MySQL 5.1.18ベータ版がリリースされました。


最も普及しているオープンソースデータベースの新ベータバージョンであるMySQL 5.1.18ベータ版がリリースされました。

これはベータ版であり、他の多くの試作商品リリースと同様、商品レベルのシステムまたは、重要なデータを含むシステムに対してインストールしないように注意してください。5.0使用の商品レベルのシステムについては、以下のMySQL Enterpriseの商品説明のページをご覧ください。

MySQL 5.1.17ベータ版は、とミラーサイトのダウンロード・ページから、ソースコード及び多くのプラットフォームのためのバイナリで現在利用可能です。





Functionality added or changed:
   * Incompatible change: The INFORMATION_SCHEMA.EVENTS and
     mysql.event tables have been changed to facilitate replication
     of events. When upgrading to MySQL 5.1.18, you must run
     mysql_upgrade prior to working with events. Until you have
     done so, any statement relating to the Event Scheduler or
     these tables (including SHOW EVENTS) will fail with the errors
     Expected field status at position 12 to have type enum
     enum('ENABLED','DISABLED') and Table mysql.event is damaged.
     Can not open.
     These changes were made as part of fixes for the following
        + The effects of scheduled events were not replicated (that
          is, binary logging of scheduled events did not work).
        + Effects of scheduled events on a replication master were
          both replicated and executed on the slave, causing double
          execution of events.
        + CREATE FUNCTION statements and their effects were not
          replicated correctly.
     For more information, see Section, "Replication of
     Invoked Features."
   * The plugin interface and its handling of server variables was
     Command line options such as '--skip-innodb' will now cause an error
     if InnoDB is not built-in/plugin-loaded. Users should use
     '--loose-skip-innodb' if they do not want any error even if InnoDB
     is not available. The 'loose' prefix modifier should be used for all
     command line options where the user is uncertain if the plugin
     exists and where they want operation to proceed even if the option
     is neccessarily ignored due to the absence of the plugin.
   * New command line options: In order to alleviate ambiguities in
     variable names, all variables related to plugins can be specified
     using a 'plugin' part in the name. For example, every time where we
     used to have 'innodb' in the command-line options, one can now write
     Furthermore, this is the preferred syntax. It helps to avoid
     ambiguities when a plugin, say, "wait", has an options called
     "timeout".  '--wait-timeout' will still set a server variable, but
     '--plugin-wait-timeout' will set the plugin variable.
     Also, there is a new command line option '--plugin-load' to
     install/load plugins at initialization time without using the
     mysql.plugin table.
   * Storage engine plugins may now be uninstalled at run time - however,
     a plugin is not actually uninstalled until after its reference
     count drops to zero. The 'default_storage_engine' server variable
     consumes a reference count, so uninstalling will not complete until
     said reference is removed.
   * The output of mysql --xml and mysqldump --xml now includes a
     valid XML namespace. (Bug#25946:
   * The mysql_create_system_tables script was removed because
     mysql_install_db no longer uses it in MySQL 5.1.
   * Renamed the old_mode system variable to old.
   * If you use SSL for a client connection, you can tell the
     client not to authenticate the server certificate by
     specifying neither --ssl-ca nor --ssl-capath. The server still
     verifies the client according to any applicable requirements
     established via GRANT statements for the client, and it still
     uses any --ssl-ca/--ssl-capath values that were passed to
     server at startup time.
   * Added a MASTER_SSL_VERIFY_SERVER_CERT option for the CHANGE
     MASTER statement, and a Master_SSL_Verify_Server_Cert output
     column to the SHOW SLAVE STATUS statement. The option value
     also is written to the file.
   * NDB Cluster (APIs): The MGM API now supports explicit setting
     of network timeouts using the ndb_mgm_set_timeout() function.
     A utility function ndb_mgm_number_of_mgmd_in_connect_string()
     is also implemented to facilitate calculation of timeouts
     based on the number of management servers in the cluster.
     For more information, see ndb_mgm_set_timeout()
     and ndb_mgm_number_of_mgmd_in_connect_string()
   * NDB Cluster: The behavior of the ndb_restore utility has been
     changed as follows:
        + It is now possible to restore selected databases or
          tables using ndb_restore.
