* UCD_Check() uses PyModule_Check()
* Simplify the internal _PyUnicode_Name_CAPI structure:
* Remove size and state members
* Remove state and self parameters of getcode() and getname()
functions
* Remove global_module_state
The private _PyUnicode_Name_CAPI structure of the PyCapsule API
unicodedata.ucnhash_CAPI moves to the internal C API. Moreover, the
structure gets a new state member which must be passed to the
getcode() and getname() functions.
* Move Include/ucnhash.h to Include/internal/pycore_ucnhash.h
* unicodedata module is now built with Py_BUILD_CORE_MODULE.
* unicodedata: move hashAPI variable into unicodedata_module_state.
If PyDict_GetItemWithError is only used to check whether the key is in dict,
it is better to use PyDict_Contains instead.
And if it is used in combination with PyDict_SetItem, PyDict_SetDefault can
replace the combination.
These functions are considered not safe because they suppress all internal errors
and can return wrong result. PyDict_GetItemString and _PyDict_GetItemId can
also silence current exception in rare cases.
Remove no longer used _PyDict_GetItemId.
Add _PyDict_ContainsId and rename _PyDict_Contains into
_PyDict_Contains_KnownHash.
Fix memory leak in subprocess.Popen() in case of uid/gid overflow
Also add a test that would catch this leak with `--huntrleaks`.
Alas, the test for `extra_groups` also exposes an inconsistency
in our error reporting: we use a custom ValueError for `extra_groups`,
but propagate OverflowError for `user` and `group`.
It should just be a syscall updating a couple of fields in the kernel side
process info. Confirming, in glibc is appears to be a shim for the setsid
syscall (based on not finding any code implementing anything special for it)
and in uclibc (*much* easier to read) it is clearly just a setsid syscall shim.
A breadcrumb _suggesting_ that it is not allowed on Darwin/macOS comes from
a commit in emacs: https://lists.gnu.org/archive/html/bug-gnu-emacs/2017-04/msg00297.html
but I don't have a way to verify if that is true or not.
As we are not supporting vfork on macOS today I just left a note in a comment.
Using POSIX_CALL() is incorrect since pthread_sigmask() returns
the error number instead of setting errno.
Also handle failure of the first call to pthread_sigmask()
in the parent process, and explain why we don't handle failure
of the second call in a comment.
* bpo-35823: subprocess: Use vfork() instead of fork() on Linux when safe
When used to run a new executable image, fork() is not a good choice
for process creation, especially if the parent has a large working set:
fork() needs to copy page tables, which is slow, and may fail on systems
where overcommit is disabled, despite that the child is not going to
touch most of its address space.
Currently, subprocess is capable of using posix_spawn() instead, which
normally provides much better performance. However, posix_spawn() does not
support many of child setup operations exposed by subprocess.Popen().
Most notably, it's not possible to express `close_fds=True`, which
happens to be the default, via posix_spawn(). As a result, most users
can't benefit from faster process creation, at least not without
changing their code.
However, Linux provides vfork() system call, which creates a new process
without copying the address space of the parent, and which is actually
used by C libraries to efficiently implement posix_spawn(). Due to sharing
of the address space and even the stack with the parent, extreme care
is required to use vfork(). At least the following restrictions must hold:
* No signal handlers must execute in the child process. Otherwise, they
might clobber memory shared with the parent, potentially confusing it.
* Any library function called after vfork() in the child must be
async-signal-safe (as for fork()), but it must also not interact with any
library state in a way that might break due to address space sharing
and/or lack of any preparations performed by libraries on normal fork().
POSIX.1 permits to call only execve() and _exit(), and later revisions
remove vfork() specification entirely. In practice, however, almost all
operations needed by subprocess.Popen() can be safely implemented on
Linux.
* Due to sharing of the stack with the parent, the child must be careful
not to clobber local variables that are alive across vfork() call.
Compilers are normally aware of this and take extra care with vfork()
(and setjmp(), which has a similar problem).
* In case the parent is privileged, special attention must be paid to vfork()
use, because sharing an address space across different privilege domains
is insecure[1].
This patch adds support for using vfork() instead of fork() on Linux
when it's possible to do safely given the above. In particular:
* vfork() is not used if credential switch is requested. The reverse case
(simple subprocess.Popen() but another application thread switches
credentials concurrently) is not possible for pure-Python apps because
subprocess.Popen() and functions like os.setuid() are mutually excluded
via GIL. We might also consider to add a way to opt-out of vfork() (and
posix_spawn() on platforms where it might be implemented via vfork()) in
a future PR.
* vfork() is not used if `preexec_fn != None`.
With this change, subprocess will still use posix_spawn() if possible, but
will fallback to vfork() on Linux in most cases, and, failing that,
to fork().
[1] https://ewontfix.com/7
Co-authored-by: Gregory P. Smith [Google LLC] <gps@google.com>
* Add F_SETPIPE_SZ and F_GETPIPE_SZ to fcntl module
* Add pipesize parameter for subprocess.Popen class
This will allow the user to control the size of the pipes.
On linux the default is 64K. When a pipe is full it blocks for writing.
When a pipe is empty it blocks for reading. On processes that are
very fast this can lead to a lot of wasted CPU cycles. On a typical
Linux system the max pipe size is 1024K which is much better.
For high performance-oriented libraries such as xopen it is nice to
be able to set the pipe size.
