PR #1638, for bpo-28411, causes problems in some (very) edge cases. Until that gets sorted out, we're reverting the merge. PR #3506, a fix on top of #1638, is also getting reverted.

The `blake2b` function does not take the `data` keyword argument.
The hex digest returned by sign was a string, whereas compare_digest expects bytes-like objects.
Typo fix: compare_digesty -> compare_digest

Update the link from Misc/NEWS to Misc/NEWS.d

Try to fix the "Warning -- threading_cleanup() failed to cleanup 1
threads" warning in test.lock_tests: wait a little bit longer to give
time to the threads to complete.

Warning seen on test_thread and test_importlib.

* Copy changes to typing from upstream repo
* Add NEWS entry

bpo-31418: Fix an assertion failure in PyErr_WriteUnraisable() in case of an exception with a bad __module__ attribute. (#3539)

support.threading_cleanup() waits for 1 second before emitting a
warning if there are threads running in the background. With this
change, it now emits the warning immediately, to be able to catch
bugs more easily.

Explicitly state that communicate() closes stdin after writing input to it.

Clear also self.server_thread attribute in TestTimeouts.tearDown().

Call POP3.close(), don't close close directly the sock attribute.

Explicitly clear the server attribute in test_ftplib and test_poplib
to prevent dangling thread.

socketserver.ThreadingMixIn now keeps a list of non-daemonic threads
to wait until all these threads complete in server_close().

Reenable test_logging skipped tests.

Fix SocketHandlerTest.tearDown(): close the socket handler before
stopping the server, so the server can join threads.

The ConfigSocketReceiver.serve_until_stopped() method from
logging.config.listen() now calls server_close() (of
socketserver.ThreadingTCPServer) rather than closing manually the
socket.

While this change has no effect yet, it will help to prevent dangling
threads once ThreadingTCPServer.server_close() will join spawned
threads (bpo-31233).

The last commit contained lines longer than 80 characters.

Make it clear that Ellipsis tokens also have type attribute set to token.OP.

ProcessPoolShutdownTest.test_del_shutdown() now closes the call queue
and joins its thread, to prevent leaking a dangling thread.

* test_thread.test_forkinthread() now waits until the thread completes.
* Check the status in the test method, not in the thread function
* Don't ignore RuntimeError anymore: since the commit
  346cbd35 (bpo-16500,
  os.register_at_fork(), os.fork() cannot fail anymore with
  RuntimeError.
* Replace 0.01 literal with a new POLL_SLEEP constant
* test_forkinthread(): test if os.fork() exists rather than testing
  the platform.

bpo-31428: Prevent raising a SystemError in case the memo arg of ElementTree.Element.__deepcopy__() isn't a dictionary. (#3512)

IDLE calls tcl/tk update in the background in order to make live
interaction and experimentatin with tkinter applications much easier.

As needed for the conversion of extensions to features.

PR #3397 introduced a large number of warnings to the Windows build. This patch fixes them.

* Drop warnoptions from PyInterpreterState.

* Drop xoptions from PyInterpreterState.

* Don't set warnoptions and _xoptions again.

* Decref after adding to sys.__dict__.

* Drop an unused macro.

* Check sys.xoptions *before* we delete it.

This fixes bpo-31407.

bpo-31416: Fix assertion failures in case of a bad warnings.filters or warnings.defaultaction. (#3496)

Patch by Oren Milman.

Adding to an int entry is not the same as deleting and inserting
because int('') will fail.

The timeout parameter was not used.

This undoes a853a8ba except for the pytime.c
parts. We want to continue to allow IEEE 754 doubles larger than FLT_MAX to be
rounded into finite floats. Tests were added to very this behavior.

This fixes an omission in the ttk conversion patch for this issue, hence no new news.
Patch by Cheryl Sabella.

About 10 IDLE features were implemented as supposedly optional
extensions.  Their different behavior could be confusing or worse for
users and not good for maintenance.  Hence the conversion.

The main difference for users is that user configurable key bindings
for builtin features are now handled uniformly.  Now, editing a binding
in a keyset only affects its value in the keyset.  All bindings are
defined together in the system-specific default keysets in config-
extensions.def.  All custom keysets are saved as a whole in config-
extension.cfg.  All take effect as soon as one clicks Apply or Ok.

The affected events are '<<force-open-completions>>', '<<expand-word>>',
'<<force-open-calltip>>', '<<flash-paren>>', '<<format-paragraph>>',
'<<run-module>>', '<<check-module>>', and '<<zoom-height>>'.  Any
(global) customizations made before 3.6.3 will not affect their keyset-
specific customization after 3.6.3. and vice versa.

Inital patch by Charles Wohlganger, revised by Terry Jan Reedy.

* Working draft without _source

* Re-use itemgetter() instances

* Speed-up calls to __new__() with a pre-bound tuple.__new__()

* Add note regarding string interning

* Remove unnecessary create function wrappers

* Minor sync-ups with PR-2736.  Mostly formatting and f-strings

* Bring-in qualname/__module fix-ups from PR-2736

* Formally remove the verbose flag and _source attribute

* Restore a test of potentially problematic field names

* Restore kwonly_args test but without the verbose option

* Adopt Inada's idea to reuse the docstrings for the itemgetters

* Neaten-up a bit

* Add news blurb

* Serhiy pointed-out the need for interning

* Jelle noticed as missing f on an f-string

* Add whatsnew entry for feature removal

* Accede to request for dict literals instead keyword arguments

* Leave the method.__module__ attribute pointing the actual location of the code

* Improve variable names and add a micro-optimization for an non-public helper function

* Simplify by in-lining reuse_itemgetter()

* Arrange steps in more logical order

* Save docstring in local cache instead of interning