[Python-checkins] bpo-37788: Fix reference leak when Thread is never joined (GH-26103)

[miss-islington]

https://github.com/python/cpython/commit/c10c2ec7a0e06975e8010c56c9c3270f8ea322ec
commit: c10c2ec7a0e06975e8010c56c9c3270f8ea322ec
branch: main
author: Antoine Pitrou <antoine at python.org>
committer: miss-islington <31488909+miss-islington at users.noreply.github.com>
date: 2021-05-14T12:37:20-07:00
summary:

bpo-37788: Fix reference leak when Thread is never joined (GH-26103)



When a Thread is not joined after it has stopped, its lock may remain in the _shutdown_locks set until interpreter shutdown.  If many threads are created this way, the _shutdown_locks set could therefore grow endlessly.  To avoid such a situation, purge expired locks each time a new one is added or removed.

files:
A Misc/NEWS.d/next/Library/2021-05-13-19-07-28.bpo-37788.adeFcf.rst
M Lib/test/test_threading.py
M Lib/threading.py

diff --git a/Lib/test/test_threading.py b/Lib/test/test_threading.py
index 546773e3329af..08c0ccd9a79b2 100644
--- a/Lib/test/test_threading.py
+++ b/Lib/test/test_threading.py
@@ -907,6 +907,13 @@ def __call__(self):
         thread.join()
         self.assertTrue(target.ran)
 
+    def test_leak_without_join(self):
+        # bpo-37788: Test that a thread which is not joined explicitly
+        # does not leak. Test written for reference leak checks.
+        def noop(): pass
+        with threading_helper.wait_threads_exit():
+            threading.Thread(target=noop).start()
+            # Thread.join() is not called
 
 
 class ThreadJoinOnShutdown(BaseTestCase):
diff --git a/Lib/threading.py b/Lib/threading.py
index fb70abd17aff8..6c3d49c2d5267 100644
--- a/Lib/threading.py
+++ b/Lib/threading.py
@@ -780,12 +780,27 @@ def _newname(name_template):
 _active = {}    # maps thread id to Thread object
 _limbo = {}
 _dangling = WeakSet()
+
 # Set of Thread._tstate_lock locks of non-daemon threads used by _shutdown()
 # to wait until all Python thread states get deleted:
 # see Thread._set_tstate_lock().
 _shutdown_locks_lock = _allocate_lock()
 _shutdown_locks = set()
 
+def _maintain_shutdown_locks():
+    """
+    Drop any shutdown locks that don't correspond to running threads anymore.
+
+    Calling this from time to time avoids an ever-growing _shutdown_locks
+    set when Thread objects are not joined explicitly. See bpo-37788.
+
+    This must be called with _shutdown_locks_lock acquired.
+    """
+    # If a lock was released, the corresponding thread has exited
+    to_remove = [lock for lock in _shutdown_locks if not lock.locked()]
+    _shutdown_locks.difference_update(to_remove)
+
+
 # Main class for threads
 
 class Thread:
@@ -968,6 +983,7 @@ def _set_tstate_lock(self):
 
         if not self.daemon:
             with _shutdown_locks_lock:
+                _maintain_shutdown_locks()
                 _shutdown_locks.add(self._tstate_lock)
 
     def _bootstrap_inner(self):
@@ -1023,7 +1039,8 @@ def _stop(self):
         self._tstate_lock = None
         if not self.daemon:
             with _shutdown_locks_lock:
-                _shutdown_locks.discard(lock)
+                # Remove our lock and other released locks from _shutdown_locks
+                _maintain_shutdown_locks()
 
     def _delete(self):
         "Remove current thread from the dict of currently running threads."
diff --git a/Misc/NEWS.d/next/Library/2021-05-13-19-07-28.bpo-37788.adeFcf.rst b/Misc/NEWS.d/next/Library/2021-05-13-19-07-28.bpo-37788.adeFcf.rst
new file mode 100644
index 0000000000000..0c33923e99245
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2021-05-13-19-07-28.bpo-37788.adeFcf.rst
@@ -0,0 +1 @@
+Fix a reference leak when a Thread object is never joined.