Android launcher: drop process-group reaping; orphan recovery stops only the app's own container

Matching the APK path in command lines could hit an unrelated process, and
the pgid file had a registration race. The launcher keeps the flock (not
inherited by Lepton) and, holding it, stops only lepton-steamlaunch-<instance>,
whose name is this app's alone. A Lepton host process may linger briefly.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

docs/vr-apks.md

@@ -178,8 +178,9 @@ The launcher supervises Lepton and handles TERM/INT/HUP and normal exit by
 stopping its own container and child process group. A lock refuses duplicate launches;
 a container still running while the lock is free was orphaned by a killed
 launcher and is stopped before the new launch. Lepton doesn't inherit the
-lock; the launcher records Lepton's process group and a later launch ends a
-recorded group still running this APK. Removing an app or title still deletes its files when Steam isn't
+lock. Orphan recovery only stops the app's own, deterministically named
+container; a Lepton host process whose launcher was killed before it created
+the container may linger briefly. Removing an app or title still deletes its files when Steam isn't
 running; tidying Steam's collections and artwork is best effort. Steam Stop uses `TerminateApp` with the exact
 64-bit game ID string. Frame Control's Stop additionally has a direct-container
 fallback. The stable instance ID and compatdata paths remain unchanged.

frame/android/lepton-app.sh

@@ -30,22 +30,10 @@ fi
 exec 9>"$DIR/launch.lock"
 flock -n 9 || { echo "Android app is already running" >&2; exit 1; }
 CONTAINER="lepton-steamlaunch-$SteamAppId"
-# Lepton doesn't hold the lock, so a launcher SIGKILLed before Lepton made its
-# container leaves a Lepton that nothing tracks: end that process group first.
-# Only a group still running this app.apk, never an unrelated reused id.
-PGID_FILE="$DIR/launch.pgid"
-if [[ -f "$PGID_FILE" ]]; then
-  old="$(cat "$PGID_FILE")"
-  if [[ "$old" =~ ^[0-9]+$ ]] && ps -A -o pgid=,args= | awk -v g="$old" '$1 == g' | grep -qF -- "$DIR/app.apk"; then
-    echo "Stopping the previous launch (process group $old)" >&2
-    kill -TERM -- "-$old" 2>/dev/null || true
-    for _ in 1 2 3 4 5 6 7 8 9 10; do kill -0 -- "-$old" 2>/dev/null || break; sleep 0.5; done
-    kill -KILL -- "-$old" 2>/dev/null || true
-  fi
-  rm -f "$PGID_FILE"
-fi
 # Holding the lock means no launcher owns a running container: it was orphaned
-# (this script SIGKILLed), so stop it rather than refuse every later Play.
+# (this script SIGKILLed), so stop it rather than refuse every later Play. The
+# name is this app's alone. A Lepton host process whose launcher was killed
+# before it made the container may linger briefly; nothing else is killed.
 if [[ "$(podman inspect --format '{{.State.Running}}' "$CONTAINER" 2>/dev/null || true)" == true ]]; then
   echo "Stopping orphaned $CONTAINER" >&2
   podman stop -t 5 "$CONTAINER" >/dev/null 2>&1 || true
@@ -74,7 +62,6 @@ cleanup() {
     kill -KILL "$child" 2>/dev/null || true
     wait "$child" 2>/dev/null || true
   fi
-  rm -f "$PGID_FILE"
 }
 trap cleanup EXIT
 trap 'exit 143' TERM
@@ -83,7 +70,6 @@ trap 'exit 129' HUP
 # 9>&-: the lock is this launcher's alone; Lepton's tree mustn't keep it held.
 setsid --wait "$LEPTON" waitforexitandrun -- "$DIR/app.apk" 9>&- &
 child=$!
-echo "$child" > "$PGID_FILE"
 rc=0
 wait "$child" || rc=$?
 child=""

tests/test_frame_android_library.py

@@ -24,7 +24,7 @@ spec.loader.exec_module(shortcuts)
 
 @unittest.skipIf(os.name == 'nt', 'POSIX launcher')
 class LauncherTests(unittest.TestCase):
-    def exercise(self, terminate, sig=signal.SIGTERM, blocked=None, orphan=False, stale=None):
+    def exercise(self, terminate, sig=signal.SIGTERM, blocked=None, orphan=False):
         with tempfile.TemporaryDirectory() as tmp:
             d = Path(tmp)
             app = d / 'Applications/Android/org.test.app'
@@ -60,13 +60,6 @@ class LauncherTests(unittest.TestCase):
             saved = d / '.local/share/Steam/steamapps/compatdata/2800000001/internal/save'
             saved.parent.mkdir(parents=True)
             saved.write_text('saved game')
-            previous = None
-            if stale:
-                # A Lepton left by a SIGKILLed launcher (its own session), or an unrelated reused id.
-                args = [sys.executable, '-c', 'import time; time.sleep(30)']
-                previous = subprocess.Popen(args + ([str(app / 'app.apk')] if stale == 'lepton' else ['other']),
-                                            start_new_session=True)
-                (app / 'launch.pgid').write_text(str(previous.pid))
             proc = subprocess.Popen(['bash', str(app / 'launch.sh')], env=env, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
             try:
                 if blocked:
@@ -80,10 +73,6 @@ class LauncherTests(unittest.TestCase):
                 while not (d / 'started').exists() and proc.poll() is None and time.monotonic() < deadline:
                     time.sleep(.02)
                 self.assertTrue((d / 'started').exists(), 'launcher did not start Lepton')
-                if previous:
-                    self.assertEqual(previous.poll() is not None, stale == 'lepton')
-                    self.assertEqual((app / 'launch.pgid').read_text().strip(),
-                                     (d / 'started').read_text().strip())  # the new group is recorded
                 self.assertFalse((d / 'inherited-lock').exists(), 'Lepton inherited the launch lock')
                 if terminate:
                     self.assertIsNone(proc.poll(), 'Steam-tracked wrapper exited during the session')
@@ -95,12 +84,7 @@ class LauncherTests(unittest.TestCase):
                 self.assertEqual(proc.returncode, 128 + sig if terminate else 23)
                 self.assertEqual(saved.read_text(), 'saved game')
                 self.assertTrue((app / 'app.apk').exists())
-                self.assertFalse((app / 'launch.pgid').exists())
             finally:
-                if previous and previous.poll() is None:
-                    previous.kill()
-                if previous:
-                    previous.wait()
                 if proc.poll() is None:
                     proc.kill()
                     proc.communicate()
@@ -121,12 +105,6 @@ class LauncherTests(unittest.TestCase):
     def test_duplicate_launch_leaves_existing_session_alone(self):
         self.exercise(False, blocked='LOCKED')
 
-    def test_previous_launch_left_by_sigkill_is_reaped_first(self):
-        self.exercise(True, stale='lepton')
-
-    def test_reused_process_group_id_is_left_alone(self):
-        self.exercise(True, stale='other')
-
     def test_orphaned_container_is_stopped_and_play_proceeds(self):
         # Container running but the lock free: its launcher was SIGKILLed.
         self.exercise(False, orphan=True)