Dear Jan,

I've added something like
$string = preg_replace('!s:(\d+):"(.*?)";!se', 
"'s:'.strlen('$2').':\"$2\";'", $string);

before each occurence of unseralize in Horde's code. Now each 
serialized array gets correctly unserialized.

Wouldn't it be possible to add a helper function horde_unserialize as 
a wrapper for the php unserialize function which does sanitize the 
returned serialized arrays from the database tables? I'm wondering why 
no other people out there were running into this issue if they use 
opensuse.

Best regards,
Steffen

Please use the mailing list to ask for support.

http://www.horde.org/mail/ contains a list of all available mailing lists.

Our production horde system is running with PHP 5.2.13 with MySQL 
5.0.67 (OpenSUSE 11.1). The development system has PHP 5.3.3 and the 
MySQL on the DB server is 5.1.46 (both OpenSUSE 11.3). I've created a 
utf8 dumpfile of the old database with phpmyadmin, changed the CHARSET 
to utf8 and imported it on the new DB server.

If I now login to the development horde, the identities aren't 
displayed at all. The reason for this is the php unserialize function, 
which reports an error at offset... It complains about the string 
length of "Standardidentität", which was previously 17 and now with 
utf8 it is 18. In the manual of unserialize at php.net 
(http://www.php.net/manual/de/function.unserialize.php#75773) there 
are a couple of comments for this issue and I was able to fix to the 
wrong string length for the utf8 thing with php's preg_replace function.

But the identities are still not displayed for those users, which have 
a signature or some alias adresses, e.g. configured. There are several 
occurrences of "\r\n" in my dumpfilere. If I apply my preg_replace 
function on a serialized array returned by MySQL, the result differs 
from applying the function to the accordant line on my dumpfile. E.g. 
the strlen functions reports a string length of 495 for my signature 
fetched from the database, but according to my dumpfile, the string 
length of my signature is 527. This results again in an error at 
offset.. reported by unserialize.

Is there any documented way on how I could first migrate the db from 
latin1 to utf8 as well as having all these serialized arrays beeing 
transformed, so that they are still usable?