Fix one-shot exit hangs by tearing down cached memory managers by Julbarth · Pull Request #40389 · openclaw/openclaw

Julbarth · 2026-03-09T00:25:06Z

Summary

Describe the problem and fix in 2–5 bullets:

Problem: One-shot CLI runs can emit successful completion events but still hang on exit.
Why it matters: Bench/automation workflows can time out even when the task itself completed.
What changed: Added explicit global teardown for cached memory search/index managers and called teardown in CLI finally; also drained in-flight index manager creations during teardown.
What did NOT change (scope boundary): No model routing/tool behavior changes, no plugin/channel behavior changes, no secret/token handling changes.

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

User-visible / Behavior Changes

One-shot CLI commands now exit cleanly after completion in scenarios where cached memory-manager watcher handles previously kept the process alive.

Security Impact (required)

New permissions/capabilities? (No)
Secrets/tokens handling changed? (No)
New/changed network calls? (No)
Command/tool execution surface changed? (No)
Data access scope changed? (No)
If any Yes, explain risk + mitigation:

Repro + Verification

Environment

OS: macOS 14.0 (local repro); CI includes Linux/Windows jobs
Runtime/container: Node.js + pnpm
Model/provider: N/A (provider-independent; reproduced with successful one-shot completion)
Integration/channel (if any): CLI one-shot run
Relevant config (redacted): default memory settings (memory search enabled)

Steps

Run one-shot CLI command on pre-fix build with default memory enabled.
Observe completion events emitted successfully.
Process remains alive instead of exiting.

Expected

Process exits after successful completion.

Actual

Process can remain alive due to lingering memory-manager resources (watcher handles), with race risk from in-flight manager creation.

Evidence

Attach at least one:

Failing test/log before + passing after
Trace/log snippets
Screenshot/recording
Perf numbers (if relevant)

Human Verification (required)

What you personally verified (not just CI), and how:

Verified scenarios:
- Reproduced one-shot hang before fix.
- Confirmed clean exit after fix.
- Ran targeted tests:
  - pnpm exec vitest run src/cli/run-main.exit.test.ts
  - pnpm exec vitest run src/memory/search-manager.test.ts
  - pnpm exec vitest run src/memory/manager.get-concurrency.test.ts
Edge cases checked:
- Teardown called through lazy runtime boundary (no eager manager import path).
- Global teardown drains pending manager creations.
- QMD manager cache teardown still closes and resets as expected.
What you did not verify:
- Long-duration stress testing across all providers/channels.
- Full production workload profiling.

Review Conversations

I replied to or resolved every bot review conversation I addressed in this PR.
I left unresolved only the conversations that still need reviewer or maintainer judgment.

If a bot review conversation is addressed by this PR, resolve that conversation yourself. Do not leave bot review conversation cleanup for maintainers.

Compatibility / Migration

Backward compatible? (Yes)
Config/env changes? (No)
Migration needed? (No)
If yes, exact upgrade steps:

Failure Recovery (if this breaks)

How to disable/revert this change quickly: Revert this PR commit set.
Files/config to restore: src/cli/run-main.ts, src/cli/run-main.exit.test.ts, src/memory/manager.ts, src/memory/manager-runtime.ts, src/memory/search-manager.ts, src/memory/search-manager.test.ts, src/memory/manager.get-concurrency.test.ts, src/memory/index.ts
Known bad symptoms reviewers should watch for: one-shot process not exiting, or teardown-time errors in memory shutdown path.

Risks and Mitigations

List only real risks for this PR. Add/remove entries as needed. If none, write None.

Risk: Teardown during in-flight creation could cause lifecycle races.
- Mitigation: Explicitly await pending creations before closing cached managers.
Risk: Accidental eager loading of memory manager module on non-memory paths.
- Mitigation: Teardown path uses lazy runtime import boundary; tests cover behavior.

greptile-apps · 2026-03-09T00:31:47Z

Greptile Summary

This PR fixes one-shot CLI exit hangs caused by lingering FSWatcher handles inside cached MemoryIndexManager instances. The fix adds explicit global teardown functions (closeAllMemorySearchManagers / closeAllMemoryIndexManagers) and calls them in the CLI finally block, ensuring all watcher handles are released after the program finishes.

