Sebastian Berg [Thu, 6 Feb 2020 14:54:05 +0000 (06:54 -0800)]
bpo-39274: Ensure Fraction.__bool__() returns a bool (GH-18017)
Some numerator types used (specifically NumPy) decides to not
return a Python boolean for the "a != b" operation. Using the equivalent
call to bool() guarantees a bool return also for such types.
Petr Viktorin [Thu, 6 Feb 2020 14:48:27 +0000 (15:48 +0100)]
bpo-39245: Make Vectorcall C API public (GH-17893)
* Add backcompat defines and move non-limited API declaration to cpython/
This partially reverts commit 2ff58a24e8a1c7e290d025d69ebaea0bbead3b8c
which added PyObject_CallNoArgs to the 3.9+ stable ABI. This should not
be done; there are enough other call APIs in the stable ABI to choose from.
* Adjust documentation
Mark all newly public functions as added in 3.9.
Add a note about the 3.8 provisional names.
Add notes on public API.
* Put PyObject_CallNoArgs back in the limited API
* Rename PyObject_FastCallDict to PyObject_VectorcallDict
Victor Stinner [Wed, 5 Feb 2020 12:12:19 +0000 (13:12 +0100)]
bpo-39542: Make PyObject_INIT() opaque in limited C API (GH-18363)
In the limited C API, PyObject_INIT() and PyObject_INIT_VAR() are now
defined as aliases to PyObject_Init() and PyObject_InitVar() to make
their implementation opaque. It avoids to leak implementation details
in the limited C API.
Exclude the following functions from the limited C API, move them
from object.h to cpython/object.h:
Victor Stinner [Wed, 5 Feb 2020 11:23:27 +0000 (12:23 +0100)]
bpo-39542: Exclude trashcan from the limited C API (GH-18362)
Exclude trashcan mechanism from the limited C API: it requires access to
PyTypeObject and PyThreadState structure fields, whereas these structures
are opaque in the limited C API.
The trashcan mechanism never worked with the limited C API. Move it
from object.h to cpython/object.h.
The macro is defined after Py_DECREF() and so is no longer used by
Py_DECREF().
Moving _Py_Dealloc() macro back from cpython/object.h to object.h
would require to move a lot of definitions as well: PyTypeObject and
many related types used by PyTypeObject.
Keep _Py_Dealloc() as an opaque function call to avoid leaking
implementation details in the limited C API (object.h): remove
_Py_Dealloc() macro from cpython/object.h.
Jakub Stasiak [Wed, 5 Feb 2020 01:10:19 +0000 (02:10 +0100)]
bpo-39491: Merge PEP 593 (typing.Annotated) support (#18260)
* bpo-39491: Merge PEP 593 (typing.Annotated) support
PEP 593 has been accepted some time ago. I got a green light for merging
this from Till, so I went ahead and combined the code contributed to
typing_extensions[1] and the documentation from the PEP 593 text[2].
My changes were limited to:
* removing code designed for typing_extensions to run on older Python
versions
* removing some irrelevant parts of the PEP text when copying it over as
documentation and otherwise changing few small bits to better serve
the purpose
* changing the get_type_hints signature to match reality (parameter
names)
I wasn't entirely sure how to go about crediting the authors but I used
my best judgment, let me know if something needs changing in this
regard.
Victor Stinner [Wed, 5 Feb 2020 00:11:10 +0000 (01:11 +0100)]
bpo-39542: Make _Py_NewReference() opaque in C API (GH-18346)
_Py_NewReference() becomes a regular opaque function, rather than a
static inline function in the C API (object.h), to better hide
implementation details.
Move _Py_tracemalloc_config from public pymem.h to internal
pycore_pymem.h header.
Philipp Gesang [Tue, 4 Feb 2020 21:25:16 +0000 (22:25 +0100)]
closes bpo-39510: Fix use-after-free in BufferedReader.readinto() (GH-18295)
When called on a closed object, readinto() segfaults on account
of a write to a freed buffer:
==220553== Process terminating with default action of signal 11 (SIGSEGV): dumping core
==220553== Access not within mapped region at address 0x2A
==220553== at 0x48408A0: memmove (vg_replace_strmem.c:1272)
==220553== by 0x58DB0C: _buffered_readinto_generic (bufferedio.c:972)
==220553== by 0x58DCBA: _io__Buffered_readinto_impl (bufferedio.c:1053)
==220553== by 0x58DCBA: _io__Buffered_readinto (bufferedio.c.h:253)
Victor Stinner [Tue, 4 Feb 2020 12:42:13 +0000 (13:42 +0100)]
Restore PyObject_IsInstance() comment (GH-18345)
Restore PyObject_IsInstance() comment explaining why only tuples of
types are accepted, but not general sequence. Comment written by
Guido van Rossum in commit 03290ecbf1661c0192e6abdbe00ae163af461d77
which implements isinstance(x, (A, B, ...)). The comment was lost in
a PyObject_IsInstance() optimization:
commit ec569b794737be248671d0dfac11b664fc930eef.