The workaround without this feature is to use my_popen_process.stdout.fileno() in
conjuction with fcntl and 1031 (value of F_SETPIPE_SZ) to acquire this behavior.
Prepare unicodedata to add a state per module: start with a global
"module" state, pass it to subfunctions which access &UCD_Type. This
change also prepares the conversion of the UCD_Type static type to a
heap type.
This API is relatively lightweight and organizationally, given that it's
used by multiple modules, it makes sense to move it to fileutils.
Requires making sure that _posixsubprocess is compiled with the appropriate
Py_BUIILD_CORE_BUILTIN macro.
This suppression is no longer needed in os_closerange_impl, as it just
invokes the internal _Py_closerange implementation. On the other hand,
consumers of _Py_closerange may not have any other reason to suppress
invalid parameter issues, so narrow the scope to here.
close_range(2) should be preferred at all times if it's available, otherwise we'll use closefrom(2) if available with a fallback to fdwalk(3) or plain old loop over fd range in order of most efficient to least.
[note that this version does check for ENOSYS, but currently ignores all other errors]
Automerge-Triggered-By: @pablogsal
Such an API can be used both for os.closerange and subprocess. For the latter, this yields potential improvement for platforms that have fdwalk but wouldn't have used it there. This will prove even more beneficial later for platforms that have close_range(2), as the new API will prefer that over all else if it's available.
The new API is structured to look more like close_range(2), closing from [start, end] rather than the [low, high) of os.closerange().
Automerge-Triggered-By: @gpshead
* bpo-26680: Adds support for int.is_integer() for compatibility with float.is_integer().
The int.is_integer() method always returns True.
* bpo-26680: Adds a test to ensure that False.is_integer() and True.is_integer() are always True.
* bpo-26680: Adds Real.is_integer() with a trivial implementation using conversion to int.
This default implementation is intended to reduce the workload for subclass
implementers. It is not robust in the presence of infinities or NaNs and
may have suboptimal performance for other types.
* bpo-26680: Adds Rational.is_integer which returns True if the denominator is one.
This implementation assumes the Rational is represented in it's
lowest form, as required by the class docstring.
* bpo-26680: Adds Integral.is_integer which always returns True.
* bpo-26680: Adds tests for Fraction.is_integer called as an instance method.
The tests for the Rational abstract base class use an unbound
method to sidestep the inability to directly instantiate Rational.
These tests check that everything works correct as an instance method.
* bpo-26680: Updates documentation for Real.is_integer and built-ins int and float.
The call x.is_integer() is now listed in the table of operations
which apply to all numeric types except complex, with a reference
to the full documentation for Real.is_integer(). Mention of
is_integer() has been removed from the section 'Additional Methods
on Float'.
The documentation for Real.is_integer() describes its purpose, and
mentions that it should be overridden for performance reasons, or
to handle special values like NaN.
* bpo-26680: Adds Decimal.is_integer to the Python and C implementations.
The C implementation of Decimal already implements and uses
mpd_isinteger internally, we just expose the existing function to
Python.
The Python implementation uses internal conversion to integer
using to_integral_value().
In both cases, the corresponding context methods are also
implemented.
Tests and documentation are included.
* bpo-26680: Updates the ACKS file.
* bpo-26680: NEWS entries for int, the numeric ABCs and Decimal.
Co-authored-by: Robert Smallshire <rob@sixty-north.com>
Use _PyType_HasFeature() in the _io module and in structseq
implementation. Replace PyType_HasFeature() opaque function call with
_PyType_HasFeature() inlined function.
The new API allows to efficiently send values into native generators
and coroutines avoiding use of StopIteration exceptions to signal
returns.
ceval loop now uses this method instead of the old "private"
_PyGen_Send C API. This translates to 1.6x increased performance
of 'await' calls in micro-benchmarks.
Aside from CPython core improvements, this new API will also allow
Cython to generate more efficient code, benefiting high-performance
IO libraries like uvloop.
* When the parameters argument is a list, correctly handle the case
of changing it during iteration.
* When the parameters argument is a custom sequence, no longer
override an exception raised in ``__len__()``.
I just realized that my recent PR with sendfile on Solaris ([PR 22040](https://github.com/python/cpython/pull/22040)) has broken error handling.
Sorry for that, this simple followup fixes that.
Automerge-Triggered-By: @1st1
siginterrupt is deprecated:
./Modules/signalmodule.c:667:5: warning: ‘siginterrupt’ is deprecated: Use sigaction with SA_RESTART instead [-Wdeprecated-declarations]
667 | if (siginterrupt(signalnum, flag)<0) {
* Fix refleak in C module __init_subclass__
This was leaking a reference to the weak cache dictionary for every
ZoneInfo subclass created.
* Fix refleak in ZoneInfo subclass's clear_cache
The previous version of the code accidentally cleared the global
ZONEINFO_STRONG_CACHE variable (and inducing `ZoneInfo` to create a new
strong cache) on calls to a subclass's `clear_cache()`. This would not
affect guaranteed behavior, but it's still not the right thing to do
(and it caused reference leaks).
[bpo-31122](): ssl.wrap_socket() now raises ssl.SSLEOFError rather than OSError when peer closes connection during TLS negotiation
Reproducer: http://tiny.cc/f4ztnz (tiny url because some bot keeps renaming b.p.o.-nnn as bpo links)
Prior to this change, attempting to subclass the C implementation of
zoneinfo.ZoneInfo gave the following error:
TypeError: unbound method ZoneInfo.__init_subclass__() needs an argument
https://bugs.python.org/issue41025
* Merge gen and frame state variables into one.
