Merge branch 'main' into issue_125038_get_iter_rework

.github/workflows/jit.yml

@@ -52,7 +52,7 @@ jobs:
           - x86_64-pc-windows-msvc/msvc
           - aarch64-pc-windows-msvc/msvc
           - x86_64-apple-darwin/clang
-          # - aarch64-apple-darwin/clang
+          - aarch64-apple-darwin/clang
           - x86_64-unknown-linux-gnu/gcc
           - x86_64-unknown-linux-gnu/clang
           - aarch64-unknown-linux-gnu/gcc
@@ -79,11 +79,10 @@ jobs:
             architecture: x86_64
             runner: macos-13
             compiler: clang
-          # GH-126464: A recent change to either GHA or LLVM broke this job:
-          # - target: aarch64-apple-darwin/clang
-          #   architecture: aarch64
-          #   runner: macos-14
-          #   compiler: clang
+          - target: aarch64-apple-darwin/clang
+            architecture: aarch64
+            runner: macos-14
+            compiler: clang
           - target: x86_64-unknown-linux-gnu/gcc
             architecture: x86_64
             runner: ubuntu-22.04
@@ -132,8 +131,8 @@ jobs:
           brew update
           find /usr/local/bin -lname '*/Library/Frameworks/Python.framework/*' -delete
           brew install llvm@${{ matrix.llvm }}
-          SDKROOT="$(xcrun --show-sdk-path)" \
-            ./configure --enable-experimental-jit ${{ matrix.debug && '--with-pydebug' || '--enable-optimizations --with-lto' }}
+          export SDKROOT="$(xcrun --show-sdk-path)"
+          ./configure --enable-experimental-jit ${{ matrix.debug && '--with-pydebug' || '--enable-optimizations --with-lto' }}
           make all --jobs 4
           ./python.exe -m test --multiprocess 0 --timeout 4500 --verbose2 --verbose3
 

Doc/c-api/typeobj.rst

@@ -2230,7 +2230,7 @@ This is done by filling a :c:type:`PyType_Spec` structure and calling
 .. _number-structs:
 
 Number Object Structures
-========================
+------------------------
 
 .. sectionauthor:: Amaury Forgeot d'Arc
 
@@ -2344,7 +2344,7 @@ Number Object Structures
 .. _mapping-structs:
 
 Mapping Object Structures
-=========================
+-------------------------
 
 .. sectionauthor:: Amaury Forgeot d'Arc
 
@@ -2381,7 +2381,7 @@ Mapping Object Structures
 .. _sequence-structs:
 
 Sequence Object Structures
-==========================
+--------------------------
 
 .. sectionauthor:: Amaury Forgeot d'Arc
 
@@ -2461,7 +2461,7 @@ Sequence Object Structures
 .. _buffer-structs:
 
 Buffer Object Structures
-========================
+------------------------
 
 .. sectionauthor:: Greg J. Stein <[email protected]>
 .. sectionauthor:: Benjamin Peterson
@@ -2556,7 +2556,7 @@ Buffer Object Structures
 
 
 Async Object Structures
-=======================
+-----------------------
 
 .. sectionauthor:: Yury Selivanov <[email protected]>
 
@@ -2624,7 +2624,7 @@ Async Object Structures
 .. _slot-typedefs:
 
 Slot Type typedefs
-==================
+------------------
 
 .. c:type:: PyObject *(*allocfunc)(PyTypeObject *cls, Py_ssize_t nitems)
 
@@ -2733,7 +2733,7 @@ Slot Type typedefs
 .. _typedef-examples:
 
 Examples
-========
+--------
 
 The following are simple examples of Python type definitions.  They
 include common usage you may encounter.  Some demonstrate tricky corner

Doc/howto/logging-cookbook.rst

@@ -1267,11 +1267,8 @@ to adapt in your own applications.
 
