the current test depended on integer sets being iterated on in a certain
fixed order. That order is different on PyPy (insertion based) and could
change in CPython in the future in theory. Make the test robust against
a different iteration order by sorting.
The "freeze" tool has been part of the repo for a long time. However, it hasn't had any tests in the test suite to guard against regressions. We add such a test here. This is especially important as there has been a lot of change recently related to frozen modules, with more to come.
Note that as part of the test we build Python out-of-tree and install it in a temp dir.
https://bugs.python.org/issue45629
The doctest module raised an error if a docstring contained an example that
attempted to access a classmethod property. (Stacking '@classmethod' on top of
`@property` has been supported since Python 3.9; see
https://docs.python.org/3/howto/descriptor.html#class-methods.)
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
Use types.GenericAlias in inspect.formatannotation to correctly add
type arguments of builtin types to the string representation of
Signatures.
Co-authored-by: Martin Rückl <martin.rueckl@codecentric.de>
(original patch by eryksun)
Correctly hand various failure modes when resizing an mmap on Windows:
* Resizing a pagefile-backed mmap now creates a new mmap and copies data
* Attempting to resize when another mapping is held on the same file raises an OSError
* Attempting to resize a nametagged mmap raises an OSError if another mapping is held with the same nametag
The recently added PyConfig.stdlib_dir was being set with ".." entries. When __file__ was added for from modules this caused a problem on out-of-tree builds. This PR fixes that by normalizing "stdlib_dir" when it is calculated in getpath.c.
https://bugs.python.org/issue45506
This is a cross-platform check that the symbols are actually
exported in the ABI, not e.g. hidden in a macro.
Caveat: PyModule_Create2 & PyModule_FromDefAndSpec2 are skipped.
These aren't exported on some of our buildbots. This is a bug
(bpo-44133). This test now makes sure all the others don't regress.
Freelists for object structs can now be disabled. A new ``configure``
option ``--without-freelists`` can be used to disable all freelists
except empty tuple singleton. Internal Py*_MAXFREELIST macros can now
be defined as 0 without causing compiler warnings and segfaults.
Signed-off-by: Christian Heimes <christian@python.org>
Rename namespace package __loader__ class to be public.
Make the old name, i.e. _NamespaceLoader, an alias for the public name, for backward compatibility.
An object implementing the os.PathLike protocol can represent a file
system path as a str or bytes object.
Therefore, _infer_return_type function should infer os.PathLike[str]
object as str type and os.PathLike[bytes] object as bytes type.
Since v8.6.11, a few configuration options seem to accept an empty value
where they did not previously; particularly the `type` of a `Menu`
widget, and the `compound` of any ttk widget with a label. Providing an
explicit expected error message to `checkEnumParam` bypasses the check
of an empty value, which no longer raises `TclError`.
Added non parallel-safe :func:`~contextlib.chdir` context manager to change
the current working directory and then restore it on exit. Simple wrapper
around :func:`~os.chdir`.
Signed-off-by: Filipe Laíns <lains@riseup.net>
Co-authored-by: Łukasz Langa <lukasz@langa.pl>
In Python 3.8 and 3.9, stacking `@functools.singledispatchmethod` on top of
`@classmethod` or `@staticmethod` caused an exception to be raised if the
method was registered using type-annotations rather than
`@method.register(int)`. This was not caught by unit tests, however, as the
tests only tested the `@method.register(int)` way of registering additional
implementations. The bug is no longer present in Python 3.10+, but
`test_functools.py` is still lacking regression tests for these cases. This
commit adds these test cases.
In gh-28954 I adjusted how test_embed determines if it should be skipped. That broke out-of-tree builds. This change fixes them.
https://bugs.python.org/issue45506
There are two errors that this commit fixes:
* The parser was not correctly computing the offset and the string
source for E_LINECONT errors due to the incorrect usage of strtok().
* The parser was not correctly unwinding the call stack when a tokenizer
exception happened in rules involving optionals ('?', [...]) as we
always make them return valid results by using the comma operator. We
need to check first if we don't have an error before continuing.
* Generate test classes at import time. It allows to filter them when
run with unittest. E.g: "./python -m unittest test.test_dbm.TestCase_gnu -v".
* Create a database class in a new directory which will be removed after
test. It guarantees that all created files and directories be removed
and will not conflict with other dbm tests.
* Restore dbm._defaultmod after tests. Previously it was set to the last
dbm module (dbm.dumb) which affected other tests.
* Enable the whichdb test for dbm.dumb.
* Move test_keys to the correct test class. It does not test whichdb().
* Remove some outdated code and comments.
I considered only falling back when both were 0, but that still seems
wrong, and the highly popular rich[1] library does it this way, so I
thought we should probably inherit that behavior.
[1] https://github.com/willmcgugan/rich
Signed-off-by: Filipe Laíns <lains@riseup.net>
Co-authored-by: Łukasz Langa <lukasz@langa.pl>
* Avoid making C calls for most calls to Python functions.
* Change initialize_locals(steal=true) and _PyTuple_FromArraySteal to consume the argument references regardless of whether they succeed or fail.
The default was "off". Switching it to "on" means users get the benefit of frozen stdlib modules without having to do anything. There's a special-case for running-in-source-tree, so contributors don't get surprised when their stdlib changes don't get used.
https://bugs.python.org/issue45020
