Fix host cleanup terminating CI runners

CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## 0.2.1
+
+- Interrupt only the configured host process during graceful shutdown so
+  Topcoat can clean up its own child server without terminating the CI runner.
+- Retain an isolated process-group kill as a bounded fallback for unresponsive
+  host commands.
+
 ## 0.2.0
 
 - Add additive, Serde-compatible manifest-v1 and frame-v1 contracts.

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "coatcheck"
-version = "0.2.0"
+version = "0.2.1"
 edition = "2024"
 license = "MIT"
 rust-version = "1.95"

src/main.rs

@@ -478,34 +478,38 @@ impl HostGuard {
         })
     }
 
-    fn stop(&mut self) {
+    fn stop(&mut self) -> Option<ExitStatus> {
         if self.stopped {
-            return;
+            return self.child.try_wait().ok().flatten();
         }
         self.stopped = true;
 
+        if let Ok(Some(status)) = self.child.try_wait() {
+            return Some(status);
+        }
+
         #[cfg(unix)]
         {
-            let group = format!("-{}", self.child.id());
-            let _ = Command::new("kill")
-                .args(["-TERM", group.as_str()])
-                .stdout(Stdio::null())
-                .stderr(Stdio::null())
-                .status();
+            signal("-INT", &self.child.id().to_string());
         }
         #[cfg(not(unix))]
         {
             let _ = self.child.kill();
         }
 
-        for _ in 0..10 {
-            if matches!(self.child.try_wait(), Ok(Some(_))) {
-                return;
+        for _ in 0..100 {
+            if let Ok(Some(status)) = self.child.try_wait() {
+                return Some(status);
             }
             thread::sleep(Duration::from_millis(50));
         }
+
+        #[cfg(unix)]
+        {
+            signal("-KILL", &format!("-{}", self.child.id()));
+        }
         let _ = self.child.kill();
-        let _ = self.child.wait();
+        self.child.wait().ok()
     }
 }
 
@@ -514,3 +518,40 @@ impl Drop for HostGuard {
         self.stop();
     }
 }
+
+#[cfg(unix)]
+fn signal(signal: &str, target: &str) {
+    let _ = Command::new("kill")
+        .args([signal, "--", target])
+        .stdout(Stdio::null())
+        .stderr(Stdio::null())
+        .status();
+}
+
+#[cfg(all(test, unix))]
+mod tests {
+    use std::os::unix::process::ExitStatusExt;
+
+    use super::{Config, HostGuard};
+
+    #[test]
+    fn host_stop_interrupts_the_child_without_terminating_its_parent() {
+        let project = tempfile::tempdir().unwrap();
+        let config: Config = toml::from_str(
+            r#"
+version = 1
+name = "Host cleanup"
+
+[host]
+command = ["sleep", "30"]
+url = "http://127.0.0.1:3000"
+"#,
+        )
+        .unwrap();
+        let mut host = HostGuard::spawn(project.path(), &config).unwrap();
+
+        let status = host.stop().expect("host should be reaped");
+
+        assert_eq!(status.signal(), Some(2));
+    }
+}