pythongh-94215: Fix reference count issue in exception_unwind

Zero-cost exception handling did not take ``frame_setlineno()`` into
account. It can pop off stacks. This can lead to a crash.

Exception unwinding now re-calculates the stack pointer.

Co-authored-by: Irit Katriel <1055913+iritkatriel@users.noreply.github.com>

Lib/test/test_pdb.py

@@ -2085,7 +2085,7 @@ def test_issue42383(self):
             expected = '(Pdb) The correct file was executed'
             self.assertEqual(stdout.split('\n')[6].rstrip('\r'), expected)
 
-    @unittest.skip("test crashes, see gh-94215")
+    # @unittest.skip("test crashes, see gh-94215")
     def test_gh_94215_crash(self):
         script = """\
             def func():

Misc/NEWS.d/next/Core and Builtins/2022-07-07-17-07-00.gh-issue-94215.cXltGH.rst

@@ -0,0 +1,4 @@
+Fix a reference counting bug in exception unwinding code for zero-cost
+exception handling. The code did not take into account that
+``frame_setlineno()`` can pop off stacks, too.
+Patch by Irit Katriel and Christian Heimes.

Python/ceval.c

@@ -5795,6 +5795,11 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, _PyInterpreterFrame *frame, int
 
                 /* Pop remaining stack entries. */
                 PyObject **stackbase = _PyFrame_Stackbase(frame);
+                if (frame->stacktop != -1) {
+                    // frame_setlineno() may have popped off additional stacks in
+                    // frame_stack_pop(). Re-calculate stack pointer.
+                    stack_pointer = _PyFrame_GetStackPointer(frame);
+                }
                 while (stack_pointer > stackbase) {
                     PyObject *o = POP();
                     Py_XDECREF(o);