When compiling 3rd party C extensions, the linker flags used by the
compiler for the interpreter and the stdlib modules, will get
leaked into distutils. In order to avoid that, the PY_CORE_LDFLAGS
and PY_LDFLAGS_NODIST are introduced to keep those flags separated.
Fix a race condition in check_interrupted_write() of test_io:
create directly the thread with SIGALRM signal blocked,
rather than blocking the signal later from the thread. Previously, it
was possible that the thread gets the signal before the signal is
blocked.
Rename test.bisect module to test.bisect_cmd to avoid conflict with
bisect module when running directly a test like
"./python Lib/test/test_xmlrpc.py".
Make platform.architecture() parsing of "file" command output more
reliable:
* Add the "-b" option to the "file" command to omit the filename;
* Force the usage of the C locale;
* Search also the "shared object" pattern.
Co-Authored-By: Serhiy Storchaka <storchaka@gmail.com>
TextTestRunner of unittest.runner now uses time.perf_counter() rather
than time.time() to measure the execution time of a test: time.time()
can go backwards, whereas time.perf_counter() is monotonic.
Similar change made in libregrtest, pprint and random.
Replace time.time() with time.monotonic() in tests to measure time
delta.
test_zipfile64: display progress every minute (60 secs) rather than
every 5 minutes (5*60 seconds).
platform.dist() is deprecated and slated for removal in Python 3.8. The
upload command itself should also not be used to upload to PyPI, but
while it continues to exist it should not use deprecated functions.
"make profile-opt" no longer replaces CFLAGS_NODIST with CFLAGS. It
now adds profile-guided optimization (PGO) flags to CFLAGS_NODIST,
existing CFLAGS_NODIST flags are kept.
bpo-34279, bpo-35412: support.run_unittest() no longer raises
TestDidNotRun if a test result contains skipped tests. The
exception is now only raised if no test have been run and no test
have been skipped.
* Fix partial and partial method signatures in mock
* Add more calls
* Add NEWS entry
* Use assertEquals and fix markup in NEWS
* Refactor branching and add markup reference for functools
* Revert partial object related changes and fix pr comments
The length check for AF_ALG salg_name and salg_type had a off-by-one
error. The code assumed that both values are not necessarily NULL
terminated. However the Kernel code for alg_bind() ensures that the last
byte of both strings are NULL terminated.
Signed-off-by: Christian Heimes <christian@python.org>
Fix xml.dom.minidom cloneNode() on a document with an entity: pass
the correct arguments to the user data handler of an entity (fix an
old copy/paste mistake).
Bug spotted and fix proposed by Charalampos Stratakis, initial
reproducer written by Petr Viktorin.
Co-Authored-By: Charalampos Stratakis <cstratak@redhat.com>
Co-Authored-By: Petr Viktorin <encukou@gmail.com>
* tests: Further validate `wraps` functionality in `unittest.mock.Mock`
Add more tests to validate how `wraps` interacts with other features of
mocks.
* Don't call the wrapped object if `side_effect` is set
When a object is wrapped using `Mock(wraps=...)`, if an user sets a
`side_effect` in one of their methods, return the value of `side_effect`
and don't call the original object.
* Refactor what to be called on `mock_call`
When a `Mock` is called, it should return looking up in the following
order: `side_effect`, `return_value`, `wraps`. If any of the first two
return `mock.DEFAULT`, lookup in the next option.
It makes no sense to check for `wraps` returning default, as it is
supposed to be the original implementation and there is nothing to
fallback to.
Replace os.popen() with subprocess.check_output() in the platform module:
* platform.uname() (its _syscmd_ver() helper function) now redirects
stderr to DEVNULL.
* Remove platform.DEV_NULL.
* _syscmd_uname() and _syscmd_file() no longer catch AttributeError.
The "except AttributeError:" was only needed in Python 2, when
os.popen() was not always available. In Python 3,
subprocess.check_output() is always available.
select() calls are retried on EINTR (per PEP 475). However, if a
timeout was provided and the deadline has passed after running the
signal handlers, rlist, wlist and xlist should be cleared since select(2)
left them unmodified.
* posixpath.expanduser() now returns the input path unchanged if
the HOME environment variable is not set and pwd.getpwuid() raises
KeyError (the current user identifier doesn't exist in the password
database).
* Add test_no_home_directory() to test_site.
platform.libc_ver() now uses os.confstr('CS_GNU_LIBC_VERSION') if
available and the *executable* parameter is not set. The default
value of the libc_ver() *executable* parameter becomes None.
