Author: Alexey Kopytov
Revision Date: 2015-02-25 13:15:34 UTC

Addendum to fix for bug #1382797 "Modify SST code to use fork()/exec() to
allow cleanup on fatal signals"

In addition to resetting ignored signals, also reset the process signal
mask for the child process.

Author: Alexey Kopytov
Revision Date: 2015-02-09 09:13:26 UTC

Empty merge from 2.2.

Author: Alexey Kopytov
Revision Date: 2015-01-11 16:34:54 UTC

Manual merge from 5.5.

Author: Alexey Kopytov
Revision Date: 2015-01-09 17:42:16 UTC

Bug #1409042: Variable length arrays in WSREP code

Removed VLAs from WSREP code.

Author: Alexey Kopytov
Revision Date: 2014-12-08 15:01:21 UTC

Bug #1382797: Modify SST code to use fork()/exec() to allow cleanup on
              fatal signals
Bug #1399175: Incorrect POSIX_SPAWN_SETSIGDEF usage in wsrep_utils.cc

Modified wsp::process() to use fork()/exec() instead of posix_spawn() to
have more control on the child process. More specifically:

- the child process now calls prctl() with appropriate arguments (if
  available) before calling exec() so that the SIGTERM signal is
  received by it in case the parent process (i.e. mysqld) is terminated
  either gracefully or ungracefully.

- the child process also call setsid() to create a new session and a
  process group. This is to simplify killing all spawned processes at
  once for child processes. Instead of keeping track of all processes
  started by an SST script and sending termination signals to the
  individually, an SST script can simply kill the entire process group.

Additionally, an optimization to speed up fork() has been implemented to
reduce page table copying overhead as compared to vfork() /
posix_spawn(). Since the InnoDB buffer pool is the major part of mysqld
RSS in most installations, use madvise(..., MADV_DONTFORK) to exclude
buffer pool from memory available to the child on fork().

This revision also fixes bug #1399175 as a side effect, since all code
related to the posix_spawn*() family of functions is replaced.

Author: Alexey Kopytov
Revision Date: 2014-10-22 19:36:52 UTC

Bug #1384413: Debug builds fail

Fixed two debug-specific compiler errors.

Author: Alexey Kopytov
Revision Date: 2014-09-26 14:13:29 UTC

Bug #1360064: LOCK TABLES FOR BACKUP should be incompatible with LOCK
              TABLES, FLUSH TABLES ... WITH READ LOCK, FLUSH TABLES
              ... FOR EXPORT in the same connection

Made sure that an error is thrown on attempt to execute LOCK TABLES FOR
BACKUP in a session which has the following active lock statements:

- LOCK TABLES READ/WRITE;
- FLUSH TABLES <table_list> WITH READ LOCK;
- FLUSH TABLES <table_list> FOR EXPORT.

And vice versa, an attempt to execute any of the above statements in a
session holding a backup tables lock, also fails with an error.

Author: Alexey Kopytov
Revision Date: 2014-09-17 09:43:06 UTC

Changed xb_galera_sst.sh to use wsrep_sst_method=xtrabackup-v2.

Author: Alexey Kopytov
Revision Date: 2014-09-17 08:26:29 UTC

Removed the check for xtrabackup_galera_info completely.

Author: Alexey Kopytov
Revision Date: 2014-09-06 16:57:28 UTC

PXC-117: Implement backup locks support in PXC

Enabled backup locks support back by reverting revision 792, since the
next PXB release will be able to auto-recover GTID info on prepare.

Author: Alexey Kopytov
Revision Date: 2014-08-11 12:34:33 UTC

Updated bootstrap.sh to use YaSSL 5.6 builds.

Author: Alexey Kopytov
Revision Date: 2014-06-26 18:25:19 UTC

Don't call post_kill_notification for wsrep_applier threads.

Author: Alexey Kopytov
Revision Date: 2014-06-26 10:26:15 UTC

Addendum to fix for bug #1333348: removed threadpool-related code from
wsrep_wait_appliers_close().

Author: Alexey Kopytov
Revision Date: 2014-03-20 12:44:36 UTC

Updated binary tarballs in bootstrap.sh to:

MySQL 5.1.73
Percona Server 5.1.73-rel14.11
MySQL 5.5.36
Percona Server 5.5.36-34.1
MySQL 5.6.16
Percona Server 5.6.16-64.1

Author: Alexey Kopytov
Revision Date: 2014-01-22 11:19:20 UTC

Test.

Author: Alexey Kopytov
Revision Date: 2013-12-18 12:32:09 UTC

Merged lp:~ignacio-nin/sysbench/0.5-bug1262178 and ixed the _WIN32
codepath to correctly handle FlushViewOfFile return value.

