@deksden-com/dd-flow-cli 0.9.0-beta.37 → 0.9.0-beta.38

package/CHANGELOG.md

@@ -1,5 +1,11 @@
 # @deksden-com/dd-flow-cli
 
+## 0.9.0-beta.38
+
+### Patch Changes
+
+- 44e3c80: Allow control operations to recheck failed provider cleanup after a leaderless process group exits naturally. Coalesce concurrent finalization retries, preserve the original exit and process lease, and keep live or unproven groups blocked without sending them signals.
+
 ## 0.9.0-beta.37
 
 ### Patch Changes
@@ -1031,6 +1037,7 @@ record` command with the bounded 15-probe contract.
 - 16f68a3: Add router-native engine snapshot commands and routed project-command dispatch.
 
   Also include the runtime/dashboard compatibility wave needed by the current Memory Bank canon:
+
   - target-based dashboard `open`/`refresh` commands and related-command help;
   - persisted FIFO lane waiters and merge wait-acquire specialization;
   - `migration plan/report/verify` evidence helpers for controlled `mb-upgrade` runtime migrations;

package/dist/build-info.json

@@ -1,8 +1,8 @@
 {
   "cli_package": "@deksden-com/dd-flow-cli",
-  "cli_version": "0.9.0-beta.37",
-  "cli_commit": "04f94a5faa271bc0e837d6d3020f3a3668835e17",
-  "built_at": "2026-09-08T02:59:50.109Z",
+  "cli_version": "0.9.0-beta.38",
+  "cli_commit": "e1d723e620ee5aeec961c2bee1de1e72fcfb71a9",
+  "built_at": "2026-09-08T03:57:03.109Z",
   "built_with_canon": {
     "version": "4.0.6",
     "commit": "c6fc50cb3b5526ea0162ee4454d1ee36f17be2da",

