fix: terminate owned runtime process tree on stop/forceStop

Add a private kill-process-tree helper to each SDK, called from the
existing owned-process termination points in stop() and forceStop().

Spawn-time isolation (POSIX):
- Node.js: detached: true
- Python: start_new_session=True
- Go: SysProcAttr.Setpgid = true
- Rust: process_group(0)

Teardown:
- Windows (all): taskkill /T /F /PID
- Node.js/Python/Go (POSIX): kill(-pid, SIGKILL) — process group signal
- Rust (POSIX): libc::kill(-pid, SIGKILL)
- Java: ProcessHandle.descendants() snapshot + destroyForcibly each
- .NET: already uses Kill(entireProcessTree: true) — no change needed

No public API changes. External-server and in-process (FFI) paths are
not affected.

Closes #1804

Author: j-zhangyiyuan

go/client.go

@@ -634,7 +634,7 @@ func (c *Client) ForceStop() {
 	// Kill the process without waiting for startStopMux, which Start may hold.
 	// This unblocks any I/O Start is doing (connect, version check).
 	if p := c.osProcess.Swap(nil); p != nil {
-		p.Kill()
+		killProcessTreeByPid(p.Pid)
 	}
 
 	// Clear sessions immediately without trying to destroy them
@@ -2189,9 +2189,7 @@ func (c *Client) killProcess() error {
 		c.ffiHost = nil
 	}
 	if p := c.osProcess.Swap(nil); p != nil {
-		if err := p.Kill(); err != nil {
-			return fmt.Errorf("failed to kill CLI process: %w", err)
-		}
+		killProcessTreeByPid(p.Pid)
 	}
 	c.process = nil
 	return nil

go/process_other.go

@@ -2,10 +2,18 @@
 
 package copilot
 
-import "os/exec"
+import (
+	"os/exec"
+	"syscall"
+)
 
-// configureProcAttr configures platform-specific process attributes.
-// On non-Windows platforms, this is a no-op.
+// configureProcAttr places the runtime in its own process group so
+// killProcessTreeByPid can signal all descendants atomically.
 func configureProcAttr(cmd *exec.Cmd) {
-	// No special configuration needed on non-Windows platforms
+	cmd.SysProcAttr = &syscall.SysProcAttr{Setpgid: true}
+}
+
+// killProcessTreeByPid signals the process group (negative PID) with SIGKILL.
+func killProcessTreeByPid(pid int) {
+	_ = syscall.Kill(-pid, syscall.SIGKILL)
 }

go/process_windows.go

@@ -3,14 +3,19 @@
 package copilot
 
 import (
+	"fmt"
 	"os/exec"
 	"syscall"
 )
 
-// configureProcAttr configures platform-specific process attributes.
-// On Windows, this hides the console window to avoid distracting users in GUI apps.
+// configureProcAttr hides the console window on Windows.
 func configureProcAttr(cmd *exec.Cmd) {
 	cmd.SysProcAttr = &syscall.SysProcAttr{
 		HideWindow: true,
 	}
 }
+
+// killProcessTreeByPid terminates the entire process tree via taskkill /T /F.
+func killProcessTreeByPid(pid int) {
+	_ = exec.Command("taskkill", "/T", "/F", "/PID", fmt.Sprintf("%d", pid)).Run()
+}

java/src/main/java/com/github/copilot/CopilotClient.java

@@ -480,19 +480,19 @@ private static void cleanupCliProcess(Process process, boolean forceImmediately)
                 // will never come just wastes time, so terminate the child
                 // immediately and only wait to reap it.
                 if (forceImmediately) {
-                    process.destroyForcibly();
+                    killProcessTree(process);
                     if (!process.waitFor(FORCE_KILL_TIMEOUT_SECONDS, TimeUnit.SECONDS)) {
                         LOG.fine("Process did not terminate within force kill timeout");
                     }
                     return;
                 }
 
-                process.destroy();
+                killProcessTree(process);
                 if (process.waitFor(FORCE_KILL_TIMEOUT_SECONDS, TimeUnit.SECONDS)) {
                     return;
                 }
 