* Replace stack pointer with depth in PyFrameObject. Makes code easier to read and saves a word of memory.
Add an accessor under SSLContext.security_level as a wrapper around
SSL_CTX_get_security_level, see:
https://www.openssl.org/docs/manmaster/man3/SSL_CTX_get_security_level.html
------
This is my first time contributing, so please pull me up on all the things I missed or did incorrectly.
Automerge-Triggered-By: @tiran
Move definition of UNUSED from modified headers of libmpdec to
_decimal.c itself. This makes the vendored source closer to the
standalone library and fixes build with --with-system-libmpdec.
Tested to build fine with either system libmpdec or the vendored one.
The running loop holder cache variable was always set to NULL when
calling set_running_loop.
Now set_running_loop saves the newly created running loop holder in the
cache variable for faster access in get_running_loop.
Automerge-Triggered-By: @1st1
Each interpreter now has its own dict free list:
* Move dict free lists into PyInterpreterState.
* Move PyDict_MAXFREELIST define to pycore_interp.h
* Add _Py_dict_state structure.
* Add tstate parameter to _PyDict_ClearFreeList() and _PyDict_Fini().
* In debug mode, ensure that the dict free lists are not used after
_PyDict_Fini() is called.
* Remove "#ifdef EXPERIMENTAL_ISOLATED_SUBINTERPRETERS".
There are a bunch of other fd: int uses in this file, I expect many if not
all of them would be better off using the fildes converter. This particular
one was flagged by Coverity as it presumably flags fpathconf as not accepting
negative fds. I'd expect the other fd's to have been flagged as well
otherwise.
I'm marking this one as skip news as it really is a no-op.
The PY_SSIZE_T_CLEAN macro must now be defined to use
PyArg_ParseTuple() and Py_BuildValue() "#" formats: "es#", "et#",
"s#", "u#", "y#", "z#", "U#" and "Z#". See the PEP 353.
Update _testcapi.test_buildvalue_issue38913().
The PyObject_INIT() and PyObject_INIT_VAR() macros become aliases to,
respectively, PyObject_Init() and PyObject_InitVar() functions.
Rename _PyObject_INIT() and _PyObject_INIT_VAR() static inline
functions to, respectively, _PyObject_Init() and _PyObject_InitVar(),
and move them to pycore_object.h. Remove their return value:
their return type becomes void.
The _datetime module is now built with the Py_BUILD_CORE_MODULE macro
defined.
Remove an outdated comment on _Py_tracemalloc_config.
On Windows, #include "pyerrors.h" no longer defines "snprintf" and
"vsnprintf" macros.
PyOS_snprintf() and PyOS_vsnprintf() should be used to get portable
behavior.
Replace snprintf() calls with PyOS_snprintf() and replace vsnprintf()
calls with PyOS_vsnprintf().
In GH-2866, _Py_Bit_Length() was added to pymath.h for lack of a better
location. GH-20518 added a more appropriate header file for bit utilities. It
also shows how to properly use intrinsics. This allows reconsidering bpo-29782.
* Move the function to the new header.
* Changed return type to match __builtin_clzl() and reviewed usage.
* Use intrinsics where available.
* Pick a fallback implementation suitable for inlining.
```
Direct leak of 8 byte(s) in 1 object(s) allocated from:
#0 0x7f008bf19667 in __interceptor_malloc (/lib64/libasan.so.6+0xb0667)
#1 0x7f007a0bee4a in subprocess_fork_exec /home/heimes/dev/python/cpython/Modules/_posixsubprocess.c:774
#2 0xe0305b in cfunction_call Objects/methodobject.c:546
```
Signed-off-by: Christian Heimes <christian@python.org>
This commit removes the old parser, the deprecated parser module, the old parser compatibility flags and environment variables and all associated support code and documentation.
Followup of bpo-40854, there is one remaining usage of PLATLIBDIR
which should be replaced by config->platlibdir.
test_sys checks that sys.platlibdir attribute exists and is a string.
Update Makefile: getpath.c and sysmodule.c no longer need PLATLIBDIR
macro, PyConfig.platlibdir member is used instead.
Co-authored-by: Sandro Mani <manisandro@gmail.com>
The PEP 353, written in 2005, introduced PY_FORMAT_SIZE_T. Python no
longer supports macOS 10.4 and Visual Studio 2010, but requires more
recent macOS and Visual Studio versions. In 2020 with Python 3.10, it
is now safe to use directly "%zu" to format size_t and "%zi" to
format Py_ssize_t.
* Rename pycore_byteswap.h to pycore_bitutils.h.
* Move popcount_digit() to pycore_bitutils.h as _Py_popcount32().
* _Py_popcount32() uses GCC and clang builtin function if available.
* Add unit tests to _Py_popcount32().
Each interpreter now has its own context free list:
* Move context free list into PyInterpreterState.
* Add _Py_context_state structure.
* Add tstate parameter to _PyContext_ClearFreeList()
and _PyContext_Fini().
* Pass tstate to clear_freelists().
Each interpreter now has its own asynchronous generator free lists:
* Move async gen free lists into PyInterpreterState.
* Move _PyAsyncGen_MAXFREELIST define to pycore_interp.h
* Add _Py_async_gen_state structure.
* Add tstate parameter to _PyAsyncGen_ClearFreeLists
and _PyAsyncGen_Fini().
Each interpreter now has its own list free list:
* Move list numfree and free_list into PyInterpreterState.
* Add _Py_list_state structure.
* Add tstate parameter to _PyList_ClearFreeList()
and _PyList_Fini().
* Remove "#ifdef EXPERIMENTAL_ISOLATED_SUBINTERPRETERS".
* _PyGC_Fini() clears gcstate->garbage list which can be stored in
the list free list. Call _PyGC_Fini() before _PyList_Fini() to
prevent leaking this list.
Each interpreter now has its own frame free list:
* Move frame free list into PyInterpreterState.
* Add _Py_frame_state structure.
* Add tstate parameter to _PyFrame_ClearFreeList()
and _PyFrame_Fini().
* Remove "#if PyFrame_MAXFREELIST > 0".
* Remove "#ifdef EXPERIMENTAL_ISOLATED_SUBINTERPRETERS".
Each interpreter now has its own float free list:
* Move tuple numfree and free_list into PyInterpreterState.
* Add _Py_float_state structure.
* Add tstate parameter to _PyFloat_ClearFreeList()
and _PyFloat_Fini().
Each interpreter now has its own tuple free lists:
* Move tuple numfree and free_list arrays into PyInterpreterState.
* Define PyTuple_MAXSAVESIZE and PyTuple_MAXFREELIST macros in
pycore_interp.h.
* Add _Py_tuple_state structure. Pass it explicitly to tuple_alloc().
* Add tstate parameter to _PyTuple_ClearFreeList()
* Each interpreter now has its own empty tuple singleton.
my_fgets() now calls _PyOS_InterruptOccurred(tstate) to check for
pending signals, rather calling PyOS_InterruptOccurred().
my_fgets() is called with the GIL released, whereas
PyOS_InterruptOccurred() must be called with the GIL held.
test_repl: use text=True and avoid SuppressCrashReport in
test_multiline_string_parsing().
Fix my_fgets() on Windows: fgets(fp) does crash if fileno(fp) is closed.
PyOS_AfterFork_Child() helper functions now return a PyStatus:
PyOS_AfterFork_Child() is now responsible to handle errors.
* Move _PySignal_AfterFork() to the internal C API
* Add #ifdef HAVE_FORK on _PyGILState_Reinit(), _PySignal_AfterFork()
and _PyInterpreterState_DeleteExceptMain().
Fix :mod:`ssl`` code to be compatible with OpenSSL 1.1.x builds that use
``no-deprecated`` and ``--api=1.1.0``.
Note: Tests assume full OpenSSL API and fail with limited API.
Signed-off-by: Christian Heimes <christian@python.org>
Co-authored-by: Mark Wright <gienah@gentoo.org>
Recent changes to _datetimemodule broke compilation on mingw; see the comments in this change for details.
FWIW, @corona10: this issue is why `PyType_FromModuleAndSpec` & friends take the `bases` argument at run time.
On macOS, socket.getaddrinfo() no longer uses an internal lock to
prevent race conditions when calling getaddrinfo(). getaddrinfo is
thread-safe is macOS 10.5, whereas Python 3.9 requires macOS 10.6 or
newer.
The lock was also used on FreeBSD older than 5.3, OpenBSD older than
201311 and NetBSD older than 4.
Previously, the result could have been an instance of a subclass of int.
Also revert bpo-26202 and make attributes start, stop and step of the range
object having exact type int.
Add private function _PyNumber_Index() which preserves the old behavior
of PyNumber_Index() for performance to use it in the conversion functions
like PyLong_AsLong().
If ctypes fails to convert the result of a callback or if a ctypes
callback function raises an exception, sys.unraisablehook is now
called with an exception set. Previously, the error was logged into
stderr by PyErr_Print().
```
D:\a\cpython\cpython\Modules\_zoneinfo.c(903,52): warning C4267: '=': conversion from 'size_t' to 'unsigned int', possible loss of data [D:\a\cpython\cpython\PCbuild\_zoneinfo.vcxproj]
D:\a\cpython\cpython\Modules\_zoneinfo.c(904,44): warning C4267: '=': conversion from 'size_t' to 'unsigned int', possible loss of data [D:\a\cpython\cpython\PCbuild\_zoneinfo.vcxproj]
D:\a\cpython\cpython\Modules\_zoneinfo.c(1772,31): warning C4244: '=': conversion from 'ssize_t' to 'uint8_t', possible loss of data [D:\a\cpython\cpython\PCbuild\_zoneinfo.vcxproj]
```
hashlib.compare_digest uses OpenSSL's CRYPTO_memcmp() function
when OpenSSL is available.
Note: The _operator module is a builtin module. I don't want to add
libcrypto dependency to libpython. Therefore I duplicated the wrapper
function and added a copy to _hashopenssl.c.
ctypes now raises an ArgumentError when a callback
is invoked with more than 1024 arguments.
The ctypes module allocates arguments on the stack in
ctypes_callproc() using alloca(), which is problematic
when large numbers of arguments are passed. Instead
of a stack overflow, this commit raises an ArgumentError
if more than 1024 parameters are passed.
Convert Py_REFCNT() and Py_SIZE() macros to static inline functions.
They cannot be used as l-value anymore: use Py_SET_REFCNT() and
Py_SET_SIZE() to set an object reference count and size.
Replace &Py_SIZE(self) with &((PyVarObject*)self)->ob_size
in arraymodule.c.
This change is backward incompatible on purpose, to prepare the C API
for an opaque PyObject structure.
The scripts in `Tools/peg_generator/scripts` mostly assume that
`ast.parse` and `compile` use the old parser, since this was the
state of things, while we were developing them. They need to be
updated to always use the correct parser. `_peg_parser` is being
extended to support both parsing and compiling with both parsers.
This updates _PyErr_ChainStackItem() to use _PyErr_SetObject()
instead of _PyErr_ChainExceptions(). This prevents a hang in
certain circumstances because _PyErr_SetObject() performs checks
to prevent cycles in the exception context chain while
_PyErr_ChainExceptions() doesn't.
The reset_peak function sets the peak memory size to the current size,
representing a resetting of that metric. This allows for recording the
peak of specific sections of code, ignoring other code that may have
had a higher peak (since the most recent `tracemalloc.start()` or
tracemalloc.clear_traces()` call).
Ifdef is not necessary, as AF_INET6 is supported from Windows Vista, and other code in overlapped.c uses AF_INET6 and is not ifdef'd.
Change the raised exception so users are not fooled to think it comes from Windows API.
Automerge-Triggered-By: @njsmith
When an asyncio.Task is cancelled, the exception traceback now
starts with where the task was first interrupted. Previously,
the traceback only had "depth one."
The internal module ``_hashlib`` wraps and exposes OpenSSL's HMAC API. The
new code will be used in Python 3.10 after the internal implementation
details of the pure Python HMAC module are no longer part of the public API.
The code is based on a patch by Petr Viktorin for RHEL and Python 3.6.
Co-Authored-By: Petr Viktorin <encukou@gmail.com>
{date, datetime}.isocalendar() now return a private custom named tuple object
IsoCalendarDate rather than a simple tuple.
In order to leave IsocalendarDate as a private class and to improve what
backwards compatibility is offered for pickling the result of a
datetime.isocalendar() call, add a __reduce__ method to the named tuples that
reduces them to plain tuples. (This is the part of this PR most likely to cause
problems — if it causes major issues, switching to a strucseq or equivalent
would be prudent).
The pure python implementation of IsoCalendarDate uses positional-only
arguments, since it is private and only constructed by position anyway; the
equivalent change in the argument clinic on the C side would require us to move
the forward declaration of the type above the clinic import for whatever
reason, so it seems preferable to hold off on that for now.
bpo-24416: https://bugs.python.org/issue24416
Original PR by Dong-hee Na with only minor alterations by Paul Ganssle.
Co-authored-by: Dong-hee Na <donghee.na92@gmail.com>
On AIX, time.thread_time() is now implemented with thread_cputime()
which has nanosecond resolution, rather than
clock_gettime(CLOCK_THREAD_CPUTIME_ID) which has a resolution of 10 ms.
This is the initial implementation of PEP 615, the zoneinfo module,
ported from the standalone reference implementation (see
https://www.python.org/dev/peps/pep-0615/#reference-implementation for a
link, which has a more detailed commit history).
This includes (hopefully) all functional elements described in the PEP,
but documentation is found in a separate PR. This includes:
1. A pure python implementation of the ZoneInfo class
2. A C accelerated implementation of the ZoneInfo class
3. Tests with 100% branch coverage for the Python code (though C code
coverage is less than 100%).
4. A compile-time configuration option on Linux (though not on Windows)
Differences from the reference implementation:
- The module is arranged slightly differently: the accelerated module is
`_zoneinfo` rather than `zoneinfo._czoneinfo`, which also necessitates
some changes in the test support function. (Suggested by Victor
Stinner and Steve Dower.)
- The tests are arranged slightly differently and do not include the
property tests. The tests live at test/test_zoneinfo/test_zoneinfo.py
rather than test/test_zoneinfo.py or test/test_zoneinfo/__init__.py
because we may do some refactoring in the future that would likely
require this separation anyway; we may:
- include the property tests
- automatically run all the tests against both pure Python and C,
rather than manually constructing C and Python test classes (similar
to the way this works with test_datetime.py, which generates C
and Python test cases from datetimetester.py).
- This includes a compile-time configuration option on Linux (though not
on Windows); added with much help from Thomas Wouters.
- Integration into the CPython build system is obviously different from
building a standalone zoneinfo module wheel.
- This includes configuration to install the tzdata package as part of
CI, though only on the coverage jobs. Introducing a PyPI dependency as
part of the CI build was controversial, and this is seen as less of a
major change, since the coverage jobs already depend on pip and PyPI.
Additional changes that were introduced as part of this PR, most / all of
which were backported to the reference implementation:
- Fixed reference and memory leaks
With much debugging help from Pablo Galindo
- Added smoke tests ensuring that the C and Python modules are built
The import machinery can be somewhat fragile, and the "seamlessly falls
back to pure Python" nature of this module makes it so that a problem
building the C extension or a failure to import the pure Python version
might easily go unnoticed.
- Adjustments to zoneinfo.__dir__
Suggested by Petr Viktorin.
- Slight refactorings as suggested by Steve Dower.
- Removed unnecessary if check on std_abbr
Discovered this because of a missing line in branch coverage.
OpenSSL can be build without support for TLS 1.0 and 1.1. The ssl module
now correctly adheres to OPENSSL_NO_TLS1 and OPENSSL_NO_TLS1_1 flags.
Also update multissltest to test with latest OpenSSL and LibreSSL
releases.
Signed-off-by: Christian Heimes <christian@python.org>
Automerge-Triggered-By: @tiran
The ``ssl`` and ``hashlib`` modules now actively check that OpenSSL is
build with thread support. Python 3.7.0 made thread support mandatory and no
longer works safely with a no-thread builds.
Signed-off-by: Christian Heimes <christian@python.org>
OpenSSL 3.0.0-alpha2 was released today. The FIPS_mode() function has
been deprecated and removed. It no longer makes sense with the new
provider and context system in OpenSSL 3.0.0.
EVP_default_properties_is_fips_enabled() is good enough for our needs in
unit tests. It's an internal API, too.
Signed-off-by: Christian Heimes <christian@python.org>
Change spelling of a #define in _tkinter.c from HAVE_LIBTOMMAMTH to HAVE_LIBTOMMATH, since this is used to keep track of tclTomMath.h, not tclTomMamth.h. No other file seems to refer to this variable.
Pass PEP 573 defining_class to os.DirEntry methods. The module state
is now retrieve from defining_class rather than Py_TYPE(self), to
support subclasses (even if DirEntry doesn't support subclasses yet).
* Pass the module rather than defining_class to DirEntry_fetch_stat().
* Only get the module state once in _posix_clear(),
_posix_traverse() and _posixmodule_exec().
Move PyInterpreterState.fs_codec into a new
PyInterpreterState.unicode structure.
Give a name to the fs_codec structure and use this structure in
unicodeobject.c.
_Py_hashtable_t values become regular "void *" pointers.
* Add _Py_hashtable_entry_t.data member
* Remove _Py_hashtable_t.data_size member
* Remove _Py_hashtable_t.get_func member. It is no longer needed
to specialize _Py_hashtable_get() for a specific value size, since
all entries now have the same size (void*).
* Remove the following macros:
* _Py_HASHTABLE_GET()
* _Py_HASHTABLE_SET()
* _Py_HASHTABLE_SET_NODATA()
* _Py_HASHTABLE_POP()
* Rename _Py_hashtable_pop() to _Py_hashtable_steal()
* _Py_hashtable_foreach() callback now gets key and value rather than
entry.
* Remove _Py_hashtable_value_destroy_func type. value_destroy_func
callback now only has a single parameter: data (void*).
Rewrite _tracemalloc to store "trace_t*" rather than directly
"trace_t" in traces hash tables. Traces are now allocated on the heap
memory, outside the hash table.
Add tracemalloc_copy_traces() and tracemalloc_copy_domains() helper
functions.
Remove _Py_hashtable_copy() function since there is no API to copy a
key or a value.
Remove also _Py_hashtable_delete() function which was commented.
Rewrite _Py_hashtable_t type to always store the key as
a "const void *" pointer. Add an explicit "key" member to
_Py_hashtable_entry_t.
Remove _Py_hashtable_t.key_size member.
hash and compare functions drop their hash table parameter, and their
'key' parameter type becomes "const void *".
Rewrite how the _tracemalloc module stores traces of other domains.
Rather than storing the domain inside the key, it now uses a new hash
table with the domain as the key, and the data is a per-domain traces
hash table.
* Add tracemalloc_domain hash table.
* Remove _Py_tracemalloc_config.use_domain.
* Remove pointer_t and related functions.
* Move Modules/hashtable.h to Include/internal/pycore_hashtable.h
* Move Modules/hashtable.c to Python/hashtable.c
* Python is now linked to hashtable.c. _tracemalloc is no longer
linked to hashtable.c. Previously, marshal.c got hashtable.c via
_tracemalloc.c which is built as a builtin module.
Convert posixmodule.c ("posix" or "nt" module) to the multiphase
initialization (PEP 489).
* Create the module using PyModuleDef_Init().
* Create ScandirIteratorType and DirEntryType with the new
PyType_FromModuleAndSpec() (PEP 573)
* Get the module state from ScandirIteratorType and DirEntryType with
the new PyType_GetModule() (PEP 573)
* Pass module to functions which access the module state.
* convert_sched_param() gets a new module parameter. It is now called
directly since Argument Clinic doesn't support passing the module
to an argument converter callback.
* Remove _posixstate_global macro.
(Note: PEP 554 is not accepted and the implementation in the code base is a private one for use in the test suite.)
If code running in a subinterpreter raises an uncaught exception then the "run" call in the calling interpreter fails. A RunFailedError is raised there that summarizes the original exception as a string. The actual exception type, __cause__, __context__, state, etc. are all discarded. This turned out to be functionally insufficient in practice. There is a more helpful solution (and PEP 554 has been updated appropriately).
This change adds the exception propagation behavior described in PEP 554 to the _xxsubinterpreters module. With this change a copy of the original exception is set to __cause__ on the RunFailedError. For now we are using "pickle", which preserves the exception's state. We also preserve the original __cause__, __context__, and __traceback__ (since "pickle" does not preserve those).
https://bugs.python.org/issue32604
Module C state is now accessible from C-defined heap type methods (PEP 573).
Patch by Marcel Plch and Petr Viktorin.
