List:General Discussion« Previous MessageNext Message »
From:Michael Widenius Date:July 31 1999 2:17am
Subject:Big query analyser problem with 3.22
View as plain text  
>>>>> "J-F" == J-F Mammet <webmaster@stripped> writes:

J-F> Hi.
J-F> We have a big problem since we migrated a machine to mysql-3.22.25 (it
J-F> was on 3.21.33b before). One query that took only .98 seconds now takes
J-F> 3.2 seconds. When I do an explain on it, it takes 2 seconds !

J-F> Here is most things you want for a bug report :
J-F> We are using Linux 2.2.10. I tried in different machines. They all used 
J-F> debian 2.1.
J-F> 128 mbytes of ram+ 128 mb of swap
J-F> the compiler is egcs-1.1.2 release
J-F> I used the --enable-assembler flag.

<cut>

Fix:

I have now fixed this for MySQL 3.23.2 (will be released withing a few
days)

Here follows a patch for MySQL 3.22:

*** /my/monty/master/mysql-3.22.24/sql/sql_select.cc    Sat Jul  3 20:29:53 1999
--- ./sql_select.cc     Fri Jul 30 14:00:30 1999
***************
*** 727,732 ****
--- 727,733 ----
        table->maybe_null=table->outer_join=1;  // Mark for send fields
        s->key_dependent=s->dependent=
        s->on_expr->used_tables() & ~(table->map);
+       s->dependent|=stat_vector[i-1]->dependent | table_vector[i-1]->map;
        outer_join|=table->map;
        continue;
      }

(The problem was that there was so many table combinations it took a
long to to find the optimal table order)

Regards,
Monty
Thread
Big query analyser problem with 3.22J-F Mammet20 Jul
  • Big query analyser problem with 3.22sinisa20 Jul
  • Big query analyser problem with 3.22Michael Widenius31 Jul