Question

I cannot pass a UTF-8 filename to move_uploaded_file() as it gets converted bytewise, resulting in a faulty name in the file system. For example:

move_uploaded_file($_FILES['userfile']['tmp_name'], '\upload\é.jpg');

creates xa9.jpg in the upload directory.

While the Windows API supports UTF-16, passing such filename (e.g., iconv('UTF-8', 'UTF-16', 'é')) to move_uploaded_file() results in an error.

It would be reasonable to percent-encode all special characters, and I definitely should do the same with the URIs, according to RFC 3986. But when I use percent-encoded URIs, Apache gives a 404 error, as it decodes the URL and can’t find anything by that name.

For example: <img src="/upload/%C3%A9.jpg" /> gives the Apache error:

File does not exist: […]/upload/\xc3\xa9.jpg.

What would be the proper solution? If I rename the file in Windows (é.jpg), the encoded HTML URI (%C3%A9.jpg) works as expected.

---

Some info on the subject: http://www.rooftopsolutions.nl/blog/filesystem-encoding-and-php

Answer 1

Passing iconv('UTF-8', 'Windows-1250', $_FILES['userfile']['name']) to move_uploaded_file(), as opposed to using UTF-16, and saving the filename for HTML as rawurlencode($_FILES['userfile']['name']) works.

If this filename is stored in a database, any file request should refer to iconv('UTF-8', 'Windows-1250', rawurldecode($filename)).

I use Windows-1250 charater set as this is the default on my system.

Additional info on MSDN:

• Character Sets Used in File Names (See: Code Pages)

• File and Directory Names (Naming Conventions)