fix: improve signal handling with process groups for hanzo net

- Use process groups (setsid) to ensure all child processes receive signals
- Send SIGTERM/SIGKILL to entire process group on shutdown
- Handles nested subprocess properly (hanzo -> hanzo/net)
- Test confirms graceful shutdown now works with Ctrl-C

Fully fixes the issue where Ctrl-C wouldn't stop hanzo net properly

pkg/hanzo/src/hanzo/cli.py

@@ -296,19 +296,33 @@ async def start_compute_node(
                     cmd_args.extend(["--default-model", models[0]])
 
                 # Run net command with detected python in a more signal-friendly way
-                process = subprocess.Popen(cmd_args, env=env)
+                # Create new process group for better signal handling
+                process = subprocess.Popen(
+                    cmd_args, 
+                    env=env,
+                    preexec_fn=os.setsid if hasattr(os, 'setsid') else None
+                )
                 
-                # Set up signal handlers to forward to subprocess
+                # Set up signal handlers to forward to subprocess group
                 def signal_handler(signum, frame):
                     if process.poll() is None:  # Process is still running
                         console.print("\n[yellow]Stopping hanzo net...[/yellow]")
-                        process.terminate()  # Try graceful termination first
                         try:
+                            # Send signal to entire process group
+                            if hasattr(os, 'killpg'):
+                                os.killpg(os.getpgid(process.pid), signal.SIGTERM)
+                            else:
+                                process.terminate()
                             process.wait(timeout=5)  # Wait up to 5 seconds
                         except subprocess.TimeoutExpired:
                             console.print("[yellow]Force stopping...[/yellow]")
-                            process.kill()  # Force kill if needed
+                            if hasattr(os, 'killpg'):
+                                os.killpg(os.getpgid(process.pid), signal.SIGKILL)
+                            else:
+                                process.kill()
                             process.wait()
+                        except ProcessLookupError:
+                            pass  # Process already terminated
                     raise KeyboardInterrupt
                 
                 # Register signal handlers

test_signal_handling.py

@@ -11,12 +11,19 @@ def test_signal_handling():
     print("Testing signal handling for 'hanzo net'...")
     print("Starting hanzo net process...")
     
-    # Start hanzo net
+    # Start hanzo net using the local module
+    import os
+    env = os.environ.copy()
+    env['PYTHONPATH'] = os.path.join(os.path.dirname(__file__), 'pkg/hanzo/src') + ':' + env.get('PYTHONPATH', '')
+    
     process = subprocess.Popen(
         [sys.executable, "-m", "hanzo", "net"],
         stdout=subprocess.PIPE,
         stderr=subprocess.PIPE,
-        text=True
+        text=True,
+        env=env,
+        cwd=os.path.dirname(__file__),
+        preexec_fn=os.setsid  # Create new process group
     )
     
     # Give it time to start
@@ -29,8 +36,38 @@ def test_signal_handling():
         print(f"Stderr: {stderr}")
         return False
     
-    print("Sending SIGINT (Ctrl-C) to process...")
-    process.send_signal(signal.SIGINT)
+    print("Process is running, checking output...")
+    # Read any initial output without blocking
+    import select
+    import fcntl
+    import os as os2
+    
+    # Make stdout non-blocking
+    fl = fcntl.fcntl(process.stdout.fileno(), fcntl.F_GETFL)
+    fcntl.fcntl(process.stdout.fileno(), fcntl.F_SETFL, fl | os2.O_NONBLOCK)
+    
+    # Try to read initial output
+    try:
+        initial_output = process.stdout.read()
+        if initial_output:
+            print(f"Initial stdout: {initial_output[:500]}")
+    except:
+        pass
+    
+    # Make stderr non-blocking too
+    fl = fcntl.fcntl(process.stderr.fileno(), fcntl.F_GETFL)
+    fcntl.fcntl(process.stderr.fileno(), fcntl.F_SETFL, fl | os2.O_NONBLOCK)
+    
+    try:
+        initial_stderr = process.stderr.read()
+        if initial_stderr:
+            print(f"Initial stderr: {initial_stderr[:500]}")
+    except:
+        pass
+    
+    print("Sending SIGINT (Ctrl-C) to process group...")
+    # Send signal to the entire process group
+    os.killpg(os.getpgid(process.pid), signal.SIGINT)
     
     # Wait for graceful shutdown (up to 10 seconds)
     try: