Auto-rebuild listener on fatal socket errors and refactor listener creation

[CypherPotato]

Motivation

• Improve robustness of the TCP listener by detecting fatal socket errors during accept and automatically rebuilding the listening socket so the host can recover without full process restart.
• Reduce duplication and make socket configuration explicit by centralizing listener creation and options.

Description

• Replace readonly _listener with a mutable socket and introduce CreateListenerSocket() to centralize socket construction and options previously in ConfigureListenerSocket().
• On Start() dispose the previous listener (if any) and recreate a fresh listener, then bind and listen before starting the accept pool.
• Add _listenerRestarting flag and logic in StartAccept(), OnAcceptCompleted() and ProcessAcceptInline() to detect fatal SocketError values via IsListenerFatalError() and trigger RebuildListenerAsync() which retries creating/binding a new listener and restarts accept loops.
• Return negative rearm values from ProcessAcceptInline() to indicate caller should stop rearming and handle restart scenarios, and adjust accept-queueing logic accordingly.

Testing

• No automated tests were executed as part of this change.

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0e5d53d365

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]

Dispose failed rebuilt listener before retry loop

When CreateListenerSocket()/Bind/Listen throws during RebuildListenerAsync, the catch (SocketException) block swallows the exception and immediately retries without disposing the just-created newListener. In a persistent failure (e.g., endpoint temporarily unavailable), this leaks one socket handle every 250ms, which can exhaust file descriptors and prevent the host from recovering.

Useful? React with 👍 / 👎.

[CypherPotato]

@codex work on this comment

[chatgpt-codex-connector]

Summary

• Fixed the listener-rebuild retry path to avoid socket handle leaks when CreateListenerSocket(), Bind, or Listen fails with a SocketException by hoisting newListener outside the try and disposing it in the catch before the next retry.
• Kept the existing successful-path behavior intact (_listener = newListener; break;) so accept-loop restart semantics are unchanged while adding cleanup only on failure.
• Committed the fix on the current branch (5875b07) and created a follow-up PR message via the make_pr tool describing the change and rationale.

Testing

• ✅ cd /workspace/core && rg -n "RebuildListenerAsync|CreateListenerSocket|SocketException|newListener" cadente/Sisk.Cadente/HttpHost.cs
• ✅ cd /workspace/core && git diff -- cadente/Sisk.Cadente/HttpHost.cs
• ✅ cd /workspace/core && git add cadente/Sisk.Cadente/HttpHost.cs && git commit -m "Dispose failed rebuilt listener before retry"
• ✅ cd /workspace/core && nl -ba cadente/Sisk.Cadente/HttpHost.cs | sed -n '258,286p'
• ✅ Web check performed: search_query("C# SocketException retry loop dispose socket on bind listen failure")

View task →