[mythtv-users] Database update schema error with trunk utf8

Michael T. Dean mtdean at thirdcontact.com
Wed Dec 17 20:33:14 UTC 2008

On 12/17/2008 09:03 AM, Janne Grunau wrote:
> On Wednesday 17 December 2008 06:02:09 Michael T. Dean wrote:
>> The default character set for the server shouldn't have any effect on Myth
>> as the proper charset for the mythconverg database will be specified by
>> mc.sql when the database is created
> Unfortunately not. MythTV sents UTF-8 encoded data over a UTF-8 connection 
> but the database charset is latin-1. The server converts now the UTF-8 
> encoded data to latin-1 and stores it in the table.

Ah, right.  I didn't realize that if configured such that the server 
default character set is utf8, then /all/ clients would get UTF-8 
connections unless they specifically ask otherwise.  I had assumed that 
the database (here, mythconverg) default character set would mean 
clients get connections using the database default character set.  In 
fact, my assumption is true /only/ if there is no specified server 
default character set (as on my system).  So, the distro's specifying a 
server default character set has made a mess...

> So the database should hold plain latin-1 strings instead of utf-8 in 
> latin-1 and the conversion will fail.

Yep.  I can now reproduce the issue (the key was understanding the 
mysqldump for MySQL 5 and above, at least, always does UTF-8 format 
dumps--even if a server default character set is specified--unless 
specifically told otherwise).  So, when I restored the backup data 
freedenizen sent me with a UTF-8 connection, I got data exactly as it's 
stored in his DB and got the exact same failure he got.


More information about the mythtv-users mailing list