Reap child processes after each of the ssh tests in case of timeouts

tests/conftest.py

@@ -5,7 +5,6 @@
 import logging
 import os
 import pathlib
-import pprint
 import re
 import shutil
 import stat
@@ -16,7 +15,6 @@ from unittest import TestCase  # pylint: disable=blacklisted-module
 import _pytest.logging
 import _pytest.skipping
 import more_itertools
-import psutil
 import pytest
 
 import salt
@@ -1299,7 +1297,6 @@ def salt_call_cli(salt_minion_factory):
 
 @pytest.fixture(scope="session", autouse=True)
 def bridge_pytest_and_runtests(
-    reap_stray_processes,
     salt_factories,
     salt_syndic_master_factory,
     salt_syndic_factory,
@@ -1336,6 +1333,8 @@ def bridge_pytest_and_runtests(
         salt_syndic_factory.config["conf_file"]
     )
     RUNTIME_VARS.TMP_SSH_CONF_DIR = str(sshd_config_dir)
+    with reap_stray_processes():
+        yield
 
 
 @pytest.fixture(scope="session")
@@ -1635,46 +1634,6 @@ def from_filenames_collection_modifyitems(config, items):
 
 
 # ----- Custom Fixtures --------------------------------------------------------------------------------------------->
-@pytest.fixture(scope="session")
-def reap_stray_processes():
-    # Run tests
-    yield
-
-    children = psutil.Process(os.getpid()).children(recursive=True)
-    if not children:
-        log.info("No astray processes found")
-        return
-
-    def on_terminate(proc):
-        log.debug("Process %s terminated with exit code %s", proc, proc.returncode)
-
-    if children:
-        # Reverse the order, sublings first, parents after
-        children.reverse()
-        log.warning(
-            "Test suite left %d astray processes running. Killing those processes:\n%s",
-            len(children),
-            pprint.pformat(children),
-        )
-
-        _, alive = psutil.wait_procs(children, timeout=3, callback=on_terminate)
-        for child in alive:
-            try:
-                child.kill()
-            except psutil.NoSuchProcess:
-                continue
-
-        _, alive = psutil.wait_procs(alive, timeout=3, callback=on_terminate)
-        if alive:
-            # Give up
-            for child in alive:
-                log.warning(
-                    "Process %s survived SIGKILL, giving up:\n%s",
-                    child,
-                    pprint.pformat(child.as_dict()),
-                )
-
-
 @pytest.fixture(scope="session")
 def sminion():
     return create_sminion()

tests/pytests/integration/ssh/conftest.py

@@ -1,5 +1,7 @@
 import pytest
 
+from tests.support.pytest.helpers import reap_stray_processes
+
 
 @pytest.fixture(scope="package", autouse=True)
 def _auto_skip_on_fedora_39(grains):
@@ -11,3 +13,12 @@ def _auto_skip_on_fedora_39(grains):
             # has been deprecated since Python 3.7, so, the logic goes into trying to import
             # backports.ssl-match-hostname which is not installed on the system.
         )
+
+
+@pytest.fixture(autouse=True)
+def _reap_stray_processes(grains):
+    # when tests timeout, we migth leave child processes behind
+    # nuke them
+    with reap_stray_processes():
+        # Run test
+        yield

tests/support/pytest/helpers.py

@@ -8,6 +8,7 @@
 import logging
 import os
 import pathlib
+import pprint
 import shutil
 import subprocess
 import tempfile
@@ -18,6 +19,7 @@ import warnings
 from contextlib import contextmanager
 
 import attr
+import psutil
 import pytest
 from saltfactories.utils import random_string
 from saltfactories.utils.tempfiles import temp_file
@@ -818,6 +820,61 @@ def change_cwd(path):
         os.chdir(old_cwd)
 
 
+@contextmanager
+def reap_stray_processes(pid: int = os.getpid()):
+
+    try:
+        pre_children = psutil.Process(pid).children(recursive=True)
+        # Do stuff
+        yield
+    finally:
+        post_children = psutil.Process(pid).children(recursive=True)
+
+    children = []
+    for process in post_children:
+        if process in pre_children:
+            # Process existed before entering the context
+            continue
+        if not psutil.pid_exists(process.pid):
+            # Process just died
+            continue
+        # This process is alive and was not running before entering the context
+        children.append(process)
+
+    if not children:
+        log.info("No astray processes found")
+        return
+
+    def on_terminate(proc):
+        log.debug("Process %s terminated with exit code %s", proc, proc.returncode)
+
+    if children:
+        # Reverse the order, sublings first, parents after
+        children.reverse()
+        log.warning(
+            "Test suite left %d astray processes running. Killing those processes:\n%s",
+            len(children),
+            pprint.pformat(children),
+        )
+
+        _, alive = psutil.wait_procs(children, timeout=3, callback=on_terminate)
+        for child in alive:
+            try:
+                child.kill()
+            except psutil.NoSuchProcess:
+                continue
+
+        _, alive = psutil.wait_procs(alive, timeout=3, callback=on_terminate)
+        if alive:
+            # Give up
+            for child in alive:
+                log.warning(
+                    "Process %s survived SIGKILL, giving up:\n%s",
+                    child,
+                    pprint.pformat(child.as_dict()),
+                )
+
+
 # Only allow star importing the functions defined in this module
 __all__ = [
     name