Deprecate _pointer_type_cache and calling POINTER on a string.
Co-authored-by: neonene <53406459+neonene@users.noreply.github.com>
Co-authored-by: Jun Komoda <45822440+junkmd@users.noreply.github.com>
Co-authored-by: Petr Viktorin <encukou@gmail.com>
After gh-130704, the interpreter replaces some uses of `LOAD_FAST` with
`LOAD_FAST_BORROW` which avoid incref/decrefs by "borrowing" references
on the interpreter stack when the bytecode compiler can determine that
it's safe.
This change broke some checks in C API extensions that relied on
`Py_REFCNT()` of `1` to determine if it's safe to modify an object
in-place. Objects may have a reference count of one, but still be
referenced further up the interpreter stack due to borrowing of
references.
This provides a replacement function for those checks.
`PyUnstable_Object_IsUniqueReferencedTemporary` is more conservative:
it checks that the object has a reference count of one and that it exists as a
unique strong reference in the interpreter's stack of temporary
variables in the top most frame.
See also:
* https://github.com/numpy/numpy/issues/28681
Co-authored-by: Pieter Eendebak <pieter.eendebak@gmail.com>
Co-authored-by: T. Wouters <thomas@python.org>
Co-authored-by: mpage <mpage@cs.stanford.edu>
Co-authored-by: Mark Shannon <mark@hotpy.org>
Co-authored-by: Victor Stinner <vstinner@python.org>
* Skip sNaN's testing in 32-bit mode.
* Drop float_set_snan() helper.
* Use memcpy() workaround for sNaN's in PyFloat_Unpack4().
* Document, that sNaN's may not be preserved by PyFloat_Pack/Unpack API.
There's some extra complexity due to making sure we we get things right when handling functions and classes defined in the __main__ module. This is also reflected in the tests, including the addition of extra functions in test.support.import_helper.
This helper is useful in a variety of ways, including in demonstrating how the different counts relate to one another.
It will be used in a later change to help identify if a function is "stateless", meaning it doesn't have any free vars or globals.
Note that a majority of this change is tests.
ntohl(), htonl(), if_indextoname(), getaddrinfo() now use __index__() if
available.
Also fix the Argument Clinic names for module-level functions (although
this does not affect the user).
Support it for the dev_t values in mknod(), major(), minor() and makedev(),
CPU numbers in sched_setaffinity(), group numbers in setgroups(),
configuration name in fpathconf(), pathconf(), confstr(), and sysconf().
The function indicates whether or not the function has a return statement.
This is used by a later change related treating some functions like scripts.
The `test_load_global_module()` test consumes a lot of dict key versions.
Skip the test if we have consumed half of the available versions that can be
used for the "load global" cache.
If `Py_IsFinalizing()` is true, non-daemon threads (other than the current one)
are done, and daemon threads are prevented from running, so they
cannot finalize themselves and become done. Joining them (without timeout)
would block forever.
Raise PythonFinalizationError instead of hanging.
Raise even when a timeout is given, for consistency with trying to join your own thread.
See gh-123940 for a use case: calling `join()` from `__del__`. This is
ill-advised, but an exception should at least make it easier to diagnose.
This was in C version from beginning, but available only
on conditional compilation (EXTRA_FUNCTIONALITY). Current
patch adds function to create IEEE contexts to the
pure-python module as well.
Co-authored-by: Bénédikt Tran <10796600+picnixz@users.noreply.github.com>
The following are added to the internal C-API:
* _PyErr_FormatV()
* _PyErr_SetModuleNotFoundError()
* _PyXIData_GetNotShareableErrorType()
* _PyXIData_FormatNotShareableError()
We also drop _PyXIData_lookup_context_t and _PyXIData_GetLookupContext().
To avoid having the debug sections being optimised away by the compiler
we use __attribute__((used)) on gcc and clang but in Windows this is
not supported by the Microsoft compiler and there is no equivalent flag.
Unfortunately Windows offers almost no alternative other than exporting
the symbol in the dynamic table or using it somehow.
OpenSSL 3.4.1 mnemonics are not compatible with OpenSSL 3.4.0 ones since
they were renumbered [1, 2]. Consequently, `_ssl_data_34.h` is renamed to
`_ssl_data_340.h` and `_ssl_data_34.h` now contains OpenSSL 3.4.1 mnemonics.
We also refine the mnemonics that are selected, discarding those that are
mnemonic-like but should not be used as such. More precisely, we remove
the ERR_LIB_MASK and ERR_LIB_OFFSET entries from OpenSSL 1.1.1 data.
[1]: https://github.com/openssl/openssl/issues/26316
[2]: https://github.com/openssl/openssl/issues/26388
* Support arbitrary bytes-like objects, not only bytes, in fcntl().
* The fcntl() buffer argument is now null-terminated.
* Automatically retry an ioctl() system calls failing with EINTR.
* Release the GIL for an ioctl() system call even for large bytes-like object.
* Do not silence arbitrary errors whet try to get a buffer.
* Optimize argument parsing, check the argument type before trying to get
a buffer or convert it to integer.
* Fix some error messages.