-                process.destroyForcibly();
+                killProcessTree(process);
                 if (!process.waitFor(FORCE_KILL_TIMEOUT_SECONDS, TimeUnit.SECONDS)) {
                     LOG.fine("Process did not terminate within force kill timeout");
                 }
@@ -505,6 +505,22 @@ private static void cleanupCliProcess(Process process, boolean forceImmediately)
         }
     }
 
+    /**
+     * Terminate the runtime's process tree: snapshot all descendants, destroy
+     * them, then destroy the root.  Uses {@link ProcessHandle#descendants()}
+     * which works cross-platform (Windows, Linux, macOS).
+     */
+    private static void killProcessTree(Process process) {
+        try {
+            process.toHandle().descendants().forEach(ph -> {
+                try { ph.destroyForcibly(); } catch (Exception ignored) {}
+            });
+        } catch (Exception e) {
+            LOG.log(Level.FINE, "Error killing process descendants", e);
+        }
+        process.destroyForcibly();
+    }
+
     /**
      * Creates a new Copilot session with the specified configuration.
      * <p>

nodejs/src/client.ts

@@ -11,7 +11,7 @@
  * @module client
  */
 
-import { spawn, type ChildProcess } from "node:child_process";
+import { spawn, execSync, type ChildProcess } from "node:child_process";
 import { randomUUID } from "node:crypto";
 import { existsSync } from "node:fs";
 import { createRequire } from "node:module";
@@ -153,6 +153,40 @@ async function waitForChildExit(child: ChildProcess, timeoutMs: number): Promise
     });
 }
 
+/**
+ * Terminate the runtime's process tree.
+ *
+ * - Windows: `taskkill /T /F` kills the entire tree rooted at `pid`.
+ * - POSIX: the runtime is spawned in its own process group (`detached: true`),
+ *   so `kill(-pid)` signals every process in that group.
+ *
+ * Falls back to `child.kill(signal)` if the tree-wide signal fails (e.g. the
+ * process already exited).
+ *
+ * @see https://github.com/github/copilot-sdk/issues/1804
+ */
+function killProcessTree(child: ChildProcess, signal: NodeJS.Signals = "SIGTERM"): boolean {
+    const pid = child.pid;
+    if (pid == null) {
+        return false;
+    }
+    if (process.platform === "win32") {
+        try {
+            execSync(`taskkill /T /F /PID ${pid}`, { stdio: "ignore", timeout: 5000 });
+            return true;
+        } catch {
+            return child.kill(signal);
+        }
+    }
+    // POSIX: signal the process group (negative PID).
+    try {
+        process.kill(-pid, signal);
+        return true;
+    } catch {
+        return child.kill(signal);
+    }
+}
+
 /**
  * Convert tool parameters to JSON schema format for sending to CLI
  */
@@ -1082,13 +1116,17 @@ export class CopilotClient {
             this.cliProcess = null;
             try {
                 if (child.exitCode == null && child.signalCode == null) {
-                    child.kill();
+                    killProcessTree(child, "SIGTERM");
                     if (!(await waitForChildExit(child, RUNTIME_SHUTDOWN_TIMEOUT_MS))) {
-                        errors.push(
-                            new Error(
-                                `Timed out waiting for CLI process to exit after kill: ${RUNTIME_SHUTDOWN_TIMEOUT_MS}ms`
-                            )
-                        );
+                        // SIGTERM-resistant descendants may survive; escalate to SIGKILL.
+                        killProcessTree(child, "SIGKILL");
+                        if (!(await waitForChildExit(child, RUNTIME_SHUTDOWN_TIMEOUT_MS))) {
+                            errors.push(
+                                new Error(
+                                    `Timed out waiting for CLI process to exit after kill: ${RUNTIME_SHUTDOWN_TIMEOUT_MS}ms`
+                                )
+                            );
+                        }
                     }
                 }
             } catch (error) {
@@ -1209,7 +1247,7 @@ export class CopilotClient {
         // Force kill CLI process (only if we spawned it)
         if (this.cliProcess && !this.isExternalServer) {
             try {
-                this.cliProcess.kill("SIGKILL");
+                killProcessTree(this.cliProcess, "SIGKILL");
             } catch {
                 // Ignore errors
             }
@@ -2468,22 +2506,33 @@ export class CopilotClient {
                     : ["ignore", "pipe", "pipe"];
 