Key changes:

src/cli/run-main.ts: Wraps the main CLI body in try/finally; calls closeCliMemoryManagers() (lazy dynamic import) on every exit path, including the tryRouteCli early-return path.
src/memory/manager.ts: closeAllMemoryIndexManagers() drains INDEX_CACHE_PENDING via Promise.allSettled before clearing INDEX_CACHE and closing each manager — correctly handles in-flight creations.
src/memory/search-manager.ts: closeAllMemorySearchManagers() closes QMD (FallbackMemoryManager) entries first, then delegates to closeAllMemoryIndexManagers(). The implementation unconditionally calls loadManagerRuntime() even when memory was never accessed, which loads the manager module purely as a teardown artifact and breaks the lazy-import boundary.
src/memory/search-manager.test.ts: Correctly migrates mockPrimary, fallbackManager, and related mocks to vi.hoisted() (required because vi.mock() factories are hoisted above module scope) and fixes the mocked module path from ./manager.js → ./manager-runtime.js to match the actual import in production code.
src/memory/manager.get-concurrency.test.ts: New test verifies that closeAllMemoryIndexManagers drains a slow in-flight get() and that a subsequent get() returns a fresh (non-cached) manager instance.

Confidence Score: 4/5

Safe to merge; fixes a real hang with correct lifecycle ordering. One minor issue with lazy-loading boundary breaks when teardown loads manager-runtime unconditionally.
The PR correctly implements global teardown with proper draining of in-flight manager creations before closure. Tests cover critical scenarios. Teardown errors are silently swallowed, preventing regressions on non-memory paths. The core logic for fixing the hang is sound. However, closeAllMemorySearchManagers unconditionally imports manager-runtime.js even when it was never loaded during execution, breaking the lazy boundary property on CLI paths that never accessed memory. This is a correctness issue with an easy fix but doesn't impact functional safety.
src/memory/search-manager.ts — the unconditional loadManagerRuntime() call should be guarded to preserve lazy-loading boundary.

_{Last reviewed commit: cec15df}

Julbarth · 2026-03-09T05:23:15Z

All feedback addressed, CI green, ready for merge.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f07f95a2a5

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-09T16:02:17Z

@@ -13,6 +13,15 @@ import { applyCliProfileEnv, parseCliProfileArgs } from "./profile.js";
 import { tryRouteCli } from "./route.js";
 import { normalizeWindowsArgv } from "./windows-argv.js";

+async function closeCliMemoryManagers(): Promise<void> {
+  try {
+    const { closeAllMemorySearchManagers } = await import("../memory/index.js");


Avoid eager memory-module import during CLI teardown

runCli now always executes closeCliMemoryManagers() in finally, and that helper imports ../memory/index.js on every command path. Because src/memory/index.ts statically re-exports MemoryIndexManager from manager.ts, even routed/fast-path commands that never used memory still load the full memory stack at shutdown, which defeats route-first startup goals and adds avoidable module-load overhead (or failures in stripped runtimes). Use a teardown import that stays on the lazy runtime boundary instead of the memory barrel export.

Useful? React with 👍 / 👎.

frankekn · 2026-03-09T23:34:49Z

Merged via squash.

Prepared head SHA: 0e600e8
Merge commit: c0cba7f

Thanks @Julbarth!

