[pull] main from python:main

Doc/c-api/list.rst

@@ -74,11 +74,25 @@ List Objects
    Like :c:func:`PyList_GetItemRef`, but returns a
    :term:`borrowed reference` instead of a :term:`strong reference`.
 
+   .. note::
+
+      In the :term:`free-threaded build`, the returned
+      :term:`borrowed reference` may become invalid if another thread modifies
+      the list concurrently. Prefer :c:func:`PyList_GetItemRef`, which returns
+      a :term:`strong reference`.
+
 
 .. c:function:: PyObject* PyList_GET_ITEM(PyObject *list, Py_ssize_t i)
 
    Similar to :c:func:`PyList_GetItem`, but without error checking.
 
+   .. note::
+
+      In the :term:`free-threaded build`, the returned
+      :term:`borrowed reference` may become invalid if another thread modifies
+      the list concurrently. Prefer :c:func:`PyList_GetItemRef`, which returns
+      a :term:`strong reference`.
+
 
 .. c:function:: int PyList_SetItem(PyObject *list, Py_ssize_t index, PyObject *item)
 
@@ -108,6 +122,14 @@ List Objects
       is being replaced; any reference in *list* at position *i* will be
       leaked.
 
+   .. note::
+
+      In the :term:`free-threaded build`, this macro has no internal
+      synchronization. It is normally only used to fill in new lists where no
+      other thread has a reference to the list. If the list may be shared,
+      use :c:func:`PyList_SetItem` instead, which uses a :term:`per-object
+      lock`.
+
 
 .. c:function:: int PyList_Insert(PyObject *list, Py_ssize_t index, PyObject *item)
 
@@ -138,6 +160,12 @@ List Objects
    Return ``0`` on success, ``-1`` on failure.  Indexing from the end of the
    list is not supported.
 
+   .. note::
+
+      In the :term:`free-threaded build`, when *itemlist* is a :class:`list`,
+      both *list* and *itemlist* are locked for the duration of the operation.
+      For other iterables (or ``NULL``), only *list* is locked.
+
 
 .. c:function:: int PyList_Extend(PyObject *list, PyObject *iterable)
 
@@ -150,6 +178,14 @@ List Objects
 
    .. versionadded:: 3.13
 
+   .. note::
+
+      In the :term:`free-threaded build`, when *iterable* is a :class:`list`,
+      :class:`set`, :class:`dict`, or dict view, both *list* and *iterable*
+      (or its underlying dict) are locked for the duration of the operation.
+      For other iterables, only *list* is locked; *iterable* may be
+      concurrently modified by another thread.
+
 
 .. c:function:: int PyList_Clear(PyObject *list)
 
@@ -168,6 +204,14 @@ List Objects
    Sort the items of *list* in place.  Return ``0`` on success, ``-1`` on
    failure.  This is equivalent to ``list.sort()``.
 
+   .. note::
+
+      In the :term:`free-threaded build`, element comparison via
+      :meth:`~object.__lt__` can execute arbitrary Python code, during which
+      the :term:`per-object lock` may be temporarily released. For built-in
+      types (:class:`str`, :class:`int`, :class:`float`), the lock is not
+      released during comparison.
+
 
 .. c:function:: int PyList_Reverse(PyObject *list)
 

Doc/c-api/typeobj.rst

@@ -3057,6 +3057,24 @@ Buffer Object Structures
 
    (5) Return ``0``.
 
+   **Thread safety:**
+
+   In the :term:`free-threaded build`, implementations must ensure:
+
+   * The export counter increment in step (3) is atomic.
+
+   * The underlying buffer data remains valid and at a stable memory
+     location for the lifetime of all exports.
+
+   * For objects that support resizing or reallocation (such as
+     :class:`bytearray`), the export counter is checked atomically before
+     such operations, and :exc:`BufferError` is raised if exports exist.
+
+   * The function is safe to call concurrently from multiple threads.
+
+   See also :ref:`thread-safety-memoryview` for the Python-level
+   thread safety guarantees of :class:`memoryview` objects.
+
    If *exporter* is part of a chain or tree of buffer providers, two main
    schemes can be used:
 
@@ -3102,6 +3120,16 @@ Buffer Object Structures
 
    (2) If the counter is ``0``, free all memory associated with *view*.
 
