Split CALL_BUILTIN_FAST decrefs into POP_NOS

Author: Fidget-Spinner

Include/internal/pycore_ceval.h

@@ -433,7 +433,7 @@ _Py_VectorCallInstrumentation_StackRefSteal(
     PyThreadState* tstate);
 
 PyAPI_FUNC(PyObject *)
-_Py_BuiltinCallFast_StackRefSteal(
+_Py_BuiltinCallFast_StackRef(
     _PyStackRef callable,
     _PyStackRef *arguments,
     int total_args);

Include/internal/pycore_magic_number.h

@@ -294,6 +294,7 @@ Known values:
     Python 3.15a4 3661 (Lazy imports IMPORT_NAME opcode changes)
     Python 3.15a8 3662 (Add counter to RESUME)
     Python 3.15a8 3663 (Merge GET_ITER and GET_YIELD_FROM_ITER. Modify SEND to make it a bit more like FOR_ITER)
+    Python 3.15b1 3664 (All code objects now have 1 extra stack slot space)
 
 
     Python 3.16 will start with 3700
@@ -307,7 +308,7 @@ PC/launcher.c must also be updated.
 
 */
 
-#define PYC_MAGIC_NUMBER 3663
+#define PYC_MAGIC_NUMBER 3664
 /* This is equivalent to converting PYC_MAGIC_NUMBER to 2 bytes
    (little-endian) and then appending b'\r\n'. */
 #define PYC_MAGIC_NUMBER_TOKEN \

Include/internal/pycore_opcode_metadata.h

@@ -845,6 +845,18 @@ _Py_TryXGetStackRef(PyObject **src, _PyStackRef *out)
         }                                                               \
     } while (0)
 
+static inline void
+_PyStackRef_CloseStack(_PyStackRef *arguments, int total_args)
+{
+    // arguments is a pointer into the GC visible stack,
+    // so we must NULL out values as we clear them.
+    for (int i = total_args-1; i >= 0; i--) {
+        _PyStackRef tmp = arguments[i];
+        arguments[i] = PyStackRef_NULL;
+        PyStackRef_CLOSE(tmp);
+    }
+}
+
 #ifdef __cplusplus
 }
 #endif