mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2026-05-15 03:17:20 +02:00
Code Issues Projects Releases Packages Wiki Activity

Applied innodb-5.0-ss677 snapshot.

Browse source 
Fixes:
- bug #19834: Using cursors when running in READ-COMMITTED can cause InnoDB to crash
- bug #21112: InnoDB slow with > 100,000 .ibd files
- bug #21113: Duplicate printout in SHOW INNODB STATUS


innobase/fil/fil0fil.c:
  Applied innodb-5.0-ss677 snapshot.
    Make the tablespace cache hash size 100 or 1000 times bigger. Fixes bug #21112.
innobase/ibuf/ibuf0ibuf.c:
  Applied innodb-5.0-ss677 snapshot.
    ibuf_print(): Don't print redundant information. Fixes bug #21113.
sql/ha_innodb.cc:
  Applied innodb-5.0-ss677 snapshot.
    ha_innobase::start_stmt(): patch from Heikki:
    Do not call read_view_close_for_mysql().  (Bug #19834)
This commit is contained in:
unknown 2006-08-01 21:59:58 +02:00
commit 75e40b161b
3 changed files with 20 additions and 37 deletions
Download patch file Download diff file Expand all files Collapse all files

26
sql/ha_innodb.cc
View file

@ -5943,14 +5943,6 @@ ha_innobase::start_stmt(
innobase_release_stat_resources(trx);
if (trx->isolation_level <= TRX_ISO_READ_COMMITTED
&& trx->global_read_view) {
/* At low transaction isolation levels we let
each consistent read set its own snapshot */
read_view_close_for_mysql(trx);
}
prebuilt->sql_stat_start = TRUE;
prebuilt->hint_need_to_fetch_extra_cols = 0;
prebuilt->read_just_key = 0;
@ -6684,17 +6676,17 @@ ha_innobase::store_lock(
&& !thd->tablespace_op
&& thd->lex->sql_command != SQLCOM_TRUNCATE
&& thd->lex->sql_command != SQLCOM_OPTIMIZE
#ifdef __WIN__
/*
for alter table on win32 for succesfull operation
completion it is used TL_WRITE(=10) lock instead of
TL_WRITE_ALLOW_READ(=6), however here in innodb handler
TL_WRITE is lifted to TL_WRITE_ALLOW_WRITE, which causes
race condition when several clients do alter table
simultaneously (bug #17264). This fix avoids the problem.
*/
&& thd->lex->sql_command != SQLCOM_ALTER_TABLE
/* For alter table on win32 for succesful operation
completion it is used TL_WRITE(=10) lock instead of
TL_WRITE_ALLOW_READ(=6), however here in innodb handler
TL_WRITE is lifted to TL_WRITE_ALLOW_WRITE, which causes
race condition when several clients do alter table
simultaneously (bug #17264). This fix avoids the problem. */
&& thd->lex->sql_command != SQLCOM_ALTER_TABLE
#endif
&& thd->lex->sql_command != SQLCOM_CREATE_TABLE) {
lock_type = TL_WRITE_ALLOW_WRITE;