Re: bk commit into 5.1 tree (svoj:1.2628) BUG#32050

[Sergei Golubchik]

Hi!

On Nov 21, Sergey Vojtovich wrote:
> ChangeSet@stripped, 2007-11-21 17:05:06+04:00, svoj@stripped +2 -0
>   BUG#32050 - table logging gone wrong.
>   
>   INSERT/UPDATE against CSV table created by MySQL earlier than 5.1.23
>   with NULL-able column results in server crash in debug builds.
>   
>   Starting with 5.1.23 CSV doesn't permit creation of NULL-able columns,
>   but it is still possible to get such table from older MySQL versions.
>   
>   Relaxed DBUG_ASSERT() so that it accepts tables with NULL-able columns
>   created by earlier versions of MySQL.
> 
> diff -Nrup a/storage/csv/ha_tina.cc b/storage/csv/ha_tina.cc
> --- a/storage/csv/ha_tina.cc	2007-10-19 02:48:00 +05:00
> +++ b/storage/csv/ha_tina.cc	2007-11-21 17:05:03 +04:00
> @@ -479,7 +479,7 @@ int ha_tina::encode_quote(uchar *buf)
>        This may only occur if the frm was created by an older version of
>        mysqld which permitted table creation with nullable columns.
>      */
> -    DBUG_ASSERT(!(*field)->maybe_null());
> +    DBUG_ASSERT(!(*field)->maybe_null() || table->s->mysql_version <=
> 50123);
>      
>      /*
>        assistance for backwards compatibility in production builds.

I don't think it's a good idea.

This assert is just wrong - one should only assert things that are
always true if the program works correctly. But I can modify frm in a
hex editor and fail the above expression even if the program itself
(mysqld) is correct. That is - you cannot assert things that are not
*completely* under your control.

Please remove the assert. If you'd like add 'if(...) sql_print_warning()'
here.
 
Regards / Mit vielen Grüssen,
Sergei

-- 
   __  ___     ___ ____  __
  /  |/  /_ __/ __/ __ \/ /   Sergei Golubchik <serg@stripped>
 / /|_/ / // /\ \/ /_/ / /__  Principal Software Developer
/_/  /_/\_, /___/\___\_\___/  MySQL GmbH, Dachauer Str. 37, D-80335 München
       <___/                  Geschäftsführer: Kaj Arnö - HRB
München 162140