No changes to the actual library code.

Provide an example usage pattern for reference stealing functions.
This way the user (hopely) sees more clearly how the reference
stealing functions are meant to be used.

When encoding an array or object ends in an error, the visited flag
wasn't zeroed, causing subsequent encoding attempts to fail. This
patch fixes the problem by always zeroing the visited flag.

Encoding an empty array or object worked, but encoding it again
(possibly after adding some items) failed, because the visited flag
(used for detecting circular references) wasn't zeroed.

The AC_TYPE_INT32_T macro first appeared in autoconf 2.60.

Use AC_C_INLINE autoconf macro, include config.h where needed, and add
a define of JSON_INLINE to jansson.h that has the correct "inline"
keyword.

Thanks to Deron Meranda for providing the initial text.

Earlier it was a decoding error.

Initialize their reference counts to (unsigned int)-1 to disable
reference counting on them. It already was meant to work like this,
but the reference counts were just initialized to 1 instead of -1.

Thanks to Andrew Thompson for reporting this issue.

This should have fixed by commit 28682322, but there was one #include
left in utf.c. It now includes utf.h instead of stdint.h.

Use AC_TYPE_INT32_T and include inttypes.h (if it exists) instead of
stdint.h for maximum portability.

The valgrind fix a while back apparently made the test system not
notice normal failures in suites/api.

This is to better catch distribution errors. It's easier to notice
that run-tests fails than to notice that one of many test suites is
silently skipped.

The problem is that Sphinx can only read input files from a single
directory. In VPATH builds, the source and build trees are separate,
and the changes.rst went into the build tree.

This patch solves the issue by using cfunc as the Sphinx default role.

Conflicts:
	Makefile.am

CHANGES is preprocessed to convert json_*() function names to Sphinx
:cfunc: cross references. This is to keep CHANGES more readable in
both plain text and HTML.

* Python is no longer required to run the tests
* Mention pkg-config support
* Fix some errors

It was accidentally left out when the test system was refactored.

There are some places where we copy a string from an existing JSON
value. In these cases the string has already been checked for valid
UTF-8.

Thanks to Sean Middleditch for contributing.

Added functions are:

* json_string_nocheck()
* json_string_set_nocheck()
* json_object_set_nocheck()
* json_object_set_new_nocheck()

These functions don't check that their string argument is valid UTF-8,
but assume that the user has already performed the check.

It's incorrect as these functions modify the value.

The tests were broken by the "%.17g" fix that was merged from the 1.1
branch.