This updates `dtoa.c` to avoid using the Bigint free-list in --disable-gil builds and
to pre-computes the needed powers of 5 during interpreter initialization.
* gh-111962: Make dtoa thread-safe in `--disable-gil` builds.
This avoids using the Bigint free-list in `--disable-gil` builds
and pre-computes the needed powers of 5 during interpreter initialization.
* Fix size of cached powers of 5 array.
We need the powers of 5 up to 5**512 because we only jump straight to
underflow when the exponent is less than -512 (or larger than 308).
* Rename Py_NOGIL to Py_GIL_DISABLED
* Changes from review
* Fix assertion placement
* gh-51944: Add some macOS constants to termios
This changeset adds all public constants in <termio.h>
and <sys/termios.h> on macOS that weren't present
already.
Based on the macOS 14.2 SDK
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
If the input prompt to the builtin input function on terminal has any null
character, then raise ValueError instead of silently truncating it.
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
Previously arbitrary errors could be cleared during formatting error
messages for ImportError or AttributeError for modules. Now all
unexpected errors are reported.
Use scanning "/dev/fd/" on macOS in support.fd_count(). That's both more efficient than scanning all possible file descriptors, and avoids crashing the interpreter when there are open "guarded" file descriptors.
"Guarded" file descriptors are a macOS feature where file descriptors used by system libraries are marked and cause hard crashes when used by "user" code.
Co-authored-by: Victor Stinner <vstinner@python.org>
Previously, "widget.unbind(sequence, funcid)" destroyed the current binding
for "sequence", leaving "sequence" unbound, and deleted the "funcid"
command.
Now it removes only "funcid" from the binding for "sequence", keeping
other commands, and deletes the "funcid" command.
It leaves "sequence" unbound only if "funcid" was the last bound command.
Co-authored-by: GiovanniL <13402461+GiovaLomba@users.noreply.github.com>
* Implement _Py_HashPointerRaw() as a static inline function.
* Add Py_HashPointer() tests to test_capi.test_hash.
* Keep _Py_HashPointer() function as an alias to Py_HashPointer().
In the slot typedefs table, the parameter of `destructor`
and the first parameter of `traverseproc` should both be
`PyObject *` rather than `void *`.
Same for `inquiry`.
* Ignore os.close() errors when ignore_errors is True.
* Pass os.close() errors to the error handler if specified.
* os.close no longer retried after error.
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
Previously a symlink attack resistant version of shutil.rmtree() could ignore
or pass to the error handler arbitrary exception when invalid arguments
were provided.
Set MAX_STRUCT_SIZE to 32 in stgdict.c when on Arm platforms.
This because on Arm platforms structs with at most 4 elements of any
floating point type values can be passed through registers. If the type
is double the maximum size of the struct is 32 bytes.
On x86-64 Linux, it's maximum 16 bytes hence we need to differentiate.
Add a track parameter to shared memory to allow resource tracking via the side-launched resource tracker process to be disabled on platforms that use it (POSIX).
This allows people who do not want automated cleanup at process exit because they are using the shared memory with processes not participating in Python's resource tracking to use the shared_memory API.
Co-authored-by: Łukasz Langa <lukasz@langa.pl>
Co-authored-by: Guido van Rossum <gvanrossum@gmail.com>
Co-authored-by: Antoine Pitrou <pitrou@free.fr>
Co-authored-by: Gregory P. Smith <greg@krypto.org>
Refactor delete-safe symbol handling in subprocess.
Only module globals are force-cleared during interpreter finalization, using a class reference instead of individually listing the constants everywhere is simpler.
Use `_from_parsed_parts()` to create a pre-joined/pre-parsed path, rather
than passing multiple arguments to `with_segments()`
Co-authored-by: Alex Waygood <Alex.Waygood@Gmail.com>
Restore `subprocess`'s intended use of `vfork()` by default for performance on Linux;
also fixes the behavior of `extra_groups=[]` which was unintentionally broken in 3.12.0:
Fixed a performance regression in 3.12's :mod:`subprocess` on Linux where it
would no longer use the fast-path ``vfork()`` system call when it could have
due to a logic bug, instead falling back to the safe but slower ``fork()``.
Also fixed a security bug introduced in 3.12.0. If a value of ``extra_groups=[]``
was passed to :mod:`subprocess.Popen` or related APIs, the underlying
``setgroups(0, NULL)`` system call to clear the groups list would not be made
in the child process prior to ``exec()``.
The security issue was identified via code inspection in the process of
fixing the first bug. Thanks to @vain for the detailed report and
analysis in the initial bug on Github.
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
Work around a macOS bug, limit zlib crc32 calls to 1GiB.
Without this, `zlib.crc32` and `binascii.crc32` could produce incorrect
results on multi-gigabyte inputs depending on the macOS version's Apple
supplied zlib implementation.
AN Long
Serhiy Storchaka
Søren Løvborg
Seth Michael Larson
Sam Gross
Łukasz Langa
Christopher Chavez
andrewluotechnologies
Mark Shannon
Tian Gao
Ronald Oussoren
Kushal Das
Pablo Galindo Salgado
Alex Waygood
Victor Stinner
Matt Prodani
Miro Hrončok
Jurjen N. E. Bos
Zackery Spytz
Ethan Furman
Diego Russo
Ken Jin
Donghee Na
Jeffrey Kintscher
Rune Tynan
pan324
Sergey B Kirpichev
Constantin Hong
Russell Keith-Magee
Barney Gale
Gregory P. Smith
Amioplk
Kamil Turek