Python 3.8.0b1

* bpo-30835: email: Fix AttributeError when parsing invalid Content-Transfer-Encoding

Parsing an email containing a multipart Content-Type, along with a
Content-Transfer-Encoding containing an invalid (non-ASCII-decodable) byte
will fail. email.feedparser.FeedParser._parsegen() gets the header and
attempts to convert it to lowercase before comparing it with the accepted
encodings, but as the header contains an invalid byte, it's returned as a
Header object rather than a str.

Cast the Content-Transfer-Encoding header to a str to avoid this.

Found using the AFL fuzzer.
Reported-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: Andrew Donnellan <andrew@donnellan.id.au>

* Add email and NEWS entry for the bugfix.

* bpo-35805: Add parser for Message-ID header.

This parser is based on the definition of Identification Fields from RFC 5322
Sec 3.6.4.

This should also prevent folding of Message-ID header using RFC 2047 encoded
words and hence fix bpo-35805.

* Prevent folding of non-ascii message-id headers.
* Add fold method to MsgID token to prevent folding.

test_venv.test_mutiprocessing() now explicitly calls pool.terminate()
to wait until the pool completes.

* Improve example on tzinfo instances

Move from GMTX to TZX when naming the classes, as GMT1 might be rather
confusing as seen in the reported issue.
In addition, move to UTC over GMT and improve the tzname implementation.

* Simplify datetime with tzinfo example

Move the example in the documentation to just use timezone.utc and a
user defined Kabul timezone rather than having two user defined
timezones with DST.

Kabul timezone is still interesting as it changes its offset but not
based on DST. This is more accurate as the previous example was missing
information about the fold attribute. Additionally, implementing the fold
attribute was rather complex and probably not relevant enough for the
section "datetime with tzinfo".

CP65001Test has been removed.

Fix this MSVC warning:

    objects\codeobject.c(264): warning C4244: '=':
    conversion from 'Py_ssize_t' to 'unsigned char', possible loss of data

Make test_stream_server_close() implementation following test_stream_server_abort().
 Add explicit timeout for tests that can hang.

This reverts commit 0f0a30f4.

bpo-37148: Fix asyncio test that check for warning when running the test suite with huntleaks (GH-13800)

This reverts commit 396e0a8d.

Revert "bpo-35621: Support running subprocesses in asyncio when loop is executed in non-main thread (#13630)" (GH-13793)



https://bugs.python.org/issue35621

--with-pydebug is commonly used to find memory leaks.
But opcache makes it harder.
So disable opcache when Py_DEBUG is defined.

Add BaseEventLoop.wait_executor_on_close attribute: true by default.

loop.close() now waits for the default executor to finish by default.
Set loop.wait_executor_on_close attribute to False to not wait for
the executor.

Signed-off-by: Christian Heimes <christian@python.org>

What's new now mentions SSLContext.hostname_checks_common_name instead of SSLContext.host_flags.


https://bugs.python.org/issue36868

Fix various compatibility issues with LibreSSL and OpenSSL 1.0.2
introduced by bpo-34271.
Signed-off-by: Christian Heimes <christian@python.org>

Signed-off-by: Christian Heimes <christian@python.org>

Revert "bpo-33608: Factor out a private, per-interpreter _Py_AddPendingCall(). (gh-13714)" (GH-13780)

This reverts commit 6a150bca.

Replace asyncio.set_event_loop() with TestCase.set_event_loop() of
test_asyncio.utils: this method calls TestCase.close_loop() which
waits until the executor completes, to avoid leaking dangling
threads.

Inherit from test_asyncio.utils.TestCase rather than
unittest.TestCase.

* Credit myself and others.
* Complete asyncio changes.

This code was for deoptimization, which is removed from
PR-12884.

This patch implements per opcode cache mechanism, and use it in
only LOAD_GLOBAL opcode.

Based on Yury's opcache3.patch in bpo-26219.