Quick benchmark on Fedora 29:
python3 -m perf command ./python -S -c 'import platform; platform.libc_ver()'
94.9 ms +- 4.3 ms -> 33.2 ms +- 1.4 ms: 2.86x faster (-65%)
When using link time optimizations, the -flto flag is passed to
BASECFLAGS, which makes it propagate to distutils. Those flags
should be reserved for the interpreter and the stdlib extension
modules only, thus moving those flags to CFLAGS_NODIST.
Modify asyncio tests to utilize the certificates from the test directory
instead of its own set, as they are the same and with each update they had
to be updated as well.
Fix an undefined behaviour in the pthread implementation of
PyThread_start_new_thread(): add a function wrapper to always return
NULL.
Add pythread_callback struct and pythread_wrapper() to thread_pthread.h.
Remove platform.popen() function, it was deprecated since Python 3.3:
use os.popen() instead.
Rename also the "Removed" section to "API and Feature Removals"
of What's New in Python 3.8.
Fix memory leak in PyUnicode_EncodeLocale() and
PyUnicode_EncodeFSDefault() on error handling.
Changes:
* Fix unicode_encode_locale() error handling
* Fix test_codecs.LocaleCodecTest
Fix mktime() overflow error in test_email: run
test_localtime_daylight_true_dst_true() and
test_localtime_daylight_false_dst_true() with a specific timezone.
The lineno and col_offset attributes of AST nodes for list comprehensions,
generator expressions and tuples are now point to the opening parenthesis or
square brace. For tuples without parenthesis they point to the position
of the first item.
Regression introduced in e3ce695 and 25b804a, where the old parameter
update_tryorder to _synthesize was first ignored, then given the opposite
value in the attempt to fix bpo-31014.
Fix str.format(), float.__format__() and complex.__format__() methods
for non-ASCII decimal point when using the "n" formatter.
Changes:
* Rewrite _PyUnicode_InsertThousandsGrouping(): it now requires
a _PyUnicodeWriter object for the buffer and a Python str object
for digits.
* Rename FILL() macro to unicode_fill(), convert it to static inline function,
add "assert(0 <= start);" and rework its code.
There are some same consts in a module. This commit merges them into
single instance. It reduces number of objects in memory after loading modules.
https://bugs.python.org/issue34100
Importing ProcessPoolExecutor may hang or cause an error when the import
accesses urandom on a low resource platform
https://bugs.python.org/issue29877
The "-I" command line option (run Python in isolated mode) is now
also copied by the multiprocessing and distutils modules when
spawning child processes. Previously, only -E and -s options (enabled
by -I) were copied.
subprocess._args_from_interpreter_flags() now copies the -I flag.
Include/*.h should be the "portable Python API", whereas
Include/cpython/*.h should be the "CPython API": CPython
implementation details.
Changes:
* Create Include/cpython/ subdirectory
* "make install" now creates $prefix/include/cpython and copy
Include/cpython/* to $prefix/include/cpython
* Create Include/cpython/objimpl.h: move objimpl.h code
surrounded by "#ifndef Py_LIMITED_API" to cpython/objimpl.h.
* objimpl.h now includes cpython/objimpl.h
* Windows installer (MSI) now also install Include/ subdirectories:
Include/cpython/ and Include/internal/.
locale.localeconv() now sets temporarily the LC_CTYPE locale to the
LC_MONETARY locale if the two locales are different and monetary
strings are non-ASCII. This temporary change affects other threads.
Changes:
* locale.localeconv() can now set LC_CTYPE to LC_MONETARY to decode
monetary fields.
* Add LocaleInfo.grouping_buffer: copy localeconv() grouping string
since it can be replaced anytime if a different thread calls
localeconv().
* _Py_GetLocaleconvNumeric() now requires a "struct lconv *"
structure, so locale.localeconv() now longer calls localeconv()
twice. Moreover, the function now requires all arguments to be
non-NULL.
* Rename STATIC_LOCALE_INFO_INIT to LocaleInfo_STATIC_INIT.
* Move _Py_GetLocaleconvNumeric() definition from fileutils.h
to pycore_fileutils.h. pycore_fileutils.h now includes locale.h.
* The _locale module is now built with Py_BUILD_CORE defined.
coro->cr_origin wasn't initialized if compute_cr_origin() failed in
PyCoro_New(), which would cause a crash during the coroutine's
deallocation.
https://bugs.python.org/issue35269
Discovered using clang's MemorySanitizer when it ran python3's
test_fstring test_misformed_unicode_character_name.
An msan build will fail by simply executing: ./python -c 'u"\N"'
The warnings module now suggests to enable tracemalloc if the source
is specified, tracemalloc module is available, but tracemalloc is not
tracing memory allocations.
Adds configure flags for msan and ubsan builds to make it easier to enable.
These also encode the detail that address sanitizer and memory sanitizer
should disable pymalloc.