 You could also write your own handler which uses the :class:`~multiprocessing.Lock`
 class from the :mod:`multiprocessing` module to serialize access to the
-file from your processes. The existing :class:`FileHandler` and subclasses do
-not make use of :mod:`multiprocessing` at present, though they may do so in the
-future. Note that at present, the :mod:`multiprocessing` module does not provide
-working lock functionality on all platforms (see
-https://bugs.python.org/issue3770).
+file from your processes. The stdlib :class:`FileHandler` and subclasses do
+not make use of :mod:`multiprocessing`.
 
 .. currentmodule:: logging.handlers
 

Include/internal/pycore_crossinterp.h

@@ -38,28 +38,28 @@ extern int _Py_CallInInterpreterAndRawFree(
 /* cross-interpreter data */
 /**************************/
 
-typedef struct _xid _PyCrossInterpreterData;
-typedef PyObject *(*xid_newobjectfunc)(_PyCrossInterpreterData *);
+typedef struct _xid _PyXIData_t;
+typedef PyObject *(*xid_newobjectfunc)(_PyXIData_t *);
 typedef void (*xid_freefunc)(void *);
 
-// _PyCrossInterpreterData is similar to Py_buffer as an effectively
+// _PyXIData_t is similar to Py_buffer as an effectively
 // opaque struct that holds data outside the object machinery.  This
 // is necessary to pass safely between interpreters in the same process.
 struct _xid {
     // data is the cross-interpreter-safe derivation of a Python object
-    // (see _PyObject_GetCrossInterpreterData).  It will be NULL if the
+    // (see _PyObject_GetXIData).  It will be NULL if the
     // new_object func (below) encodes the data.
     void *data;
     // obj is the Python object from which the data was derived.  This
     // is non-NULL only if the data remains bound to the object in some
     // way, such that the object must be "released" (via a decref) when
     // the data is released.  In that case the code that sets the field,
-    // likely a registered "crossinterpdatafunc", is responsible for
+    // likely a registered "xidatafunc", is responsible for
     // ensuring it owns the reference (i.e. incref).
     PyObject *obj;
     // interp is the ID of the owning interpreter of the original
     // object.  It corresponds to the active interpreter when
-    // _PyObject_GetCrossInterpreterData() was called.  This should only
+    // _PyObject_GetXIData() was called.  This should only
     // be set by the cross-interpreter machinery.
     //
     // We use the ID rather than the PyInterpreterState to avoid issues
@@ -77,96 +77,77 @@ struct _xid {
     // okay (e.g. bytes) and for those types this field should be set
     // to NULL.  However, for most the data was allocated just for
     // cross-interpreter use, so it must be freed when
-    // _PyCrossInterpreterData_Release is called or the memory will
+    // _PyXIData_Release is called or the memory will
     // leak.  In that case, at the very least this field should be set
     // to PyMem_RawFree (the default if not explicitly set to NULL).
     // The call will happen with the original interpreter activated.
     xid_freefunc free;
 };
 
-PyAPI_FUNC(_PyCrossInterpreterData *) _PyCrossInterpreterData_New(void);
-PyAPI_FUNC(void) _PyCrossInterpreterData_Free(_PyCrossInterpreterData *data);
+PyAPI_FUNC(_PyXIData_t *) _PyXIData_New(void);
+PyAPI_FUNC(void) _PyXIData_Free(_PyXIData_t *data);
 
-#define _PyCrossInterpreterData_DATA(DATA) ((DATA)->data)
-#define _PyCrossInterpreterData_OBJ(DATA) ((DATA)->obj)
-#define _PyCrossInterpreterData_INTERPID(DATA) ((DATA)->interpid)
+#define _PyXIData_DATA(DATA) ((DATA)->data)
+#define _PyXIData_OBJ(DATA) ((DATA)->obj)
+#define _PyXIData_INTERPID(DATA) ((DATA)->interpid)
 // Users should not need getters for "new_object" or "free".
 