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn (cherry picked from commit c0cba7f)

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn (cherry picked from commit c0cba7f)

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…law#40389) Merged via squash. Prepared head SHA: 0e600e8 Co-authored-by: Julbarth <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

Cached MemoryIndexManager instances live in a module-level INDEX_CACHE that grows for the entire gateway daemon lifetime. Each instance owns a chokidar FSWatcher recursively watching the memory directory (under the default sync.watch=true), so every cache entry holds an FSEventWrap native handle plus one awaitWriteFinish file descriptor per touched memory/*.md file. Cached managers are torn down at CLI exit (openclaw#40389), but daemons never reach that exit path until restart. A production daemon (OpenClaw 2026.4.15) measured +1.6 GB RSS over 11 h and 612 accumulated FSEventWrap handles before crashing. This change adds a periodic runtime sweep that closes idle managers during normal daemon operation, so restart is no longer the only way to release these resources. Approach -------- ManagedCache: - Track lastAccessAt per cache key (refreshed on hit and on initial set). - closeIdleManagedCacheEntries walks lastAccessAt, closes stale entries, and isolates close() errors so one bad entry does not stop the rest. Reuses the per-manager close() introduced in openclaw#40389 — no new teardown code path. In-flight protection (always-on safety net): - inflightCount map keyed the same way as cache / lastAccessAt. - acquireManagedCacheKey() increments the count, refreshes lastAccessAt, and returns an idempotent release() function. - isManagedCacheKeyBusy() reports whether the count is > 0. - closeIdleManagedCacheEntries consults isManagedCacheKeyBusy in both the survey loop and immediately before close(); busy entries are skipped, counted, and surfaced via onSkipBusy / skippedBusy. - A revalidation gate (just before close) also re-checks lastAccessAt freshness and cache identity, catching entries that became active or were swapped during the scan-to-close gap. Reported as skippedRevalidated. MemoryIndexManager: - Protected withBusy(fn) helper wraps acquire/release in try/finally. - Guards sync(), search(), and ensureProviderInitialized() so the busy ref is held across provider init as well as the manager work itself. - close() and closeMemoryIndexManagersForAgent clear any residual inflightCount entry alongside the cache slot. release() is idempotent so a release that fires after close() is a no-op. Gateway sidecar: - scheduleMemoryIndexIdleEvict registers a gateway-lifetime sidecar that calls closeIdleMemorySearchManagers every idleEvictScanMs (default 5 min), evicting managers idle for idleEvictMs (default 15 min). Setting either to 0 disables the sweep entirely. Configuration ------------- Two new optional fields under agents.defaults.memorySearch.sync: - idleEvictMs (default 900000): idle TTL before a manager is eligible for eviction. 0 disables. - idleEvictScanMs (default 300000): sweep cadence. 0 disables. Documented in the config schema (zod), the MemorySearchConfig type, and the UI field labels map. Validation ---------- Live soak on the production daemon (OpenClaw 2026.4.15 + this patch backported, 13 h+ continuous, sampling every 10 min): | Metric | Without patch (11 h) | With patch (13 h soak) | |-----------------|-------------------------|-------------------------------| | FSEventWrap | 612 native handles | bounded, returns to ~3 baseline | | memory_fd | 488 (chokidar holds) | 0 between sweeps | | RSS growth rate | ~1 GB/h | ~150 MB/h (~7x improvement) | | Outcome at 13 h | OOM within 4 to 6 h | 3.25 GB stable, no OOM | Sidecar observed behavior: - fd accumulates during memory_search activity (up to ~600 handles) - next sweep evicts idle managers, fd returns to baseline within 10 min - skippedBusy and skippedRevalidated stayed at 0 throughout the soak — the workload burst rate did not exercise the deferral paths, but the unit tests cover them deterministically Tests ----- - extensions/memory-core/src/memory/manager-cache.test.ts: 21 tests (was 11). New coverage: acquireManagedCacheKey, isManagedCacheKeyBusy, busy deferral, refcount, lastAccessAt refresh, inflightCount cleanup, revalidation gate for cache-hit / busy / identity-swap / slot- disappearance gaps. - extensions/memory-core/src/memory/manager.idle-gc.test.ts: 10 tests (was 5). New coverage: in-flight protection at manager-level entry points (search, sync, ensureProviderInitialized). - extensions/memory-core/: 718/718 across 57 files. - src/gateway/server-startup-post-attach.test.ts: 41 tests pass. - src/agents/memory-search.test.ts: 27 tests pass. - pnpm run lint, tsgo:core, tsgo:extensions, tsgo:test, build all green. Risk / Compatibility -------------------- - Default idle threshold (15 min) is conservative — comfortably outlives typical bursts of memory_search calls within a session. - Worst-case dwell is idleMs + scanMs + max in-flight op duration (was unbounded; idleMs + scanMs if all ops are short). - Short-lived CLI invocations are unchanged: closeAllMemoryIndexManagers remains the exit path from openclaw#40389, and it clears inflightCount alongside the rest of the cache. - Public MemoryPluginRuntime.closeIdleMemorySearchManagers return type grows two additive fields (skippedBusy, skippedRevalidated). Existing callers reading evicted / remaining continue to work unchanged via TypeScript structural typing. - No new dependencies, no schema migration, no protocol change. Out of scope ------------ - Does NOT change sync.watch default value (openclaw#71335 tracks that separately). - Does NOT refactor the watcher to be singleton-per-workspace (much larger architectural change). - Does NOT introduce an opt-out for the in-flight protection — it's a safety net that should always be on. The only configurable knobs are the sweep cadence and idle threshold. Refs openclaw#71335, openclaw#40389.