Author: Alexey Kopytov
Revision Date: 2013-12-05 11:13:48 UTC

Merged 5.6 trunk.

Author: Alexey Kopytov
Revision Date: 2013-11-27 17:25:14 UTC

Implemented thread_pool_high_prio_mode.

Author: Alexey Kopytov
Revision Date: 2013-11-08 10:18:25 UTC

Port of rev. 5743 ("WL#6899 - Reduce lock_sys_t::mutex contention when
converting implicit lock to an explicit lock") from MySQL 5.7.

Author: Alexey Kopytov
Revision Date: 2013-11-08 10:11:26 UTC

Fixed a deadlock in the WL6899 port.

Author: Alexey Kopytov
Revision Date: 2013-10-23 13:22:39 UTC

Replaced mach_*() primitives with bswap builtins.

Author: Alexey Kopytov
Revision Date: 2013-09-07 17:22:50 UTC

Don’t assign undo records from the same page to different worker threads.

Author: Alexey Kopytov
Revision Date: 2013-09-04 15:48:58 UTC

Merged ahi-fixes-5.6.

Author: Alexey Kopytov
Revision Date: 2013-08-30 09:34:55 UTC

Bug #1218347: Unnecessary overhead from persistent adaptive hash index
              latches

Disabled persistent AHI latch code so that if row_search_for_mysql()
acquires an S latch on AHI, it is guaranteed to release it before
return. This also makes trx_search_latch_release_if_reserved() and some
ugly code in row_search_for_mysql() redundant.

Author: Alexey Kopytov
Revision Date: 2013-06-28 11:26:05 UTC

Various test suite cleanups:

* Re-implemented testrun.c in testrun.sh. This allows cleaner and more
  efficient implementations of parallel tests execution and status
  aggregation. Unlike previously, a server instance to detect version
  is now started just once before tests execution, which should save a
  lot of time. And we don't need to hash the binary. Progress is now
  reported and stats are also collected in the main script. And we
  also don't need multuple subunit files.

* As we don't need the run.sh wrapper now either, renamed testrun.sh
  back to its original name.

* Fixed bug #1097434 "CTRL-C or kill run.sh only terminates current
  test, not entire run" by defining appropriate signal handler in the script.

* Fixed bug #1097444 "xtrabackup test suite leaves xtrabackup.mysql.sock
  and xtrabackup_port_lock files behind in /tmp" by making sure
  everything is killed and cleaned up in all circumstances

* Fixed bug #1194813 "The port lock mechanism in XB test suite does not work"

* Time taken by each test is now reported

* The summary is now a lot more similar to MTR and includes test
  execution time stats and skipped tests summary.

* When a test fails in debug mode, information on where the artifacts
  can be found is now displayed in the summary

* record_db_state / verify_db_state now pass --skip-extended-insert to
  mysqldump, which makes identifying difference in data rows much
  easier.

* reduced the time we are waiting for a server to start by tweaking
  mysql_ping()

Author: Alexey Kopytov
Revision Date: 2013-06-14 07:24:28 UTC

Bug #1190876: Update links to server tarballs in bootstrap.sh

Updated links along with some whitespace cleanups.

Author: Alexey Kopytov
Revision Date: 2013-06-12 13:15:54 UTC

Manual merge from 2.0.

Author: Alexey Kopytov
Revision Date: 2013-06-12 12:31:04 UTC

Fix for bugs #1187071 and #1177206.

Author: Alexey Kopytov
Revision Date: 2013-06-12 11:35:29 UTC

Fix for bugs #1187071 and #1177206.

Author: Alexey Kopytov
Revision Date: 2013-03-15 09:24:26 UTC

Removed read view allocation diagnostics.

Author: Alexey Kopytov
Revision Date: 2013-02-08 11:45:53 UTC

Exclude innodb_extend_slow.patch

Author: Alexey Kopytov
Revision Date: 2013-01-30 08:22:49 UTC

Enabled and revived UNIV_LOG_ARCHIVE code.

Author: Alexey Kopytov
Revision Date: 2013-01-10 19:17:15 UTC

Fix for bug#1093787

Author: Alexey Kopytov
Revision Date: 2012-11-30 08:41:50 UTC

innodb_instant_drop_table implementation.

Author: Alexey Kopytov
Revision Date: 2012-10-19 16:05:44 UTC

Automerge from 2.0.

Author: Alexey Kopytov
Revision Date: 2012-09-27 18:14:06 UTC

Fixed out-of-source builds in 5.1.

Author: Alexey Kopytov
Revision Date: 2012-04-18 07:54:25 UTC

Add more diagnostics to xtrabackup_copy_datafile() to track down an
issue with unexpected EOFs.

