gh-104341: Adjust tstate_must_exit() to Respect Interpreter Finalization (gh-104437)

With the move to a per-interpreter GIL, this check slipped through the cracks.

1 files changed, 7 insertions, 5 deletions

diff --git a/Python/pystate.c b/Python/pystate.c
index 26debf1..25e655a 100644
--- a/Python/pystate.c
+++ b/Python/pystate.c
@@ -1436,11 +1436,13 @@ PyThreadState_Clear(PyThreadState *tstate)
 
     if (verbose && tstate->cframe->current_frame != NULL) {
         /* bpo-20526: After the main thread calls
-           _PyRuntimeState_SetFinalizing() in Py_FinalizeEx(), threads must
-           exit when trying to take the GIL. If a thread exit in the middle of
-           _PyEval_EvalFrameDefault(), tstate->frame is not reset to its
-           previous value. It is more likely with daemon threads, but it can
-           happen with regular threads if threading._shutdown() fails
+           _PyInterpreterState_SetFinalizing() in Py_FinalizeEx()
+           (or in Py_EndInterpreter() for subinterpreters),
+           threads must exit when trying to take the GIL.
+           If a thread exit in the middle of _PyEval_EvalFrameDefault(),
+           tstate->frame is not reset to its previous value.
+           It is more likely with daemon threads, but it can happen
+           with regular threads if threading._shutdown() fails
            (ex: interrupted by CTRL+C). */
         fprintf(stderr,
           "PyThreadState_Clear: warning: thread still has a frame\n");