gh-115999: Specialize `CALL_KW` in free-threaded builds (#127713)

* Enable specialization of CALL_KW

* Fix bug pushing frame in _PY_FRAME_KW

`_PY_FRAME_KW` pushes a pointer to the new frame onto the stack for
consumption by the next uop. When pushing the frame fails, we do not
want to push the result, `NULL`, to the stack because it is not
a valid stackref. This works in the default build because `PyStackRef_NULL`
 and `NULL` are the same value, so the `PyStackRef_XCLOSE()` in the error
handler ignores it. In the free-threaded build the values are not the same;
`PyStackRef_XCLOSE()` will attempt to decref a null pointer.

1 files changed, 10 insertions, 14 deletions

diff --git a/Python/generated_cases.c.h b/Python/generated_cases.c.h
index 33f32ab..fc0f555 100644
--- a/Python/generated_cases.c.h
+++ b/Python/generated_cases.c.h
@@ -1895,7 +1895,7 @@
                 callable = &stack_pointer[-3 - oparg];
                 uint16_t counter = read_u16(&this_instr[1].cache);
                 (void)counter;
-                #if ENABLE_SPECIALIZATION
+                #if ENABLE_SPECIALIZATION_FT
                 if (ADAPTIVE_COUNTER_TRIGGERS(counter)) {
                     next_instr = this_instr;
                     _PyFrame_SetStackPointer(frame, stack_pointer);
@@ -1905,7 +1905,7 @@
                 }
                 OPCODE_DEFERRED_INC(CALL_KW);
                 ADVANCE_ADAPTIVE_COUNTER(this_instr[1].counter);
-                #endif  /* ENABLE_SPECIALIZATION */
+                #endif  /* ENABLE_SPECIALIZATION_FT */
             }
             /* Skip 2 cache entries */
             // _MAYBE_EXPAND_METHOD_KW
@@ -2093,7 +2093,7 @@
                 int code_flags = ((PyCodeObject*)PyFunction_GET_CODE(callable_o))->co_flags;
                 PyObject *locals = code_flags & CO_OPTIMIZED ? NULL : Py_NewRef(PyFunction_GET_GLOBALS(callable_o));
                 _PyFrame_SetStackPointer(frame, stack_pointer);
-                new_frame = _PyEvalFramePushAndInit(
+                _PyInterpreterFrame *temp = _PyEvalFramePushAndInit(
                     tstate, callable[0], locals,
                     args, positional_args, kwnames_o, frame
                 );
@@ -2101,12 +2101,12 @@
                 PyStackRef_CLOSE(kwnames);
                 // The frame has stolen all the arguments from the stack,
                 // so there is no need to clean them up.
-                stack_pointer[-3 - oparg].bits = (uintptr_t)new_frame;
-                stack_pointer += -2 - oparg;
+                stack_pointer += -3 - oparg;
                 assert(WITHIN_STACK_BOUNDS());
-                if (new_frame == NULL) {
+                if (temp == NULL) {
                     goto error;
                 }
+                new_frame = temp;
             }
             // _SAVE_RETURN_OFFSET
             {
@@ -2123,8 +2123,6 @@
                 // Eventually this should be the only occurrence of this code.
                 assert(tstate->interp->eval_frame == NULL);
                 _PyInterpreterFrame *temp = new_frame;
-                stack_pointer += -1;
-                assert(WITHIN_STACK_BOUNDS());
                 _PyFrame_SetStackPointer(frame, stack_pointer);
                 assert(new_frame->previous == frame || new_frame->previous->previous == frame);
                 CALL_STAT_INC(inlined_py_calls);
@@ -2271,7 +2269,7 @@
                 int code_flags = ((PyCodeObject*)PyFunction_GET_CODE(callable_o))->co_flags;
                 PyObject *locals = code_flags & CO_OPTIMIZED ? NULL : Py_NewRef(PyFunction_GET_GLOBALS(callable_o));
                 _PyFrame_SetStackPointer(frame, stack_pointer);
-                new_frame = _PyEvalFramePushAndInit(
+                _PyInterpreterFrame *temp = _PyEvalFramePushAndInit(
                     tstate, callable[0], locals,
                     args, positional_args, kwnames_o, frame
                 );
@@ -2279,12 +2277,12 @@
                 PyStackRef_CLOSE(kwnames);
                 // The frame has stolen all the arguments from the stack,
                 // so there is no need to clean them up.
-                stack_pointer[-3 - oparg].bits = (uintptr_t)new_frame;
-                stack_pointer += -2 - oparg;
+                stack_pointer += -3 - oparg;
                 assert(WITHIN_STACK_BOUNDS());
-                if (new_frame == NULL) {
+                if (temp == NULL) {
                     goto error;
                 }
+                new_frame = temp;
             }
             // _SAVE_RETURN_OFFSET
             {
@@ -2301,8 +2299,6 @@
                 // Eventually this should be the only occurrence of this code.
                 assert(tstate->interp->eval_frame == NULL);
                 _PyInterpreterFrame *temp = new_frame;
-                stack_pointer += -1;
-                assert(WITHIN_STACK_BOUNDS());
                 _PyFrame_SetStackPointer(frame, stack_pointer);
                 assert(new_frame->previous == frame || new_frame->previous->previous == frame);
                 CALL_STAT_INC(inlined_py_calls);