Author: Sergei Glushchenko
Revision Date: 2015-02-20 19:21:56 UTC

all tests passed on galera 5.6 except backup locks and SST

Author: Sergei Glushchenko
Revision Date: 2015-02-12 12:32:28 UTC

Automerge fix for bug 1414221 from 2.3 series.

Author: Sergei Glushchenko
Revision Date: 2015-02-12 12:29:28 UTC

Bug 1414221: Warning "The log was not applied to the intended LSN"
should optionally be an error

Instead of warning print an error message.
Print value of srv_start_lsn right after the error message.

Author: Sergei Glushchenko
Revision Date: 2015-02-02 00:07:01 UTC

2.2.8 with fix for bug1368846

Author: Sergei Glushchenko
Revision Date: 2015-01-28 17:52:24 UTC

Bug 1368846: xtrabackup Assertion failure with Apply log process

Looking at the stack trace I see pending IO complete operations
which led ibuf_merge after page_cleaner has stopped. We should not
stop page_cleaner if there are pending read IOs.

Author: Sergei Glushchenko
Revision Date: 2015-01-16 23:27:35 UTC

exclude client API

Author: Sergei Glushchenko
Revision Date: 2014-12-08 07:10:34 UTC

Automerge fix for bug 1388122 from 2.2 series.

Author: Sergei Glushchenko
Revision Date: 2014-12-08 07:07:59 UTC

Bug 1388122: Innobackupex does not work with .mylogin.cnf on some
Linux versions

Use my_print_defaults to receive login options stored into
.mylogin.cnf. This login method will only be used if
my_print_defaults which comes with server package is installed.

Author: Sergei Glushchenko
Revision Date: 2014-09-22 02:35:22 UTC

Bug 1240352: Sporadic t/ib_part_include_stream.sh failures in Jenkins

Test fails with checksum = 0 before backup and checksum != 0 after
restore. Checksum value of zero indicates that data haven't made it
to the table yet. Since table is MyISAM, it can be effect of
concurrent inserts for example.

The fix is to FLUSH TABLES in order to make sure all data are
written into the table.

Author: Sergei Glushchenko
Revision Date: 2014-08-31 08:49:54 UTC

test deadlock fix on concurrent delete

Author: Sergei Glushchenko
Revision Date: 2014-08-25 04:35:22 UTC

test written in python

Author: Sergei Glushchenko
Revision Date: 2014-08-22 14:36:16 UTC

Bug 569387: innobackupex ignores --databases without --stream
Changes:
- two options have been added for xtrabackup:
- "--databases" is a space separated list of entries
database_name[.table_name]
- "--databases-file" is a name of file which contains entries
database_name[.table_name] one entry per line
- check_if_table_matches_filters has been modified to check whether
database is enabled first and then check the table
- databases and tables filters initialization has been refactored in
order to avoid code duplication
- innnobackupex's --databases option is mapped to one of these
options, depending on value.

Author: Sergei Glushchenko
Revision Date: 2014-08-04 09:24:16 UTC

attempt to fix

Author: Sergei Glushchenko
Revision Date: 2014-04-21 12:11:22 UTC

Merge audit log plugin from 5.5

Author: Sergei Glushchenko
Revision Date: 2014-03-19 09:33:43 UTC

Merge plugin for scalability metrics from 5.5 series

Author: Sergei Glushchenko
Revision Date: 2014-03-16 07:21:40 UTC

Merge test case for bug 1289754 from 5.5

Author: Sergei Glushchenko
Revision Date: 2014-03-15 16:58:09 UTC

Bug 1289754: table_id is defined differently in sql/table.h vs sql/log_event.h
Straightforward backport from MySQL 5.6

Author: Sergei Glushchenko
Revision Date: 2014-02-09 18:50:28 UTC

Plugin for scalability metrics.
This is an audit plugin which provides following status variables
once loaded and enabled:
- scalability_metrics_elapsedtime - total time elapsed since starting point
- scalability_metrics_queries - number of completed queries since starting point
- scalability_metrics_concurrency - number of queries currently executed
- scalability_metrics_totaltime - total execution time of all queries
- scalability_metrics_busytime - server's total busy time
time is counted in microseconds.
There is also the global control variable named scalability_metrics_control
with possible values ON, OFF, RESET.
Setting the value to ON enables counting of metrics. Setting it to OFF disables
counting. By setting the value to RESET one can reset counters (status
variables) while continuing to count metrics.

