fix(infra): treat undici terminated aborts as transient by liuxiaopai-ai · Pull Request #31860 · openclaw/openclaw

liuxiaopai-ai · 2026-03-02T15:55:08Z

Summary

Describe the problem and fix in 2–5 bullets:

Problem: Gateway could exit on unhandled rejection when Node/undici surfaced TypeError: terminated from Fetch.onAborted.
Why it matters: transient upstream disconnect/abort paths should not crash the whole gateway process.
What changed: added targeted detection for undici-style TypeError("terminated") (stack/cause-aware) and classify it as transient network error.
What did NOT change (scope boundary): no retry policy, no provider-specific request logic, no change to fatal/config error handling paths.

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

User-visible / Behavior Changes

Gateway no longer exits on undici TypeError: terminated abort/disconnect rejections; it logs as non-fatal transient network error and continues running.

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
Runtime/container: Node 25.x local test run
Model/provider: N/A
Integration/channel (if any): N/A
Relevant config (redacted): N/A

Steps

Trigger installUnhandledRejectionHandler with an unhandled rejection reason matching undici abort shape: TypeError("terminated") with Fetch.onAborted stack.
Observe gateway unhandled rejection handling path.
Run infra unhandled-rejection unit tests.

Expected

Reason is classified transient; process is not exited.

Actual

Before fix: generic unhandled rejection branch could exit process.
After fix: non-fatal branch logs warning and continues.

Evidence

Attach at least one:

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

pnpm test src/infra/unhandled-rejections.test.ts src/infra/unhandled-rejections.fatal-detection.test.ts

Passed: 2 files, 36 tests
Added regression cases for TypeError: terminated and fatal-detector non-exit behavior

Human Verification (required)

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

Verified scenarios:
- isTransientNetworkError returns true for undici TypeError: terminated with Fetch.onAborted stack.
- plain TypeError: terminated without undici/abort context remains non-transient.
- installUnhandledRejectionHandler does not exit for undici-terminated transient case.
Edge cases checked:
- Existing TypeError: fetch failed and nested-cause transient paths still pass.
- Non-transient generic errors still exit.
What you did not verify:
- Live upstream reproduction from real gateway runtime logs.

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 commit 28d93597a.
Files/config to restore:
- src/infra/unhandled-rejections.ts
- src/infra/unhandled-rejections.test.ts
- src/infra/unhandled-rejections.fatal-detection.test.ts
Known bad symptoms reviewers should watch for:
- Unexpected suppression of truly fatal non-network TypeError: terminated errors (guarded by stack/cause checks).

Risks and Mitigations

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

Risk: over-classifying generic TypeError: terminated as transient.
- Mitigation: detection requires undici-specific stack marker or abort/UND_ERR cause evidence; added negative regression test for plain TypeError: terminated.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 28d93597a4

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 · 2026-03-02T15:57:56Z

+    return true;
+  }
+  const causeMessage = getErrorMessage(cause).toLowerCase();
+  if (causeMessage.includes("aborted") || causeMessage.includes("terminated")) {


Tighten cause-message heuristic for terminated errors

The new isUndiciTerminatedFetchError path treats TypeError("terminated") as transient whenever cause.message contains "aborted" or "terminated", even without undici-specific evidence. That over-matches non-network failures (for example, an app-level TypeError("terminated") caused by a business-rule cancellation message), which then makes installUnhandledRejectionHandler suppress process exit for a genuinely unhandled runtime error. Please limit this fallback to stronger undici indicators (stack markers, AbortError, or UND_ERR_* codes) instead of generic message substrings.

Useful? React with 👍 / 👎.

greptile-apps · 2026-03-02T16:09:44Z

Greptile Summary

This PR addresses a bug where the gateway could crash on undici TypeError: terminated from Fetch.onAborted during client disconnects. The implementation adds targeted detection in isUndiciTerminatedFetchError() with multiple fallback mechanisms: stack trace inspection for undici markers, AbortError cause detection, UND_ERR_ error code checking, and cause message pattern matching. The detection is comprehensive but not overly strict - it will classify TypeError("terminated") as transient if it has abort-related evidence in the stack or cause chain, which appropriately balances safety (preventing crashes on transient network issues) with specificity. Test coverage includes both positive cases (undici terminated errors) and negative cases (plain TypeError without undici context).

Confidence Score: 4/5

Safe to merge with minimal risk - addresses a real crash scenario with appropriate guardrails
The implementation is well-structured with comprehensive detection logic and good test coverage. The risk of over-classifying generic TypeError("terminated") is mitigated by requiring undici-specific evidence (stack markers, UND_ERR_ codes, or abort-related cause patterns). The broader cause message check (lines 191-194) could theoretically catch non-undici errors, but in the context of a gateway handling HTTP requests, this is likely intentional and acceptable.
No files require special attention - changes are focused and well-tested

_{Last reviewed commit: 28d9359}

Infra: suppress undici terminated unhandled rejections

28d9359

openclaw-barnacle Bot added size: S experienced-contributor labels Mar 2, 2026

chatgpt-codex-connector Bot reviewed Mar 2, 2026

View reviewed changes

thewilloftheshadow closed this Mar 6, 2026

msobrosa mentioned this pull request Jul 4, 2026

Gateway crashes with uncaught TypeError: terminated (undici) — still occurring on 2026.6.11 despite prior transient-handling fixes; related wedged-but-alive episodes #100094

Closed

clawsweeper Bot mentioned this pull request Jul 4, 2026

fix: keep undici terminated exceptions non-fatal #100098

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(infra): treat undici terminated aborts as transient#31860

fix(infra): treat undici terminated aborts as transient#31860
liuxiaopai-ai wants to merge 1 commit into
openclaw:mainfrom
liuxiaopai-ai:codex/undici-terminated-nonfatal-31855

liuxiaopai-ai commented Mar 2, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Mar 2, 2026

Uh oh!

greptile-apps Bot commented Mar 2, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

liuxiaopai-ai commented Mar 2, 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)

Compatibility / Migration

Failure Recovery (if this breaks)

Risks and Mitigations

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 2, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps Bot commented Mar 2, 2026

Greptile Summary

Confidence Score: 4/5

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants