cmd/tailscale/cli: start WatchIPNBus before initial Start

This partially reverts f3d2fd2.

When that patch was written, the goroutine that responds to IPN notifications
could call `StartLoginInteractive`, creating a race condition that led to
flaky integration tests. We no longer call `StartLoginInteractive` in that
goroutine, so the race is now impossible.

Moving the `WatchIPNBus` call earlier ensures the CLI gets all necessary
IPN notifications, preventing a reauth from hanging.

Updates tailscale/corp#31476

Signed-off-by: Alex Chan <alexc@tailscale.com>

cmd/tailscale/cli/up.go

@@ -540,8 +540,18 @@ func runUp(ctx context.Context, cmd string, args []string, upArgs upArgsT) (retE
 		}
 	}()
 
-	running := make(chan bool, 1) // gets value once in state ipn.Running
+	// Start watching the IPN bus before we call Start() or StartLoginInteractive(),
-	watchErr := make(chan error, 1)
+	// or we could miss IPN notifications.
+	//
+	// In particular, if we're doing a force-reauth, we could miss the
+	// notification with the auth URL we should print for the user.  The
+	// initial state could contain the auth URL, but only if IPN is in the
+	// NeedsLogin state -- sometimes it's in Starting, and we don't get the URL.
+	watcher, err := localClient.WatchIPNBus(watchCtx, ipn.NotifyInitialState)
+	if err != nil {
+		return err
+	}
+	defer watcher.Close()
 
 	// Special case: bare "tailscale up" means to just start
 	// running, if there's ever been a login.
@@ -587,11 +597,8 @@ func runUp(ctx context.Context, cmd string, args []string, upArgs upArgsT) (retE
 		}
 	}
 
-	watcher, err := localClient.WatchIPNBus(watchCtx, ipn.NotifyInitialState)
+	running := make(chan bool, 1)
-	if err != nil {
+	watchErr := make(chan error, 1)
-		return err
-	}
-	defer watcher.Close()
 
 	go func() {
 		var printed bool // whether we've yet printed anything to stdout or stderr