Unblock session calls when broker workers shut down

[RiSKeD]

All session methods (Print, Printf, Println, RequestFile, SendFile) blocked indefinitely on unbuffered channels if the companion toClientWorker had already exited — for example when fromClientWorker returned an error first and cancelled the shared worker context.

Add a done <-chan struct{} field to session, set to workerCtx.Done() in Broker.Start. Every blocking channel send/receive in the session now selects on done and returns immediately (or with an error for file operations) when the broker is shutting down.

[Copilot]

Pull request overview

This PR prevents module session calls from blocking indefinitely when the broker’s worker goroutines have already exited (e.g., due to early cancellation), by introducing a shutdown signal and making session operations cancellation-aware.

Changes:

• Added a done <-chan struct{} to session, wired to the broker worker context’s Done().
• Updated Print, Printf, Println to avoid blocking sends by selecting on done.
• Updated RequestFile/SendFile to select on done for blocking channel operations and return errors on shutdown.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
internal/dutagent/session.go	Adds done and uses it to unblock or fail fast in session print and file-transfer APIs.
internal/dutagent/broker.go	Wires session.done to the broker worker context so session calls can observe shutdown.

Comments suppressed due to low confidence (1)

internal/dutagent/session.go:21

• This PR unblocks Print* and file operations when broker workers shut down, but Console() still returns chanio.ChanWriter instances that write to unbuffered channels (internal/chanio/chanio.go:110-115). If toClientWorker has exited, module writes to stdout/stderr can still block indefinitely, which seems to contradict the PR intent of avoiding deadlocks when workers stop. Consider wrapping the console writers with a session-aware writer that selects on s.done, or otherwise ensuring stdout/stderr writes cannot block once done is closed.

	done      <-chan struct{} // closed when broker workers shut down; unblocks pending session calls
	printCh   chan string
	stdinCh   chan []byte
	stdoutCh  chan []byte
	stderrCh  chan []byte

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

SendFile now returns an error when done is closed, but the method still log.Fatals if s.currentFile != "" before it ever checks done. With the new shutdown-unblocking behavior, currentFile can remain set during cancellation (e.g., an in-flight RequestFile), which can turn a shutdown into a process exit. Consider checking done (and returning an error) before the currentFile != "" fatal, or ensuring currentFile is always cleared on shutdown.

[Copilot]

If ctx.Done() fires here, fromClientWorker returns without clearing s.currentFile. In cancellation scenarios where toClientWorker has already set currentFile for an outstanding request, this can leave the session in an inconsistent state and may cause later calls (notably SendFile, which checks currentFile) to log.Fatal. Consider resetting s.currentFile before returning on the cancellation path.

Suggested change

	case <-ctx.Done():
	case <-ctx.Done():
	s.currentFile = ""