        + Several options have been added for use with ndb_restore
          --print_data to facilitate the creation of structured
          data dump files. These options can be used to make dumps
          made using ndb_restore more like those produced by
          mysqldump. (Bug#26900:
     For details of these changes, see Section 15.9.3, "ndb_restore
     --- Restore a Cluster Backup."
   * NDB Cluster (Replication)/Incompatible Change: The definition
     of the mysql.ndb_apply_status table has changed such that an
     online upgrade is not possible from MySQL 5.1.17 or earlier
     for a replication slave cluster; you must shut down all SQL
     nodes as part of the upgrade procedure. See Section 15.6.2,
     "Cluster Upgrade and Downgrade Compatibility" before upgrading
     for details.
     For more information about the changes to
     mysql.ndb_apply_status see Section 15.11.4, "Cluster
     Replication Schema and Tables."
   * NDB Cluster (Replication): Some circular replication setups
     are now supported for MySQL Cluster. See Section 15.11.3,
     "Known Issues," for detailed information.
   * NDB Cluster: The following changes were made for the utility:
        + When calculates a value for a given
          configuration parameter that is less than the default
          value, it now suggests the default value instead.
        + The dependency on HTML::Template was removed.
   * Several additional data types are supported for columns in
     integer types. (Bug#27047:
   * NDB Cluster: The internal specifications for columns in NDB
     tables has changed to allow compatibility with future MySQL
     Cluster releases that are expected to implement online adding
     and dropping of columns. This change is not backwards
     compatible with earlier MySQL versions.
     See [1]the related note in Section 15.6.2, "Cluster Upgrade
     and Downgrade Compatibility," for important information prior
     to upgrading a MySQL Cluster to MySQL 5.1.18 or later from
     MySQL 5.1.17 or earlier.
     See alsoBug#28205:
   * If a set function S with an outer reference S(outer_ref)
     cannot be aggregated in the outer query against which the
     outer reference has been resolved, MySQL interprets
     S(outer_ref) the same way that it would interpret S(const).
     However, standard SQL requires throwing an error in this
     situation. An error now is thrown for such queries if the ANSI
     SQL mode is enabled. (Bug#27348:

Bugs fixed:
   * Security fix: If a stored routine was declared using SQL
     SECURITY INVOKER, a user who invoked the routine could gain
     privileges. (Bug#27337:
   * Security fix: The requirement of the DROP privilege for RENAME
     TABLE was not being enforced.
   * Security fix: A user with only the ALTER privilege on a
     partitioned table could obtain information about the table
     that should require the SELECT privilege.
   * The patches forBug#19370:
    Bug#21789: reverted.
   * ON DUPLICATE KEY UPDATE failed for a table partitioned by KEY
     on a primary key VARCHAR column.
   * NDB Cluster (Disk Data): It was possible to drop the last
     remaining datafile in a tablespace (using an ALTER TABLESPACE
     statement), even though there was still an empty table using
     the tablespace. (Bug#21699:
     It should be noted that the datafile could be not dropped if
     the table still contained any rows, so this bug involved no
     loss of data.
   * NDB Cluster: INSERT IGNORE wrongly ignored NULL values in
     unique indexes. (Bug#27980:
   * NDB Cluster: The name of the month "March" was given
     incorrectly in the cluster error log.
   * NDB Cluster (Cluster Replication / Disk Data): An issue with
     replication of Disk Data tables could in some cases lead to
     node failure. (Bug#28161:
   * NDB Cluster (APIs): For BLOB reads on operations with lock
     mode LM_CommittedRead, the lock mode was not upgraded to
     LM_Read before the state of the BLOB had already been
     calculated. The NDB API methods affected by this problem
     included the following:
        + NdbOperation::readTuple()
        + NdbScanOperation::readTuples()
        + NdbIndexScanOperation::readTuples()
   * NDB Cluster (Disk Data): Setting the value of the UNDO BUFFER
     SIZE to 64K or less in a CREATE LOGFILE GROUP statement led to
     failure of cluster data nodes.