+   **Thread safety:**
+
+   In the :term:`free-threaded build`:
+
+   * The export counter decrement in step (1) must be atomic.
+
+   * Resource cleanup when the counter reaches zero must be done atomically,
+     as the final release may race with concurrent releases from other
+     threads and dellocation must only happen once.
+
    The exporter MUST use the :c:member:`~Py_buffer.internal` field to keep
    track of buffer-specific resources. This field is guaranteed to remain
    constant, while a consumer MAY pass a copy of the original buffer as the

Doc/data/threadsafety.dat

@@ -17,3 +17,59 @@
 PyMutex_Lock:shared:
 PyMutex_Unlock:shared:
 PyMutex_IsLocked:atomic:
+
+# List objects (Doc/c-api/list.rst)
+
+# Type checks - read ob_type pointer, always safe
+PyList_Check:atomic:
+PyList_CheckExact:atomic:
+
+# Creation - pure allocation, no shared state
+PyList_New:atomic:
+
+# Size - uses atomic load on free-threaded builds
+PyList_Size:atomic:
+PyList_GET_SIZE:atomic:
+
+# Strong-reference lookup - lock-free with atomic ops
+PyList_GetItemRef:atomic:
+
+# Borrowed-reference lookups - no locking; returned borrowed
+# reference is unsafe in free-threaded builds without
+# external synchronization
+PyList_GetItem:compatible:
+PyList_GET_ITEM:compatible:
+
+# Single-item mutations - hold per-object lock for duration;
+# appear atomic to lock-free readers
+PyList_SetItem:atomic:
+PyList_Append:atomic:
+
+# Insert - protected by per-object critical section; shifts
+# elements so lock-free readers may observe intermediate states
+PyList_Insert:shared:
+
+# Initialization macro - no synchronization; normally only used
+# to fill in new lists where there is no previous content
+PyList_SET_ITEM:compatible:
+
+# Bulk operations - hold per-object lock for duration
+PyList_GetSlice:atomic:
+PyList_AsTuple:atomic:
+PyList_Clear:atomic:
+
+# Reverse - protected by per-object critical section; swaps
+# elements so lock-free readers may observe intermediate states
+PyList_Reverse:shared:
+
+# Slice assignment - lock target list; also lock source when it
+# is a list
+PyList_SetSlice:shared:
+
+# Sort - per-object lock held; comparison callbacks may execute
+# arbitrary Python code
+PyList_Sort:shared:
+
+# Extend - lock target list; also lock source when it is a
+# list, set, or dict
+PyList_Extend:shared:

Doc/library/stdtypes.rst

@@ -5045,6 +5045,9 @@ copying.
 
       .. versionadded:: 3.3
 
+For information on the thread safety of :class:`memoryview` objects in
+the :term:`free-threaded build`, see :ref:`thread-safety-memoryview`.
+
 
 .. _types-set:
 

Doc/library/threadsafety.rst

@@ -548,3 +548,59 @@ Thread safety for bytearray objects
 
    Consider external synchronization when sharing :class:`bytearray` instances
    across threads.  See :ref:`freethreading-python-howto` for more information.
+
+
+.. _thread-safety-memoryview:
+
+Thread safety for memoryview objects
+====================================
+
+:class:`memoryview` objects provide access to the internal data of an
+underlying object without copying. Thread safety depends on both the
+memoryview itself and the underlying buffer exporter.
+
+The memoryview implementation uses atomic operations to track its own
+exports in the :term:`free-threaded build`. Creating and
+releasing a memoryview are thread-safe. Attribute access (e.g.,
+:attr:`~memoryview.shape`, :attr:`~memoryview.format`) reads fields that
+are immutable for the lifetime of the memoryview, so concurrent reads
+are safe as long as the memoryview has not been released.
+
+However, the actual data accessed through the memoryview is owned by the
+underlying object. Concurrent access to this data is only safe if the
+underlying object supports it:
+
+* For immutable objects like :class:`bytes`, concurrent reads through
+  multiple memoryviews are safe.
+
+* For mutable objects like :class:`bytearray`, reading and writing the
+  same memory region from multiple threads without external
+  synchronization is not safe and may result in data corruption.
+  Note that even read-only memoryviews of mutable objects do not
+  prevent data races if the underlying object is modified from
+  another thread.
+
+.. code-block::
+   :class: bad
+
+   # NOT safe: concurrent writes to the same buffer
+   data = bytearray(1000)
+   view = memoryview(data)
+   # Thread 1: view[0:500] = b'x' * 500
+   # Thread 2: view[0:500] = b'y' * 500
+
+.. code-block::
+   :class: good
+
+   # Safe: use a lock for concurrent access
+   import threading
+   lock = threading.Lock()
+   data = bytearray(1000)
+   view = memoryview(data)
+
+   with lock:
+       view[0:500] = b'x' * 500
+
+Resizing or reallocating the underlying object (such as calling
+:meth:`bytearray.resize`) while a memoryview is exported raises
+:exc:`BufferError`. This is enforced regardless of threading.