Define MEMORY_SANITIZER when appropriate at build time and adds workarounds
to existing code to mark things as initialized where the sanitizer is otherwise unable to
determine that. This lets our build succeed under the memory sanitizer. not all tests
pass without sanitizer failures yet but we're in pretty good shape after this.
The System Preferences Dock "prefer tabs always" setting disables some
IDLE features. Menus are a bit different than as described for Windows
and Linux.
Fix an off by one error in the peephole optimizer when checking for unreachable code beyond a return.
Do a bounds check within find_op so it can return before going past the end as a safety measure.
7db3c48833 (diff-a33329ae6ae0bb295d742f0caf93c137)
introduced this off by one error while fixing another one nearby.
This bug was shipped in all Python 3.6 and 3.7 releases.
The included unittest won't fail unless you do a clang msan build.
The call to `_untrack_reader` is performed too soon, causing the protocol
to forget about the reader before `connection_lost` can run and feed the
EOF to the reader. See bpo-35065.
Current support for hash-based bytecode files in `zipimport` is rather
sparse, which leads to test failures when the test suite is ran with
the ``SOURCE_DATE_EPOCH`` environment variable set.
This teaches zipimport to handle hash-based pycs properly.
Handle Unicode contents on localised Windows systems when activating a
venv. activate.bat currently breaks on German Windows systems, as chcp.com does
not return a plain number as on English systems, but (arbitrarily) appends a dot at the end
(for example "Aktive Codepage: 850." instead of "Active Codepage: 850"). The
dependency to chcp.com is removed and ctypes is used to get, set and restore the
console output code page. The code page for console input is not changed.
We can't use __VENV_PYTHON__ to find python.exe, since it's UTF-8. cmd.exe decodes
the script using the console output code page.
Without setting mtime, time.time() will be used as the timestamp which will
end up in the compressed data and each invocation of the compress() function
will vary over time.
The new option in the CLI of the profile module allow to profile
executable modules. This change follows the same implementation as the
one already present in `cProfile`.
As the argument is now present on both modules, move the tests to the
common test case to be run with profile as well.
Some methods in the os module can accept path-like objects. This is documented in the general documentation but not in the function docstrings. To keep both in sync, the docstrings need to be updated to reflect that path-like objects are also accepted.
This implements getstate and setstate for the cjkcodecs multibyte incremental encoders/decoders, primarily to fix issues with seek/tell.
The encoder getstate/setstate is slightly tricky as the "state" is pending bytes + MultibyteCodec_State but only an integer can be returned. The approach I've taken is to encode this data into a long, similar to how .tell() encodes a "cookie_type" as a long.
https://bugs.python.org/issue33578
Modules imported last are now cleared first at interpreter shutdown.
A newly imported module is moved to the end of sys.modules, behind
modules on which it depends.
The list() constructor isn't taking full advantage of known input
lengths or length hints. This commit makes the constructor
pre-size and not over-allocate when the input size is known (the
input collection implements __len__). One on the main advantages is
that this provides 12% difference in memory savings due to the difference
between overallocating and allocating exactly the input size.
For efficiency purposes and to avoid a performance regression for small
generators and collections, the size of the input object is calculated using
__len__ and not __length_hint__, as the later is considerably slower.
There is only one trivial change to idle.rst. Nearly all the changes to help.html are the elimination of chapter and section numbers on headers due to changes in the build system. help.py no longer requires header numbering.
Since `SourceFileLoader.set_data()` catches exceptions raised by `_write_atomic()` and logs an informative message consequently, always logging successful outcome in 'SourceLoader.get_code()' seems redundant.
https://bugs.python.org/issue35024
Prior to this revision, after the shutdown of a `BaseServer`,
the server accepted a last single request
if it was sent between the server socket polling
and the polling timeout.
This can be problematic for instance for a server restart
for which you do not want to interrupt the service,
by not closing the listening socket during the restart.
One request failed because of this behavior.
Note that only one request failed,
following requests were not accepted, as expected.
Visual Studio solution: Set InlineFunctionExpansion to
OnlyExplicitInline ("/Ob1" option) on all projects (in
pyproject.props) in Debug mode on Win32 and x64 platforms to expand
functions marked as inline.
This change should make Python compiled in Debug mode a little bit
faster on Windows. On Unix, GCC uses -Og optimization level for
./configure --with-pydebug.
inspect.isfunction() processes both inspect.isfunction(func) and
inspect.isfunction(partial(func, arg)) correctly but some other functions in the
inspect module (iscoroutinefunction, isgeneratorfunction and isasyncgenfunction)
lack this functionality. This commits adds a new check in the mentioned functions
in the inspect module so they can work correctly with arbitrarily nested partial
functions.
The MagicMock class supports many magic methods, but not __fspath__. To ease
testing with modules such as os.path, this function is now supported by default.
tracemalloc now tries to update the traceback when an object is
reused from a "free list" (optimization for faster object creation,
used by the builtin list type for example).
