Fix process tree cancellation for tools

src/agent_tools.py

@@ -6,6 +6,7 @@ import os
 import re
 import selectors
 import shutil
+import signal
 import subprocess
 import sys
 import time
@@ -2087,9 +2088,9 @@ def _run_python_exec(arguments: dict[str, Any], context: ToolExecutionContext) -
             stderr=subprocess.PIPE,
             text=True,
             env=_build_subprocess_env(context),
+            **_subprocess_isolation_kwargs(),
         )
         _register_child_process(context, process)
-        _register_child_process(context, process)
         if stdin and process.stdin is not None:
             process.stdin.write(stdin)
             process.stdin.close()
@@ -2108,7 +2109,7 @@ def _run_python_exec(arguments: dict[str, Any], context: ToolExecutionContext) -
     except subprocess.TimeoutExpired:
         if process is not None:
             _terminate_process(process)
-            stdout, stderr = process.communicate()
+            stdout, stderr = _communicate_after_stop(process)
         else:
             stdout, stderr = '', ''
         payload = [
@@ -2200,19 +2201,67 @@ def _unregister_child_process(
         registry.discard(process)
 
 
+def _subprocess_isolation_kwargs() -> dict[str, Any]:
+    if os.name == 'posix':
+        return {'start_new_session': True}
+    creation_flags = getattr(subprocess, 'CREATE_NEW_PROCESS_GROUP', 0)
+    if creation_flags:
+        return {'creationflags': creation_flags}
+    return {}
+
+
 def _terminate_process(process: subprocess.Popen[Any]) -> None:
     if process.poll() is not None:
         return
+    _signal_process_tree(process, signal.SIGTERM)
+    try:
+        process.wait(timeout=1)
+        return
+    except (OSError, subprocess.TimeoutExpired):
+        pass
+    _signal_process_tree(process, signal.SIGKILL)
     try:
-        process.terminate()
         process.wait(timeout=1)
     except (OSError, subprocess.TimeoutExpired):
-        try:
+        pass
+
+
+def _signal_process_tree(process: subprocess.Popen[Any], sig: signal.Signals) -> None:
+    try:
+        if os.name == 'posix':
+            pgid = os.getpgid(process.pid)
+            if pgid == process.pid:
+                os.killpg(pgid, sig)
+                return
+        if sig == signal.SIGTERM:
+            process.terminate()
+        else:
             process.kill()
+    except ProcessLookupError:
+        return
+    except OSError:
+        try:
+            if sig == signal.SIGTERM:
+                process.terminate()
+            else:
+                process.kill()
         except OSError:
             pass
 
 
+def _communicate_after_stop(process: subprocess.Popen[str]) -> tuple[str, str]:
+    try:
+        stdout, stderr = process.communicate(timeout=1)
+        return stdout or '', stderr or ''
+    except subprocess.TimeoutExpired:
+        _terminate_process(process)
+        try:
+            stdout, stderr = process.communicate(timeout=1)
+            return stdout or '', stderr or ''
+        except (OSError, subprocess.TimeoutExpired):
+            return '', ''
+
+
 def _run_python_package(arguments: dict[str, Any], context: ToolExecutionContext) -> str:
     _ensure_process_execution_allowed(context, 'Python package management')
     action = _require_string(arguments, 'action')
@@ -3633,7 +3682,9 @@ def _stream_bash(
             text=True,
             bufsize=1,
             env=_build_subprocess_env(context),
+            **_subprocess_isolation_kwargs(),
         )
+        _register_child_process(context, process)
     except (ToolPermissionError, ToolExecutionError, OSError, subprocess.SubprocessError) as exc:
         yield ToolStreamUpdate(
             kind='result',
@@ -3663,7 +3714,7 @@ def _stream_bash(
                 timeout_error = (
                     f'Command timed out after {context.command_timeout_seconds:.1f}s: {command}'
                 )
-                process.kill()
+                _terminate_process(process)
                 break
             events = selector.select(timeout=min(remaining, 0.1))
             if not events and process.poll() is not None: