From https://github.com/pfalcon/berkeley-db-1.xx, which so far contains
pristine 1.85, but will get patches and compile warning fixes going
forward.

Berkeley DB 1.xx is BSD-licensed, and will form the basis of "btree"
simple database module.

This implements basic wrapping of native get/put/seq API, and then dictionary
access protocol. Native API is intended to be superceded going forward.

Docs are now by default rebuilt from scratch, as required to build
conditionalized (i.e. using only:: directive) docs across different
output types. We have pretty small docset, so that's still rather fast.
However, if that's a concern, incremental rebuilds can be used by
passing "FORCE=" (nothing after =) as a make parameter. This will work
when using the same output type (e.g. only "html").

Based on my experience, there's rather non-zero chance to have an image be
flashed incorrectly. As --verify option is now works well in teh latest
esptool.py, enable it by default.

There's a need for .exp file because CPython renders OrderedDict's as:

OrderedDict([('b', 2)])

while MicroPython as:

OrderedDict({'b': 2})

For modindex_exclude extension, per-port module excludes are also added.
With these changes, it's possible to generate docs for a particular port
devoid of any superfluous and unrelated content, including in indexes and
full-text search - with small caveat: when generating PDF docs after HTML,
or vice-versa cached internal doctree representation (build/*/doctrees/)
must be removed first.

Designed specifically to workaround issues we were facing with generating
multiple conditionalized output docsets from a single master doctree.
Extensions were factored out into a separate project, based on the fact
that many other Sphinx users experience similar or related problems:
https://github.com/pfalcon/sphinx_selective_exclude

Corresponds to the 182f4a8da57 upstream revision.

Seealso and Limitations sectiosn added, better formatting and grammar.

With currentmodule:: set properly, none are needed. Extra "machine." prefix
produces wrong indexing data.

Based on unix version. No descriptions so far.

Class designator will be used as is in indexes, so must match actual class
name.

Based on python-dev discussion regarding PEP467.

This makes rendered docs to not provide incorrect information and
consistent with how it's down for other machine classes.

Class designator will be used as is in indexes, so must match actual class
name.

With currentmodule:: support, all targets should be generated properly by
Sphinx.

Makes sure that classes described in these separate files are properly
designated as belonging to "pyb" module in indexes.

Makes sure that classes described in these separate files are properly
designated as belonging to "machine" module in indexes.

Assignments of the form "_id = const(value)" are treated as private
(following a similar CPython convention) and code is no longer emitted
for the assignment to a global variable.

See issue #2111.

Requires a link of the form: <BASEURL>/<lang>/<ver>/micropython-<port>.pdf

By design, at wake up from deepsleep, the RTC timer will be reset, but
the data stored in RTC memory will not [1]. Therefore, we have to adjust
delta in RTC memory before going into deepsleep to get almost correct
time after waking up.

[1] http://bbs.espressif.com/viewtopic.php?t=1184#p4082

Instead of calling strlen(), which won't work if there're 32 chars in
returned ESSID. struct bss_info::ssid_len is not documented in SDK API
Guide, but is present in SDK headers since 1.4.0. Just in case, previous
code is left commented.