   * NDB Cluster: The cluster waited 30 seconds instead of 30
     milliseconds before reading table statistics.
   * NDB Cluster: It was not possible to add a unique index to an
     NDB table while in single user mode.
   * NDB Cluster: Performing a delete followed by an insert during
     a local checkpoint could cause a Rowid already allocated
     error. (Bug#27205:
   * NDB Cluster (Disk Data): When restarting a data node following
     the creation of a large number (~200) of Disk Data objects,
     the cluster could not assign a node ID to the restarting node.
   * NDB Cluster: Adding of indexes online failed for NDB tables
     having BLOB or TEXT columns.
   * NDB Cluster: The utility did not calculate
     memory usage for indexes correctly.
   * NDB Cluster: While a data node was stopped, dropping a table
     then creating an index on a different table caused that node
     to fail during restart. This was due to the re-use of the
     dropped table's internal ID for the index without verifying
     that the index now referred to a different database object.
   * NDB Cluster (Disk Data): When in single user mode, it was
     possible to create log file groups and tablespaces from any
     SQL node connected to the cluster.
   * NDB Cluster (Disk Data): Changing a column specification or
     issuing a TRUNCATE statement on a Disk Data table caused the
     table to become an in-memory table.
     This fix supersedes an incomplete fix that was made for this
     issue in MySQL 5.1.15.
   * NDB Cluster (Replication): Setting SQL_LOG_BIN to zero did not
     disable binary logging.
     This issue affected only the NDB storage engine.
   * NDB Cluster (Replication): It was possible for API nodes to
     begin interacting with the cluster subscription manager before
     they were fully connected to the cluster.
   * NDB Cluster (Replication): Under very high loads, checkpoints
     could be read or written with checkpoint indexes out of order.
   * NDB Cluster: NDB tables having MEDIUMINT AUTO_INCREMENT
     columns were not restored correctly by ndb_restore, causing
     spurious duplicate key errors. This issue did not affect
   * NDB Cluster: NDB tables with indexes whose names contained
     space characters were not restored correctly by ndb_restore
     (the index names were truncated).
   * NDB Cluster (Disk Data): Changes to a Disk Data table made as
     part of a transaction could not be seen by the client
     performing the changes until the transaction had been
     committed. (Bug#27757:
   * NDB Cluster: An INSERT followed a delete DELETE on the same
     NDB table caused a memory leak.
     (This bug was introduced by the fix for
   * NDB Cluster: Some queries that updated multiple tables were
     not backed up or replicated correctly.
   * NDB Cluster: Joins on multiple tables containing BLOB columns
     could cause data nodes run out of memory, and to crash with
     the error NdbObjectIdMap::expand unable to expand.
   * NDB Cluster: When trying to create an NDB table after the
     server was started with --ndbcluster but without
     --ndb-connectstring, mysqld produced a memory allocation
     error. (Bug#27359:
   * NDB Cluster (APIs): Using NdbBlob::writeData() to write data
     in the middle of an existing blob value (that is, updating the
     value) could overwrite some data past the end of the data to
     be changed. (Bug#27018:
   * NDB Cluster (Replication): Trying to replicate a large number
     of frequent updates with a relatively small relay log
     (max-relay-log-size set to 1M or less) could cause the slave
     to crash. (Bug#27529:
   * NDB Cluster (Replication): An SQL node acting as a replication
     master server could be a single point of failure; that is, if
     it failed, the replication slave had no way of knowing this,
     which could result in a mismatch of data between the master
     and the slave. (Bug#21494:
   * NDB Cluster (Disk Data): CREATE TABLE ... LIKE disk_data_table
     created an in-memory NDB table.
   * NDB Cluster (Disk Data): Creating an excessive number of data
     files for a single tablespace caused data nodes to crash.
   * NDB Cluster: Under certain rare circumstances, DROP TABLE or
     TRUNCATE of an NDB table could cause a node failure or forced
     cluster shutdown. (Bug#27581:
   * NDB Cluster: Memory usage of a mysqld process grew even while
     idle. (Bug#27560:
   * NDB Cluster: Using more than 16GB for DataMemory caused
     problems with variable-size columns.
