On 7/25/05, Sun.betty <alanxzq at yahoo.com.cn> wrote:
>    in addition , I test for 
>    createdb -E UNICODE test
>    master and slave as the same.
>    export PGCLIENTENCODING=UNICODE
>    I find as the same question.
>    at master :
>    insert into T_NUM_CENTER (TYPE , CURRVAL , CURRDATE , MEMO) values
> ('?????' , 1 , now() , '???');
>    at slave
>    select * from T_NUM_CENTER ;
>    type    | currval |  currdate  |          memo          
> ------------+---------+------------+------------------------
>  ????   |       1 | 2005-07-25 | ??? 

That is the sympton of the bug.  The bug in the trigger truncates
strings with multi-byte characters.  It ends up removing one byte for
each multi-byte character in the strings.  My guess is that this was
enough to remove the last Chinese character.

 - Ian