Original patches for code and tests by Guilherme Polo and
Cheryl Sabella, respectively.

* Maintain a list of BufferedWriter objects.  Flush them on exit.

In Python 3, the buffer and the underlying file object are separate
and so the order in which objects are finalized matters.  This is
unlike Python 2 where the file and buffer were a single object and
finalization was done for both at the same time.  In Python 3, if
the file is finalized and closed before the buffer then the data in
the buffer is lost.

This change adds a doubly linked list of open file buffers.  An atexit
hook ensures they are flushed before proceeding with interpreter
shutdown.  This is addition does not remove the need to properly close
files as there are other reasons why buffered data could get lost during
finalization.

Initial patch by Armin Rigo.

* Use weakref.WeakSet instead of WeakKeyDictionary.

* Simplify buffered double-linked list types.

* In _flush_all_writers(), suppress errors from flush().

* Remove NEWS entry, use blurb.

* Take more care when flushing file buffers from atexit.

The previous implementation was not careful enough to avoid
causing issues in multi-threaded cases.  Check for buf->ok
and buf->finalizing before actually doing the flush.  Also,
increase the refcnt to ensure the object does not disappear.

Use prefixed macro names for the `authoraddress` function, add T2A to the font encoding in LaTeX sources to support Cyrillic characters in the PDF documentation, and replace the deprecated `font_size` config option with `pointsize`.

* Give pdb.set_trace() an optional `header` argument

* What's new.

* Give pdb.set_trace() an optional `header` argument

* What's new.

Affected classes are bytearray, array, deque, defaultdict, count and repeat.

Previously ensurepip would always report success, even if the
pip installation failed.

Fix a memory corruption in getpath.c due to mixed memory allocators
between Py_GetPath() and Py_SetPath().

The fix use the Raw allocator to mimic the windows version.

This patch should be used from python3.6 to the current version

for more details, see the bug report and
  https://github.com/pyinstaller/pyinstaller/issues/2812

* bpo-31536: Avoid wholesale rebuild after `make regen-all`

* Add NEWS

This fixes a regression from Python 2.  To get optional subparsers,
use the new parameter ``add_subparsers(required=False)``.

Patch by Anthony Sottile.

This is based on
https://github.com/ghaering/pysqlite/commit/40b349cadbd87c42f70fc92e5e1aee6d02564c6d#diff-0489411409cd2934730e88bf7767790,
though we can be a bit more aggressive about deleting code.

Don't export the cell_set_contents() symbol in the C API.

This one line of Serhiy Storchacka's bpo-31500 patch for is needed for other issues.

* bpo-31479: Always reset the signal alarm in tests

Use "try: ... finally: signal.signal(0)" pattern to make sure that
tests don't "leak" a pending fatal signal alarm.

* Move two more alarm() calls into the try block

Fix also typo: replace signal.signal(0) with signal.alarm(0)

* Move another signal.alarm() into the try block

bpo-31293: Fix crashes in truediv and mul of a timedelta by a float with a bad as_integer_ratio() method. (#3227)

bpo-31315: Fix an assertion failure in imp.create_dynamic(), when spec.name is not a string. (#3257)

* Trivial cleanups following bpo-31370

* Also cleanup the "importlib._bootstrap_external" module

* bpo-31499, xml.etree: Fix xmlparser_gc_clear() crash

xml.etree: xmlparser_gc_clear() now sets self.parser to NULL to prevent a
crash in xmlparser_dealloc() if xmlparser_gc_clear() was called previously
by the garbage collector, because the parser was part of a reference cycle.
Co-Authored-By: Serhiy Storchaka <storchaka@gmail.com>

This reverses a never-released regression resulting from bpo-31287.

This function returns the last component of tp_name after a dot.
Returns tp_name itself if it doesn't contain a dot.

bpo-31482: Missing bytes support for random.seed() version 1 #3614