fix(websocket): avoid asyncSpawn and make ws shutdown cleanup robust

[Copilot]

Summary

Reworks WebSocket connection cleanup to avoid untracked asyncSpawn usage and to eliminate flaky shutdown behavior where a connection can remain open after transport stop.

In addition to tracking per-connection cleanup futures, this updates WsTransport.stop() to wait for all close/cleanup futures via allFinished, so one failing close does not short-circuit remaining cleanup work.

Affected Areas

• Gossipsub

• Transports

  WebSocket transport cleanup/shutdown flow in libp2p/transports/wstransport.nim.

• Peer Management / Discovery

• Protocol Logic

• Build / Tooling

• Other

Compatibility & Downstream Validation

Reference PRs / branches / commits demonstrating successful integration:

• Nimbus:
  N/A

• Waku:
  N/A

• Codex:
  N/A

Impact on Library Users

No public API changes.

Behavioral impact is internal to WebSocket transport shutdown: stop now drains cleanup more reliably under dial-cancellation/close-error timing, reducing flaky leaked-stream outcomes in tests and CI.

Risk Assessment

Low to medium risk.

Changes are limited to WebSocket transport stop/cleanup sequencing. Main risk is shutdown-path timing differences, but logic is constrained to teardown and intended to improve determinism and resource cleanup.

References

• Original flaky behavior and rationale: connection can remain open after transport close
• Related prior discussion/PR context: feat: use BoringSSL for noise crypto primitives #2469

Additional Notes

Follow-up fix after CI feedback: handle dial cancellation was still failing with tracker mismatch (Opened stream.transport vs Closed stream.transport) on macOS CI. This was addressed by replacing allFutures with allFinished in the stop cleanup path.

[richard-ramos]

ah interesting! I had seen this happening yesterday in #2468 and was wondering what was that about 🤔

[Copilot]

Pull request overview

This PR updates the WebSocket transport to track per-connection cleanup futures instead of using untracked asyncSpawn, aligning the transport with repository async task management rules and improving shutdown behavior.

Changes:

• Adds connectionCleanupFuts to WsTransport.
• Tracks WebSocket connection cleanup futures created in connHandler.
• Updates stop() to close active connections and wait for pending cleanup futures.

[vladopajic]

ah interesting! I had seen this happening yesterday in #2468 and was wondering what was that about 🤔

yeah, i took it form your pr, but it was #2469

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.47%. Comparing base (6e876df) to head (3ce82d9).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2471      +/-   ##
==========================================
- Coverage   84.48%   84.47%   -0.01%     
==========================================
  Files         154      154              
  Lines       26393    26396       +3     
  Branches       12       12              
==========================================
+ Hits        22298    22299       +1     
- Misses       4095     4097       +2     

Files with missing lines	Coverage Δ
libp2p/transports/wstransport.nim	81.69% <100.00%> (-0.11%)	⬇️

... and 3 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated no new comments.

[vladopajic]

not fixed: https://github.com/vacp2p/nim-libp2p/actions/runs/25933002749/job/76231265652
failed again with current code

[vladopajic]

@copilot tests have failed with same error; please analyze current code and provide details why it fails

[Copilot]

not fixed: https://github.com/vacp2p/nim-libp2p/actions/runs/25933002749/job/76231265652
failed again with current code

I tracked this to the same flaky path in handle dial cancellation and pushed a fix in 3ce82d9: WsTransport.stop() now uses allFinished for both connection closes and cleanup futures so one close error cannot abort the remaining shutdown cleanup.

@copilot tests have failed with same error; please analyze current code and provide details why it fails

Analysis from run 25933002749 / job 76231265652: only test failure was WebSocket transport -> handle dial cancellation, with tracker mismatch Opened stream.transport: 1096 | Closed stream.transport: 1095. The shutdown path in WsTransport.stop() was using allFutures, so a single close failure could short-circuit the rest of cleanup and leave one stream unclosed. I fixed that in 3ce82d9 by switching those waits to allFinished so shutdown still drains all close/cleanup futures.