDIO, timeout: PS_TIMEOUT_MS, maxBuffer: 16 * 1024 * 1024 });
-        return descendantActivity(table, rootPid);
-    }
-    catch {
-        return 'unknown';
+    // One transient `ps` failure under CI load used to freeze the snapshot at
+    // `unknown`, which made the Darwin integration test (and a real supervisor
+    // tick) treat a later concrete `0:none` as "activity changed" without ever
+    // seeing the grandchild. Retry once before degrading.
+    let lastError;
+    for (let attempt = 0; attempt < 2; attempt++) {
+        try {
+            const table = (0, child_process_1.execFileSync)('ps', ['-A', '-o', 'pid=,ppid=,time='], {
+                encoding: 'utf8', stdio: exports.EXEC_STDIO, timeout: PS_TIMEOUT_MS, maxBuffer: 16 * 1024 * 1024,
+            });
+            return descendantActivity(table, rootPid);
+        }
+        catch (error) {
+            lastError = error;
+        }
     }
+    void lastError;
+    return 'unknown';
 }
 /** Llamada UNA VEZ POR TICK por el supervisor (ver superviseController en
  *  commands/watch/supervisor.ts) mientras un controlador este activo — el

package/dist/tests/core/journal/process.test.js

@@ -202,28 +202,51 @@ describe('process identity', () => {
         test('ignores malformed ps lines instead of inventing processes', () => {
             expect((0, process_1.descendantActivity)('garbage\n  20    10 00:00:01\n\n  x y z', 10)).toMatch(/^1:/);
         });
+        test('keeps Darwin-style rows with blank or dotted cputime in the tree', () => {
+            // macOS `ps -o time=` is often `0:00.01`; a brand-new process can also
+            // omit the third column entirely on some samples.
+            expect((0, process_1.descendantActivity)('  10     1 0:00.05\n  20    10 0:00.01\n  30    20\n', 10)).toMatch(/^2:/);
+            expect((0, process_1.descendantActivity)('  10     1\n  20    10\n', 10)).toMatch(/^1:/);
+        });
     });
     test('activitySnapshot ve un nieto fuera del process group del controlador', async () => {
         if ((0, paths_1.isWindowsNative)())
             return; // win32 degrada activity sin ps (ver activitySnapshot)
-        const script = "setTimeout(() => { require('child_process').spawn(process.execPath, ['-e', 'setTimeout(()=>{}, 4000)'], { detached: true, stdio: 'ignore' }).unref(); }, 400); setTimeout(()=>{}, 4000)";
+        const script = "setTimeout(() => { require('child_process').spawn(process.execPath, ['-e', 'setTimeout(()=>{}, 15000)'], { detached: true, stdio: 'ignore' }).unref(); }, 200); setTimeout(()=>{}, 15000)";
         const { child, ref } = (0, process_1.spawnStructured)(['node', '-e', script], process.cwd(), 'n-desc');
         try {
-            const before = (0, process_1.activitySnapshot)(ref);
-            expect(before).not.toBeNull();
-            let after = (0, process_1.activitySnapshot)(ref);
-            for (let i = 0; i < 40 && after !== null && after.descendants === before.descendants; i++) {
+            // Wait for a concrete baseline (not `unknown` from a transient ps miss).
+            let before = (0, process_1.activitySnapshot)(ref);
+            for (let i = 0; i < 50 && (before === null || before.descendants === 'unknown'); i++) {
                 await new Promise((resolve) => setTimeout(resolve, 50));
-                after = (0, process_1.activitySnapshot)(ref);
+                before = (0, process_1.activitySnapshot)(ref);
+            }
+            expect(before).not.toBeNull();
+            expect(before.descendants).not.toBe('unknown');
+            const beforeGroup = before.groupSize;
+            let after = before;
+            const deadline = Date.now() + 8000;
+            while (Date.now() < deadline) {
+                const snap = (0, process_1.activitySnapshot)(ref);
+                if (snap !== null)
+                    after = snap;
+                // Success: same process group, at least one PPID-descendant appeared.
+                if (after.groupSize === beforeGroup && /^[1-9]\d*:/.test(after.descendants))
+                    break;
+                await new Promise((resolve) => setTimeout(resolve, 100));
             }
-            expect(after).not.toBeNull();
             // The detached grandchild lives in its own process group: the group
             // signal cannot see it, the descendant signal must.
-            expect(after.groupSize).toBe(before.groupSize);
-            expect(after.descendants).not.toBe(before.descendants);
-            expect(after.descendants).toMatch(/^1:/);
+            expect(after.groupSize).toBe(beforeGroup);
+            expect(after.descendants).toMatch(/^[1-9]\d*:/);
         }
         finally {
+            // Detached grandchildren are outside `-PGID`; reap by parent pid too.
+            try {
+                const { execFileSync } = require('child_process');
+                execFileSync('pkill', ['-P', String(ref.pid)], { stdio: 'ignore', timeout: 1000 });
+            }
+            catch { /* ya ausente */ }
             try {
                 process.kill(-ref.processGroup, 'SIGKILL');
             }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentic-workflow-manager",
-  "version": "9.16.0",
+  "version": "9.16.1",
   "main": "dist/src/index.js",
   "bin": {
     "awm": "./dist/src/index.js"