Notes on implementation.
Three types of notifications are used to count these metrics (all three are of
MYSQL_AUDIT_GENERAL_CLASS class).
MYSQL_AUDIT_GENERAL_LOG with command 'Query' and 'Execute' is considered a
query start event
MYSQL_AUDIT_GENERAL_RESULT and MYSQL_AUDIT_GENERAL_ERROR are considered a
query end event
Statistics calculated for each THD and stored in data structure
associated with THD. The only synchronization point is the atomic
variable used to track number of currently running queries. It is
used for the purpose of calculating server busy time. Once
variable becomes non-zero we start tracking busy time, once it
becomes zero, we stop doing it.

Author: Sergei Glushchenko
Revision Date: 2014-02-09 14:20:31 UTC

fix some more warnings

Author: Sergei Glushchenko
Revision Date: 2013-12-16 11:43:17 UTC

Bug 1182535: no MYSQL_AUDIT_GENERAL_LOG notifications with general log off
The reason is that MYSQL_AUDIT_GENERAL_LOG events are emitted in the same
method which does general log writes. Method is not called either if
general_log is turned off or if we don't want to log specific command.

Fixed my moving invocation of mysql_audit_general_log from
LOGGER::general_log_write() to general_log_write(). user_name
and time needed for both general log and audit log are
calculated twice if both enabled. Further optimization would be
to calculate user_name upon user login and/or change by change
user command.

Author: Sergei Glushchenko
Revision Date: 2013-12-10 15:41:50 UTC

change test case

Author: Sergei Glushchenko
Revision Date: 2013-12-06 11:55:52 UTC

test

Author: Sergei Glushchenko
Revision Date: 2013-11-14 14:05:39 UTC

add debug info

Author: Sergei Glushchenko
Revision Date: 2013-09-26 12:39:55 UTC

valgrind suppressions

Author: Sergei Glushchenko
Revision Date: 2013-09-24 11:09:38 UTC

Merge fix for #1206065 from 5.5.
It is empty because was already fixed in
lp:~sergei.glushchenko/percona-server/5.6-ps-bug1194097.

Author: Sergei Glushchenko
Revision Date: 2013-09-19 12:53:10 UTC

print debug info to help debugging the bug https://bugs.launchpad.net/percona-server/+bug/1193319

Author: Sergei Glushchenko
Revision Date: 2013-08-12 12:51:45 UTC

Fix for bug #1206065: Memory corruption at server shutdown when using threadpool.
Memory corruption is caused by concurrent writing to vio structure and releasing
memory occupied by very same structure. This is possible when close_sockets calls
vio_shutdown via post_kill_notification and epoll event handler thread destroys
THD because it's vio is closed.
Other invocations of post_kill_notification are protected by LOCK_thd_data mutex,
we doing same here.

Author: Sergei Glushchenko
Revision Date: 2013-08-05 12:01:40 UTC

port MDEV-4566

Author: Sergei Glushchenko
Revision Date: 2013-07-30 12:53:13 UTC

Bug 1191375: Percona Server 5.6 RC with thread pool crashes under load
Attached gdb stack trace points that SIGSEGV is emitted in pfs_spawn_thread,
when newly created thread tries to initialize some of it's instrumentation
properties by copying them prom parent's thread instrumentation properties.
This likely caused by the fact that parent thread finished shortly after
it spawned child thread and before child thread accessed parent's properties.
This patch fixes race by copying all needed properties from parent's
structures to the argument structure for child thread before the child
thread is spawned. This makes child thread initialization immune to parent's
thread death.

Author: Sergei Glushchenko
Revision Date: 2013-07-16 21:48:06 UTC

do not notify wait when timeout is 0

Author: Sergei Glushchenko
Revision Date: 2013-07-01 12:01:12 UTC

Bug 1196460: Percona Server won't compile with dtrace: Sun D 1.11
`dtrace -G` invoked for every plugin to produce object file containing
instrumentation code. However for dtrace version "Sun D 1.11" invocation
`dtrace -G` for non-instrumented code filing with error.
Workaround is to not to invoke `dtrace -G` for non-instrumented plugins.
This is done by adding extra argument "DTRACE_INSTRUMENTED" to
MYSQL_ADD_PLUGIN macro. All plugins which need instrumentation been marked
as "DTRACE_INSTRUMENTED".
Also -DDISABLE_DTRACE=1 (--disabe-dtrace for ./configure) has been added to
omit dtrace instrumentation even when dtrace support is found.

