Fix hanging on exec for missing command (#33)

Author: sjmiller609

cmd/api/api/exec.go

@@ -138,7 +138,11 @@ func (s *ApiService) ExecHandler(w http.ResponseWriter, r *http.Request) {
 			"duration_ms", duration.Milliseconds(),
 		)
 		// Send error message over WebSocket before closing
-		ws.WriteMessage(websocket.TextMessage, []byte(fmt.Sprintf("Error: %v", err)))
+		// Use BinaryMessage so the CLI writes it to stdout (it ignores TextMessage for output)
+		// Use \r\n so it displays properly when client terminal is in raw mode
+		ws.WriteMessage(websocket.BinaryMessage, []byte(fmt.Sprintf("Error: %v\r\n", err)))
+		// Send exit code 127 (command not found - standard Unix convention)
+		ws.WriteMessage(websocket.TextMessage, []byte(`{"exitCode":127}`))
 		return
 	}
 
@@ -199,4 +203,3 @@ func (w *wsReadWriter) Write(p []byte) (n int, err error) {
 	}
 	return len(p), nil
 }
-

cmd/exec/main.go

@@ -143,22 +143,22 @@ func main() {
 
 	// Use custom WebSocket dialer
 	dialer := &websocket.Dialer{}
-	
+
 	// Set up headers for WebSocket connection (body was already sent)
 	headers := http.Header{}
 	headers.Set("Authorization", fmt.Sprintf("Bearer %s", jwtToken))
 
 	// Make HTTP POST with body
 	client := &http.Client{}
-	
+
 	// Actually, we need a custom approach. Let me use a modified request
 	// that sends body AND upgrades to WebSocket.
 	// For simplicity, let's POST the JSON as the Sec-WebSocket-Protocol header value (hacky but works)
 	// OR we can encode params in URL query string
-	
+
 	// Actually, the simplest approach: POST the body first, get a session ID, then connect WebSocket
 	// But that requires server changes.
-	
+
 	// Let's use the approach where we send JSON as first WebSocket message after connect
 	ws, resp, err := dialer.Dial(wsURL.String(), headers)
 	if err != nil {
@@ -240,7 +240,10 @@ func runInteractive(ws *websocket.Conn) (int, error) {
 		for {
 			msgType, message, err := ws.ReadMessage()
 			if err != nil {
-				if !websocket.IsUnexpectedCloseError(err, websocket.CloseNormalClosure, websocket.CloseGoingAway) {
+				if websocket.IsUnexpectedCloseError(err, websocket.CloseNormalClosure, websocket.CloseGoingAway) {
+					// Unexpected close - report as error to avoid hanging
+					errCh <- fmt.Errorf("connection closed unexpectedly: %w", err)
+				} else {
 					// Normal close, default exit code 0
 					exitCodeCh <- 0
 				}
@@ -310,8 +313,8 @@ func runNonInteractive(ws *websocket.Conn) (int, error) {
 			msgType, message, err := ws.ReadMessage()
 			if err != nil {
 				// Connection closed is normal - default exit code 0
-				if websocket.IsCloseError(err, websocket.CloseNormalClosure, websocket.CloseGoingAway, websocket.CloseAbnormalClosure) || 
-				   err == io.EOF {
+				if websocket.IsCloseError(err, websocket.CloseNormalClosure, websocket.CloseGoingAway, websocket.CloseAbnormalClosure) ||
+					err == io.EOF {
 					exitCodeCh <- 0
 					return
 				}

lib/instances/exec_test.go

@@ -9,6 +9,7 @@ import (
 	"testing"
 	"time"
 
+	"github.com/onkernel/hypeman/lib/exec"
 	"github.com/onkernel/hypeman/lib/images"
 	"github.com/onkernel/hypeman/lib/paths"
 	"github.com/onkernel/hypeman/lib/system"
@@ -214,4 +215,43 @@ func TestExecConcurrent(t *testing.T) {
 		"streams appear serialized - took %v, expected < %v", streamElapsed, maxExpected)
 
 	t.Logf("Long-running streams completed in %v (concurrent OK)", streamElapsed)
+
+	// Phase 3: Test command not found returns quickly (no hang)
+	// Regression test for a hang that occurred when the command wasn't found.
+	t.Log("Phase 3: Testing exec with non-existent command...")
+
+	// Test without TTY
+	start := time.Now()
+	var stdout, stderr strings.Builder
+	_, err = exec.ExecIntoInstance(ctx, inst.VsockSocket, exec.ExecOptions{
+		Command: []string{"nonexistent_command_asdfasdf"},
+		Stdout:  &stdout,
+		Stderr:  &stderr,
+		TTY:     false,
+	})
+	elapsed := time.Since(start)
+	t.Logf("Exec (no TTY) completed in %v (error: %v)", elapsed, err)
+
+	require.Error(t, err, "exec should fail for non-existent command")
+	require.Contains(t, err.Error(), "executable file not found", "error should mention command not found")
+	require.Less(t, elapsed, 5*time.Second, "exec should not hang, took %v", elapsed)
+
+	// Test with TTY
+	start = time.Now()
+	stdout.Reset()
+	stderr.Reset()
+	_, err = exec.ExecIntoInstance(ctx, inst.VsockSocket, exec.ExecOptions{
+		Command: []string{"nonexistent_command_xyz123"},
+		Stdout:  &stdout,
+		Stderr:  &stderr,
+		TTY:     true,
+	})
+	elapsed = time.Since(start)
+	t.Logf("Exec (with TTY) completed in %v (error: %v)", elapsed, err)
+
+	require.Error(t, err, "exec with TTY should fail for non-existent command")
+	require.Contains(t, err.Error(), "executable file not found", "error should mention command not found")
+	require.Less(t, elapsed, 5*time.Second, "exec with TTY should not hang, took %v", elapsed)
+
+	t.Log("Command not found tests passed - exec does not hang")
 }