Cleanup also the code. recursive_isinstance() is no longer recursive,
so rename it to object_isinstance(), whereas object_isinstance() is
recursive and so rename it to object_recursive_isinstance().
Eddie Elizondo [Tue, 4 Feb 2020 10:29:25 +0000 (02:29 -0800)]
bpo-38076 Clear the interpreter state only after clearing module globals (GH-18039)
Currently, during runtime destruction, `_PyImport_Cleanup` is clearing the interpreter state before clearing out the modules themselves. This leads to a segfault on modules that rely on the module state to clear themselves up.
For example, let's take the small snippet added in the issue by @DinoV :
```
import _struct
The module `_struct` uses the module state to run `pack`. Therefore, the module state has to be alive until after the module has been cleared out to successfully run `C.__del__`. This happens at line 606, when `_PyImport_Cleanup` calls `_PyModule_Clear`. In fact, the loop that calls `_PyModule_Clear` has in its comments:
> Now, if there are any modules left alive, clear their globals to minimize potential leaks. All C extension modules actually end up here, since they are kept alive in the interpreter state.
That means that we can't clear the module state (which is used by C Extensions) before we run that loop.
Moving `_PyInterpreterState_ClearModules` until after it, fixes the segfault in the code snippet.
Finally, this updates a test in `io` to correctly assert the error that it now throws (since it now finds the io module state). The test that uses this is: `test_create_at_shutdown_without_encoding`. Given this test is now working is a proof that the module state now stays alive even when `__del__` is called at module destruction time. Thus, I didn't add a new tests for this.
Victor Stinner [Mon, 3 Feb 2020 16:55:05 +0000 (17:55 +0100)]
bpo-39542: Simplify _Py_NewReference() (GH-18332)
* Remove _Py_INC_REFTOTAL and _Py_DEC_REFTOTAL macros: modify
directly _Py_RefTotal.
* _Py_ForgetReference() is no longer defined if the Py_TRACE_REFS
macro is not defined.
* Remove _Py_NewReference() implementation from object.c:
unify the two implementations in object.h inline function.
* Fix Py_TRACE_REFS build: _Py_INC_TPALLOCS() macro has been removed.
PyThreadState.on_delete is a callback used to notify Python when a
thread completes. _thread._set_sentinel() function creates a lock
which is released when the thread completes. It sets on_delete
callback to the internal release_sentinel() function. This lock is
known as Threading._tstate_lock in the threading module.
The release_sentinel() function uses the Python C API. The problem is
that on_delete is called late in the Python finalization, when the C
API is no longer fully working.
The PyThreadState_Clear() function now calls the
PyThreadState.on_delete callback. Previously, that happened in
PyThreadState_Delete().
The release_sentinel() function is now called when the C API is still
fully working.
Adds an additional assertion check based on a race condition for `test__xxsubinterpreters.DestroyTests.test_still_running` discovered in the bpo issue.
Victor Stinner [Thu, 30 Jan 2020 11:18:32 +0000 (12:18 +0100)]
bpo-38631: Avoid Py_FatalError() in unicodeobject.c (GH-18281)
Replace Py_FatalError() calls with _PyErr_WriteUnraisableMsg(),
_PyObject_ASSERT_FAILED_MSG() or Py_UNREACHABLE()
in unicode_dealloc() and unicode_release_interned().
Victor Stinner [Thu, 30 Jan 2020 08:02:14 +0000 (09:02 +0100)]
bpo-38631: Avoid Py_FatalError() in init_slotdefs() (GH-18263)
Rename init_slotdefs() to _PyTypes_InitSlotDefs() and add a return
value of type PyStatus. The function is now called exactly once from
_PyTypes_Init(). Replace calls to init_slotdefs() with an assertion
checking that slotdefs is initialized.
Victor Stinner [Thu, 30 Jan 2020 08:01:07 +0000 (09:01 +0100)]
bpo-38631: Replace Py_FatalError() with _PyObject_ASSERT_FAILED_MSG() (GH-18258)
Replace Py_FatalError() with _PyObject_ASSERT_FAILED_MSG() in
object.c and typeobject.c to also dump the involved Python object on
a fatal error. It should ease debug when such fatal error occurs.