Author: Sergei Glushchenko
Revision Date: 2013-07-01 11:02:12 UTC

Merge fix for bug 1196460 from 5.5 series.

Author: Sergei Glushchenko
Revision Date: 2013-06-29 22:21:59 UTC

This patch adds option --disable-dtrace (-DDISABLE_DTRACE=1 for cmake),
also this patch adds DTRACE_INSTRUMENTED argument for MYSQL_ADD_PLUGIN
macro. This allows to tell build system to perform actions needed to
instrument the plugin. Plugins which are needed to be instumented
declared as DTRACE_INSTRUMENTED.

Author: Laurynas Biveinis
Revision Date: 2013-06-19 11:34:32 UTC

Automerge lp:~sergei.glushchenko/percona-server/5.6-merge-pam

Author: Sergei Glushchenko
Revision Date: 2013-06-23 16:00:11 UTC

monkeyfix of failing test

Author: Sergei Glushchenko
Revision Date: 2013-06-23 15:10:31 UTC

test cases

Author: Sergei Glushchenko
Revision Date: 2013-06-20 17:54:55 UTC

test fix for bug 1192354

Author: Sergei Glushchenko
Revision Date: 2013-06-13 18:26:15 UTC

merge scalability metrics plugin

Author: Sergei Glushchenko
Revision Date: 2013-06-13 18:22:52 UTC

merge scalability metrics plugin

Author: Sergei Glushchenko
Revision Date: 2013-06-12 14:47:40 UTC

Merge fix and test case for bug 1182949 from 5.5 series.

Author: Sergei Glushchenko
Revision Date: 2013-06-11 19:42:19 UTC

Merge fix for bug 1182949 from 5.1 series.
Add testcase for this bug with using audit_null plugin.

Author: Sergei Glushchenko
Revision Date: 2013-06-11 19:29:13 UTC

Bug 1182949: Prepared statements are written to general query log after being executed
Fix is to move invocation of general_log_write before the statement execution. This
keeps the feature that expanded statements are logged by execute event, but
makes metrics less precise since time taken by query expansion won't be accounted.

Author: Sergei Glushchenko
Revision Date: 2013-06-11 17:19:01 UTC

Plugin for scalability metrics.
This is an audit plugin which provides following status variables
once loaded and enabled:
- scalability_metrics_elapsedtime - total time elapsed since starting point
- scalability_metrics_queries - number of completed queries since starting point
- scalability_metrics_concurrency - number of queries currently executed
- scalability_metrics_totaltime - total execution time of all queries
- scalability_metrics_busytime - server's total busy time
time is counted in microseconds.
There is also the global control variable named scalability_metrics_ctl
with possible values ON, OFF, RESET.
Setting the value to ON enables counting of metrics. Setting it to OFF disables
counting. By setting the value to RESET one can reset counters (status
variables) while continuing to count metrics.

Notes on implementation.
Three types of notifications are used to count these metrics (all three are of
MYSQL_AUDIT_GENERAL_CLASS class).
MYSQL_AUDIT_GENERAL_LOG with command 'Query' and 'Execute' is considered a
query start event
MYSQL_AUDIT_GENERAL_RESULT and MYSQL_AUDIT_GENERAL_ERROR are considered a
query end event
Time intervals are calculated with mutex held. Pressure on this mutex
however is nearly not as much as pressure on LOCK_plugin used by MySQL.
Timestamps are gathered on Linux by using clock_gettime(CLOCK_MONOTONIC, )
which should use TSC CPU register to receive time and also should be
fully executed in user-space via VDSO. Since OS X lacks clock_gettime,
mach_absolute_time was used on this platform. QueryPerformanceCounter
can be used on Windows, but it is not implemented since author donesn't
have access to Windows machine.