   * NDB Cluster: In an NDB table having a TIMESTAMP column using
     DEFAULT CURRENT_TIMESTAMP, that column would assume a random
     value when another column in the same row was updated.
   * NDB Cluster: Performing ALTER TABLE ... ENGINE=MERGE on an NDB
     table caused mysqld to crash.
   * NDB Cluster: The Cluster table handler did not set bits in
     null bytes correctly. (Bug#26591:
   * NDB Cluster: In some cases, AFTER UPDATE and AFTER DELETE
     triggers on NDB tables that referenced subject table did not
     see the results of operation which caused invocation of the
     trigger, but rather saw the row as it was prior to the update
     or delete operation.
     This was most noticeable when an update operation used a
     subquery to obtain the rows to be updated. An example would be
     UPDATE tbl1 SET col2 = val1 WHERE tbl1.col1 IN (SELECT col3
     FROM tbl2 WHERE c4 = val2) where there was an AFTER UPDATE
     trigger on table tbl1. In such cases, the trigger would fail
     to execute.
     The problem occurred because the actual update or delete
     operations were deferred to be able to perform them later as
     one batch. The fix for this bug solves the problem by
     disabling this optimization for a given update or delete if
     the table has an AFTER trigger defined for this operation.
   * NDB Cluster: START BACKUP NOWAIT caused a spurious Out of
     backup record error in the management client (START BACKUP and
     START BACKUP WAIT STARTED performed normally).
   * NDB Cluster: When trying to create tables on an SQL node not
     connected to the cluster, a misleading error message Table
     'tbl_name' already exists was generated. The error now
     generated is Could not connect to storage engine.
   * NDB Cluster: A data node failing while another data node was
     restarting could leave the cluster in an inconsistent state.
     In certain rare cases, this could lead to a race condition and
     the eventual forced shutdown of the cluster.
   * NDB Cluster: When using the MemReportFrequency configuration
     parameter to generate periodic reports of memory usage in the
     cluster log, DataMemory usage was not always reported for all
     data nodes. (Bug#27444:
   * NDB Cluster: Error messages displayed when running in single
     user mode were inconsistent.
   * NDB Cluster: On Solaris, the value of an NDB table column
     declared as BIT(33) was always displayed as 0.
   * NDB Cluster (Replication): An UPDATE on the master became a
     DELETE on slaves. (Bug#27378:
   * The fix forBug#17212:
     correct sort order for misordered output of certain queries,
     but caused significant overall query performance degradation.
     (Results were correct (good), but returned much more slowly
     (bad).) The fix also affected performance of queries for which
     results were correct. The performance degradation has been
     addressed. (Bug#27531:
   * On Windows, connection handlers did not properly decrement the
     server's thread count when exiting.
   * SELECT COUNT(*) from a table containing a DATETIME NOT NULL
     column could produce spurious warnings with the NO_ZERO_DATE
     SQL mode enabled. (Bug#22824:
   * Nested aggregate functions could be improperly evaluated.
   * Using CAST() to convert DATETIME values to numeric values did
     not work. (Bug#23656:
   * Early NULL-filtering optimization did not work for eq_ref
     table access. (Bug#27939:
   * Non-grouped columns were allowed by * in ONLY_FULL_GROUP_BY
     SQL mode. (Bug#27874:
   * Debug builds on Windows generated false alarms about
     uninitialized variables with some Visual Studio runtime
     libraries. (Bug#27811:
   * An error message suggested the use of
     mysql_fix_privilege_tables after an upgrade, but the
     recommended program is now mysql_upgrade.
   * mysqld did not check the length of option values and could
     crash with a buffer overflow for long values.
   * Index hints (USE INDEX, IGNORE INDEX, FORCE INDEX) cannot be
     used with FULLTEXT indexes, but were not being ignored.
   * mysql_upgrade did not detect failure of external commands that
     it runs. (Bug#26639:
   * mysql_upgrade did not pass a password to mysqlcheck if one was
     given. (Bug#25452:
   * On Windows, mysql_upgrade was sensitive to lettercase of the
     names of some required components.
