Treat zombie test processes as exited

archivebox/tests/conftest.py

@@ -577,12 +577,11 @@ def wait_for_process(predicate: Callable[[psutil.Process, str], bool], *, timeou
 
 def pid_is_alive(pid: int) -> bool:
     try:
-        os.kill(pid, 0)
-    except ProcessLookupError:
+        return psutil.Process(pid).status() != psutil.STATUS_ZOMBIE
+    except (psutil.NoSuchProcess, psutil.ZombieProcess):
         return False
-    except PermissionError:
+    except psutil.AccessDenied:
         return True
-    return True
 
 
 def wait_for_pid_to_disappear(pid: int, *, timeout: float = 20.0) -> None:

archivebox/tests/test_takeover_util.py

@@ -10,6 +10,7 @@ import time
 from pathlib import Path
 
 import pytest
+import psutil
 
 from archivebox.core.models import ArchiveResult, Snapshot
 from archivebox.crawls.models import Crawl
@@ -38,6 +39,19 @@ from archivebox.tests.test_orm_helpers import use_archivebox_db
 pytestmark = pytest.mark.django_db(transaction=True)
 
 
+def test_pid_is_alive_treats_unreaped_zombie_as_exited():
+    proc = subprocess.Popen([sys.executable, "-c", "pass"])
+    try:
+        deadline = time.time() + 5
+        while time.time() < deadline and psutil.Process(proc.pid).status() != psutil.STATUS_ZOMBIE:
+            time.sleep(0.01)
+
+        assert psutil.Process(proc.pid).status() == psutil.STATUS_ZOMBIE
+        assert not pid_is_alive(proc.pid)
+    finally:
+        proc.wait(timeout=5)
+
+
 def _require_sonic_binary() -> None:
     assert shutil.which("sonic") is not None, "sonic server binary is required for Sonic worker takeover tests"