Author: Sergei Glushchenko
Revision Date: 2013-06-03 16:54:39 UTC

test

Author: Sergei Glushchenko
Revision Date: 2013-04-25 12:51:29 UTC

Blueprint: Support InnoDB buffer pool dumps in MySQL 5.6
https://blueprints.launchpad.net/percona-xtrabackup/+spec/support-56-buffer-pool-dumps
MySQL 5.6 adds an ability to dump the buffer pool into file for it to be
possible to load it into memory at startup or on demand. The location of
buffer pool file is set by global variable innodb_buffer_pool_filename
and is relative to MySQL data directory.
The patch adds an ability to take a backup of buffer pool dump. It has been
made exactly the same way as XtraDB LRU dump is backed up.
Three tests have been added for this feature. One takes regular backup and
after takes streaming backup. Another one takes backup with --rsync
option. Third test tests feature with incremental backups.

Author: Sergei Glushchenko
Revision Date: 2013-04-19 15:47:39 UTC

Bug 1092593: crash resistant replication doesn't work correctly after
             change master or reset slave
    - further might look rather unstructured; so one might benefit from
      refreshing discussion here https://code.launchpad.net/~laurynas-biveinis/percona-server/bug1012715-5.1/+merge/120042
      before reading this
    - this is how I imagine XA-transactions flow
      * BEGIN do something PREPARE COMMIT BEGIN do something PREPARE COMMIT
      * on each state ROLLBACK can be issued except after
        the COMMIT has been completed successfully and BEGIN is not issued
        (it this case ROLLBACK will just be noop)
      * ROLLBACK should leave us somewhere before the BEGIN
    - when InnoDB performs recovery it takes binlog position
      from the "prepare" point.
    - after this XA transaction can be reverted or committed
    - if XA transaction is reverted then we take binlog position
      from "commit" point which in this case is older than "prepare"
    - if XA transaction is committed then we continue to use "prepare"
      which in this case of same age as "commit" point
    - this however does not work when XA transactions are not in use;
      in this case we never write "prepare" points
    - this might be masked by the fact that we never have real info
      in "prepare" points so it's content is not overwrite MySQL *-info
      files; it is still bad because we don't really have transactional
      writes for binlog position as has been claimed in the documentation
    - solution of this problem looks trivial for me; on commit we should
      overwrite both "prepare" and "commit"; this doesn't make any harm
      as after commit there is no way back to prepare anyway, so XA-case
      continues to work
    - non XA case will look as following
      BEGIN do something COMMIT BEGIN do something COMMIT
      * commit and prepare become the one commit point after which
        there is no way back and it is reflected in our "commit" and
        "prepare" points holding the same binlog position; rollback
        is possible to the previous commit point only

Author: Sergei Glushchenko
Revision Date: 2013-04-17 15:15:31 UTC

repro scripts for bug #1092593

Author: Sergei Glushchenko
Revision Date: 2013-03-30 09:27:22 UTC

Determining offset by reading achive header on startup has been implemented

Author: Sergei Glushchenko
Revision Date: 2013-03-29 06:38:46 UTC

Implementation of log archiving based on branch by KAlexey
which enables UNIV_LOG_ARCHIVE
* variables and startup parameters have following names:
  - innodb_log_archive = ON/OFF
  - innodb_log_arch_dir = DIR
  - innodb_log_arch_expire_sec = N
* archive log files naming has been changed from original
  zero-based sequence (0, 1, 2, 3) to sequence of start LSNs
  of archived file. Also prefix has been changed to be
  the same as in previous implementation. Numeric part of
  archived log file name has been changed to be
  20 characters width, zero padded
* purge commands have been implemented
  (PURGE ARCHIVED LOGS BEFORE/TO)
* innodb_log_arch_expire_sec has been implemented
* innodb_xtra_log_archive_throttle has not been implemented
  as it is not applicable
* type of archived_file_no, archived_offset, next_archived_file_no,
  next_archived_offset members of log_group_t has been changed to
  lsn_t
* two additional fields in chechpoint record have been introduced
  to store high 4 bytes of archived file number and offset in
  archived file

Author: Sergei Glushchenko
Revision Date: 2013-03-28 15:35:09 UTC

4-bit archived offset and file number; long lines

Author: Sergei Glushchenko
Revision Date: 2013-03-25 13:33:45 UTC

Port workaround for epoll bug in Linux kernels from MariaDB.
New global boolean option threadpool_workaround_epoll_bug has been
added which if is set changes io_poll_start_read implementation
from using EPOLL_CTL_MOD to EPOLL_CTL_DEL and EPOLL_CTL_ADD.

