>>>>> "Phuc" == Phuc Truong <ptruong@stripped> writes:
Phuc> Hi guys,
Phuc> I'm experiencing a really strange problem with a table that stores html
Phuc> documents. There are only two records (if you need the dump of the
Phuc> record, I could e-mail it to you it's only 64k). If you dump into the
Phuc> database from the dump file or insert these html documents into the
Phuc> database, it works fine but when you do this selection on the table, it
Phuc> gives you this error:
Phuc> select distinct announcement.* from announcement;
Phuc> ERROR 2013: Lost connection to MySQL server during query
Phuc> This error then causes mysql to restart because of hanging processes:
Phuc> Number of processes running now: 1
Phuc> mysqld process hanging, pid 830 - killed
Phuc> mysqld restarted on Wed Sep 29 03:11:32 UTC 1999
Phuc> /usr/local/mysql/bin/mysqld: ready for connections
Phuc> I'm running:
Phuc> mysqladmin Ver 7.11 Distrib 3.22.25, for pc-linux-gnu on i686
Phuc> linux 2.2.10
Phuc> I also tried both binary and statically compiled versions of mysql. Also
Phuc> tried increasing the max_allowed_packet like the manual says but this
Phuc> didn't help. Below are the settings I used for my safe_mysql script:
Phuc> nohup $ledir/mysqld --basedir=$MY_BASEDIR_VERSION --datadir=$DATADIR \
Phuc> -O max_allowed_packet=32m -O key_buffer=16m \
Phuc> -O sort_buffer=16m -O record_buffer=16m -O max_connections=160 \
Phuc> -O table_cache=500 -O tmp_table_size=16m --skip-locking >> $err_log
2> &1
Phuc> Note that only certain html files causes these errors and it happens only
Phuc> on a select. If you do a select * from announcement, it also works
Phuc> (perhaps the "distinct" screws it up). Also, if you want to know,
Phuc> the html documents were generated in front page. Thanks for the help.
Hi!
This is a bug in the DISTINCT handling of BLOBS in 3.22.25 and 3.22.26
Anyway, here is the patch for this:
*** /my/monty/master/mysql-3.22.26a/sql/sql_select.cc Fri Sep 3 00:39:51 1999
--- ./sql_select.cc Wed Sep 15 13:41:44 1999
***************
*** 4045,4056 ****
if (error == 0)
{
ha_info(entry,1);
- memcpy(new_record,org_record,reclength);
if (copy_blobs(first_field))
{
error=ENOMEM;
break;
}
while ((error=info.read_record(&info)) <= 0)
{
if (error == 0)
--- 4045,4056 ----
if (error == 0)
{
ha_info(entry,1);
if (copy_blobs(first_field))
{
error=ENOMEM;
break;
}
+ memcpy(new_record,org_record,reclength);
while ((error=info.read_record(&info)) <= 0)
{
if (error == 0)
Regards,
Monty