Changes:
* Add _PyTraceMalloc_NewReference() function which tries to update
the Python traceback of a Python object.
* _Py_NewReference() now calls _PyTraceMalloc_NewReference().
* Add an unit test.
.o generated by clang in LTO mode actually are LLVM bitcode files, which
leads to a few errors during configure/build step:
- add lto flags to the BASECFLAGS instead of CFLAGS, as CFLAGS are used
to build autoconf test case, and some are not compatible with clang LTO
(they assume binary in the .o, not bitcode)
- force llvm-ar instead of ar, as ar is not aware of .o files generated
by clang -flto
Raise ValueError OverflowError in case of a negative
_length_ in a ctypes.Array subclass. Also raise TypeError
instead of AttributeError for non-integer _length_.
Co-authored-by: Oren Milman <orenmn@gmail.com>
For builtin types with builtin subclasses, help() on the type now shows up
to 4 of the subclasses. This partially replaces the exception hierarchy
information previously displayed in Python 2.7.
path_error() uses GetLastError() on Windows, but some os functions
are implemented via CRT APIs which report errors via errno.
This may result in raising OSError with invalid error code (such
as zero).
Introduce posix_path_error() function and use it where appropriate.
If buffering=1 is specified for open() in binary mode, it is silently
treated as buffering=-1 (i.e., the default buffer size).
Coupled with the fact that line buffering is always supported in Python 2,
such behavior caused several issues (e.g., bpo-10344, bpo-21332).
Warn that line buffering is not supported if open() is called with
binary mode and buffering=1.
The reprlib code was copied here instead of importing reprlib. I'm not sure if we really need to avoid the import, but since I expect dataclasses to be more common that reprlib, it seems wise. Plus, the code is small.
Adding `max_num_fields` to `cgi.FieldStorage` to make DOS attacks harder by
limiting the number of `MiniFieldStorage` objects created by `FieldStorage`.
Restores the use of pyexpatns.h to isolate our embedded copy of the expat C
library so that its symbols do not conflict at link or dynamic loading time
with an embedding application or other extension modules with their own
version of libexpat.
5dc3f23b5f (diff-3afaf7274c90ce1b7405f75ad825f545) inadvertently removed it when upgrading expat.
python-gdb.py now handles errors on computing the line number
of a Python frame.
Changes:
* PyFrameObjectPtr.current_line_num() now catchs any Exception on
calling addr2line(), instead of failing with a surprising "<class
'TypeError'> 'FakeRepr' object is not subscriptable" error.
* All callers of current_line_num() now handle current_line_num()
returning None.
* PyFrameObjectPtr.current_line() now also catchs IndexError on
getting a line from the Python source file.
Allow annotated global names in the module namespace after the symbol is
declared as global. Previously, only symbols annotated before they are declared
as global (i.e. inside a function) were allowed. This change allows symbols to be
declared as global before the annotation happens in the global scope.
Arthur Neufeld
Victor Stinner
Cheryl Sabella
stratakis
Chris Rands
Vladimir Matveev
Serhiy Storchaka
Zackery Spytz
Dima Tisnek
Paul Ganssle
Beomsoo Kim
Steve Dower
Xiang Zhang
Anthony Sottile
Xtreak
Andrew Svetlov
Ned Deily
Christian Heimes
jdemeyer
Mario Corchero
Tal Einat
Oran Avraham
William Grzybowski
Sergey Fedoseev
Andrew Dunai
Chris Withers
Ismo Toijala
Siddhesh Poyarekar
Pablo Galindo
INADA Naoki
Zhiming Wang
Lisa Roach
Dustin Spicuzza
nierob
Zhiming Wang
Brendan Gerrity
Donald Stufft
Stefano Rivera
Terry Jan Reedy
Julien Palard
Gregory P. Smith
Eddie Elizondo
Giampaolo Rodola
Srinivas Thatiparthy (శ్రీనివాస్ తాటిపర్తి)
Vincent Michel
Elvis Pranskevichus
samstagern
Diego Rojas
HongWeipeng
guoci
Alexander Buchkovsky
Rémi Lapeyre
Max Bélanger
Stéphane Wirtel
Lysandros Nikolaou
BNMetrics
Christopher Thorne
Petter Strandmark
Raymond Hettinger
Sebastian Rittau
Quentin Agren
Andrés Delfino
Denis Ledoux
Tim Graham
madman-bob
serge-sans-paille
Sanyam Khurana
Alexey Izbyshev
Matthias Bussonnier
David Herberth
Srinivas Thatiparthy (శ్రీనివాస్ తాటిపర్తి)
matthewbelisle-wf
Juliette Monsel