fix(renderer): re-arm runtime event gate when reconcile throws

Co-authored-by: Orca <help@stably.ai>

src/renderer/src/hooks/useIpcEvents.test.ts

@@ -58,6 +58,45 @@ describe('buildRuntimeClientEventEnvironmentKey', () => {
   })
 })
 
+describe('runtime client-event subscription gate', () => {
+  it('re-arms after a failed reconcile so the next publication retries', async () => {
+    let failNextStatusRead = false
+    const statusReads: string[] = []
+    const runtimeStatusByEnvironmentId = new Map<string, { status: string } | null>()
+    runtimeStatusByEnvironmentId.get = (environmentId: string) => {
+      statusReads.push(environmentId)
+      if (failNextStatusRead) {
+        failNextStatusRead = false
+        throw new Error('reconcile failed')
+      }
+      return undefined
+    }
+    const storeState = createHarnessStoreState({
+      tabsByWorktree: {},
+      runtimeEnvironments: [{ id: 'env-1' }],
+      runtimeStatusByEnvironmentId
+    })
+    const harness = await loadIpcEventsHarness(storeState)
+    const { useAppStore } = await import('../store')
+    const { bumpRuntimeClientEventSubscriptionGeneration } =
+      await import('@/runtime/runtime-client-event-subscription-invalidation')
+    harness.useIpcEvents()
+    const reconcile = vi.mocked(useAppStore.subscribe).mock.calls[0]?.[0] as (
+      state: unknown
+    ) => void
+
+    bumpRuntimeClientEventSubscriptionGeneration()
+    failNextStatusRead = true
+    expect(() => reconcile(storeState)).toThrow('reconcile failed')
+    const readsAfterFailure = statusReads.length
+
+    // Same store references as the failed publication: only the failure re-arm can drive this.
+    reconcile(storeState)
+
+    expect(statusReads.length).toBeGreaterThan(readsAfterFailure)
+  })
+})
+
 describe('getNewlyConnectedRuntimeEnvironmentIds', () => {
   it('returns only environments that became connected', () => {
     expect(getNewlyConnectedRuntimeEnvironmentIds(['env-a'], ['env-a', 'env-b'])).toEqual(['env-b'])

src/renderer/src/hooks/useIpcEvents.ts

@@ -183,7 +183,10 @@ import {
   type DirectSshConnectedStateOrigin
 } from './direct-ssh-state-routing'
 import { isDirectSshReconnectCoordinatorRoutingEnabled } from './direct-ssh-reconnect-rollout'
-import { createRuntimeClientEventSubscriptionInvalidationGate } from '@/runtime/runtime-client-event-subscription-invalidation'
+import {
+  bumpRuntimeClientEventSubscriptionGeneration,
+  createRuntimeClientEventSubscriptionInvalidationGate
+} from '@/runtime/runtime-client-event-subscription-invalidation'
 
 function getShortcutPlatform(): NodeJS.Platform {
   if (navigator.userAgent.includes('Mac')) {
@@ -1037,41 +1040,47 @@ export function useIpcEvents(): void {
       if (!runtimeClientEventSubscriptionInvalidation.changed(state)) {
         return
       }
-      const nextEnvironmentIds = getRuntimeClientEventEnvironmentIds()
-      const nextKey = buildRuntimeClientEventEnvironmentKey(nextEnvironmentIds)
-      const nextReachableEnvironmentIds = getReachableRuntimeEnvironmentIds()
-      const nextReachableKey = buildRuntimeClientEventEnvironmentKey(nextReachableEnvironmentIds)
-      if (
-        nextKey === runtimeClientEventEnvironmentKey &&
-        nextReachableKey === reachableRuntimeEnvironmentKey
-      ) {
-        return
+      try {
+        const nextEnvironmentIds = getRuntimeClientEventEnvironmentIds()
+        const nextKey = buildRuntimeClientEventEnvironmentKey(nextEnvironmentIds)
+        const nextReachableEnvironmentIds = getReachableRuntimeEnvironmentIds()
+        const nextReachableKey = buildRuntimeClientEventEnvironmentKey(nextReachableEnvironmentIds)
+        if (
+          nextKey === runtimeClientEventEnvironmentKey &&
+          nextReachableKey === reachableRuntimeEnvironmentKey
+        ) {
+          return
+        }
+        for (const environmentId of getRuntimeProjectRefreshEnvironmentIds({
+          previousDesired: runtimeClientEventEnvironmentIds,
+          nextDesired: nextEnvironmentIds,
+          previousReachable: reachableRuntimeEnvironmentIds,
+          nextReachable: nextReachableEnvironmentIds
+        })) {
+          runtimeProjectRefreshScheduler.request(environmentId)
+        }
+        for (const environmentId of getNewlyDisconnectedRuntimeEnvironmentIds(
+          reachableRuntimeEnvironmentIds,
+          nextReachableEnvironmentIds
+        )) {
+          // No-op when the environment has no SSH bucket (e.g. web client).
+          useAppStore.getState().markEnvironmentSshStateStale(environmentId)
+        }
+        runtimeClientEventEnvironmentIds = nextEnvironmentIds
+        reachableRuntimeEnvironmentIds = nextReachableEnvironmentIds
+        // Nested stale-state writes can advance the out-of-store SSH generation.
+        runtimeClientEventEnvironmentKey = buildRuntimeClientEventEnvironmentKey(
+          runtimeClientEventEnvironmentIds
+        )
+        reachableRuntimeEnvironmentKey = buildRuntimeClientEventEnvironmentKey(
+          reachableRuntimeEnvironmentIds
+        )
+        runtimeClientEventsSync.sync()
+      } catch (error) {
+        // Why: the gate already consumed this publication, so re-arm it or the subscription stays stale forever.
+        bumpRuntimeClientEventSubscriptionGeneration()
+        throw error
       }
-      for (const environmentId of getRuntimeProjectRefreshEnvironmentIds({
-        previousDesired: runtimeClientEventEnvironmentIds,
-        nextDesired: nextEnvironmentIds,
-        previousReachable: reachableRuntimeEnvironmentIds,
-        nextReachable: nextReachableEnvironmentIds
-      })) {
-        runtimeProjectRefreshScheduler.request(environmentId)
-      }
-      for (const environmentId of getNewlyDisconnectedRuntimeEnvironmentIds(
-        reachableRuntimeEnvironmentIds,
-        nextReachableEnvironmentIds
-      )) {
-        // No-op when the environment has no SSH bucket (e.g. web client).
-        useAppStore.getState().markEnvironmentSshStateStale(environmentId)
-      }
-      runtimeClientEventEnvironmentIds = nextEnvironmentIds
-      reachableRuntimeEnvironmentIds = nextReachableEnvironmentIds
-      // Nested stale-state writes can advance the out-of-store SSH generation.
-      runtimeClientEventEnvironmentKey = buildRuntimeClientEventEnvironmentKey(
-        runtimeClientEventEnvironmentIds
-      )
-      reachableRuntimeEnvironmentKey = buildRuntimeClientEventEnvironmentKey(
-        reachableRuntimeEnvironmentIds
-      )
-      runtimeClientEventsSync.sync()
     })
     unsubs.push(runtimeClientEventsSync.stop)
     unsubs.push(runtimeProjectRefreshScheduler.stop)