Serhiy Storchaka [Wed, 18 Feb 2026 11:20:31 +0000 (13:20 +0200)]
gh-135573: Make pickled lists, sets and dicts a tiny bit smaller (GH-144162)
Ensure that APPENDS and SETITEMS are never used for a batch of size 1.
Ensure that ADDITEMS and SETITEMS are never used for a batch of size 0.
This harmonizes the C implementation with the Python implementation
which already guarantees this and makes a pickle a tiny bit smaller
with a tiny chance (about 0.1%).
Saves 1 byte for list and dict with size 1001, 2001, ...
Saves 2 bytes for set and dict with size 1000, 2000, ...
Serhiy Storchaka [Wed, 18 Feb 2026 10:58:21 +0000 (12:58 +0200)]
gh-144882: Optimize name tables in unicodedata by excluding names derived by rule NR2 (GH-144883)
Since the code for rule NR2 is already here, to support CJK unified
ideographs and Tangut ideographs, it can also be used for other names
derived by rule NR2.
Sam Gross [Tue, 17 Feb 2026 16:12:25 +0000 (09:12 -0700)]
gh-144438: Fix false sharing between QSBR and tlbc_index (gh-144554)
Align the QSBR thread state array to a 64-byte cache line boundary
and add padding at the end of _PyThreadStateImpl. Depending on heap
layout, the QSBR array could end up sharing a cache line with a
thread's tlbc_index, causing QSBR quiescent state updates to contend
with reads of tlbc_index in RESUME_CHECK. This is sensitive to
earlier allocations during interpreter init and can appear or
disappear with seemingly unrelated changes.
Either change alone is sufficient to fix the specific issue, but both
are worthwhile to avoid similar problems in the future.
Co-authored-by: Hugo van Kemenade <1324225+hugovk@users.noreply.github.com> Co-authored-by: Adam Johnson <me@adamj.eu> Co-authored-by: Benedikt Johannes <benedikt.johannes.hofer@gmail.com>
gh-142349: Fix ast.unparse for lazy import statements (#144893)
The unparser was not handling the `is_lazy` attribute on Import and
ImportFrom AST nodes, causing lazy imports to be unparsed as regular
imports. This broke the round-trip (parse → unparse → reparse) for
any file containing `lazy import` statements.
Jason Yalim, PhD [Mon, 16 Feb 2026 22:20:37 +0000 (15:20 -0700)]
gh-140715: Add `%D` format code support to `strptime()` (GH-144819)
* %D support for strptime, including test and Doc update
* additional %D test
* change documentation example date for %D so it is more legible to non-US readers
* change testing date for %D so it is more legible to non-US readers
* mv News blurb to Library, consistent with previous %F change
* change invalid format code from %D to C-standard unused %!
* Fix erroneous and misleading example Doc to %y from %Y, use correct C99+ definition for C99 %D; update additional tests
gh-144833: Fix use-after-free in SSL module when SSL_new() fails (GH-144843)
In newPySSLSocket(), when SSL_new() returns NULL, Py_DECREF(self)
was called before _setSSLError(get_state_ctx(self), ...), causing
a use-after-free. Additionally, get_state_ctx() was called with
self (PySSLSocket*) instead of sslctx (PySSLContext*), which is
a type confusion bug.
Fix by calling _setSSLError() before Py_DECREF() and using
sslctx instead of self for get_state_ctx().
Gregory P. Smith [Mon, 16 Feb 2026 01:43:39 +0000 (17:43 -0800)]
gh-86519: Add prefixmatch APIs to the re module (GH-31137)
Adds `prefixmatch` APIs to the re module as an alternate name for our long existing `match` APIs to help alleviate a common Python confusion for those coming from other languages regular expression libraries.
These alleviate common confusion around what "match" means as Python is different than other popular languages regex libraries in our use of the term as an API name. The original `match` names are **NOT being deprecated**. Source tooling like linters, IDEs, and LLMs could suggest using `prefixmatch` instead of match to improve code health and reduce cognitive burden of understanding the intent of code when configured for a modern minimum Python version.
See the documentation changes for a better description.
Discussions took place in the PR, in the issue, and finally at https://discuss.python.org/t/add-re-prefixmatch-deprecate-re-match/105927
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> Co-authored-by: Shantanu <12621235+hauntsaninja@users.noreply.github.com> Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
gh-144727: Add test for circular lazy import crash (#144727) (#144838)
Add a regression test ensuring that circular lazy imports raise a proper
error (ImportCycleError) instead of crashing with a segfault. The crash
was fixed in gh-144733 but no test was added at the time.
gh-144759: Fix undefined behavior from NULL pointer arithmetic in lexer (#144788)
Guard against NULL pointer arithmetic in `_PyLexer_remember_fstring_buffers`
and `_PyLexer_restore_fstring_buffers`. When `start` or `multi_line_start`
are NULL (uninitialized in tok_mode_stack[0]), performing `NULL - tok->buf`
is undefined behavior. Add explicit NULL checks to store -1 as sentinel
and restore NULL accordingly.
