Neil Schemenauer [Tue, 20 Jan 2026 18:01:09 +0000 (10:01 -0800)]
gh-144054: no deferred refcount for untracked (gh-144081)
This reverts gh-144055 and fixes the bug in a different way. Deferred
reference counting relies on the object being tracked by the GC,
otherwise the object will live until interpreter shutdown. So, take
care that we do not enable deferred reference counting for objects that
are untracked. Also, if a tuple has deferred reference counting
enabled, don't untrack it.
Neil Schemenauer [Tue, 20 Jan 2026 13:45:12 +0000 (05:45 -0800)]
gh-144054: shutdown fix for deferred ref counting (GH-144055)
When shutting down, disable deferred refcounting for all GC objects. It
is important to do this also for untracked objects, which before this
change were getting missed.
Small code cleanup:
We can remove the shutdown case disable_deferred_refcounting() call
inside scan_heap_visitor() if we are careful about it. The key is
that frame_disable_deferred_refcounting() might fail if the object
is untracked.
Terry Jan Reedy [Tue, 20 Jan 2026 09:07:29 +0000 (04:07 -0500)]
gh-143774 - Improve IDLE Format Paragraph doc (#143775)
Add a reminder to not rewrap code line to the Menu => Format => Reformat Paragraph entry.
In Editing and Nagivagion, add a new 'Format block' subsection that defines 'paragraph'
to better match what is dependably handled as more or less expected.
In particular, specify equal indents and that the resulting indent equals original indent.
Also mention that selections are expanded to complete lines and how to modify max length.
(Also fix a couple case errors in cross references.)
gh-143935: Email preserve parens when folding comments (#143936)
Fix a bug in the folding of comments when flattening an email message
using a modern email policy. Comments consisting of a very long sequence of
non-foldable characters could trigger a forced line wrap that omitted the
required leading space on the continuation line, causing the remainder of
the comment to be interpreted as a new header field. This enabled header
injection with carefully crafted inputs.
Gregory P. Smith [Sun, 18 Jan 2026 22:04:18 +0000 (14:04 -0800)]
gh-141860: Add on_error= keyword arg to `multiprocessing.set_forkserver_preload` (GH-141859)
Add a keyword-only `on_error` parameter to `multiprocessing.set_forkserver_preload()`. This allows the user to have exceptions during optional `forkserver` start method module preloading cause the forkserver subprocess to warn (generally to stderr) or exit with an error (preventing use of the forkserver) instead of being silently ignored.
This _also_ fixes an oversight, errors when preloading a `__main__` module are now treated the similarly. Those would always raise unlike other modules in preload, but that had gone unnoticed as up until bug fix PR GH-135295 in 3.14.1 and 3.13.8, the `__main__` module was never actually preloaded.
Based on original work by Nick Neumann @aggieNick02 in GH-99515.
If we are specializing to `LOAD_GLOBAL_MODULE` or `LOAD_ATTR_MODULE`, try
to enable deferred reference counting for the value, if the object is owned by
a different thread. This applies to the free-threaded build only and should
improve scaling of multi-threaded programs.
Muneeb Ullah [Fri, 16 Jan 2026 17:09:56 +0000 (22:09 +0500)]
gh-143674: Document F/D complex format characters in struct module (#143675)
Add documentation for the F (complex from two floats) and D (complex
from two doubles) format characters in the struct module docstring.
These format characters were implemented but not documented.
Co-authored-by: Sergey B Kirpichev <skirpichev@gmail.com>
Mark Shannon [Fri, 16 Jan 2026 09:11:58 +0000 (09:11 +0000)]
GH-143493: Conform to spec for generator expressions while supporting virtual iterators (GH-143569)
* Moves the `GET_ITER` instruction into the generator function preamble.
This means the the iterable is converted into an iterator during generator
creation, as documented, but keeps it in the same code object allowing
optimization.
zhong [Thu, 15 Jan 2026 15:08:55 +0000 (07:08 -0800)]
gh-143602: Fix duplicate buffer exports in io.BytesIO.write (#143629)
Fix an inconsistency issue in io.BytesIO.write() where the buffer was exported
twice, which could lead to unexpected data overwrites and position drift when
the buffer changes between exports.