Cached MemoryIndexManager instances live in a module-level INDEX_CACHE that grows for the entire gateway daemon lifetime. Each instance owns a chokidar FSWatcher recursively watching the memory directory (under the default sync.watch=true), so every cache entry holds an FSEventWrap native handle plus one awaitWriteFinish file descriptor per touched memory/*.md file. Cached managers are torn down at CLI exit (openclaw#40389), but daemons never reach that exit path until restart. A production daemon (OpenClaw 2026.4.15) measured +1.6 GB RSS over 11 h and 612 accumulated FSEventWrap handles before OOM. This change adds a periodic runtime sweep that closes idle managers during normal daemon operation, so restart is no longer the only way to release these resources. Approach -------- ManagedCache: - Track lastAccessAt per cache key (refreshed on hit and on initial set). - closeIdleManagedCacheEntries walks lastAccessAt, closes stale entries, and isolates close() errors so one bad entry does not stop the rest. Reuses the per-manager close() introduced in openclaw#40389 - no new teardown code path. In-flight protection (always-on safety net): - inflightCount map keyed the same way as cache / lastAccessAt. - acquireManagedCacheKey() increments the count, refreshes lastAccessAt, and returns an idempotent release() function. - isManagedCacheKeyBusy() reports whether the count is > 0. - closeIdleManagedCacheEntries consults isManagedCacheKeyBusy in both the survey loop and immediately before close(); busy entries are skipped, counted, and surfaced via onSkipBusy / skippedBusy. - A revalidation gate (just before close) also re-checks lastAccessAt freshness and cache identity, catching entries that became active or were swapped during the scan-to-close gap. Reported as skippedRevalidated. - Cache metadata cleanup in MemoryIndexManager.close() and closeMemoryIndexManagersForAgent is now scoped to the cache identity check that previously guarded only the INDEX_CACHE delete, so a replacement manager installed in the same key during the close() await window keeps its own lastAccessAt and inflightCount entries. MemoryIndexManager: - Protected withBusy(fn) helper wraps acquire/release in try/finally. - Guards sync(), search(), and ensureProviderInitialized() so the busy ref is held across provider init as well as the manager work itself. - close() and closeMemoryIndexManagersForAgent clear residual cache metadata only when this instance is still the cache occupant. release() is idempotent so a release that fires after close() is a no-op. Gateway sidecar: - scheduleMemoryIndexIdleEvict registers a gateway-lifetime sidecar that calls closeIdleMemorySearchManagers every idleEvictScanMs (default 5 min), evicting managers idle for idleEvictMs (default 15 min). Setting either to 0 disables the sweep entirely. - Resolved from agents.defaults.memorySearch.sync only. The cache is a process-wide singleton and the sweep runs once per cadence over the entire cache; there is no per-agent sweep timer, and no coherent per-agent answer when two agents share a workspace and disagree on the threshold. Per-agent values under agents.list[].memorySearch.sync are accepted by the (shared) schema but ignored at runtime; this is documented on the TS doc, zod comments, UI labels, FIELD_HELP, and in docs/reference/memory-config.md. Configuration ------------- Two new optional fields under agents.defaults.memorySearch.sync: - idleEvictMs (default 900000): idle TTL before a manager is eligible for eviction. 