fix: Windows daemon survives session jobs; quieter wrap note; correct mute error

Verified on a Windows 11 VM over WinRM: doctor, SAPI synthesis, Edge,
oto/WASAPI playback, autostart, mute file, wrap exit codes, doctor --probe
and toast mode all work. Three findings fixed:

- The autostarted daemon died with the WinRM shell: Windows session hosts
  (WinRM, SSH, CI runners) put children into a job object that is
  terminated with the session. Spawn now starts the daemon with
  CREATE_BREAKAWAY_FROM_JOB first and falls back to a start without it when
  the job forbids breakaway. Confirmed: the daemon now outlives the WinRM
  shell and `talk daemon stop` still ends it.
- `talk mute` without a daemon printed "daemon call failed (<nil>)": the
  outer err was reported instead of the call's. Now "no daemon (<reason>);
  writing the mute file directly".
- `talk wrap` printed "command did not exit cleanly" for every non-zero
  exit code. The note is now reserved for commands that produced no exit
  code of their own (not found, signal); a plain failure is carried by the
  announcement and the exit code alone.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

README.md

@@ -113,9 +113,10 @@ one announcement at a time, in the order `now` > `high` > `normal` > `low`
 `7800`, config `[server] port`, env `TALK_PORT`.
 
 **Autostart:** when no daemon runs, `talk say` (and `talk wrap`) start it on
-the first speak command, detached from the terminal, logging to `daemon.log`
+the first speak command, detached from the terminal (on Windows also broken
-with its PID in `daemon.pid` in the config directory; stderr says "daemon
+away from the caller's job object, so a WinRM, SSH or CI session does not
-started" once. The daemon then keeps running (idle costs next to nothing)
+take it down), logging to `daemon.log` with its PID in `daemon.pid` in the
+config directory; stderr says "daemon started" once. The daemon then keeps running (idle costs next to nothing)
 until `talk daemon stop`, logout or reboot. Parallel first calls are
 harmless: whoever loses the port race uses the other one's daemon. If you
 manage the daemon yourself (systemd, launchd, Task Scheduler with `talk

cmd/talk/control.go

@@ -41,14 +41,15 @@ func setMute(muted bool, stdout, stderr io.Writer) int {
 	client := api.NewClient(cfg.Server.Port)
 	ctx, cancel := context.WithTimeout(context.Background(), daemonCallTimeout)
 	defer cancel()
-	if err := client.SetMute(ctx, muted); err == nil {
+	callErr := client.SetMute(ctx, muted)
+	if callErr == nil {
 		printMuteState(muted, stdout)
 		return exitOK
 	}
 
 	// Daemon call failed (usually: not running): write the file directly;
 	// the daemon (and the direct path) re-read it on every announcement.
-	fmt.Fprintf(stderr, "talk: daemon call failed (%v); writing the mute file directly\n", err)
+	fmt.Fprintf(stderr, "talk: no daemon (%v); writing the mute file directly\n", callErr)
 	dir, err := config.Dir(getenv)
 	if err != nil {
 		fmt.Fprintf(stderr, "talk: %v\n", err)

cmd/talk/wrap.go

@@ -50,9 +50,11 @@ func runWrap(args []string, opts *sayOptions, stdin io.Reader, stdout, stderr io
 	runErr := execChild(args, stdin, stdout, stderr)
 	duration := time.Since(start)
 	code := childExitCode(runErr)
-	if runErr != nil {
+	if runErr != nil && !isPlainExit(runErr) {
-		// For example: binary not found or killed by a signal. The note goes
+		// Binary not found or killed by a signal: the command never produced
-		// to stderr; the announcement names the same code wrap exits with.
+		// an exit code of its own, so say why wrap exits with the one it
+		// does. A plain non-zero exit code needs no note; the announcement
+		// and the exit code carry it.
 		fmt.Fprintf(stderr, "talk wrap: command did not exit cleanly: %v\n", runErr)
 	}
 
@@ -121,6 +123,13 @@ func childExitCode(runErr error) int {
 	return exitNotFound
 }
 
