RE: bk commit into 5.1 tree (mats:1.2559) BUG#29020

[Chuck Bell]

Mats,

I had a test failure on Windows and Linux. 

Run with "./mysql-test-run.pl --mysqld=--binlog-format=row"

binlog.binlog_row_ctype_ucs    [ fail ]

Warning: Not freed memory segments: 2
Warning: Memory that was not free'ed (559 bytes):
           103 bytes at 0x001152bb8, allocated at line 1018 in
'D:\source\c++\my
sql-5.1_BUG_29020\sql\log_event.cc'
           456 bytes at 0x001152350, allocated at line  201 in
'..\mysys\my_allo
c.c'
---
d:/source/c++/mysql-5.1_BUG_29020/mysql-test/suite/binlog/r/binlog_row_ctype
_ucs.result     2008-03-26 17:59:10.209765000 +0300
+++
d:\source\c++\mysql-5.1_BUG_29020\mysql-test\suite\binlog\r\binlog_row_ctype
_ucs.reject     2008-03-26 18:29:48.191765000 +0300
@@ -16,10 +16,14 @@
 ROLLBACK/*!*/;
 use test/*!*/;
 SET TIMESTAMP=10000/*!*/;
+SET @@session.pseudo_thread_id=999999999/*!*/;
 SET @@session.foreign_key_checks=1, @@session.sql_auto_is_null=1,
@@session.uni
que_checks=1/*!*/;
 SET @@session.sql_mode=0/*!*/;
+SET @@session.auto_increment_increment=1,
@@session.auto_increment_offset=1/*!*
/;
 /*!\C latin1 *//*!*/;
 SET
@@session.character_set_client=8,@@session.collation_connection=8,@@session
.collation_server=8/*!*/;
+SET @@session.lc_time_names=0/*!*/;
+SET @@session.collation_database=DEFAULT/*!*/;
 BEGIN
 /*!*/;
 SET TIMESTAMP=10000/*!*/;

mysqltest: Result content mismatch

Other than that, there were only very minor changes below. 


[...]

> +
> +    /*
> +      The only valid positions that can be truncated to are at the
> +      beginning of a statement. We are relying on this fact 
> to be able
> +      to set the at_least_one_stmt flag correctly. In other word, if
> +      we are truncating to the beginning of the transaction cache,
> +      there will be no statements in the cache, otherwhise, we will
> +      have at least one statement in the transaction cache.
> +     */
> +    at_least_one_stmt= (pos > 0);
>    }

Minor spelling issues in comments. Should read:

"In other words, if we are truncating..."
             ^^^
and:

"...cache, otherwise, we will..."
               ^^^

Or possibly this, but above is ok, just being picky. :)

"...cache. Otherwise we will..."
        ^^^^^^^^^^^^

[...]

>    /*
> -    Write commit event if at least one of the following holds:
> -     - the user sends an explicit COMMIT; or
> -     - the autocommit flag is on, and we are not inside a BEGIN.
> -    However, if the user has not sent an explicit COMMIT, and we are
> -    either inside a BEGIN or run with autocommit off, then 
> this is not
> -    the end of a transaction and we should not write a commit event.
> +    Decision table for committing a transaction:
> +
> +    =============================  = = = = = = = = = = = = = = = =
> +    Real transaction               N N N N N N N N N N N N N N N N
> +    Statement in cache             N N N N N N N N Y Y Y Y Y Y Y Y
> +    In transaction                 N N N N Y Y Y Y N N N N Y Y Y Y
> +    Stmt modified non-trans        N N Y Y N N Y Y N N Y Y N N Y Y
> +    All modified non-trans         N Y N Y N Y N Y N Y N Y N Y N Y
> +
> +    Action: (C)ommit/(A)ccumulate  C C C C A C C C - - - - A A A A
> +    =============================  = = = = = = = = = = = = = = = =
> +
> +
> +    =============================  = = = = = = = = = = = = = = = =
> +    Real transaction               Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
> +    Statement in cache             N N N N N N N N Y Y Y Y Y Y Y Y
> +    In transaction                 N N N N Y Y Y Y N N N N Y Y Y Y
> +    Stmt modified non-trans        N N Y Y N N Y Y N N Y Y N N Y Y
> +    All modified non-trans         N Y N Y N Y N Y N Y N Y N Y N Y
> +
> +    (C)ommit/(A)ccumulate/(-)      - - - - C C C C - - - - C C C C
> +    =============================  = = = = = = = = = = = = = = = =
> +
> +    In other words, we commit the transaction if and only if:
> +     - The keep log option is set
> +     - We are in a transaction and:
> +       - A full transaction is committed,
> +       - A non-transactional statement is committed and there is
> +         no statement cached, or
> +     - We are not in a transaction and committing a statement
> +
> +    Otherwise, we accumulate the statement
>    */

Excellent comment!!! This needs to go in the documentation!