0 disables. - idleEvictScanMs (default 300000): sweep cadence. 0 disables. Documented in: - src/config/zod-schema.agent-runtime.ts (schema with scope contract) - src/config/types.tools.ts (MemorySearchConfig TS doc) - src/config/schema.labels.ts (UI labels, tagged "gateway-wide") - src/config/schema.help.ts (FIELD_HELP user-facing copy) - docs/reference/memory-config.md (reference docs with ParamField blocks) Validation ---------- Live soak on the production daemon (OpenClaw 2026.4.15 + this patch backported, 17 h+ continuous at PR submission, sampling every 10 min): | Metric | Without patch (11 h) | With patch (17 h soak) | |-----------------|-------------------------|-------------------------------| | FSEventWrap | 612 native handles | bounded, returns to ~3 baseline | | memory_fd | 488 (chokidar holds) | 0 between sweeps | | RSS growth rate | ~1 GB/h | ~150 MB/h (~7x improvement) | | Outcome | OOM within 4 to 6 h | 3.16 GB stable, no OOM | Sidecar observed behavior: - fd accumulates during memory_search activity (up to ~600 handles) - next sweep evicts idle managers, fd returns to baseline within 10 min - skippedBusy and skippedRevalidated stayed at 0 throughout the soak - the workload burst rate did not exercise the deferral paths, but the unit tests cover them deterministically Tests ----- - extensions/memory-core/src/memory/manager-cache.test.ts: 21 tests (was 11). New coverage: acquireManagedCacheKey, isManagedCacheKeyBusy, busy deferral, refcount, lastAccessAt refresh, inflightCount cleanup, revalidation gate for cache-hit / busy / identity-swap / slot- disappearance gaps. - extensions/memory-core/src/memory/manager.idle-gc.test.ts: 12 tests (was 5). New coverage: in-flight protection at manager-level entry points (search, sync, ensureProviderInitialized), plus cache-occupant identity-check coverage for the close() metadata-cleanup path. - extensions/memory-core/: 720/720 across 57 files. - src/gateway/server-startup-post-attach.test.ts: 41 tests pass. - src/agents/memory-search.test.ts: 27 tests pass. - src/config/: 1493/1493 across all config schema tests pass. - pnpm run lint, tsgo:core, tsgo:extensions, tsgo:test, build all green. Risk / Compatibility -------------------- - Default idle threshold (15 min) is conservative - comfortably outlives typical bursts of memory_search calls within a session. - Worst-case dwell is idleMs + scanMs + max in-flight op duration (was unbounded; idleMs + scanMs if all ops are short). - Short-lived CLI invocations are unchanged: closeAllMemoryIndexManagers remains the exit path from openclaw#40389, and it clears inflightCount alongside the rest of the cache. - Public MemoryPluginRuntime.closeIdleMemorySearchManagers return type grows two additive fields (skippedBusy, skippedRevalidated). Existing callers reading evicted / remaining continue to work unchanged via TypeScript structural typing. - No new dependencies, no schema migration, no protocol change. Out of scope ------------ - Does NOT change sync.watch default value (openclaw#71335 tracks that separately). - Does NOT refactor the watcher to be singleton-per-workspace (much larger architectural change). - Does NOT introduce an opt-out for the in-flight protection - it's a safety net that should always be on. The only configurable knobs are the sweep cadence and idle threshold. Refs openclaw#71335, openclaw#40389.