 
+/* getting cross-interpreter data */
+
+typedef int (*xidatafunc)(PyThreadState *tstate, PyObject *, _PyXIData_t *);
+
+typedef struct _xid_lookup_state _PyXIData_lookup_t;
+
+PyAPI_FUNC(xidatafunc) _PyXIData_Lookup(PyObject *);
+PyAPI_FUNC(int) _PyObject_CheckXIData(PyObject *);
+PyAPI_FUNC(int) _PyObject_GetXIData(PyObject *, _PyXIData_t *);
+
+
+/* using cross-interpreter data */
+
+PyAPI_FUNC(PyObject *) _PyXIData_NewObject(_PyXIData_t *);
+PyAPI_FUNC(int) _PyXIData_Release(_PyXIData_t *);
+PyAPI_FUNC(int) _PyXIData_ReleaseAndRawFree(_PyXIData_t *);
+
+
 /* defining cross-interpreter data */
 
-PyAPI_FUNC(void) _PyCrossInterpreterData_Init(
-        _PyCrossInterpreterData *data,
+PyAPI_FUNC(void) _PyXIData_Init(
+        _PyXIData_t *data,
         PyInterpreterState *interp, void *shared, PyObject *obj,
         xid_newobjectfunc new_object);
-PyAPI_FUNC(int) _PyCrossInterpreterData_InitWithSize(
-        _PyCrossInterpreterData *,
+PyAPI_FUNC(int) _PyXIData_InitWithSize(
+        _PyXIData_t *,
         PyInterpreterState *interp, const size_t, PyObject *,
         xid_newobjectfunc);
-PyAPI_FUNC(void) _PyCrossInterpreterData_Clear(
-        PyInterpreterState *, _PyCrossInterpreterData *);
+PyAPI_FUNC(void) _PyXIData_Clear( PyInterpreterState *, _PyXIData_t *);
 
 // Normally the Init* functions are sufficient.  The only time
 // additional initialization might be needed is to set the "free" func,
 // though that should be infrequent.
-#define _PyCrossInterpreterData_SET_FREE(DATA, FUNC) \
+#define _PyXIData_SET_FREE(DATA, FUNC) \
     do { \
         (DATA)->free = (FUNC); \
     } while (0)
 // Additionally, some shareable types are essentially light wrappers
-// around other shareable types.  The crossinterpdatafunc of the wrapper
+// around other shareable types.  The xidatafunc of the wrapper
 // can often be implemented by calling the wrapped object's
-// crossinterpdatafunc and then changing the "new_object" function.
-// We have _PyCrossInterpreterData_SET_NEW_OBJECT() here for that,
+// xidatafunc and then changing the "new_object" function.
+// We have _PyXIData_SET_NEW_OBJECT() here for that,
 // but might be better to have a function like
-// _PyCrossInterpreterData_AdaptToWrapper() instead.
-#define _PyCrossInterpreterData_SET_NEW_OBJECT(DATA, FUNC) \
+// _PyXIData_AdaptToWrapper() instead.
+#define _PyXIData_SET_NEW_OBJECT(DATA, FUNC) \
     do { \
         (DATA)->new_object = (FUNC); \
     } while (0)
 
 
-/* using cross-interpreter data */
-
-PyAPI_FUNC(int) _PyObject_CheckCrossInterpreterData(PyObject *);
-PyAPI_FUNC(int) _PyObject_GetCrossInterpreterData(PyObject *, _PyCrossInterpreterData *);
-PyAPI_FUNC(PyObject *) _PyCrossInterpreterData_NewObject(_PyCrossInterpreterData *);
-PyAPI_FUNC(int) _PyCrossInterpreterData_Release(_PyCrossInterpreterData *);
-PyAPI_FUNC(int) _PyCrossInterpreterData_ReleaseAndRawFree(_PyCrossInterpreterData *);
-
-
 /* cross-interpreter data registry */
 