             // For .js files, spawn node explicitly; for executables, spawn directly
+            // Place the runtime in its own process group so killProcessTree()
+            // can signal all descendants atomically. On Windows detached has
+            // no effect — taskkill /T handles tree termination instead.
+            const detached = process.platform !== "win32";
             const isJsFile = this.resolvedCliPath.endsWith(".js");
             if (isJsFile) {
                 this.cliProcess = spawn(getNodeExecPath(), [this.resolvedCliPath, ...args], {
                     stdio: stdioConfig,
                     cwd: this.options.workingDirectory,
                     env: envWithoutNodeDebug,
                     windowsHide: true,
+                    detached,
                 });
             } else {
                 this.cliProcess = spawn(this.resolvedCliPath, args, {
                     stdio: stdioConfig,
                     cwd: this.options.workingDirectory,
                     env: envWithoutNodeDebug,
                     windowsHide: true,
+                    detached,
                 });
             }
+            // Prevent the detached child from keeping the parent's event loop
+            // alive when the embedder exits without calling stop().
+            if (detached) {
+                this.cliProcess.unref();
+            }
 
             let stdout = "";
             let resolved = false;

python/copilot/client.py

@@ -1200,6 +1200,44 @@ def _session_lifecycle_event_from_dict(data: dict) -> SessionLifecycleEvent:
 _CLI_PROCESS_EXIT_TIMEOUT_SECONDS = 5
 
 
+def _kill_process_tree(proc: subprocess.Popen[Any]) -> None:
+    """Terminate the runtime's process tree.
+
+    Windows: ``taskkill /T /F`` kills the entire tree rooted at *pid*.
+    POSIX:   the runtime is spawned with ``start_new_session=True``, so
+             ``os.killpg(pid, signal)`` signals every process in that group.
+
+    Falls back to ``proc.kill()`` if the tree-wide signal fails.
+
+    See: https://github.com/github/copilot-sdk/issues/1804
+    """
+    pid = proc.pid
+    if pid is None:
+        return
+    if sys.platform == "win32":
+        try:
+            result = subprocess.run(
+                ["taskkill", "/T", "/F", "/PID", str(pid)],
+                capture_output=True,
+                timeout=5,
+            )
+            if result.returncode != 0:
+                proc.kill()
+        except Exception:
+            try:
+                proc.kill()
+            except Exception:
+                pass
+    else:
+        try:
+            os.killpg(pid, 9)  # SIGKILL to the runtime's process group
+        except (ProcessLookupError, PermissionError, OSError):
+            try:
+                proc.kill()
+            except Exception:
+                pass
+
+
 def _get_or_download_cli(*, include_runtime_lib: bool = False) -> str | None:
     """Get the cached CLI binary, downloading if necessary.
 
@@ -1910,14 +1948,14 @@ async def stop(self) -> None:
             poll = getattr(self._cli_process, "poll", None)
             is_running = poll is None or poll() is None
             if is_running:
-                self._cli_process.terminate()
+                _kill_process_tree(self._cli_process)
                 try:
                     await asyncio.to_thread(
                         self._cli_process.wait,
                         timeout=_CLI_PROCESS_EXIT_TIMEOUT_SECONDS,
                     )
                 except subprocess.TimeoutExpired:
-                    self._cli_process.kill()
+                    _kill_process_tree(self._cli_process)
                     try:
                         await asyncio.to_thread(
                             self._cli_process.wait,
@@ -1976,7 +2014,7 @@ async def force_stop(self) -> None:
                     if self._process is not None and self._process is not self._cli_process:
                         self._process.terminate()
                     if self._cli_process is not None:
-                        self._cli_process.kill()
+                        _kill_process_tree(self._cli_process)
                     self._process = None
                     self._cli_process = None
             except Exception:
@@ -4027,6 +4065,9 @@ async def _start_cli_server(self) -> None:
                 cwd=cwd,
                 env=env,
                 creationflags=creationflags,
+                # Place the runtime in its own process group so
+                # _kill_process_tree() can signal all descendants.
+                start_new_session=(sys.platform != "win32"),
             )
             self._cli_process = self._process
         else:
@@ -4040,6 +4081,7 @@ async def _start_cli_server(self) -> None:
                 cwd=cwd,
                 env=env,
                 creationflags=creationflags,
+                start_new_session=(sys.platform != "win32"),
             )
             self._cli_process = self._process
         log_timing(