   * The result set of a query that used WITH ROLLUP and DISTINCT
     could lack some rollup rows (rows with NULL values for
     grouping attributes) if the GROUP BY list contained constant
     expressions. (Bug#24856:
   * Some upgrade problems are detected and better error messages
     suggesting that mysql_upgrade be run are produced.
   * A performance degradation was observed for outer join queries
     to which a not-exists optimization was applied.
   * The LEAST() and GREATEST() functions compared DATE and
     DATETIME values as strings, which in some cases could lead to
     an incorrect result. (Bug#27759:
     failed with an Access denied error, even for a user who has
     the FILE privilege. (Bug#28181:
   * Certain queries that used uncorrelated scalar subqueries
     caused EXPLAIN to to crash.
   * INSERT...ON DUPLICATE KEY UPDATE could cause Error 1032: Can't
     find record in ... for inserts into an InnoDB table unique
     index using key column prefixes with an underlying utf8 string
     column. (Bug#13191:
   * On Linux, the server could not create temporary tables if
     lower_case_table_names was set to 1 and the value of tmpdir
     was a directory name containing any uppercase letters.
   * A slave that used --master-ssl-cipher could not connect to the
     master. (Bug#21611:
   * mysqldump crashed if it got no data from SHOW CREATE PROCEDURE
     (for example, when trying to dump a routine defined by a
     different user and for which the current user had no
     privileges). Now it prints a comment to indicate the problem.
     It also returns an error, or continues if the --force option
     is given. (Bug#27293:
   * Several math functions produced incorrect results for large
     unsigned values. ROUND() produced incorrect results or a crash
     for a large number-of-decimals argument.
   * For storage engines that allow the current auto-increment
     value to be set, using ALTER TABLE ... ENGINE to convert a
     table from one such storage engine to another caused loss of
     the current value. (For storage engines that do not support
     setting the value, it cannot be retained anyway when changing
     the storage engine.) (Bug#25262:
   * Comparison of a DATE with a DATETIME did not treat the DATE as
     having a time part of 00:00:00.
   * A multiple-table UPDATE could return an incorrect rows-matched
     value if, during insertion of rows into a temporary table, the
     table had to be converted from a MEMORY table to a MyISAM
     table. (Bug#22364:
   * The omission of leading zeros in dates could lead to erroneous
     results when these were compared with the output of certain
     date and time functions.
   * If CREATE TABLE t1 LIKE t2 failed due to a full disk, an empty
     t2.frm file could be created but not removed. This file then
     caused subsequent attempts to create a table named t2 to fail.
     This is easily corrected at the filesystem level by removing
     the t2.frm file manually, but now the server removes the file
     if the create operation does not complete successfully.
   * The MERGE storage engine could return incorrect results when
     several index values that compare equality were present in an
     index (for example, 'gross' and 'gross ', which are considered
     equal but have different lengths).
   * For InnoDB tables, a multiple-row INSERT of the form INSERT
     id=VALUES(id), where id is an AUTO_INCREMENT column, could
     cause ERROR 1062 (23000): Duplicate entry... errors or lost
     rows. (Bug#27650:
   * mysql_install_db is supposed to detect existing system tables
     and create only those that do not exist. Instead, it was
     exiting with an error if tables already existed.
   * Failure to allocate memory associated with
     transaction_prealloc_size could cause a server crash.
   * Aborting a statement on the master that applied to a
     non-transactional statement broke replication. The statement
     was written to the binary log but not completely executed on
     the master. Slaves receiving the statement executed it
     completely, resulting in loss of data synchrony. Now an error
     code is written to the error log so that the slaves stop
     without executing the aborted statement. (That is, replication
     stops, but synchrony to the point of the stop is preserved and
     you can investigate the problem.)
   * The AUTO_INCREMENT value would not be correctly reported for
     InnoDB tables when using SHOW CREATE TABLE statement or
     mysqldump command. (Bug#23313:
   * Creating a temporary table with InnoDB when using the
     one-file-per-table setting, when the host filesystem for
     temporary tables is tmpfs would cause an assertion in mysqld.
     This was due to the use of O_DIRECT when opening the temporary
     table file. (Bug#26662:
   * An interaction between SHOW TABLE STATUS and other concurrent
     statements that modify the table could result in a
     divide-by-zero error and a server crash.
