test: handle aggregate balanced pool errors

[marko1olo]

Summary

• Handle request errors whose port or meaningful code is wrapped inside AggregateError.errors in the balanced-pool weighted round-robin tests.
• Stop assuming every thrown request error has error.socket.remotePort.
• Keep the existing expected request ordering and ratio assertions unchanged.

Why

On Windows, recent Node versions can surface localhost connection failures as aggregate errors with nested ECONNREFUSED details. That made the test either throw while reading error.socket.remotePort or fail to count the intended connection-refused event.

Test

• node --test test\node-test\balanced-pool.js --test-name-pattern "weighted round robin"
• npx eslint test\node-test\balanced-pool.js
• git diff --check

AI-assisted contribution: implementation and verification were performed with OpenAI Codex under my direction.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.49%. Comparing base (1089808) to head (8c0848d).
⚠️ Report is 5 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5377      +/-   ##
==========================================
- Coverage   93.50%   93.49%   -0.02%     
==========================================
  Files         110      110              
  Lines       38303    38303              
==========================================
- Hits        35816    35810       -6     
- Misses       2487     2493       +6     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[marko1olo]

Rebased on main and cut down. This had drifted into conflict, and re-checking it against current main showed most of it was no longer needed.

What changed:

• Dropped the e.socket.remotePort guard. test: guard balanced-pool error port lookup #5463 landed the equivalent guard on 2 July as getErrorPort(), so there was nothing left to do there.

• Dropped getRequestErrorCode(). I had assumed the ECONNREFUSED / UND_ERR_SOCKET branches needed the code unwrapped out of the AggregateError. That assumption was wrong — Node sets code on the aggregate as well as on the causes:

  constructor : AggregateError
  e.code      : "ECONNREFUSED"
  

  so the existing branches already classify correctly and that helper was dead weight.

• Kept the port lookup, now recursing into err.errors. This is the part still broken on main. getErrorPort() avoids the crash but returns undefined for an aggregate, because the aggregate carries no port of its own:

  getErrorPort(e)       : undefined      <- main
  servers.find(...)     : NO MATCH  -> falls through to `throw e`
  recursing equivalent  : matched server A
  

  So test: guard balanced-pool error port lookup #5463 converted the TypeError into a rethrow rather than making the case pass.

Why an aggregate turns up here at all: the cases build their upstreams as http://localhost:${port}, and localhost resolves to both ::1 and 127.0.0.1. When a test server is down, both families are attempted and the two refusals come back wrapped by the happy eyeballs attempt. That is also why this surfaces on the Windows jobs specifically.

The added test does not depend on a downed server's timing — nothing listens on port 1, so the refusal is deterministic. Against the previous helper it fails with actual: undefined, expected: 1.

Locally node --test test/node-test/balanced-pool.js gives 23/23 and eslint is clean. That is Node 24 on Windows, so I have not reproduced the Node 22 matrix cell that originally went red — the argument for the fix rests on the aggregate shape above, not on my run.