+// isPlainExit reports whether runErr is just a non-zero exit code of a
+// command that ran and exited normally.
+func isPlainExit(runErr error) bool {
+	var ee *exec.ExitError
+	return errors.As(runErr, &ee) && ee.ExitCode() >= 0
+}
+
 // wrapLabel resolves the announcement label: config `wrap.label` (env
 // override TALK_WRAP_LABEL) beats the command's base name
 // (`/usr/bin/make` → "make").

internal/daemon/spawn.go

@@ -17,7 +17,6 @@ import (
 // port in use) without waiting for a timeout. If the caller exits first,
 // the daemon keeps running (that is the point).
 func Spawn(exe, logPath string, env []string) (<-chan error, error) {
-	var logFile *os.File
 	if logPath == "" {
 		logPath = os.DevNull
 	}
@@ -27,13 +26,24 @@ func Spawn(exe, logPath string, env []string) (<-chan error, error) {
 	}
 	defer logFile.Close() // the child holds its own handle
 
-	cmd := exec.Command(exe, "daemon", "run")
+	// First attempt: fully detached, including a breakaway from the parent's
-	cmd.Env = env
+	// job object (Windows). Shells such as WinRM, SSH servers or CI runners
-	cmd.Stdin = nil
+	// put children into a job that is terminated with the session; the
-	cmd.Stdout = logFile
+	// breakaway lets the daemon outlive it. If the job forbids breakaway,
-	cmd.Stderr = logFile
+	// CreateProcess fails, and the second attempt starts without it.
-	detach(cmd)
+	var cmd *exec.Cmd
-	if err := cmd.Start(); err != nil {
+	for _, breakaway := range []bool{true, false} {
+		cmd = exec.Command(exe, "daemon", "run")
+		cmd.Env = env
+		cmd.Stdin = nil
+		cmd.Stdout = logFile
+		cmd.Stderr = logFile
+		detach(cmd, breakaway)
+		if err = cmd.Start(); err == nil {
+			break
+		}
+	}
+	if err != nil {
 		return nil, fmt.Errorf("starting daemon (%s daemon run): %w", exe, err)
 	}
 	exited := make(chan error, 1)

internal/daemon/spawn_other.go

@@ -5,4 +5,4 @@ package daemon
 import "os/exec"
 
 // detach is a no-op on this platform (no Setsid, no console).
-func detach(*exec.Cmd) {}
+func detach(*exec.Cmd, bool) {}

internal/daemon/spawn_unix.go

@@ -7,8 +7,9 @@ import (
 	"syscall"
 )
 
-// detach detaches the child from the terminal: own session (no SIGHUP when
+// detach detaches the child from the terminal (the breakaway flag is a
+// Windows concept and ignored here): own session (no SIGHUP when
 // the terminal closes, no Ctrl-C from the shell).
-func detach(cmd *exec.Cmd) {
+func detach(cmd *exec.Cmd, _ bool) {
 	cmd.SysProcAttr = &syscall.SysProcAttr{Setsid: true}
 }

internal/daemon/spawn_windows.go

@@ -11,10 +11,18 @@ import (
 // binding to the caller's console.
 const detachedProcess = 0x00000008
 
+// createBreakawayFromJob is the Win32 flag CREATE_BREAKAWAY_FROM_JOB: the
+// child leaves the parent's job object, so a session-bound job (WinRM, SSH,
+// CI) does not take the daemon down with it. Only works when the job allows
+// it; Spawn falls back to a start without the flag otherwise.
+const createBreakawayFromJob = 0x01000000
+
 // detach detaches the child from the console: own process group (no Ctrl-C
 // from the shell) and no console (no flashing window).
-func detach(cmd *exec.Cmd) {
+func detach(cmd *exec.Cmd, breakaway bool) {
-	cmd.SysProcAttr = &syscall.SysProcAttr{
+	flags := uint32(syscall.CREATE_NEW_PROCESS_GROUP | detachedProcess)
-		CreationFlags: syscall.CREATE_NEW_PROCESS_GROUP | detachedProcess,
+	if breakaway {
+		flags |= createBreakawayFromJob
 	}
+	cmd.SysProcAttr = &syscall.SysProcAttr{CreationFlags: flags}
 }