   * The decimal.h header file was incorrectly omitted from binary
     distributions. (Bug#27456:
   * mysqldump could not connect using SSL.
   * yaSSL crashed on pre-Pentium Intel CPUs.
   * The test for the MYSQL_OPT_SSL_VERIFY_SERVER_CERT option for
     mysql_options() was performed incorrectly. Also changed as a
     result of this bugfix: The arg option for the mysql_options()
     C API function was changed from char * to void *.
   * Comparisons using row constructors could fail for rows
     containing NULL values.
   * Performing a UNION on two views that had had ORDER BY clauses
     resulted in an Unknown column error.
   * The CRC32() function returns an unsigned integer, but the
     metadata was signed, which could cause certain queries to
     return incorrect results. (For example, queries that selected
     a CRC32() value and used that value in the GROUP BY clause.)
   * A race condition between DROP TABLE and SHOW TABLE STATUS
     could cause the latter to display incorrect information.
   * mysqldump would not dump a view for which the DEFINER no
     longer exists. (Bug#26817:
   * Changing a utf8 column in an InnoDB table to a shorter length
     did not shorten the data values.
   * The server did not shut down cleanly.
   * Using SET GLOBAL to change the lc_time_names system variable
     had no effect on new connections.
   * The XML output representing an empty result was an empty
     string rather than an empty <resultset/> element.
   * The range optimizer could consume a combinatorial amount of
     memory for certain classes of WHERE clauses.
   * mysqlbinlog produced different output with the -R option than
     without it. (Bug#27171:
   * A stored function invocation in the WHERE clause was treated
     as a constant. (Bug#27354:
   * mysqldump could not dump the log tables.
   * Implicit conversion of 9912101 to DATE did not match
     CAST(9912101 AS DATE).
   * CURDATE() is less than NOW(), either when comparing CURDATE()
     directly (CURDATE() < NOW() is true) or when casting CURDATE()
     to DATE (CAST(CURDATE() AS DATE) < NOW() is true). However,
     storing CURDATE() in a DATE column and comparing col_name <
     NOW() incorrectly yielded false. This is fixed by comparing a
     DATE column as DATETIME for comparisons to a DATETIME
     constant. (Bug#21103:
   * For dates with 4-digit year parts less than 200, an incorrect
     implicit conversion to add a century was applied for date
     arithmetic performed with DATE_ADD(), DATE_SUB(), + INTERVAL,
     and - INTERVAL. (For example, DATE_ADD('0050-01-01 00:00:00',
     INTERVAL 0 SECOND) became '2050-01-01 00:00:00'.)
   * Some equi-joins containing a WHERE clause that included a NOT
     IN subquery caused a server crash.
   * A memory leak in the event scheduler that was uncovered by
     Valgrind was fixed. (Bug#27733:
   * For queries that used ORDER BY with InnoDB tables, if the
     optimizer chose an index for accessing the table but found a
     covering index that enabled the ORDER BY to be skipped, no
     results were returned.
   * Group relay log rotation updated only the log position and not
     the name, causing the slave to stop.
   * Conversion of DATETIME values in numeric contexts sometimes
     did not produce a double (YYYYMMDDHHMMSS.uuuuuu) value.
   * Passing nested row expressions with different structures to an
     IN predicate caused a server crash.
   * SELECT DISTINCT could return incorrect results if the select
     list contained duplicated columns.
   * A subquery could get incorrect values for references to outer
     query columns when it contained aggregate functions that were
     aggregated in outer context.
   * In some cases, the optimizer preferred a range or full index
     scan access method over lookup access methods when the latter
     were much cheaper. (Bug#19372:
   * Duplicates were not properly identified among (potentially)
     long strings used as arguments for GROUP_CONCAT(DISTINCT).
   * For InnoDB, fixed consistent-read behavior of the first read
     statement, if the read was served from the query cache, for
     the READ COMMITTED isolation level.
   * The binary log incompatiblity introduced by the fix for
    Bug#22583: corrected.
