fix(ext+hub): v0.5.2 Idle→Resume 신호 소실 3중 버그 수정 — auth_fail 재연결 + pending_owners 보존 + step-probe 리셋

.agents/references/known-issues.md

@@ -53,6 +53,12 @@
 - **해결** (v0.5.1): `approval-handler.ts` L384에 `browser_subagent` 추가, `step-probe.ts` L481/L549에 `browser_subagent`/`open_browser_url` step_type 분류 추가 (`549af6d`)
 - **주의**: 새로운 AG 도구 추가 시 반드시 (1) step-probe step_type 매핑 (2) approval-handler RPC payload 매핑 양쪽 모두 업데이트
 
+### [2026-03-21] Idle→Resume 신호 소실 — 3중 버그
+- **증상**: AG 장시간 idle 후 작업 재개 시 Discord 승인 신호가 전달되지 않음
+- **원인**: (1) `ws-client.ts` `auth_fail` 시 `shouldReconnect=false` — JWT 24h 만료 시 WS 영구 종료. (2) `hub.py` `_disconnect`에서 유일 연결 시 `pending_owners` 삭제 — 재연결 후 Discord 버튼 무효. (3) `step-probe.ts` `stallProbed=true` + `lastPendingStepIndex=N`이 WS 재연결 시 리셋 안 됨 — WAITING step 재전송 영구 차단
+- **해결** (v0.5.2): (1) `auth_fail` → `registrationCode` 재시도. (2) `pending_owners` orphan 마커로 보존+재할당. (3) `resetPendingStateForReconnect()` + `onConnected`에서 호출
+- **주의**: WS `onConnected`에서 반드시 step-probe 상태 리셋 필수. `stallProbed`/`lastPendingStepIndex`는 TTL 없는 영구 값
+
 ---
 
 > [!NOTE]
@@ -79,3 +85,4 @@
 | 10 | **diff_review는 VS Code 커맨드만 유효** — RPC 3개 전략 모두 실패 확정 | diff_review RPC dead-end |
 | 11 | **HttpBridgeContext에 프리미티브 by-value 복사 금지** — 별도 객체 생성 시 getter 사용 | HttpBridgeContext stale primitive |
 | 12 | **새 AG 도구 추가 시 step-probe step_type 매핑 + approval-handler RPC payload 매핑 양쪽 필수** | browser_subagent Allow |
+| 13 | **WS `onConnected`에서 step-probe 상태 리셋 필수** — `stallProbed`/`lastPendingStepIndex`는 TTL 없는 영구 값 | Idle→Resume 신호 소실 |

docs/devlog/2026-03-19.md

@@ -2,4 +2,5 @@
 
 | # | 시간 | 작업 | 커밋 | 상태 |
 |---|------|------|------|------|
-| 1 | 07:30 | v0.5.0 이후 이슈 조사 — 외부 네트워크 접근(서버 정상 확인), browser_subagent Allow RPC 매핑 수정, .env 정리 | `549af6d` | ✅ |
+| 1 | 07:30 | v0.5.1 browser_subagent Allow RPC 매핑 수정 + .env 정리 | `549af6d` | ✅ |
+| 2 | 10:35 | v0.5.2 Idle→Resume 신호 소실 3중 버그 수정: auth_fail 재연결, pending_owners 보존, step-probe 리셋 | — | 🔧 |

extension/out/extension.js

@@ -443,6 +443,8 @@ async function activate(context) {
                 logToFile(`[WS] Connected: ${connId} instance=#${instanceNum}`);
                 statusBar.text = '$(check) Bridge WS';
                 statusBar.tooltip = `Gravity Bridge: ${projectName} (WS #${instanceNum})`;
+                // Reset step-probe state so WAITING steps are re-detected after reconnect
+                (0, step_probe_1.resetPendingStateForReconnect)();
             },
             onDisconnected: () => {
                 logToFile('[WS] Disconnected — using file fallback');

extension/package.json

@@ -2,7 +2,7 @@
     "name": "gravity-bridge",
     "displayName": "Gravity Bridge",
     "description": "Antigravity ↔ Discord 브리지 연동 확장",
-    "version": "0.5.1",
+    "version": "0.5.2",
     "publisher": "variet",
     "engines": {
         "vscode": "^1.100.0"