-// For now we use a global registry of shareable classes.  An
-// alternative would be to add a tp_* slot for a class's
-// crossinterpdatafunc. It would be simpler and more efficient.
-
-typedef int (*crossinterpdatafunc)(PyThreadState *tstate, PyObject *,
-                                   _PyCrossInterpreterData *);
-
-struct _xidregitem;
-
-struct _xidregitem {
-    struct _xidregitem *prev;
-    struct _xidregitem *next;
-    /* This can be a dangling pointer, but only if weakref is set. */
-    PyTypeObject *cls;
-    /* This is NULL for builtin types. */
-    PyObject *weakref;
-    size_t refcount;
-    crossinterpdatafunc getdata;
-};
-
-struct _xidregistry {
-    int global;  /* builtin types or heap types */
-    int initialized;
-    PyMutex mutex;
-    struct _xidregitem *head;
-};
-
-PyAPI_FUNC(int) _PyCrossInterpreterData_RegisterClass(PyTypeObject *, crossinterpdatafunc);
-PyAPI_FUNC(int) _PyCrossInterpreterData_UnregisterClass(PyTypeObject *);
-PyAPI_FUNC(crossinterpdatafunc) _PyCrossInterpreterData_Lookup(PyObject *);
+#define Py_CORE_CROSSINTERP_DATA_REGISTRY_H
+#include "pycore_crossinterp_data_registry.h"
+#undef Py_CORE_CROSSINTERP_DATA_REGISTRY_H
 
 
 /*****************************/
@@ -175,22 +156,19 @@ PyAPI_FUNC(crossinterpdatafunc) _PyCrossInterpreterData_Lookup(PyObject *);
 
 struct _xi_runtime_state {
     // builtin types
-    // XXX Remove this field once we have a tp_* slot.
-    struct _xidregistry registry;
+    _PyXIData_lookup_t data_lookup;
 };
 
 struct _xi_state {
     // heap types
-    // XXX Remove this field once we have a tp_* slot.
-    struct _xidregistry registry;
+    _PyXIData_lookup_t data_lookup;
 
     // heap types
     PyObject *PyExc_NotShareableError;
 };
 
 extern PyStatus _PyXI_Init(PyInterpreterState *interp);
 extern void _PyXI_Fini(PyInterpreterState *interp);
-
 extern PyStatus _PyXI_InitTypes(PyInterpreterState *interp);
 extern void _PyXI_FiniTypes(PyInterpreterState *interp);
 

Include/internal/pycore_crossinterp_data_registry.h

@@ -0,0 +1,36 @@
+#ifndef Py_CORE_CROSSINTERP_DATA_REGISTRY_H
+#  error "this header must not be included directly"
+#endif
+
+
+// For now we use a global registry of shareable classes.  An
+// alternative would be to add a tp_* slot for a class's
+// xidatafunc. It would be simpler and more efficient.
+
+struct _xidregitem;
+
+struct _xidregitem {
+    struct _xidregitem *prev;
+    struct _xidregitem *next;
+    /* This can be a dangling pointer, but only if weakref is set. */
+    PyTypeObject *cls;
+    /* This is NULL for builtin types. */
+    PyObject *weakref;
+    size_t refcount;
+    xidatafunc getdata;
+};
+
+struct _xidregistry {
+    int global;  /* builtin types or heap types */
+    int initialized;
+    PyMutex mutex;
+    struct _xidregitem *head;
+};
+
+PyAPI_FUNC(int) _PyXIData_RegisterClass(PyTypeObject *, xidatafunc);
+PyAPI_FUNC(int) _PyXIData_UnregisterClass(PyTypeObject *);
+
+struct _xid_lookup_state {
+    // XXX Remove this field once we have a tp_* slot.
+    struct _xidregistry registry;
+};

Include/internal/pycore_runtime_init.h

@@ -50,8 +50,10 @@ extern PyTypeObject _PyExc_MemoryError;
             .next_id = -1, \
         }, \
         .xi = { \
-            .registry = { \
-                .global = 1, \
+            .data_lookup = { \
+                .registry = { \
+                    .global = 1, \
+                }, \
             }, \
         }, \
         /* A TSS key must be initialized with Py_tss_NEEDS_INIT \