bpo-38707: Fix for multiprocessing.Process MainThread.native_id (GH-17088)

This PR implements a fix for `multiprocessing.Process` objects; the error occurs when Processes are created using either `fork` or `forkserver` as the `start_method`.

In these instances, the `MainThread` of the newly created `Process` object retains all attributes from its parent's `MainThread` object, including the `native_id` attribute. The resulting behavior is such that the new process' `MainThread` captures an incorrect/outdated `native_id` (the parent's instead of its own). 

This change forces the Process object to update its `native_id` attribute during the bootstrap process.

cc @vstinner





https://bugs.python.org/issue38707



Automerge-Triggered-By: @pitrou

3 files changed, 25 insertions, 0 deletions

diff --git a/Lib/multiprocessing/process.py b/Lib/multiprocessing/process.py
index c62c826..be13c07 100644
--- a/Lib/multiprocessing/process.py
+++ b/Lib/multiprocessing/process.py
@@ -301,6 +301,8 @@ class BaseProcess(object):
             _current_process = self
             _parent_process = _ParentProcess(
                 self._parent_name, self._parent_pid, parent_sentinel)
+            if threading._HAVE_THREAD_NATIVE_ID:
+                threading.main_thread()._set_native_id()
             try:
                 util._finalizer_registry.clear()
                 util._run_after_forkers()
diff --git a/Lib/test/_test_multiprocessing.py b/Lib/test/_test_multiprocessing.py
index f7bebc6..611291c 100644
--- a/Lib/test/_test_multiprocessing.py
+++ b/Lib/test/_test_multiprocessing.py
@@ -361,6 +361,28 @@ class _TestProcess(BaseTestCase):
         self.assertNotIn(p, self.active_children())
         close_queue(q)
 
+    @unittest.skipUnless(threading._HAVE_THREAD_NATIVE_ID, "needs native_id")
+    def test_process_mainthread_native_id(self):
+        if self.TYPE == 'threads':
+            self.skipTest('test not appropriate for {}'.format(self.TYPE))
+
+        current_mainthread_native_id = threading.main_thread().native_id
+
+        q = self.Queue(1)
+        p = self.Process(target=self._test_process_mainthread_native_id, args=(q,))
+        p.start()
+
+        child_mainthread_native_id = q.get()
+        p.join()
+        close_queue(q)
+
+        self.assertNotEqual(current_mainthread_native_id, child_mainthread_native_id)
+
+    @classmethod
+    def _test_process_mainthread_native_id(cls, q):
+        mainthread_native_id = threading.main_thread().native_id
+        q.put(mainthread_native_id)
+
     @classmethod
     def _sleep_some(cls):
         time.sleep(100)
diff --git a/Misc/NEWS.d/next/Core and Builtins/2019-11-08-00-36-10.bpo-38707.SZL036.rst b/Misc/NEWS.d/next/Core and Builtins/2019-11-08-00-36-10.bpo-38707.SZL036.rst
new file mode 100644
index 0000000..4ef9ed8
--- /dev/null
+++ b/Misc/NEWS.d/next/Core and Builtins/2019-11-08-00-36-10.bpo-38707.SZL036.rst
@@ -0,0 +1 @@
+``MainThread.native_id`` is now correctly reset in child processes spawned using :class:`multiprocessing.Process`, instead of retaining the parent's value.