Mark threads as "done" *before* GC is finalized.

Author: ericsnowcurrently

Python/pylifecycle.c

@@ -1133,16 +1133,16 @@ Py_FinalizeEx(void)
         return status;
     }
 
+    /* Get current thread state and interpreter pointer */
+    PyThreadState *tstate = _PyThreadState_GET();
+    PyInterpreterState *interp = tstate->interp;
+
     // Wrap up existing "threading"-module-created, non-daemon threads.
     wait_for_thread_shutdown();
 
     // Make any remaining pending calls.
     _Py_FinishPendingCalls();
 
-    /* Get current thread state and interpreter pointer */
-    PyThreadState *tstate = _PyThreadState_GET();
-    PyInterpreterState *interp = tstate->interp;
-
     /* The interpreter is still entirely intact at this point, and the
      * exit funcs may be relying on that.  In particular, if some thread
      * or exit func is still waiting to do an import, the import machinery
@@ -2194,6 +2194,9 @@ wait_for_thread_shutdown(void)
         Py_DECREF(result);
     }
     Py_DECREF(threading);
+
+    // All threading module threads are marked as "done" later
+    // in PyThreadState_Clear().
 }
 
 #define NEXITFUNCS 32

Python/pystate.c

@@ -781,6 +781,13 @@ PyThreadState_Clear(PyThreadState *tstate)
     Py_CLEAR(tstate->async_gen_finalizer);
 
     Py_CLEAR(tstate->context);
+
+    if (tstate->on_delete != NULL) {
+        // This will unblock any joining threads.
+        tstate->on_delete(tstate->on_delete_data);
+        tstate->on_delete = NULL;
+        tstate->on_delete_data = NULL;
+    }
 }
 
 
@@ -804,6 +811,8 @@ tstate_delete_common(_PyRuntimeState *runtime, PyThreadState *tstate)
         tstate->next->prev = tstate->prev;
     HEAD_UNLOCK(runtime);
     if (tstate->on_delete != NULL) {
+        // This will unblock any joining threads.
+        // We also do this in PyThreadState_Clear(), but do it here to be sure.
         tstate->on_delete(tstate->on_delete_data);
     }
     PyMem_RawFree(tstate);