openclaw-barnacle Bot added cli CLI command changes size: M labels Mar 9, 2026

Julbarth mentioned this pull request Mar 9, 2026

Fix one-shot exit hangs by tearing down cached memory managers #40367

Closed

20 tasks

greptile-apps Bot reviewed Mar 9, 2026

View reviewed changes

Comment thread src/memory/search-manager.ts Outdated

Julbarth force-pushed the fix/one-shot-memory-teardown-clean branch 2 times, most recently from 0bc26dd to 524f80e Compare March 9, 2026 01:37

frankekn self-assigned this Mar 9, 2026

frankekn force-pushed the fix/one-shot-memory-teardown-clean branch 2 times, most recently from 141558b to f07f95a Compare March 9, 2026 15:53

chatgpt-codex-connector Bot reviewed Mar 9, 2026

View reviewed changes

frankekn force-pushed the fix/one-shot-memory-teardown-clean branch 3 times, most recently from 8bdc1f2 to 6b02314 Compare March 9, 2026 23:28

Julbarth and others added 5 commits March 10, 2026 07:29

Fix one-shot exit hangs by tearing down cached memory managers

dd76334

Harden memory teardown lazy boundary and pending drain

a657cfa

Avoid runtime import on teardown when memory was unused

38387da

fix: document one-shot memory teardown

9f0071b

fix: preserve lazy memory teardown boundary

0e600e8

frankekn force-pushed the fix/one-shot-memory-teardown-clean branch from 6b02314 to 0e600e8 Compare March 9, 2026 23:34

frankekn merged commit c0cba7f into openclaw:main Mar 9, 2026
25 of 26 checks passed

github-actions Bot mentioned this pull request Mar 10, 2026

📡 Upstream Digest — 2026-03-10 01:16 UTC curtismercier/openclaw-mods#224

Open

gemini-code-assist Bot mentioned this pull request Mar 10, 2026

Sync/upstream 2026 03 10 QVerisAI/QVerisBot#102

Merged

jiangwenwei818 mentioned this pull request Mar 11, 2026

fix(cli): close cached memory managers on exit to prevent process hang #40532

Closed

alexey-pelykh mentioned this pull request Mar 22, 2026

Cherry-pick issue #919: model context, telegram timeout, web search, gateway fixes remoteclaw/remoteclaw#1808

Merged

quengh mentioned this pull request May 25, 2026

fix(memory): bound INDEX_CACHE lifetime in long-running gateway daemons #86293

Closed

quengh mentioned this pull request May 25, 2026

fix(memory): bound INDEX_CACHE lifetime in long-running gateway daemons #86337

Closed

quengh mentioned this pull request May 25, 2026

fix(memory): bound INDEX_CACHE lifetime in long-running gateway daemons #86345

Closed

RomneyDa mentioned this pull request Jun 1, 2026

fix(memory): warn on gateway watcher FD risk #89185

Merged

clawsweeper Bot mentioned this pull request Jun 12, 2026

openclaw agent --local --json can emit final JSON but keep process/workers alive #92464

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

Fix one-shot exit hangs by tearing down cached memory managers#40389

Fix one-shot exit hangs by tearing down cached memory managers#40389
frankekn merged 5 commits into
openclaw:mainfrom
Julbarth:fix/one-shot-memory-teardown-clean

Julbarth commented Mar 9, 2026

Uh oh!

greptile-apps Bot commented Mar 9, 2026

Uh oh!

Uh oh!

Julbarth commented Mar 9, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Mar 9, 2026

Uh oh!

Uh oh!

frankekn commented Mar 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

Julbarth commented Mar 9, 2026

Summary

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

User-visible / Behavior Changes

Security Impact (required)

Repro + Verification

Environment

Steps

Expected

Actual

Evidence

Human Verification (required)

Review Conversations

Compatibility / Migration

Failure Recovery (if this breaks)

Risks and Mitigations

Uh oh!

greptile-apps Bot commented Mar 9, 2026

Greptile Summary

Confidence Score: 4/5

Uh oh!

Uh oh!

Julbarth commented Mar 9, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

frankekn commented Mar 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants