This PR adds the ability to enable the GIL if it was disabled at
interpreter startup, and modifies the multi-phase module initialization
path to enable the GIL when loading a module, unless that module's spec
includes a slot indicating it can run safely without the GIL.
PEP 703 called the constant for the slot `Py_mod_gil_not_used`; I went
with `Py_MOD_GIL_NOT_USED` for consistency with gh-104148.
A warning will be issued up to once per interpreter for the first
GIL-using module that is loaded. If `-v` is given, a shorter message
will be printed to stderr every time a GIL-using module is loaded
(including the first one that issues a warning).
Move the following private functions and structures to
pycore_modsupport.h internal C API:
* _PyArg_BadArgument()
* _PyArg_CheckPositional()
* _PyArg_NoKeywords()
* _PyArg_NoPositional()
* _PyArg_ParseStack()
* _PyArg_ParseStackAndKeywords()
* _PyArg_Parser structure
* _PyArg_UnpackKeywords()
* _PyArg_UnpackKeywordsWithVararg()
* _PyArg_UnpackStack()
* _Py_ANY_VARARGS()
Changes:
* Python/getargs.h now includes pycore_modsupport.h to export
functions.
* clinic.py now adds pycore_modsupport.h when one of these functions
is used.
* Add pycore_modsupport.h includes when a C extension uses one of
these functions.
* Define Py_BUILD_CORE_MODULE in C extensions which now include
directly or indirectly (via code generated by Argument Clinic)
pycore_modsupport.h:
* _csv
* _curses_panel
* _dbm
* _gdbm
* _multiprocessing.posixshmem
* _sqlite.row
* _statistics
* grp
* resource
* syslog
* _testcapi: bad_get() no longer uses METH_FASTCALL calling
convention but METH_VARARGS. Replace _PyArg_UnpackStack() with
PyArg_ParseTuple().
* _testcapi: add PYTESTCAPI_NEED_INTERNAL_API macro which is defined
by _testcapi sub-modules which need the internal C API
(pycore_modsupport.h): exceptions.c, float.c, vectorcall.c,
watchers.c.
* Remove Include/cpython/modsupport.h header file.
Include/modsupport.h no longer includes the removed header file.
* Fix mypy clinic.py
Argument Clinic now only includes pycore_gc.h if PyGC_Head is needed,
and only includes pycore_runtime.h if _Py_ID() is needed.
* Add 'condition' optional argument to Clinic.add_include().
* deprecate_keyword_use() includes pycore_runtime.h when using
the _PyID() function.
* Fix rendering of includes: comments start at the column 35.
* Mark PC/clinic/_wmimodule.cpp.h and
"Objects/stringlib/clinic/*.h.h" header files as generated in
.gitattributes.
Effects:
* 42 header files generated by AC no longer include the internal C
API, instead of 4 header files before. For example,
Modules/clinic/_abc.c.h no longer includes the internal C API.
* Fix _testclinic_depr.c.h: it now always includes pycore_runtime.h
to get _Py_ID().
Move these private functions to the internal C API (pycore_long.h):
* _PyLong_UnsignedInt_Converter()
* _PyLong_UnsignedLongLong_Converter()
* _PyLong_UnsignedLong_Converter()
* _PyLong_UnsignedShort_Converter()
Argument Clinic now emits #include "pycore_long.h" when these
functions are used.
Here we are doing no more than adding the value for Py_mod_multiple_interpreters and using it for stdlib modules. We will start checking for it in gh-104206 (once PyInterpreterState.ceval.own_gil is added in gh-104204).
We only statically initialize for core code and builtin modules. Extension modules still create
the tuple at runtime. We'll solve that part of interpreter isolation separately.
This change includes generated code. The non-generated changes are in:
* Tools/clinic/clinic.py
* Python/getargs.c
* Include/cpython/modsupport.h
* Makefile.pre.in (re-generate global strings after running clinic)
* very minor tweaks to Modules/_codecsmodule.c and Python/Python-tokenize.c
All other changes are generated code (clinic, global strings).
setup.py no longer defines Py_BUILD_CORE_MODULE. Instead every
module defines the macro before #include "Python.h" unless
Py_BUILD_CORE_BUILTIN is already defined.
Py_BUILD_CORE_BUILTIN is defined for every module that is built by
Modules/Setup.
The PR also simplifies Modules/Setup. Makefile and makesetup
already define Py_BUILD_CORE_BUILTIN and include Modules/internal
for us.