If the double linked list is inconsistent, _Py_ForgetReference() no
longer dumps previous and next objects in the fatal error, it now
only dumps the current object. It ensures that the error message
is displayed even if dumping the object does crash Python.
Enhance _Py_ForgetReference() error messages;
_PyObject_ASSERT_FAILED_MSG() logs the "_Py_ForgetReference" function
name.
Chris Withers [Wed, 29 Jan 2020 16:24:54 +0000 (16:24 +0000)]
Get mock coverage back to 100% (GH-18228)
* use the `: pass` and `: yield` patterns for code that isn't expected to ever be executed.
* The _Call items passed to _AnyComparer are only ever of length two, so assert instead of if/else
* fix typo
* Fix bug, where stop-without-start patching dict blows up with `TypeError: 'NoneType' object is not iterable`, highlighted by lack of coverage of an except branch.
* The fix for bpo-37972 means _Call.count and _Call.index are no longer needed.
* add coverage for calling next() on a mock_open with readline.return_value set.
* __aiter__ is defined on the Mock so the one on _AsyncIterator is never called.
bpo-39485: fix corner-case in method-detection of mock (GH-18252)
Replace check for whether something is a method in the mock module. The
previous version fails on PyPy, because there no method wrappers exist
(everything looks like a regular Python-defined function). Thus the
isinstance(getattr(result, '__get__', None), MethodWrapperTypes) check
returns True for any descriptor, not just methods.
This condition could also return erroneously True in CPython for
C-defined descriptors.
Instead to decide whether something is a method, just check directly
whether it's a function defined on the class. This passes all tests on
CPython and fixes the bug on PyPy.
XFS filesystem is limited to 32-bit timestamp, but the utimensat()
syscall doesn't fail. Moreover, there is a VFS bug which returns
a cached timestamp which is different than the value on disk.
Joannah Nanjekye [Wed, 29 Jan 2020 11:20:53 +0000 (07:20 -0400)]
bpo-39153: Clarify C API *SetItem refcounting semantics (GH-18220)
Some of the *SetItem methods in the C API steal a reference to the
given value. This annotates the better behaved ones to assure the
reader that these are not the ones with the inconsistent behaviour.
Bruce Merry [Wed, 29 Jan 2020 07:09:24 +0000 (09:09 +0200)]
bpo-36051: Drop GIL during large bytes.join() (GH-17757)
Improve multi-threaded performance by dropping the GIL in the fast path
of bytes.join. To avoid increasing overhead for small joins, it is only
done if the output size exceeds a threshold.
Steve Dower [Wed, 29 Jan 2020 02:46:33 +0000 (13:46 +1100)]
bpo-39401: Avoid unsafe DLL load on Windows 7 and earlier (GH-18231)
As Windows 7 is not supported by Python 3.9, we just replace the dynamic load with a static import. Backports will have a different fix to ensure they continue to behave the same.
Christoph Reiter [Tue, 28 Jan 2020 09:41:50 +0000 (10:41 +0100)]
bpo-38883: Don't use POSIX `$HOME` in `pathlib.Path.home/expanduser` on Windows (GH-17961)
In bpo-36264 os.path.expanduser was changed to ignore HOME on Windows.
Path.expanduser/home still honored HOME despite being documented as behaving the same
as os.path.expanduser. This makes them also ignore HOME so that both implementations
behave the same way again.
bpo-36018: Minor fixes to the NormalDist() examples and recipes. (GH-18226)
* Change the source for the SAT data to a primary source.
* Fix typo in the standard deviation
* Clarify that the binomial probabalities are just for the Python room.
Terry Jan Reedy [Mon, 27 Jan 2020 23:41:18 +0000 (18:41 -0500)]
bpo-39392: Turtle overlap fill depends on OS (#18223)
Whether or not overlap regions for self-intersecting polygons
or multiple shapes are filled depends on the operating system graphics,
typeof overlap, and number of overlaps.
Victor Stinner [Mon, 27 Jan 2020 21:37:44 +0000 (22:37 +0100)]
bpo-38631: Avoid Py_FatalError() in _memory_release() (GH-18214)
If the export count is negative, _memory_release() now raises a
SystemError and returns -1, rather than calling Py_FatalError()
which aborts the process.
Victor Stinner [Mon, 27 Jan 2020 21:37:05 +0000 (22:37 +0100)]
bpo-38631: Avoid Py_FatalError() in PyModule_Create2() (GH-18212)
If PyModule_Create2() is called when the Python import machinery is
not initialized, it now raises a SystemError and returns NULL,
instead of calling Py_FatalError() which aborts the process.
The caller must be prepared to handle NULL anyway.