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.

Use year 1990 instead of year 1970 to avoid issues with negative timestamps.

Use year 1990 instead of year 1970 to avoid issues with negative timestamps.

os.listdir() now returns the original path in OSError.filename (on any
platform), even if "*.*" was added to the path (on Windows).

os.open() uses _wopen() which sets errno, not the Windows error code.

Issue #16350: Fix zlib decompressor handling of unused_data with multiple calls to decompress() after EOF.

Patch by Serhiy Storchaka.

Issue #16350: Fix zlib decompressor handling of unused_data with multiple calls to decompress() after EOF.

Patch by Serhiy Storchaka.

Issue #16350: Fix zlib decompressor handling of unused_data with multiple calls to decompress() after EOF.

Patch by Serhiy Storchaka.

#5057: the peepholer no longer optimizes subscription on unicode literals (e.g. u"foo"[0]) in order to produce compatible pyc files between narrow and wide builds.

#8271: the utf-8 decoder now outputs the correct number of U+FFFD  characters when used with the "replace" error handler on invalid utf-8 sequences.  Patch by Serhiy Storchaka, tests by Ezio Melotti.