PyObject_GetAttrString intern temporary key string.
It's completely redudant.

use "union" merge strategy for Misc/NEWS.

As of Python 3.6 **kwargs are ordered, thus, remove the paragraph stating that
ordering is undefined and change snippet to remove the unecessary sorted call.

* Add sentence mentioning guaranteed output order of kwargs

Clarify that methods take a string which is interpreted as a regex,
not a regex object.

Also clarify what the old `-1`, `0`, `1` and `2` options were.

* Fix incorrect handling of signed zeros for complex-related classes.

* Add Misc/NEWS entry.

lookdict_unicode() and lookdict_unicode_nodummy() may raise exception
when key is not unicode.

no longer affects a partial object after creation.

3eea8c67 missed dot at the end of paragraph.

* Fixed bpo-29565: Corrected ctypes passing of large structs by value.

Added code and test to check that when a structure passed by value
is large enough to need to be passed by reference, a copy of the
original structure is passed. The callee updates the passed-in value,
and the test verifies that the caller's copy is unchanged. A similar
change was also added to the test added for bpo-20160 (that test was
passing, but the changes should guard against regressions).

* Reverted unintended whitespace changes.

* Drop duplicate work 'object' in lzma docs
* Fix typo in os docs: fpr -> for

bpo-22807: Expose platform UUID generation safety information.

``local.getlocale(locale.LC_CTYPE)`` and
``locale.getpreferredencoding(False)`` may give different answers
in some cases (such as the ``en_IN`` locale).

``re.LOCALE`` uses the latter, so update the test case to match.

* bpo-29026: Clarity documentation of time.time

Clarify the documentation of time.time by more
precisely defining what is meant by "seconds since
the epoch" on most platforms. Additionally explain
how gmtime and localtime may be used to extract
calendar components and convert to a more common
date format.

* bpo-29026: Minor improvements for time.time doc

* bpo-29026: Consistency fixes for time.time doc

bltinmodule.c: Added in b744ba1d and no longer necessary since d64e8a75
posixmodule.c: Added in d1cd4d41 and no longer necessary since efb00c0c
pythonrun.c:   Added in 73d538b9 and no longer necessary since d6009517
sysmodule.c:   Added in 5467d4c0 and no longer necessary since a2c17c58

* Reformat Doc/README.txt to Doc/README.rst

* Update mention of Doc/README.rst

* Update mention of README.txt to README.rst

* Make line fold

* rstlint ignore Doc/README.rst

* conf.py ignore Doc/README.rst

* Update issue tracker url in Docs/README.rst

rstlint.py run by "make check" doesn't support Python 2.

"make venv" runs "$(PYTHON) -m venv", whereas Python 2 doens't
provide the venv module: it's a module of Python 3 standard library.

- Return to always building the docs, it's a relatively cheap operation,
  the check had a bug, and the docs build depends on other files.
- Add `make suspicious`
- Do all three checks as one `make check suspicious html` invocation for
  earliest possible exit in case of issues.

If someone pushes a branch to python/cpython and then creates a PR
it will cause Travis to run tests needlessly, once for the PR and
once for the push. This will limit the branches that Travis will
run tests for to the `master` branch and branches that match the regex
`^\d\.\d$`.

This will have the effect that if someone purposely makes another branch
they won't get tests to run, but in that rare case they can adjust this
themselves.

Without this flag, the coverage test will hold up the status of the pull request.

IPPROTO_IPV6 is already defined further above in the same way.

"appears to not work in all cases" does not inspire confidence in this
module. I can find no context for what bug this was referencing so it
should be removed.