Author: Sergei Glushchenko
Revision Date: 2013-03-21 14:46:55 UTC

Bug 1157037 create_duplicate_weedout_tmp_table() leaves key_part_flag uninitialized
Manual initialization of key_part_info fields has been replaced with
KEY_PART_INFO::init_from_field invocation.

Author: Sergei Glushchenko
Revision Date: 2013-03-16 23:10:56 UTC

fix derived.test

Author: Sergei Glushchenko
Revision Date: 2013-03-16 12:27:34 UTC

Merge from latest feature branch

Author: Sergei Glushchenko
Revision Date: 2013-03-08 09:54:10 UTC

BUG#1132194: UPDATE/DELETE with LIMIT clause unsafe for SBR
even when ORDER BY PK is present

MYSQL BUGUG#42415 UPDATE/DELETE with LIMIT clause unsafe for SBL
even with ORDER BY PK clause

This is an adoption of http://lists.mysql.com/commits/126382

This is only a partial solution which allows to suppress warning
in many cases when statement with LIMIT isn't really nondeterministic.
Partial means that in mixed mode all statements with LIMIT will be
switched to ROW unconditionally. To determine whether the staments is
really unsafe some analysis of query should be performed. This requires
in particular all tables should be opened and all fields should be
fixed. However decision to switch to ROW mode is made before this
happens.

Original description:
UPDATE/DELETE/INSERT..SELECT with LIMIT clause were considered unsafe
unconditionally, even if there is an ORDER BY PK or WHERE condition
that can guarantee the result to be deterministic.

The problem is fixed by implementing a algorithm to do more elaborate
analyzed on the nature of the query to determine whether the query
will cause uncertainty for replication or not.

The statement will not be considered unsafe for following cases:
 - single table UPDATE/DELETE/INSERT..SELECT with ORDER BY
   <non null unique key>
 - single table UPDATE/DELETE/INSERT..SELECT with WHERE clause
   that include <non null unique key> = <const value>, if it is
   a multi-part key, then it must be <keypart1>=<const1> AND
   <keypart2>=<const2> ..., and this condition is ANDed with
   other conditions if there is any.
 - single table INSERT..SELECT with WHERE clause
   that include some parts of the non null unique key compare to
   const values, and the ORDER BY clause includes all the other
   key pars of the same non null unique key. for example (a,b) is
   a non null unique key, then WHERE a=<const> ORDER b will make
   the query result deterministic.
 - for INSERT..SELECT ... JOIN ..., the join condition (ON,
   USING, or WHERE) must include equations between columns of
   non null unique key of tables from both side of the join.
   For example t1 JOIN t2 USING (a,b), if (a,b) is not a non null
   unique key for both t1 and t2, then the result will be non-
   deterministic. otherwise the result can be deterministic with
   appropirate WHERE/ORDER clauses, and in this case, the same
   rule for single table above applys. But there is a difference
   for INNER JOIN with OUTER JOIN, for OUTER JOIN, only one table
   of the two JOIN tables will be used when checking the WHERE/ORDER
   conditions, it's the left table for LEFT JOIN and the right one
   for RIGHT JOIN when checking the keys. On the other hand, for
   INNER JOIN, keys from both tables can be used when checking
   the conditions. For example:
     INSERT..SELECT * FROM t1 INNER JOIN t2 USING(pk) ORDER BY nnuk LIMIT 1;
   This can be safe if nnuk is a non null unique key of either
   t1 or t2. But if we change the INNER JOIN to LEFT JOIN or
   RIGHT JOIN, then nnuk must be a non null unique key key of
   t1 (LEFT JOIN) or t2 (RIGHT JOIN) respectively.
 - If JOIN are nested, the will be handled recursively from inner
   outside.
 - UNIONs without global LIMIT are unsafe if any of sub-selects
   is unsafe, and safe if all sub-selects are safe. UNIONs with
   global LIMIT are marked unsafe no matter this is ORDER BY or
   not.

If the key field length is longer than MAX_SORT_LENGTH, then it
will be ignore and not treated as a key field.