Signed-off-by: Christian Heimes <christian@python.org>
Move Include/pystrhex.h to Include/internal/pycore_strhex.h.
The header file only contains private functions.
The following C extensions are now built with Py_BUILD_CORE_MODULE
macro defined to get access to the internal C API:
* _blake2
* _hashopenssl
* _md5
* _sha1
* _sha3
* _ssl
* binascii
No longer use deprecated aliases to functions:
* Replace PyObject_MALLOC() with PyObject_Malloc()
* Replace PyObject_REALLOC() with PyObject_Realloc()
* Replace PyObject_FREE() with PyObject_Free()
* Replace PyObject_Del() with PyObject_Free()
* Replace PyObject_DEL() with PyObject_Free()
Update _asyncio, _bz2, _csv, _curses, _datetime,
_io, _operator, _pickle, _queue, blake2,
multibytecodec and overlapped C extension modules
to use PyModule_AddType().
The usedforsecurity keyword only argument added to the hash constructors is useful for FIPS builds and similar restrictive environment with non-technical requirements that legacy algorithms be forbidden by their implementations without being explicitly annotated as not being used for any security related purposes. Linux distros with FIPS support benefit from this being standard rather than making up their own way(s) to do it.
Contributed and Signed-off-by: Christian Heimes christian@python.org
* help(hashlib) didn't work because of incorrect module name in blake2b and
blake2s classes.
* Constructors blake2*(), sha3_*(), shake_*() and keccak_*() incorrectly
accepted keyword argument "string" for binary data, but documented as
accepting the "data" keyword argument. Now this parameter is positional-only.
* Keyword-only parameters in blake2b() and blake2s() were not documented as
keyword-only.
* Default value for some parameters of blake2b() and blake2s() was None,
which is not acceptable value.
* The length argument for shake_*.digest() was wrapped out to 32 bits.
* The argument for shake_128.digest() and shake_128.hexdigest() was not
positional-only as intended.
* TypeError messages for incorrect arguments in all constructors sha3_*(),
shake_*() and keccak_*() incorrectly referred to sha3_224.
Also made the following enhancements:
* More accurately specified input and result types for strings, bytes and
bytes-like objects.
* Unified positional parameter names for update() and constructors.
* Improved formatting.
All Blake2 params have to be encoded in little-endian byte order. For
the two multi-byte integer params, leaf_length and node_offset, that
means that assigning a native-endian integer to them appears to work on
little-endian platforms, but gives the wrong result on big-endian. The
current libb2 API doesn't make that very clear, and @sneves is working
on new API functions in the GH issue above. In the meantime, we can work
around the problem by explicitly assigning little-endian values to the
parameter block.
See https://github.com/BLAKE2/libb2/issues/12.
Replace occurence of nested comments in blake2 reference implementation
with preprocessor directive for disabling unused code.
`blake2s-load-xop.h` is conditionally pulled in only on chips with XOP
support, among others the AMD Bulldozer. The malformed comments in the
source file breaks the build of `hashlib`'s `_blake2` on GCC 6.3.0.
Official reference code on github uses `#if` so this change should be
uncontroversial.
Rework the code choosing BLAKE2 code paths from using the optimized
variant on all x86_64 machines to using it when SSSE3 or better
supported instructions sets are available.
Firstly, this solves the problem of using pure SSE2 code path on x86_64
machines. As reported in the bug, this code is slower than the reference
code on all tested x86_64 machines. Furthermore, on Athlon64 that lacks
SSSE3, it is even 2.5 times slower than the reference code! Checking
for SSSE3 therefore ensures that the optimized implementation will only
be used when it has a chance of performing better.
Secondly, this makes it possible to use SSSE3+ optimizations on 32-bit
x86 systems. This allows for even 2 times speed gain on modern 32-bit
x86 systems (tested in a 32-bit chroot).
Brett Simmers
Yuriy Chernyshov
Tomas R
Nikita Sobolev
Furkan Onder
Victor Stinner
Hugo van Kemenade
Eric Snow
Max Bachmann
Erlend E. Aasland
Christian Heimes
Serhiy Storchaka
Inada Naoki
Mohamed Koubaa
Dong-hee Na
Rémi Lapeyre
Jeroen Demeyer
David Carlier
Peter Eisentraut
Jack O'Connor
Lars Viklund
Michał Górny
Antoine Pitrou
Benjamin Peterson