Author: Alexey Kopytov
Revision Date: 2012-04-05 10:27:15 UTC

Backport of the fix for bug #13704145 / #64284 "Eliminate LRU scan when
dropping a table" from upstream 5.5.23.

Author: Alexey Kopytov
Revision Date: 2012-01-13 16:29:23 UTC

Bug #891406: tar4ibd fails on datafiles generated on 5.0

The problem was that tar4ibd incorrectly read flags from system
tablespaces generated on MySQL/PS 5.0.

Fixed by copying the behavior of fil_space_get_flags() in InnoDB: return
0 as the flags value for the system tablespace.

Author: Alexey Kopytov
Revision Date: 2011-12-07 18:34:20 UTC

Don't use `cat ...` in mysql_ping() because that makes other system()
calls return -1 i it is interrupted by a SIGALRM handler.

Author: Alexey Kopytov
Revision Date: 2011-09-30 08:42:41 UTC

Bug #858945: innodb_expand_fast_index_creation should be optional

This fix makes all functionality provided by
innodb_expand_fast_index_creation.patch disabled by default.

It can be enabled by using a new system variable,
innodb_expand_fast_index_creation, that has both global or session
scopes, and can also be specified in a configuration file or on the
command line. The variable is of the boolean type and is FALSE by
default. The session value of TRUE enables optimizations and
improvements provided by the patch. The variable also has no effect when
fast index creation is disabled with the fast_index_creation option
provided by control_online_alter_index.patch.

Since innodb_expand_fast_index_creation is now disabled by default, some
test suite adjustments that were previously required have been reverted.

Author: Alexey Kopytov
Revision Date: 2011-09-29 10:08:42 UTC

Bug #858945: innodb_expand_fast_index_creation should be optional

This fix makes all functionality provided by
innodb_expand_fast_index_creation.patch disabled by default.

It can be enabled by using a new system variable,
innodb_expand_fast_index_creation, that has both global or session
scopes, and can also be specified in a configuration file or on the
command line. The variable is of the boolean type and is FALSE by
default. The session value of TRUE enables optimizations and
improvements provided by the patch. The variable also has no effect when
fast index creation is disabled with the fast_index_creation option
provided by control_online_alter_index.patch.

Since innodb_expand_fast_index_creation is now disabled by default, some
test suite adjustments that were previously required have been reverted.

Author: Alexey Kopytov
Revision Date: 2011-08-01 18:38:21 UTC

Added relay log coordinates to xid diagnostic info.2

Author: Alexey Kopytov
Revision Date: 2011-07-09 16:07:01 UTC

Implementation of
https://blueprints.launchpad.net/percona-server/+spec/utf8-general50-ci-5.5

The patch is intended to help users:
- Having indexes on columns with collations
utf8_general_ci or ucs2_general_ci
- Having German letter SHARP S (SZLIG) in these columns
- Upgrading from MySQL from versions 5.0.x or
5.1.23 (and earlier) to version 5.1.24 (and higher).

This patch introduces new collations utf8_general50_ci
and ucs2_general50_ci which reproduce the "old"
sorting order provided by pre-5.1.24 versions of xxx_general_ci.

In order to start using new MySQL-5.1.24+ please do the following:

- Start new version of mysqld
- Convert all affected tables using this query (in case of utf8):

ALTER TABLE t1 CONVERT TO CHARACTER SET utf8 COLLATE utf8_general50_ci;

Or if you need to apply changes per-column level, use this example:

ALTER TABLE t1 MODIFY c1 CHAR(N) CHARACTER SET utf8 COLLATE utf8_general50_ci;

(Make sure you're using the old data type and size,
NULL/NOT NULL constraints, etc).

Author: Alexey Kopytov
Revision Date: 2011-07-09 15:54:48 UTC

Duplicated utf8_general_ci and ucs2_general_ci to utf8_general50_ci and
ucs2_general50_ci, respectively.

This is the 5.0 part of
https://blueprints.launchpad.net/percona-server/+spec/utf8-general50-ci-5.1
and
https://blueprints.launchpad.net/percona-server/+spec/utf8-general50-ci-5.5

Author: Alexey Kopytov
Revision Date: 2011-06-23 06:21:03 UTC

Attempt to fix CentOS Jenkins slaves. Simplified mysql_ping() in common.sh.

Author: Alexey Kopytov
Revision Date: 2011-06-17 11:31:00 UTC

Another attempt to diagnose bug #796125.

Print return value and errno when calling read().

Author: Alexey Kopytov
Revision Date: 2011-06-16 11:07:26 UTC

Attempt to fix bug #797944 (since the test failure is sporadic and I
can't reproduce it locally).

Use replication "macros" from include/ rather than custom error
checking code.

Author: Alexey Kopytov
Revision Date: 2011-06-15 07:13:32 UTC

Manual merge from 1.6.