package/dist/harness-runtime/lib/dd-agy-daemon.mjs

@@ -212,12 +212,16 @@ export class Runtime {
           if (this.active) { this.active.reject(error); this.active = null; }
           const patch = beforeInit ? { shutdown_state: "failed", startup_error: errorPayload(error), provider_exit: { code, signal }, active_tree: false } : { provider_exit: { code, signal }, active_tree: interrupted };
           await providerRegistered;
-          try {
-            await stopProcessGroup(providerChild);
-            await this.persist(patch);
-            await finishDaemonProcess(this.config, providerProcess, beforeInit || interrupted ? "failed" : "stopped", beforeInit || interrupted ? "provider_interrupted" : "provider_stopped");
-          } catch (cleanup) { await this.persist({ shutdown_state: "cleanup_failed", active_tree: true, cleanup_error: errorPayload(cleanup) }); throw cleanup; }
-          finally { if (beforeInit) this.server?.close(); }
+          this.retryProviderFinalization = async (graceMs = 1_000) => {
+            try {
+              await stopProcessGroup(providerChild, graceMs);
+              await this.persist(patch);
+              await finishDaemonProcess(this.config, providerProcess, beforeInit || interrupted ? "failed" : "stopped", beforeInit || interrupted ? "provider_interrupted" : "provider_stopped");
+              this.retryProviderFinalization = null;
+            } catch (cleanup) { await this.persist({ shutdown_state: "cleanup_failed", active_tree: true, cleanup_error: errorPayload(cleanup) }); throw cleanup; }
+            finally { if (beforeInit) this.server?.close(); }
+          };
+          await this.retryProviderFinalization();
         }).then(resolve, reject);
       });
     });
@@ -328,13 +332,25 @@ export class Runtime {
     await observeModel({ journal: this.config.journal, harness: "antigravity-cli", sessionId: conversationId, parentSessionId: parent, requested: this.config, observed, source: "native.hook", evidence: payload.model ? "configured" : "unavailable", reason: payload.model ? null : "hook_omits_current_model" });
     const eventId = createHash("sha256").update(JSON.stringify({ daemon: this.config.daemonId, event, conversationId, turn_generation: this.turnGeneration, fully_idle: payload.fullyIdle ?? null, termination_reason: payload.terminationReason ?? null, execution: payload.executionNum ?? null, step: payload.stepIdx ?? null, tool: payload.toolCall?.name ?? null, args: payload.toolCall?.args ?? null })).digest("hex"); await this.persist(completed ? { active_tree: false } : {}); await this.journal("hook", { stale, hook_event: event, event_id: eventId, conversation_id: conversationId, turn_generation: this.turnGeneration, execution_num: execution, step_index: step, termination_reason: payload.terminationReason ?? null, fully_idle: payload.fullyIdle ?? null }); completed?.resolve(settled); return { event_id: eventId, daemon_id: this.config.daemonId, parent_conversation_id: parent, profile: observed, requested_profile: { provider: this.config.provider, model: this.config.model, reasoning: this.config.reasoning, mode: this.config.mode }, project_root: this.config.projectRoot, dd_flow_bin: this.config.ddFlowBin, dd_flow_home: this.config.ddFlowHome };
   }
-  async cancel() { cancelPendingDispatch(this); if (!this.child) return { ...this.receipt(), settled: Boolean(this.exited) }; if (!this.exited) { try { process.kill(-this.child.pid, "SIGINT"); } catch (error) { if (error.code !== "ESRCH") throw error; } const deadline = Date.now() + 5000; while (!this.exited && Date.now() < deadline) await new Promise(resolve => setTimeout(resolve, 100)); if (!this.exited) await stopProcessGroup(this.child, 5_000); } await this.providerFinalization; await this.persist({ active_tree: false }); return { ...this.receipt(), cancelled: true, settled: Boolean(this.exited) };
+  async finalizeProviderForControl() {
+    const pending = this.providerFinalization;
+    try { await pending; }
+    catch (error) {
+      // A leaderless group may have exited naturally since the last check.
+      // Only control may recheck it; startup never retries failed cleanup.
+      if (this.providerFinalization !== pending) return await this.providerFinalization;
+      if (error.code !== "process_group_ownership_unknown" || !this.retryProviderFinalization) throw error;
+      this.providerFinalization = this.retryProviderFinalization(5_000);
+      await this.providerFinalization;
+    }
+  }
+  async cancel() { cancelPendingDispatch(this); if (!this.child) return { ...this.receipt(), settled: Boolean(this.exited) }; if (!this.exited) { try { process.kill(-this.child.pid, "SIGINT"); } catch (error) { if (error.code !== "ESRCH") throw error; } const deadline = Date.now() + 5000; while (!this.exited && Date.now() < deadline) await new Promise(resolve => setTimeout(resolve, 100)); if (!this.exited) await stopProcessGroup(this.child, 5_000); } await this.finalizeProviderForControl(); await this.persist({ active_tree: false }); return { ...this.receipt(), cancelled: true, settled: Boolean(this.exited) };
   }
   async close(cancelTree) { if (cancelTree) cancelPendingDispatch(this); if (!cancelTree && (this.active || this.state.active_tree || this.state.unclaimed_activity || (this.lastResult && !this.receipt().settled))) throw new DaemonError("tree_not_settled", "Antigravity root or child tree is still running"); if (cancelTree) await this.cancel(); else if (this.child && !this.exited) this.child.stdin.end(); const deadline = Date.now() + 5000; while (!this.exited && Date.now() < deadline) await new Promise(resolve => setTimeout(resolve, 100)); if (!this.exited) await this.cancel(); await this.providerFinalization; }
   async abortProvider(reason) {
     const child = this.child, providerProcess = this.providerProcess;
     await stopProcessGroup(child);
-    await this.providerFinalization;
+    await this.finalizeProviderForControl();
     await finishDaemonProcess(this.config, providerProcess, "failed", reason);
   }
   async finishResource(state = "stopped", reason = "daemon_stopped") { await finishDaemonProcess(this.config, this.state.resource_process, state, reason); }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@deksden-com/dd-flow-cli",
-  "version": "0.9.0-beta.37",
+  "version": "0.9.0-beta.38",
   "description": "Mechanical runtime CLI for dd-flow workflows.",
   "type": "module",
   "bin": {