from a Decimal argument.

Patch by Victor Stinner.

Patch by Victor Stinner.

Patch by Victor Stinner.

 * Remove unicode_widen(): replaced with _PyUnicodeWriter_Prepare()
 * Remove unicode_putchar(): replaced with
   PyUnicodeWriter_Prepare() + PyUnicode_WRITER()
 * When handling an decoding error, only overallocate the buffer by +25%
   instead of +100%

With the ANSI code page 932, os.listdir(b'\xe7') return an empty list (instead
of failing), whereas os.listdir(b'\xff') raises a FileNotFoundError.

It looks like a Windows bug: b'\xe7' directory does not exist,
FindFirstFileA(b'\xe7') fails with ERROR_FILE_NOT_FOUND (2), instead of
ERROR_PATH_NOT_FOUND (3).

It should increase the probability of finding a non-ASCII character on any
locale encoding.

 * support.TESTFN_UNDECODABLE was decodable if the filesystem encoding was
   cp932
 * test_genericpath.test_nonascii_abspath() didn't work on Windows if the
   path was not decodable (ex: with cp932)

These constants are used to test functions with non-ASCII data, especially
filenames.

If anyone finds another recursive C path that bypasses the recursion
limiting, they can add a new crasher example.

Patch by Taras Lyapun.