The `allow_abbrev` option for ArgumentParser is documented and intended to disable support for unique prefixes of --options, which may sometimes be ambiguous due to deferred parsing.

However, the initial implementation also broke parsing of grouped short flags, such as `-ab` meaning `-a -b` (or `-a=b`).  Checking the argument for a leading `--` before rejecting it fixes this.

This was prompted by pytest-dev/pytest#5469, so a backport to at least 3.8 would be great 😄  
And this is my first PR to CPython, so please let me know if I've missed anything!


https://bugs.python.org/issue26967

Hi,

I've faced an issue w/ `mailbox.Maildir()`. The case is following:
1. I create a folder with `tempfile.TemporaryDirectory()`, so it's empty
2. I pass that folder path as an argument when instantiating `mailbox.Maildir()`
3. Then I receive an exception happening because "there's no such file or directory" (namely `cur`, `tmp` or `new`) during interaction with Maildir

**Expected result:** subdirs are created during `Maildir()` instance creation.

**Actual result:** subdirs are assumed as existing which leads to exceptions during use.

**Workaround:** remove the actual dir before passing the path to `Maildir()`. It will be created automatically with all subdirs needed.

**Fix:** This PR. Basically it adds creation of subdirs regardless of whether the base dir existed before.


https://bugs.python.org/issue30088

https://bugs.python.org/issue37358

Returns NotImplemented for timedelta and time in __eq__ for different types in Python implementation, which matches the C implementation.

This also adds tests to enforce that these objects will fall back to the right hand side's __eq__ and/or __ne__ implementation.

bpo-37579

[bpo-37580](https://bugs.python.org/issue37580): Markup typo in http.cookiejar doc





https://bugs.python.org/issue37580

Fix importlib examples to insert any newly created modules via importlib.util.module_from_spec() immediately into sys.modules instead of after calling loader.exec_module().

Thanks to Benjamin Mintz for finding the bug.


https://bugs.python.org/issue37521

Move `threaded_import_hangers`, a dependency of `test_threaded_import`, to the directory `test_importlib/`. Also update the import references for `threaded_import_hangers` in `test_threaded_import`.





https://bugs.python.org/issue19696

The `arraymodule`'s `b_getitem` function returns a `PyLong` converted 
from `arrayobject`'s array, by dereferencing a pointer to `char`.

When the `char` type is `signed char`, the `if (x >= 128) x -= 256;` comparison/code is redundant because a `signed char` will have a value of `[-128, 127]` and so `x` will never be greater or equal than 128.

This check was indeed needed for situations where a given compiler would assume `char` being `unsigned char` which would make `x` in `[0, 256]` range.

However, the check can be removed if we cast the `ob_item` into a signed char pointer (`signed char*`) instead of `char*`.

This PR/commit introduces this change.

With the addition of shared memory into Python 3.8, we now have three tests failing on Solaris, namely `test_multiprocessing_fork`, `test_multiprocessing_forkserver` and `test_multiprocessing_spawn`. The reason seems to be incorrect name handling which results in two slashes being prepended.


https://bugs.python.org/issue37558

https://bugs.python.org/issue37554

uintptr_t is an integer type, and can't be compared to NULL directly.

Keeping an account of allocated blocks slows down _PyObject_Malloc()
and _PyObject_Free() by a measureable amount.  Have
_Py_GetAllocatedBlocks() iterate over the arenas to sum up the
allocated blocks for pymalloc.

After fcf1d003, IDLE startup failed with python compiled without docstrings.

test_ssl.test_pha_required_nocert() now uses
support.catch_threading_exception() to ignore the expected SSLError
in ConnectionHandler of ThreadedEchoServer (it is only raised
sometimes on Windows).

This reverts commit cf9c41c4.

Replace PyLong_FromLong() with PyLong_FromSize_t():
SSL_CTX_get_num_tickets() return type is size_t.


https://bugs.python.org/issue37120

Context manager catching threading.Thread exception using
threading.excepthook.

Close the TLS connection in test_pha_required_nocert() of test_ssl to
fix a ResourceWarning.

Nested BinOp instances (e.g. a+b+c) had a wrong col_offset for the
second BinOp (e.g. 2 instead of 0 in the example). Fix it by using the
correct st node to copy the line and col_offset from in ast.c.

Initial report by Michael Blankenship on docs@

test_concurrent_futures now explicitly stops the ForkServer instance
if it's running.