Below is the list of changes that have just been committed into a local
5.0 repository of mydev. When mydev does a push these changes will
be propagated to the main repository and, within 24 hours after the
push, to the public repository.
For information on how to access the public repository
see http://dev.mysql.com/doc/mysql/en/installing-source-tree.html
ChangeSet
1.1967 05/09/07 15:34:30 ingo@stripped +2 -0
Bug#12845 - Stress test: Server crashes on CREATE .. SELECT statement
Solution for 5.0.
Changed calls to open_table(). Requested to ignore
flush at places where the command did already lock tables.
This could happen in CREATE ... SELECT and ALTER TABLE.
No test case. The bug can only be triggered by true concurrency.
The stress test suite provides a test case for this.
sql/sql_table.cc
1.270 05/09/07 15:34:25 ingo@stripped +4 -2
Bug#12845 - Stress test: Server crashes on CREATE .. SELECT statement
Solution for 5.0.
Changed calls to open_table(). Requested to ignore
flush at places where the command did already lock tables.
This could happen in CREATE ... SELECT and ALTER TABLE.
sql/sql_base.cc
1.297 05/09/07 15:34:24 ingo@stripped +15 -13
Bug#12845 - Stress test: Server crashes on CREATE .. SELECT statement
Solution for 5.0.
Changed open_table() so that ignoring flush requests
is dependend on the 'flags' parameter as ignoring drop
request was already. This aims for consistent behaviour.
# This is a BitKeeper patch. What follows are the unified diffs for the
# set of deltas contained in the patch. The rest of the patch, the part
# that BitKeeper cares about, is below these diffs.
# User: ingo
# Host: chilla.local
# Root: /home/mydev/mysql-5.0-5000
--- 1.296/sql/sql_base.cc 2005-09-02 16:06:09 +02:00
+++ 1.297/sql/sql_base.cc 2005-09-07 15:34:24 +02:00
@@ -1030,23 +1030,23 @@
SYNOPSIS
open_table()
- thd Thread context
- table_list Open first table in list
- refresh Pointer to memory that will be set to 1 if
- we need to close all tables and reopen them
- If this is a NULL pointer, then the is no version
- number checking and the table is not put in the
- thread-open-list
- flags Bitmap of flags to modify how open works:
- MYSQL_LOCK_IGNORE_FLUSH - Open table even if someone
- has done a flush or namelock on it.
+ thd Thread context.
+ table_list Open first table in list.
+ refresh INOUT Pointer to memory that will be set to 1 if
+ we need to close all tables and reopen them.
+ If this is a NULL pointer, then the table is not
+ put in the thread-open-list.
+ flags Bitmap of flags to modify how open works:
+ MYSQL_LOCK_IGNORE_FLUSH - Open table even if
+ someone has done a flush or namelock on it.
+ No version number checking is done.
IMPLEMENTATION
Uses a cache of open tables to find a table not in use.
RETURN
NULL Open failed. If refresh is set then one should close
- all other tables and retry the open
+ all other tables and retry the open.
# Success. Pointer to TABLE object for open table.
*/
@@ -1201,10 +1201,12 @@
if (!thd->open_tables)
thd->version=refresh_version;
- else if (thd->version != refresh_version && refresh)
+ else if ((thd->version != refresh_version) &&
+ ! (flags & MYSQL_LOCK_IGNORE_FLUSH))
{
/* Someone did a refresh while thread was opening tables */
- *refresh=1;
+ if (refresh)
+ *refresh=1;
VOID(pthread_mutex_unlock(&LOCK_open));
DBUG_RETURN(0);
}
--- 1.269/sql/sql_table.cc 2005-09-01 09:05:39 +02:00
+++ 1.270/sql/sql_table.cc 2005-09-07 15:34:25 +02:00
@@ -1753,7 +1753,8 @@
create_info, *extra_fields, *keys, 0,
select_field_count))
{
- if (!(table= open_table(thd, create_table, thd->mem_root, (bool*)0, 0)))
+ if (! (table= open_table(thd, create_table, thd->mem_root, (bool*) 0,
+ MYSQL_LOCK_IGNORE_FLUSH)))
quick_rm_table(create_info->db_type, create_table->db,
table_case_name(create_info, create_table->table_name));
}
@@ -3576,7 +3577,8 @@
bzero((void*) &tbl, sizeof(tbl));
tbl.db= new_db;
tbl.table_name= tbl.alias= tmp_name;
- new_table= open_table(thd, &tbl, thd->mem_root, 0, 0);
+ new_table= open_table(thd, &tbl, thd->mem_root, (bool*) 0,
+ MYSQL_LOCK_IGNORE_FLUSH);
}
else
{
| Thread |
|---|
| • bk commit into 5.0 tree (ingo:1.1967) BUG#12845 | ingo | 7 Sep |