[mythtv-commits] Ticket #6013: restore script--fix partial restore to account for character set

MythTV mythtv at cvs.mythtv.org
Thu Dec 18 01:33:49 UTC 2008


#6013: restore script--fix partial restore to account for character set
----------------------------------------------+-----------------------------
 Reporter:  sphery <mtdean at thirdcontact.com>  |       Owner:  ijr 
     Type:  patch                             |      Status:  new 
 Priority:  minor                             |   Milestone:      
Component:  mythtv                            |     Version:  head
 Severity:  medium                            |     Mlocked:  0   
----------------------------------------------+-----------------------------
 The partial/"new host" restore needs to account for the character set used
 in the backup file to prevent data corruption.  The attached patch simply
 adds a new condition to the filter for a "white-listed" partial restore to
 ensure the line:

  /*!40101 SET NAMES utf8 */;

 is included, to properly set the character set for the restore.  It will
 work even if the backup uses a different character set--it doesn't have to
 be utf8--so it should work for any backup file (even one created by a user
 who explicily specifies a different character set).  Note that
 --partial_restore --with_plugin_data (a blacklisted restore) works before
 the patch as the SET NAMES line is not excluded.

 Since both this patch and the patch in #5975, update the script version in
 the restore script to 1.0.3, please either commit them together or give me
 time to update the other patch to get a new script version or update the
 script version when committing the other.  Thanks.

 Background information:

 mysqldump (from MySQL 5.0 and above, at least) always uses UTF-8 for the
 dump unless otherwise instructed.  Therefore, since a MySQL server may be
 configured with a server-default character set (which would cause client
 connections to use that character set) that differs from the character set
 used in the backup data could be corrupted (twice-converted).  Likewise,
 on a system without a specified server-default character set, the
 connection will use the database-default character set.  Therefore, on any
 0.21-fixes and below system that's properly configured for Myth (using
 either a server default charset of latin1 or without a specified charset),
 a partial restore that does not include the SET NAMES line will cause data
 corruption.  The same issue could happen on trunk/post-0.21-fixes systems,
 depending on configuration.

 In addition, any partial restore using the instructions at
 http://mythtv.org/docs/mythtv-HOWTO-23.html#ss23.7 on a system properly-
 configured for 0.21-fixes and below (i.e. using a latin1 charset) will
 cause data corruption.  I left a note in the already-open ticket at #3762.

-- 
Ticket URL: <http://svn.mythtv.org/trac/ticket/6013>
MythTV <http://www.mythtv.org/>
MythTV


More information about the mythtv-commits mailing list