Backport of r65032 from trunk

Fixes Issue #874900: after an os.fork() call the threading module state is cleaned
up in the child process to prevent deadlock and report proper thread counts
if the new process uses the threading module.

Lib/test/test_threading.py

@@ -220,8 +220,85 @@ class ThreadTests(unittest.TestCase):
             sys.setcheckinterval(old_interval)
 
 
+class ThreadJoinOnShutdown(unittest.TestCase):
+
+    def _run_and_join(self, script):
+        script = """if 1:
+            import sys, os, time, threading
+
+            # a thread, which waits for the main program to terminate
+            def joiningfunc(mainthread):
+                mainthread.join()
+                print 'end of thread'
+        \n""" + script
+
+        import subprocess
+        p = subprocess.Popen([sys.executable, "-c", script], stdout=subprocess.PIPE)
+        rc = p.wait()
+        self.assertEqual(p.stdout.read(), "end of main\nend of thread\n")
+        self.failIf(rc == 2, "interpreter was blocked")
+        self.failUnless(rc == 0, "Unexpected error")
+
+    def test_1_join_on_shutdown(self):
+        # The usual case: on exit, wait for a non-daemon thread
+        script = """if 1:
+            import os
+            t = threading.Thread(target=joiningfunc,
+                                 args=(threading.currentThread(),))
+            t.start()
+            time.sleep(0.1)
+            print 'end of main'
+            """
+        self._run_and_join(script)
+
+
+    def test_2_join_in_forked_process(self):
+        # Like the test above, but from a forked interpreter
+        import os
+        if not hasattr(os, 'fork'):
+            return
+        script = """if 1:
+            childpid = os.fork()
+            if childpid != 0:
+                os.waitpid(childpid, 0)
+                sys.exit(0)
+
+            t = threading.Thread(target=joiningfunc,
+                                 args=(threading.currentThread(),))
+            t.start()
+            print 'end of main'
+            """
+        self._run_and_join(script)
+
+    def test_3_join_in_forked_from_thread(self):
+        # Like the test above, but fork() was called from a worker thread
+        # In the forked process, the main Thread object must be marked as stopped.
+        import os
+        if not hasattr(os, 'fork'):
+            return
+        script = """if 1:
+            main_thread = threading.currentThread()
+            def worker():
+                childpid = os.fork()
+                if childpid != 0:
+                    os.waitpid(childpid, 0)
+                    sys.exit(0)
+
+                t = threading.Thread(target=joiningfunc,
+                                     args=(main_thread,))
+                print 'end of main'
+                t.start()
+                t.join() # Should not block: main_thread is already stopped
+
+            w = threading.Thread(target=worker)
+            w.start()
+            """
+        self._run_and_join(script)
+
+
 def test_main():
-    test.test_support.run_unittest(ThreadTests)
+    test.test_support.run_unittest(ThreadTests,
+                                   ThreadJoinOnShutdown)
 
 if __name__ == "__main__":
     test_main()

Lib/threading.py

@@ -766,6 +766,40 @@ except ImportError:
     from _threading_local import local
 
 
+def _after_fork():
+    # This function is called by Python/ceval.c:PyEval_ReInitThreads which
+    # is called from PyOS_AfterFork.  Here we cleanup threading module state
+    # that should not exist after a fork.
+
+    # Reset _active_limbo_lock, in case we forked while the lock was held
+    # by another (non-forked) thread.  http://bugs.python.org/issue874900
+    global _active_limbo_lock
+    _active_limbo_lock = _allocate_lock()
+
+    # fork() only copied the current thread; clear references to others.
+    new_active = {}
+    current = currentThread()
+    _active_limbo_lock.acquire()
+    try:
+        for ident, thread in _active.iteritems():
+            if thread is current:
+                # There is only one active thread.
+                new_active[ident] = thread
+            else:
+                # All the others are already stopped.
+                # We don't call _Thread__stop() because it tries to acquire
+                # thread._Thread__block which could also have been held while
+                # we forked.
+                thread._Thread__stopped = True
+
+        _limbo.clear()
+        _active.clear()
+        _active.update(new_active)
+        assert len(_active) == 1
+    finally:
+        _active_limbo_lock.release()
+
+
 # Self-test code
 
 def _test():

Misc/NEWS

@@ -144,6 +144,10 @@ Library
 - Issue #2113: Fix error in subprocess.Popen if the select system call is
   interrupted by a signal.
 
+- Issue #874900: after an os.fork() call the threading module state is cleaned
+  up in the child process to prevent deadlock and report proper thread counts
+  if the new process uses the threading module.
+
 
 Extension Modules
 -----------------

Python/ceval.c

@@ -272,6 +272,9 @@ PyEval_ReleaseThread(PyThreadState *tstate)
 void
 PyEval_ReInitThreads(void)
 {
+	PyObject *threading, *result;
+	PyThreadState *tstate;
+
 	if (!interpreter_lock)
 		return;
 	/*XXX Can't use PyThread_free_lock here because it does too
@@ -281,6 +284,23 @@ PyEval_ReInitThreads(void)
 	interpreter_lock = PyThread_allocate_lock();
 	PyThread_acquire_lock(interpreter_lock, 1);
 	main_thread = PyThread_get_thread_ident();
+
+	/* Update the threading module with the new state.
+	 */
+	tstate = PyThreadState_GET();
+	threading = PyMapping_GetItemString(tstate->interp->modules,
+					    "threading");
+	if (threading == NULL) {
+		/* threading not imported */
+		PyErr_Clear();
+		return;
+	}
+	result = PyObject_CallMethod(threading, "_after_fork", NULL);
+	if (result == NULL)
+		PyErr_WriteUnraisable(threading);
+	else
+		Py_DECREF(result);
+	Py_DECREF(threading);
 }
 #endif