Petr Viktorin [Thu, 15 Jan 2026 11:37:30 +0000 (12:37 +0100)]
gh-141004: Mark up docs of old PyMem macros (GH-143783)
These had a docs-only deprecation notice since the first version of the docs in this repo. Nowadays we call things “soft deprecated” if there's just a note in the docs.
The deprecated directive needs a version, I went with the first one that had the notice (2.0; it's not in 1.6):
- https://docs.python.org/release/2.0/api/memoryInterface.html
- https://docs.python.org/release/1.6/api/memoryInterface.html
Since PEP 445, they are now direct aliases; there are no (additional) binary compatibility concerns over the preferred names.
Peter Gessler [Thu, 15 Jan 2026 09:51:11 +0000 (03:51 -0600)]
gh-80620: Support negative timestamps on windows in `time.gmtime`, `time.localtime`, and `datetime` module (#143463)
Previously, negative timestamps (representing dates before 1970-01-01) were
not supported on Windows due to platform limitations. The changes introduce a
fallback implementation using the Windows FILETIME API, allowing negative
timestamps to be correctly handled in both UTC and local time conversions.
Additionally, related test code is updated to remove Windows-specific skips
and error handling, ensuring consistent behavior across platforms.
Co-authored-by: Victor Stinner <vstinner@python.org>
Add the unicodedata.iter_graphemes() function to iterate over grapheme
clusters according to rules defined in Unicode Standard Annex #29.
Add unicodedata.grapheme_cluster_break(), unicodedata.indic_conjunct_break()
and unicodedata.extended_pictographic() functions to get the properties
of the character which are related to the above algorithm.
Duane Griffin [Wed, 14 Jan 2026 11:25:46 +0000 (00:25 +1300)]
gh-106287: Do not write objects after an unmarshalling error (GH-132715)
Writing out an object may involve a slot lookup, which is not safe to do with
an exception raised. In debug mode an assertion failure will occur if this
happens.
Petr Viktorin [Wed, 14 Jan 2026 09:54:45 +0000 (10:54 +0100)]
gh-141004: Document old provisional names for vectorcall/code API (GH-143568)
gh-141004: Document old provisional names for vectorcall & code API
For vectorcall API, the underscore-prefixed provisional names
continue to be available per PEP 590:
https://peps.python.org/pep-0590/#finalizing-the-api
Use `versionadded`, `versionchanged`, and `:no-typesetting:`
to mark this up.
For PyCode API, use `:no-typesetting:` rather than `index::` so that
these are semantically documented as C functions.
Petr Viktorin [Tue, 13 Jan 2026 12:21:59 +0000 (13:21 +0100)]
gh-143054: Disallow non-top-level Cut for now (GH-143622)
The behaviour of Cut in nested parentheses, Repeat, Opt, and similar
is somewhat chaotic. Apparently even the academic papers on PEG aren't
as clear as they could be.
And it doesn't really matter. Python only uses top-level cuts.
When that changes, we can clarify as much as necessary (and even
change the implementation to make sense for what we'll need).
Document that this is deliberately unspecified, and add a test to
make sure any decision is deliberate, tested and documented.
Gregory P. Smith [Mon, 12 Jan 2026 23:47:21 +0000 (15:47 -0800)]
gh-143706: Fix sys.argv not set during multiprocessing forkserver `__main__` preload (#143717)
The forkserver was not passing sys.argv to its main() function, causing
sys.argv to be empty during `__main__` module import in child processes. This
fixes a non-obvious regression inadvertently introduced by the gh-126631 main
preloading fix.
Łukasz Langa [Mon, 12 Jan 2026 23:01:01 +0000 (00:01 +0100)]
Bump automatic certificate update on Windows during builds (GH-143741)
Without this, OpenSSL that we use to download external dependencies might use a stale certificate store and be unable to connect to servers. We need to use a Windows-specific HTTP client that uses CryptoAPI directly to trigger certificate updates.
We only do it on failure to avoid hitting servers twice. And we only do it once per each URL.