[wp-trac] Re: [WordPress Trac] #6532: db collation and db charset break serialize and unserialize on some sites

[WordPress Trac]

#6532: db collation and db charset break serialize and unserialize on some sites
--------------------------+-------------------------------------------------
 Reporter:  tandilboy     |        Owner:  anonymous
     Type:  defect (bug)  |       Status:  reopened 
 Priority:  normal        |    Milestone:  2.9      
Component:  General       |      Version:  2.5      
 Severity:  normal        |   Resolution:           
 Keywords:  dev-feedback  |  
--------------------------+-------------------------------------------------
Changes (by Denis-de-Bernardy):

  * keywords:  => dev-feedback

Comment:

 No no, it's charset-related all right. I've seen this happen often enough
 to know, and I ran into the "real" issue while investigating this in a
 separate app.

 unserialize() uses strlen() to determine the length of strings, and
 strlen() occasionally outputs erroneous string lengths when using
 multibyte characters. Trying to overload strlen() using the ini setup is
 hopeless to make it mb safe, because unserialize() will still use,
 internally, the non-mb safe version of strlen().

 If my memory serves me well the php devs have no intention to fix it. So I
 resolved (in the other app) to never serialize anything that may contain
 captions or user-editable text.

 This means that fixing the text widget (and all sorts of other plugins,
 widgets, meta options, etc.) would mean introducing separate options to
 store user-editable text, rather than storing them within an array.

-- 
Ticket URL: <http://trac.wordpress.org/ticket/6532#comment:11>
WordPress Trac <http://trac.wordpress.org/>
WordPress blogging software