Author: Sergei Glushchenko
Revision Date: 2013-02-25 08:21:30 UTC

Port of fix for bug 1131856 from 5.5 series

Author: Sergei Glushchenko
Revision Date: 2013-02-25 08:10:47 UTC

Bug 1131856
dec_connection_count() definition doesn't match its declaration
Declaration has been changed to match definition.

Author: Sergei Glushchenko
Revision Date: 2013-02-25 07:57:21 UTC

Bug 569387
xtrabackup ignores --databases, i.e. when --databases
is specified it affects only innobackupex script,
InnoDB data will be copied fully.
two options have been added for xtrabackup
  * --databases is a space separated list of entries database_name[.table_name]
  * --databases-file is a name of file which contains
          entries database_name[.table_name] one entry per line
  * check_if_skip_table has been modified to check whether database is
    enabled first and after check the table
  * databases and tables filters initialization has been refactored
    in order to avoid code duplication
innnobackupex's --databases option is mapped to one of these options,
depending on value.
Bug 1131084
Unneccessary/debug print in xtrabackup output.
There was msg("xtrabackup: tables regcomp(%s): %s\n", p, errbuf);
which has been removed in this branch.

Author: Sergei Glushchenko
Revision Date: 2013-02-20 05:52:53 UTC

Port of fix for bug 1127008 from 5.5 series

Author: Sergei Glushchenko
Revision Date: 2013-02-13 15:17:37 UTC

fix mp

Author: Sergei Glushchenko
Revision Date: 2013-02-12 11:55:22 UTC

More accurate LSN in file number (matches with LSN in file number)

Author: Sergei Glushchenko
Revision Date: 2013-02-01 15:31:40 UTC

merge with threadpool

Author: Sergei Glushchenko
Revision Date: 2013-01-17 08:43:57 UTC

Add extra log file to log file group when log archiving
is turned on. innodb_xtra_log_archive has been made
read only.

Author: Sergei Glushchenko
Revision Date: 2013-01-14 05:58:15 UTC

Bug 569387
xtrabackup ignores --databases, i.e. when --databases
is specified it affects only innobackupex script,
InnoDB data will be copied fully.
--databases and --databases-file options been introduced
for xtrabackup binary, which cover two modes of --databases
option of innobackupex.

Author: Sergei Glushchenko
Revision Date: 2013-01-09 16:55:47 UTC

fix main.percona_query_response_time-replication

Author: Sergei Glushchenko
Revision Date: 2013-01-02 14:46:32 UTC

test

Author: Sergei Glushchenko
Revision Date: 2012-12-19 19:31:48 UTC

Fix issue 27220, bug 1042946

Author: Sergei Glushchenko
Revision Date: 2012-12-19 11:22:24 UTC

Fix issue 27220, bug 1042946

Author: Sergei Glushchenko
Revision Date: 2012-12-14 14:51:37 UTC

Stronger assertion. Check that mdl_ticket == 0 only for temporary table

Author: Sergei Glushchenko
Revision Date: 2012-12-06 14:47:25 UTC

Test rpl_rotate_logs fails with asssertion
Assertion failed: (thread_group->connection_count > 0), function wait_begin, file sql/threadpool_unix.cc, line 1151.
The reason is that when threadpool_remove_connection performs THD cleanup,
it calls close_temporary_tables which calls MYSQL_BIN_LOG::write, which calls
MYSQL_BIN_LOG::rotate and eventually my_sync is called which calls wait_begin,
which deal with connection's group. However connection count in group been
alreadyn decreased in connection_abort before threadpool_remove_connection been
called.

Author: Sergei Glushchenko
Revision Date: 2012-10-22 14:18:04 UTC

basic testing for crashing

Author: Sergei Glushchenko
Revision Date: 2012-10-22 13:16:00 UTC

Merge with 5.1 trunk

Author: Sergei Glushchenko
Revision Date: 2012-10-04 15:30:46 UTC

made extra initialization for innodb51 only

Author: Sergei Glushchenko
Revision Date: 2012-10-04 15:30:46 UTC

made extra initialization for innodb51 only

Author: Sergei Glushchenko
Revision Date: 2012-09-18 13:02:37 UTC

Partitioned tables are not correctly handled by the --databases, --include,
--tables-file options of innobackupex, and by the --tables and
--tables-file options of xtrabackup.
Solution is to remove partition suffix (#P#...) before doing filtering.
Testcases cover variants of using filtering options with MyISAM and
InnoDB tables (to test both innobackupex and xtrabackup) with either stream
mode turned on and turned off.