Add test_lexer_buffer_realloc_with_null_start to test_repl.py that
exercises the code path where the lexer buffer is reallocated while
tok_mode_stack[0] has NULL start/multi_line_start pointers. This
triggers _PyLexer_remember_fstring_buffers and verifies the NULL
checks prevent undefined behavior.
gh-136672: Docs: Move parts of Enum HOWTO to API Docs (GH-139176)
To avoid duplicate content in the Enum HOWTO and
API documentation which is not automatically synced,
the section about supported __dunder__ and _sunder
names is moved from HOWTO to API docs.
See also https://github.com/python/cpython/pull/136791
gh-142349: Fix refcount corruption in lazy import specialization (#144733)
Remove spurious Py_DECREF on borrowed ref in LOAD_GLOBAL specialization
_PyDict_LookupIndexAndValue() returns a borrowed reference via
_Py_dict_lookup(), but specialize_load_global_lock_held() called
Py_DECREF(value) on it when bailing out for lazy imports. Each time
the adaptive counter fired while a lazy import was still in globals,
this stole one reference from the dict's object. With 8+ threads
racing through LOAD_GLOBAL during concurrent lazy import resolution,
enough triggers accumulated to drive the refcount to zero while the
dict and other threads still referenced the object, causing
use-after-free.
Victor Stinner [Tue, 10 Feb 2026 19:31:12 +0000 (20:31 +0100)]
gh-141563: Fix test_cppext on macOS (#144685)
Don't test internal header files including mimalloc on macOS since
mimalloc emits compiler warnings:
In file included from extension.cpp:21:
In file included from Include/internal/pycore_backoff.h:15:
In file included from Include/internal/pycore_interp_structs.h:15:
In file included from Include/internal/pycore_tstate.h:14:
In file included from Include/internal/pycore_mimalloc.h:43:
Include/internal/mimalloc/mimalloc.h:464:85: error: defaulted
function definitions are a C++11 extension
[-Werror,-Wc++11-extensions]
mi_stl_allocator() mi_attr_noexcept = default;
^
Include/internal/mimalloc/mimalloc.h:465:85: error: defaulted
function definitions are a C++11 extension
[-Werror,-Wc++11-extensions]
mi_stl_allocator(const mi_stl_allocator&) mi_attr_noexcept = default;
Log also CXX and CXXFLAGS env vars in test_cppext. Log also CPPFLAGS
in test_cext.
Gregory P. Smith [Tue, 10 Feb 2026 13:08:33 +0000 (05:08 -0800)]
gh-143650: Fix importlib race condition on import failure (GH-143651)
Fix a race condition where a thread could receive a partially-initialized
module when another thread's import fails. The race occurs when:
1. Thread 1 starts importing, adds module to sys.modules
2. Thread 2 sees the module in sys.modules via the fast path
3. Thread 1's import fails, removes module from sys.modules
4. Thread 2 returns a stale module reference not in sys.modules
The fix adds verification after the "skip lock" optimization in both Python
and C code paths to check if the module is still in sys.modules. If the
module was removed (due to import failure), we retry the import so the
caller receives the actual exception from the import failure rather than
a stale module reference.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
gh-144563: Fix remote debugging with duplicate libpython mappings from ctypes (#144595)
When _ctypes is imported, it may call dlopen on the libpython shared
library, causing the dynamic linker to load a second mapping of the
library into the process address space. The remote debugging code
iterates memory regions from low addresses upward and returns the first
mapping whose filename matches libpython. After _ctypes is imported, it
finds the dlopen'd copy first, but that copy's PyRuntime section was
never initialized, so reading debug offsets from it fails.
Fix this by validating each candidate PyRuntime address before accepting
it. The validation reads the first 8 bytes and checks for the "xdebugpy"
cookie that is only present in an initialized PyRuntime. Uninitialized
duplicate mappings will fail this check and be skipped, allowing the
search to continue to the real, initialized PyRuntime.
Steve Dower [Tue, 10 Feb 2026 00:01:17 +0000 (00:01 +0000)]
gh-144278: Enable overriding sys.implementation's name and cache_tag when building sysmodule.c (GH-144293)
Changing the values requires forking and patching, which is intentional. Simply rebuilding from source does not change the implementation enough to justify changing these values - they would still be `cpython` and compatible with existing `.pyc` files. But people who maintain forks are better served by being able to easily override these values in a place that can be forward-ported reliably.
projects / thirdparty / Python / cpython.git / log