Co-authored-by: Marcel Plch <mplch@redhat.com>
Co-authored-by: Victor Stinner <vstinner@python.org>
When Python is built with experimental isolated interpreters, a
garbage collection now does nothing in an isolated interpreter.
Temporary workaround until subinterpreters stop sharing Python
objects.
This commit makes both APIs more consistent by doing the following:
- Remove the `PyPegen_CodeObjectFrom*` functions, which weren't used
and will probably not be needed. Functions like `Py_CompileStringObject`
can be used instead.
- Include a `const char *filename` parameter in `PyPegen_ASTFromString`.
- Rename `PyPegen_ASTFromFile` to `PyPegen_ASTFromFilename`, because
its signature is not the same with `PyParser_ASTFromFile`.
An isolated subinterpreter cannot spawn threads, spawn a child
process or call os.fork().
* Add private _Py_NewInterpreter(isolated_subinterpreter) function.
* Add isolated=True keyword-only parameter to
_xxsubinterpreters.create().
* Allow again os.fork() in "non-isolated" subinterpreters.
Now that the default parser is the new PEG parser, ast.parse uses it, which means that we don't actually test something in test_peg_parser. This commit introduces a new keyword argument (`oldparser`) for `_peg_parser.parse_string` for specifying that a string needs to be parsed with the old parser. This keyword argument is used in the tests to actually compare the ASTs the new parser generates with those generated by the old parser.
Remove _random.Random.randbytes(): the C implementation of
randbytes(). Implement the method in Python to ease subclassing:
randbytes() now directly reuses getrandbits().
test.pythoninfo logs OpenSSL FIPS_mode() and Linux
/proc/sys/crypto/fips_enabled in a new "fips" section.
Co-Authored-By: Petr Viktorin <encukou@gmail.com>
New PyFrame_GetBack() function: get the frame next outer frame.
Replace frame->f_back with PyFrame_GetBack(frame) in most code but
frameobject.c, ceval.c and genobject.c.
Remove the following function from the C API:
* PyAsyncGen_ClearFreeLists()
* PyContext_ClearFreeList()
* PyDict_ClearFreeList()
* PyFloat_ClearFreeList()
* PyFrame_ClearFreeList()
* PyList_ClearFreeList()
* PySet_ClearFreeList()
* PyTuple_ClearFreeList()
Make these functions private, move them to the internal C API and
change their return type to void.
Call explicitly PyGC_Collect() to free all free lists.
Note: PySet_ClearFreeList() did nothing.
This allows the caller to avoid creation of an exception when the channel is empty (just like `dict.get()` works). `ChannelEmptyError` is still raised if no default is provided.
Automerge-Triggered-By: @ericsnowcurrently
PyFrame_GetCode(frame): return a borrowed reference to the frame
code.
Replace frame->f_code with PyFrame_GetCode(frame) in most code,
except in frameobject.c, genobject.c and ceval.c.
Also add PyFrame_GetLineNumber() to the limited C API.
This commit also allows to pass flags to the new parser in all interfaces and fixes a bug in the parser generator that was causing to inline rules with actions, making them disappear.
Optimize the subprocess module on FreeBSD using closefrom().
A single close(fd) syscall is cheap, but when sysconf(_SC_OPEN_MAX)
is high, the loop calling close(fd) on each file descriptor can take
several milliseconds.
The workaround on FreeBSD to improve performance was to load and
mount the fdescfs kernel module, but this is not enabled by default.
Initial patch by Ed Maste (emaste), Conrad Meyer (cem), Kyle Evans
(kevans) and Kubilay Kocak (koobs):
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=242274
On FreeBSD, os.closerange(fd_low, fd_high) now calls
closefrom(fd_low) if fd_high is greater than or equal to
sysconf(_SC_OPEN_MAX).
Initial patch by Ed Maste (emaste), Conrad Meyer (cem), Kyle Evans
(kevans) and Kubilay Kocak (koobs):
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=242274
Fix the Windows implementation of os.waitpid() for exit code
larger than "INT_MAX >> 8". The exit status is now interpreted as an
unsigned number.
os.waitstatus_to_exitcode() now accepts wait status larger than
INT_MAX.
* Replace PY_INT64_T with int64_t
* Replace PY_UINT32_T with uint32_t
* Replace PY_UINT64_T with uint64_t
sha3module.c no longer checks if PY_UINT64_T is defined since it's
always defined and uint64_t is always available on platforms
supported by Python.
Add random.randbytes() function and random.Random.randbytes()
method to generate random bytes.
Modify secrets.token_bytes() to use SystemRandom.randbytes()
rather than calling directly os.urandom().
Rename also genrand_int32() to genrand_uint32(), since it returns an
unsigned 32-bit integer, not a signed integer.
The _random module is now built with Py_BUILD_CORE_MODULE defined.
Add a new internal pycore_byteswap.h header file with the following
functions:
* _Py_bswap16()
* _Py_bswap32()
* _Py_bswap64()
Use these functions in _ctypes, sha256 and sha512 modules,
and also use in the UTF-32 encoder.
sha256, sha512 and _ctypes modules are now built with the internal
C API.
Rename _PyInterpreterState_GET_UNSAFE() to _PyInterpreterState_GET()
for consistency with _PyThreadState_GET() and to have a shorter name
(help to fit into 80 columns).
Add also "assert(tstate != NULL);" to the function.
