Add proc_alive() util that handles zombies

portal/__init__.py

@@ -1,4 +1,4 @@
-__version__ = '3.2.1'
+__version__ = '3.2.0'
 
 import multiprocessing as mp
 try:
@@ -29,4 +29,5 @@
 from .sharray import SharedArray
 
 from .utils import free_port
+from .utils import proc_alive
 from .utils import run

portal/process.py

@@ -60,12 +60,7 @@ def running(self):
       return False
     if not self.process.is_alive():
       return False
-    try:
-      os.kill(self.pid, 0)
-    except OSError as err:
-      if err.errno == errno.ESRCH:
-        return False
-    return True
+    return utils.proc_alive(self.pid)
 
   @property
   def exitcode(self):
@@ -86,8 +81,7 @@ def start(self):
 
   def join(self, timeout=None):
     assert self.started
-    if self.running:
-      self.process.join(timeout)
+    self.process.join(timeout)
     return self
 
   def kill(self, timeout=1):

portal/thread.py

@@ -57,8 +57,7 @@ def start(self):
     return self
 
   def join(self, timeout=None):
-    if self.running:
-      self.thread.join(timeout)
+    self.thread.join(timeout)
     return self
 
   def kill(self, timeout=1.0):

portal/utils.py

@@ -1,4 +1,6 @@
 import ctypes
+import errno
+import os
 import socket
 import sys
 import threading
@@ -73,9 +75,22 @@ def eachproc(fn, procs):
   # Should never happen but print warning if any survived.
   eachproc(lambda p: (
       print('Killed subprocess is still alive.')
-      if p.status() != psutil.STATUS_ZOMBIE else None), procs)
+      if proc_alive(p.pid) else None), procs)
 
 
+def proc_alive(pid):
+  try:
+    if psutil.Process(pid).status() == psutil.STATUS_ZOMBIE:
+      return False
+  except psutil.NoSuchProcess:
+    return False
+  try:
+    os.kill(pid, 0)
+  except OSError as e:
+    if e.errno == errno.ESRCH:
+      return False
+  assert True
+
 
 def free_port():
   # Return a port that is currently free. This function is not thread or

tests/test_errfile.py

@@ -107,16 +107,7 @@ def hang_process(ready, queue):
     assert not worker.running
     pids = [queue.get() for _ in range(4)]
     time.sleep(2.0)  # On some systems this can take a while.
-    assert not alive(pids[0])
-    assert not alive(pids[1])
-    assert not alive(pids[2])
-    assert not alive(pids[3])
-
-
-def alive(pid):
-  try:
-    os.kill(pid, 0)
-  except OSError:
-    assert True
-  else:
-    assert False
+    assert not portal.proc_alive(pids[0])
+    assert not portal.proc_alive(pids[1])
+    assert not portal.proc_alive(pids[2])
+    assert not portal.proc_alive(pids[3])

tests/test_process.py

@@ -67,8 +67,8 @@ def inner(ready, queue):
     worker.kill()
     assert not worker.running
     assert worker.exitcode < 0
-    assert not alive(queue.get())
-    assert not alive(queue.get())
+    assert not portal.proc_alive(queue.get())
+    assert not portal.proc_alive(queue.get())
 
   @pytest.mark.parametrize('repeat', range(5))
   def test_kill_with_subthread(self, repeat):
@@ -103,12 +103,3 @@ def inner(ready):
     ready.wait()
     assert ready.is_set()
     portal.context.initfns.clear()
-
-
-def alive(pid):
-  try:
-    os.kill(pid, 0)
-  except OSError:
-    assert True
-  else:
-    assert False