Doc/library/wave.rst

@@ -9,14 +9,19 @@
 --------------
 
 The :mod:`!wave` module provides a convenient interface to the Waveform Audio
-"WAVE" (or "WAV") file format. Only uncompressed PCM encoded wave files are
-supported.
+"WAVE" (or "WAV") file format.
+
+The module supports uncompressed PCM and IEEE floating-point WAV formats.
 
 .. versionchanged:: 3.12
 
    Support for ``WAVE_FORMAT_EXTENSIBLE`` headers was added, provided that the
    extended format is ``KSDATAFORMAT_SUBTYPE_PCM``.
 
+.. versionchanged:: next
+
+   Support for reading and writing ``WAVE_FORMAT_IEEE_FLOAT`` files was added.
+
 The :mod:`!wave` module defines the following function and exception:
 
 
@@ -60,6 +65,21 @@ The :mod:`!wave` module defines the following function and exception:
    specification or hits an implementation deficiency.
 
 
+.. data:: WAVE_FORMAT_PCM
+
+   Format code for uncompressed PCM audio.
+
+
+.. data:: WAVE_FORMAT_IEEE_FLOAT
+
+   Format code for IEEE floating-point audio.
+
+
+.. data:: WAVE_FORMAT_EXTENSIBLE
+
+   Format code for WAVE extensible headers.
+
+
 .. _wave-read-objects:
 
 Wave_read Objects
@@ -98,6 +118,14 @@ Wave_read Objects
       Returns number of audio frames.
 
 
+   .. method:: getformat()
+
+      Returns the frame format code.
+
+      This is one of :data:`WAVE_FORMAT_PCM`,
+      :data:`WAVE_FORMAT_IEEE_FLOAT`, or :data:`WAVE_FORMAT_EXTENSIBLE`.
+
+
    .. method:: getcomptype()
 
       Returns compression type (``'NONE'`` is the only supported type).
@@ -112,8 +140,8 @@ Wave_read Objects
    .. method:: getparams()
 
       Returns a :func:`~collections.namedtuple` ``(nchannels, sampwidth,
-      framerate, nframes, comptype, compname)``, equivalent to output of the
-      ``get*()`` methods.
+      framerate, nframes, comptype, compname)``, equivalent to output
+      of the ``get*()`` methods.
 
 
    .. method:: readframes(n)
@@ -190,6 +218,9 @@ Wave_write Objects
 
       Set the sample width to *n* bytes.
 
+      For :data:`WAVE_FORMAT_IEEE_FLOAT`, only 4-byte (32-bit) and
+      8-byte (64-bit) sample widths are supported.
+
 
    .. method:: getsampwidth()
 
@@ -238,11 +269,32 @@ Wave_write Objects
       Return the human-readable compression type name.
 
 
+   .. method:: setformat(format)
+
+      Set the frame format code.
+
+      Supported values are :data:`WAVE_FORMAT_PCM` and
+      :data:`WAVE_FORMAT_IEEE_FLOAT`.
+
+      When setting :data:`WAVE_FORMAT_IEEE_FLOAT`, the sample width must be
+      4 or 8 bytes.
+
+
+   .. method:: getformat()
+
+      Return the current frame format code.
+
+
    .. method:: setparams(tuple)
 
-      The *tuple* should be ``(nchannels, sampwidth, framerate, nframes, comptype,
-      compname)``, with values valid for the ``set*()`` methods.  Sets all
-      parameters.
+      The *tuple* should be
+      ``(nchannels, sampwidth, framerate, nframes, comptype, compname, format)``,
+      with values valid for the ``set*()`` methods. Sets all parameters.
+
+      For backwards compatibility, a 6-item tuple without *format* is also
+      accepted and defaults to :data:`WAVE_FORMAT_PCM`.
+
+      For ``format=WAVE_FORMAT_IEEE_FLOAT``, *sampwidth* must be 4 or 8.
 
 
    .. method:: getparams()
@@ -279,3 +331,6 @@ Wave_write Objects
       Note that it is invalid to set any parameters after calling :meth:`writeframes`
       or :meth:`writeframesraw`, and any attempt to do so will raise
       :exc:`wave.Error`.
+
+      For :data:`WAVE_FORMAT_IEEE_FLOAT` output, a ``fact`` chunk is written as
+      required by the WAVE specification for non-PCM formats.