Move the PyGC_Head structure and the following private macros to the
internal C API:
* _PyGCHead_FINALIZED()
* _PyGCHead_NEXT()
* _PyGCHead_PREV()
* _PyGCHead_SET_FINALIZED()
* _PyGCHead_SET_NEXT()
* _PyGCHead_SET_PREV()
* _PyGC_FINALIZED()
* _PyGC_PREV_MASK
* _PyGC_PREV_MASK_COLLECTING
* _PyGC_PREV_MASK_FINALIZED
* _PyGC_PREV_SHIFT
* _PyGC_SET_FINALIZED()
* _PyObject_GC_IS_TRACKED()
* _PyObject_GC_MAY_BE_TRACKED()
* _Py_AS_GC(o)
Keep the private _PyGC_FINALIZED() macro in the public C API for
backward compatibility with Python 3.8: make it an alias to the new
PyObject_GC_IsFinalized() function.
Move the SIZEOF_PYGC_HEAD constant from _testcapi module to
_testinternalcapi module.
Don't access PyInterpreterState.config member directly anymore, but
use new functions:
* _PyInterpreterState_GetConfig()
* _PyInterpreterState_SetConfig()
* _Py_GetConfig()
Add the functions PyObject_GC_IsTracked and PyObject_GC_IsFinalized to the public API to allow to query if Python objects are being currently tracked or have been already finalized by the garbage collector respectively.
Fix the signal handler: it now always uses the main interpreter,
rather than trying to get the current Python thread state.
The following function now accepts an interpreter, instead of a
Python thread state:
* _PyEval_SignalReceived()
* _Py_ThreadCanHandleSignals()
* _PyEval_AddPendingCall()
* COMPUTE_EVAL_BREAKER()
* SET_GIL_DROP_REQUEST(), RESET_GIL_DROP_REQUEST()
* SIGNAL_PENDING_CALLS(), UNSIGNAL_PENDING_CALLS()
* SIGNAL_PENDING_SIGNALS(), UNSIGNAL_PENDING_SIGNALS()
* SIGNAL_ASYNC_EXC(), UNSIGNAL_ASYNC_EXC()
Py_AddPendingCall() now uses the main interpreter if it fails to the
current Python thread state.
Convert _PyThreadState_GET() and PyInterpreterState_GET_UNSAFE()
macros to static inline functions.
The PyObject_NEW() macro becomes an alias to the PyObject_New()
macro, and the PyObject_NEW_VAR() macro becomes an alias to the
PyObject_NewVar() macro, to hide implementation details. They no
longer access directly the PyTypeObject.tp_basicsize member.
Exclude _PyObject_SIZE() and _PyObject_VAR_SIZE() macros from
the limited C API.
Replace PyObject_NEW() with PyObject_New() and replace
PyObject_NEW_VAR() with PyObject_NewVar().
Add a private _at_fork_reinit() method to _thread.Lock,
_thread.RLock, threading.RLock and threading.Condition classes:
reinitialize the lock after fork in the child process; reset the lock
to the unlocked state.
Rename also the private _reset_internal_locks() method of
threading.Event to _at_fork_reinit().
* Add _PyThread_at_fork_reinit() private function. It is excluded
from the limited C API.
* threading.Thread._reset_internal_locks() now calls
_at_fork_reinit() on self._tstate_lock rather than creating a new
Python lock object.
This implements things like `list[int]`,
which returns an object of type `types.GenericAlias`.
This object mostly acts as a proxy for `list`,
but has attributes `__origin__` and `__args__`
that allow recovering the parts (with values `list` and `(int,)`.
There is also an approximate notion of type variables;
e.g. `list[T]` has a `__parameters__` attribute equal to `(T,)`.
Type variables are objects of type `typing.TypeVar`.
Convert the PyObject_GET_WEAKREFS_LISTPTR() macro to a function to
hide implementation details: the macro accessed directly to the
PyTypeObject.tp_weaklistoffset member.
Add _PyObject_GET_WEAKREFS_LISTPTR() static inline function to the
internal C API.
Add os.waitstatus_to_exitcode() function to convert a wait status to an
exitcode.
Suggest waitstatus_to_exitcode() usage in the documentation when
appropriate.
Use waitstatus_to_exitcode() in:
* multiprocessing, os, subprocess and _bootsubprocess modules;
* test.support.wait_process();
* setup.py: run_command();
* and many tests.
Changes on 7dd549eb08 made _functools compatible with
PEP-489 and we could have multiple modules instances loaded.
But, right now there is no way to make `kwd_mark` global into
a per module instance variable. kwd_mark is used on lru_cache_new
which does not have a reference to a PyModule*, necessary to use
PyModule_GetState.
PEP-573 will solve this problem and will allow us to move the global
state to per-module data and properly clear the state when unloading
a module instance.
This change temporarily disable cleaning of kwd_mark to avoid NULL
pointer dereference if we clear kwd_mark and other module instances
still alive use it.
Replace statically allocated types with heap allocated types:
use PyType_FromSpec().
Add a module state to store the _abc_data_type.
Add traverse, clear and free functions to the module.
Replace statically allocated types with heap allocated types:
use PyType_FromSpec().
Add a module state to store the Scanner and Encoder types.
Add traverse, clear and free functions to the module.
_tracemalloc.c and _xxsubinterpretersmodule.c use
PyThreadState_GetFrame() and PyThreadState_GetInterpreter() to no
longer depend on the PyThreadState structure.
Victor Stinner