extension/src/extension.ts

@@ -16,7 +16,7 @@ import * as path from 'path';
 import * as os from 'os';
 import * as cp from 'child_process';
 import { WSBridgeClient, WSResponseData, WSCommandData } from './ws-client';
-import { initStepProbe, BridgeContext, writePendingApproval, tryApprovalStrategies, writeRegistration, getApprovalContext, resetPendingState, handleDiffReviewResponse, getActiveSessionId as getStepProbeSessionId, getStepProbeContext } from './step-probe';
+import { initStepProbe, BridgeContext, writePendingApproval, tryApprovalStrategies, writeRegistration, getApprovalContext, resetPendingState, resetPendingStateForReconnect, handleDiffReviewResponse, getActiveSessionId as getStepProbeSessionId, getStepProbeContext } from './step-probe';
 import { startHttpBridge, getDeterministicPort, HttpBridgeContext } from './http-bridge';
 import { setupApprovalObserver } from './html-patcher';
 import { watchCommandsDir, handleWSCommand, disposeCommandsWatcher, CommandHandlerContext } from './command-handler';
@@ -433,6 +433,8 @@ export async function activate(context: vscode.ExtensionContext) {
                 logToFile(`[WS] Connected: ${connId} instance=#${instanceNum}`);
                 statusBar.text = '$(check) Bridge WS';
                 statusBar.tooltip = `Gravity Bridge: ${projectName} (WS #${instanceNum})`;
+                // Reset step-probe state so WAITING steps are re-detected after reconnect
+                resetPendingStateForReconnect();
             },
             onDisconnected: () => {
                 logToFile('[WS] Disconnected — using file fallback');

extension/src/step-probe.ts

@@ -85,6 +85,19 @@ export function resetPendingState(): void {
     ctx.sawRunningAfterPending = false;
 }
 
+/**
+ * Reset step-probe state after WS reconnection.
+ * Without this, stallProbed=true + lastPendingStepIndex=N permanently block
+ * re-detection of WAITING steps whose pending was lost during disconnect.
+ */
+export function resetPendingStateForReconnect(): void {
+    ctx.lastPendingStepIndex = -1;
+    ctx.stallProbed = false;
+    ctx.sawRunningAfterPending = false;
+    recentPendingSteps.clear();
+    ctx.logToFile('[STEP-PROBE] Reset pending state for WS reconnect');
+}
+
 // handleDiffReviewResponse → moved to ./approval-handler.ts
 
 /**

extension/src/ws-client.ts

@@ -359,14 +359,21 @@ export class WSBridgeClient {
             case 'auth_fail': {
                 const reason = (msg as any).reason || 'Unknown';
                 this.logFn(`[WS] Auth failed: ${reason}`);
-                // Clear session token if it was rejected
+                // Clear session token if it was rejected (e.g. expired after 24h)
                 this.sessionToken = '';
+                if (this.registrationCode) {
+                    // Token expired → retry with registration code on next reconnect
+                    this.logFn('[WS] Retrying with registration code...');
+                    this._cleanup();
+                    this._scheduleReconnect();
+                } else {
+                    // No registration code available → permanent failure
                     // MUST set shouldReconnect=false BEFORE _cleanup(), because _cleanup()
                     // closes the WS → triggers close event → _onDisconnect() → _scheduleReconnect().
-                // Without this, auth failures cause infinite reconnect loops.
                     this.shouldReconnect = false;
                     this._cleanup();
                     this.handlers.onError?.(`Auth failed: ${reason}`);
+                }
                 break;
             }
 

hub.py

@@ -269,11 +269,17 @@ class WSHub:
             self.project_connections[conn.project] = set()
         self.project_connections[conn.project].add(conn.conn_id)
 
-        # FIX: Reassign orphaned pending_owners (from dead conn_ids) to this new connection.
+        # FIX: Reassign orphaned pending_owners (from dead conn_ids or orphan markers)
-        # When Extension reconnects, old conn_id entries become stale.
+        # to this new connection. When Extension reconnects, old entries become stale.
         reassigned = 0
         for rid, cid in list(self.pending_owners.items()):
-            if cid not in self.connections:
+            if cid.startswith("orphan:"):
+                # Only reassign orphans from the same project
+                if cid != f"orphan:{conn.project}":
+                    continue
+                self.pending_owners[rid] = conn.conn_id
+                reassigned += 1
+            elif cid not in self.connections:
                 self.pending_owners[rid] = conn.conn_id
                 reassigned += 1
         if reassigned:
@@ -324,7 +330,13 @@ class WSHub:
                         f"(disconnected {conn_id})"
                     )
                 else:
-                    del self.pending_owners[rid]
+                    # Preserve as orphan marker instead of deleting —
+                    # will be reassigned when Extension reconnects
+                    self.pending_owners[rid] = f"orphan:{project}"
+                    logger.info(
+                        f"[HUB] Orphaned pending {rid[:12]} "
+                        f"(no active connections in {project})"
+                    )
 
         # Close WebSocket if still open
         if not conn.ws.closed: