Reported on docs@ by Keith Briggs.

Use time.monotonic instead of time.clock.

Use time.monotonic or time.time instead of time.clock.

failure when the test is ran the second time.

The root attribute was set in parent class in setUpClass and then
overridded in child class in tearDownClass.

failure when the test is ran the second time.

The root attribute was set in parent class in setUpClass and then
overridded in child class in tearDownClass.

I forgot to rename it in my previous refactoring of pytime.c.

From patch by Xavier de Gaye.

This avoids possible buffer overreads when int(), float(), compile(), exec()
and eval() are passed bytes-like objects. Similar code is removed from the
complex() constructor, where it was not reachable.

Patch by John Leitch, Serhiy Storchaka and Martin Panter.

compiling AST from Python objects.

The auth method tests simply weren't adequate because of the fact that
smtpd doesn't support authentication.  I borrowed some of Milan's
code for that from issue #21935 and added it to the smtplib tests.
Also discovered that the direct test for the 'auth' method wasn't actually
testing anything and fixed it.

The fix makes the new authobject mechanism work the way it is
documented...the problem was that wasn't checking for a 334 return code
if an initial-response was provided, which works fine for auth plain
and cram-md5, but not for auth login.

The mkstemp() and mkdtemp() signatures have already been updated, but the
higher-level functions still suggest that the default values are forced to
text strings. Also merged some paragraphs describing the "suffix" and
"prefix" parameters, and pointed out that mktemp() does not support the new
changes.

reliable.

Thanks Марк Коренберг.

_odict_add_new_node() fails.

Implementation by Antoine Pitrou.