The characters in Actual results got converted into html entities in this bug tracking system. The script has output actual characters, not entities.

First of all, don't use utf8_encode... does it work then?

No, it does not, it outputs some garbled characters :(

Ditch those mb_* calls  and remove that setlocale() call too. Then replace utf8_encode() with utf8_decode()..

Jani, tried it with no luck again, this was the output:
http://obrazok.eu/files/bg05v7lyfto1i883czzs.png
I have tried many more things, such as iconv and mb_convert_encoding but nothing helped. And I cannot replace those mangled characters by the desirable ones since zip_entry_name() returns the same mangled character for different national characters.

Please post a link to the zip archive you use for your tests as well as an image to show us how the text should be displayed (if it is using non ascii chars :).

You can find the zip file here:
http://kotuha.com/file/zWio7-aaa.html
There is a xls file inside with the filename ??軾???????????????? ??ȫ?????????????????.xls and it should display the xls filename exactly like it is (with national characters).
With the code originally provided it looks like this:
http://obrazok.eu/files/6hhmttlsmiwezc8gmo6m.png
All non-national characters (if there were any in the xls filename) are retained, national characters get garbled.

Here is a screen shot:

http://pierre.libgd.org/zip/46456.png

the console output uses the script below, the GUI is Winrar (same with Winzip or windows compressed folder).

<?php
$somezipfile = '46456.zip';
$zip = zip_open(realpath($somezipfile));
while ($zip_entry = zip_read($zip))
{
	echo zip_entry_name($zip_entry);
}
zip_close($zip);


There is no difference, no bug > bogus. As Derick said earlier, you are messing with the encoding and try to encode something that it is already UTF-8.

I tried the same script as pajoye sent here. Only added a second line with original filename for comparison:

<?php
$somezipfile = '46456.zip';
$zip = zip_open(realpath($somezipfile));
while ($zip_entry = zip_read($zip))
{
	echo zip_entry_name($zip_entry);
}
zip_close($zip);
?>
<br>
&#318;?&#269;&#357;???????&#328;??&#283;&#271;&#345;&#314;??? &#317;?&#268;&#356;???????&#327;??&#282;&#270;&#344;&#313;???.xls

Be sure to save it as ANSI (so we are not messing with utf-8 now). What it outputs is: http://obrazok.eu/files/79gro4t39dzzzr3mk6u7.png
The strings should be the same. I do not think what it outputs is okay.

My god, the string got converted to html entities again :-/

> Be sure to save it as ANSI (so we are not messing with utf-8 now).

Thing is Zip does not convert anything, it gives you what the zip entry contains. If the console or your html page are not using the same encoding, then it is likely to be displayed badly. 

Can you tell me what encoding does the zip format use? Maybe I could use iconv then... it must be possible somehow since the national chars are all okay when reopening the zip file in winrar.

Zip supports UTF-8 if some specific flag is set. But 99.999% of the time, the strings are stored as is, without any conversion (by many tools).