Author: Sergei Glushchenko
Revision Date: 2012-09-17 14:28:01 UTC

Merge fix for bug 1042517 from 5.1

Author: Sergei Glushchenko
Revision Date: 2012-09-17 14:24:16 UTC

This is a merge of fix for --with-max-indexes=128 by Alexey Kopytov.
This fix originally was committed to lp:percona-server/rnt-5.1.
Original commit message is:
Fix for MySQL bug #54127: mysqld segfaults when built using
--with-max-indexes=128

When using a non-default MAX_KEY value, a different code path is used
when processing index bitmaps. With the default value of 64, the
optimized "template <> class Bitmap<64>" is used which represents and
processes bitmaps as 64-bit integeres. Otherwise, "template <uint
default_width> class Bitmap" is used in which case bitmaps are
represented as arrays.

Multiple problems with the "non-optimized" Bitmap class were discovered
when testing a server binary built with --with-max-indexes=128:

1. bitmap_set_prefix() could overrun the internal buffer when resetting
the remainder of the buffer after setting the prefix due to an
incorrectly calculated remainder's length. This was the reason for the
crash on startup in MySQL bug #54127.

2. Bitmap::intersect() did not take into account that bitmap_init()
resets the supplied buffer, so an intersection with a zero bitmap was
always calculated (reported as MySQL #61178). This led to numerous test
failures due to different execution plans produced by the optimizer.

3. Bitmap::to_ulonglong() incorrectly calculated the result value due to
serious bugs in [u]int*korr/[u]int*store set of macros in
my_global.h (reported as MySQL bugs #61179 and #61180). This led to test
failures in distinct.test and group_min_max.test.

There are still a number of failing tests when running the test suite
with --with-max-indexes=128:

- create.test contains a test case explicitly testing the 64-bit index
limit;

- the ps_N* series of tests verifies the metadata sent by EXPLAIN, where
the field length of "possible_keys" and "key_len" columns depends on
the MAX_KEY value and hence, is different for a binary built with a
non-default value of --with-max-indexes.

The workaround was implemented for failing tests.
include/have_64_keys.inc verify MAX_INDEXES by requesting EXPLAIN
metadata and if MAX_INDEXES is different from default value the
test is skipped.

Author: Sergei Glushchenko
Revision Date: 2012-09-13 17:47:42 UTC

InnoDB coding conventions, comments for functions and arguments.
Fixed some warnings.

Author: Sergei Glushchenko
Revision Date: 2012-03-05 04:31:19 UTC

Bug 711166
Partitioned tables are not correctly handled by the --databases, --include,
--tables-file options of innobackupex, and by the --tables and
--tables-file options of xtrabackup.
Solution is to remove partition suffix (#P#...) before doing filtering.
Testcases cover variants of using filtering options with MyISAM and
InnoDB tables (to test both innobackupex and xtrabackup) with either stream
mode turned on and turned off.

Author: Sergei Glushchenko
Revision Date: 2012-08-29 09:44:41 UTC

apply fix for bug 747032 to release-5.1.57-12.8

Author: Sergei Glushchenko
Revision Date: 2012-08-28 16:30:17 UTC

implementation of datasink_popen

Author: Sergei Glushchenko
Revision Date: 2012-08-27 14:43:46 UTC

new iteration

Author: Sergei Glushchenko
Revision Date: 2012-08-19 12:13:34 UTC

BP https://blueprints.launchpad.net/percona-xtrabackup/+spec/xbstream-decompress
Support for decompression in xbstream. New option -z been added to
automatically decompress .qp files inside the stream. decompress datasink
been implemented for qpress10 one-file archives extraction.

Author: Sergei Glushchenko
Revision Date: 2012-08-15 14:54:29 UTC

merge decompression from 2.0 series

Author: Sergei Glushchenko
Revision Date: 2012-07-23 17:34:00 UTC

implemetation with single thread for archived file

Author: Sergei Glushchenko
Revision Date: 2012-07-23 15:39:56 UTC

initial implementation of decompression

Author: Sergei Glushchenko
Revision Date: 2012-07-11 08:30:26 UTC

merge 2.0 trunk