improved tunnel cleanup

[kcoopermiller]

Note

Low Risk
Small defensive lifecycle change reusing existing sync_stop; atexit may run extra HTTP deletes on abrupt exit but does not alter normal start/stop behavior.

Overview
Adds an atexit backstop so tunnels that were started but never stopped still run cleanup when the process exits.

After a successful start(), each Tunnel is tracked in a module-level _active_tunnels set. On normal shutdown paths, sync_stop and _cleanup remove the instance from that set. At process exit, _stop_active_tunnels iterates any remaining entries and calls sync_stop() (errors ignored), which should tear down frpc, delete the backend registration, and remove the config file—reducing leaked registrations when cleanup is skipped due to crashes or forgotten stop().

^{Reviewed by Cursor Bugbot for commit 7784d73. Bugbot is set up for automated code reviews on this repo. Configure here.}

[kcoopermiller]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7784d73bc1

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".

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 7784d73. Configure here.}

[cursor]

Premature discard defeats atexit backstop on cancellation

Low Severity

_active_tunnels.discard(self) at the top of _cleanup() removes the tunnel from the atexit tracking set before any actual cleanup runs. If the async cleanup is interrupted by CancelledError at an await point (e.g., delete_tunnel or client.close), the error escapes the except Exception handlers (since CancelledError is a BaseException), skipping remaining cleanup steps. Because the tunnel was already discarded, the atexit backstop in _stop_active_tunnels can never retry the incomplete cleanup — defeating its stated purpose of catching "forgotten cleanup" and "unhandled exception" scenarios. Moving the discard to the end of _cleanup() would let the atexit handler pick up partially-cleaned tunnels via sync_stop.

 

^{Reviewed by Cursor Bugbot for commit 7784d73. Configure here.}