Description:
------------
On an NTFS-formatted drive, Windows XP stores every character of a file name in two bytes. Specifically, the encoding is UTF-16LE.

PHP's "readdir" has no problem reading NTFS file names that consist only of characters included in the ISO-8859-1 repertoire. For these characters, the first UTF-16LE byte is the same as the ISO-8859-1 representation, and the second byte is simply 0x00, the null byte, which causes no damage.

However, for two-byte characters at higher code points - Chinese characters, for example - "readdir" fails.

To illustrate, I created a directory with two files. One of the files is named with Chinese characters; the other is named with ISO-8859-1 compatible characters. The following code:

if ($dir = opendir('.\TestFiles')) {
  while (false !== ($file = readdir($dir))) {
    echo $file . '<br>';
  }
}

produces this output:

.
..
???????.m4a
Greensleeves.m4a

Clearly, "readdir" does not handle the two-byte Chinese characters correctly. Instead, it returns question marks. This is a big problem if one wants to use a PHP script to back up an NTFS-formatted drive.

(And, by the way, using mb_internal_encoding("UTF-16LE") does not solve the problem; this seems to have no impact on "readdir".)

Expected result:
----------------
"readdir" should be able to handle UTF-16 (i.e., two-byte) characters

Actual result:
--------------
"readdir" does not interpret two-byte characters correctly