   * Row-based replication of MyISAM to non-MyISAM tables did not
     work correctly for BIT columns. This has been corrected, but
     the fix introduces an incompatibility into the binary log
     format. (Bug#22583: (The
     incompatibility is corrected by the fix for
   * Duplicate members in SET definitions were not detected. Now
     they result in a warning; if strict SQL mode is enabled, an
     error occurs instead. (Bug#27069:
   * If the name of a table given to myisamchk -rq was a packed
     table and the name included the .MYI extension, myisamchk
     incorrectly created a file with a .MYI.MYI extension.
   * For INSERT INTO ... SELECT where index searches used column
     prefixes, insert errors could occur when key value type
     conversion was done. (Bug#26207:
     was not always cleared properly.
   * On Windows, if the server was installed as a service, it did
     not auto-detect the location of the data directory.
   * The FEDERATED engine did not allow the local and remote tables
     to have different names.
   * The NO_DIR_IN_CREATE server SQL mode was not enforced for
     partitioned tables. (Bug#24633:
   * On Windows, trying to use backslash (\) characters in paths
     for DATA DIRECTORY and INDEX DIRECTORY when creating
     partitioned tables caused MySQL to crash.
     (You must use / characters when specifying paths for these
     options, regardless of platform. See Section 16.1, "Overview
     of Partitioning in MySQL," for an example using absolute paths
     for DATA DIRECTORY and INDEX DIRECTORY when creating a
     partitioned table on Windows.)
   * A damaged or missing mysql.event table caused SHOW VARIABLES
     to fail. (Bug#23631:
   * Database and table names have a maximum length of 64
     characters (even if they contain multi-byte characters), but
     were being truncated to 64 bytes.
   * If a rotate event occured in the middle of a non-transaction
     group, the group position would be updated by the rotate event
     indicating an illegal group start position that was
     effectively inside a group. This can happen if, for example, a
     rotate occurs between an Intvar event and the associated Query
     event, or between the table map events and the rows events
     when using row-based replication.
   * mysqldump could crash or exhibit incorrect behavior when some
     options were given very long values, such as
     --fields-terminated-by="some very long string". The code has
     been cleaned up to remove a number of fixed-sized buffers and
     to be more careful about error conditions in memory
     allocation. (Bug#26346:
   * Setting a column to NOT NULL with an ON DELETE SET NULL clause
     foreign key crashes the server.
   * The values displayed for the Innodb_row_lock_time,
     Innodb_row_lock_time_avg, and Innodb_row_lock_time_max status
     variables were incorrect.
   * COUNT(decimal_expr) sometimes generated a spurious truncation
     warning. (Bug#21976:
   * With NO_AUTO_VALUE_ON_ZERO SQL mode enabled, LOAD DATA
     operations could assign incorrect AUTO_INCREMENT values.
   * Incorrect results could be returned for some queries that
     contained a select list expression with IN or BETWEEN together
     with an ORDER BY or GROUP BY on the same expression using NOT
     IN or NOT BETWEEN. (Bug#27532:
     tables, some status values were incorrectly converted to the
     data type of the VARIABLE_VALUE column.
   * Queries containing subqueries with COUNT(*) aggregated in an
     outer context returned incorrect results. This happened only
     if the subquery did not contain any references to outer
     columns. (Bug#27257:
   * Use of an aggregate function from an outer context as an
     argument to GROUP_CONCAT() caused a server crash.
   * Restoration of the default database after stored routine or
     trigger execution on a slave could cause replication to stop
     if the database no longer existed.
   * String truncation upon insertion into an integer or year
     column did not generate a warning (or an error in strict
     mode). (Bug#26359:,
   * In out-of-memory conditions, the server might crash or
     otherwise not report an error to the Windows event log.
   * The temporary file-creation code was cleaned up on Windows to
     improve server stability.
   * Out-of-memory errors were not reported. Now they are written
     to the error log. (Bug#26844:
   * mysqldump crashed for MERGE tables if the --complete-insert
     (-c) option was given.
   * Corrupted MyISAM tables that have different definitions in the
     .frm and .MYI tables might cause a server crash.
   * In certain situations, MATCH ... AGAINST returned false hits
     for NULL values produced by LEFT JOIN when no full-text index
     was available. (Bug#25729:
   * OPTIMIZE TABLE might fail on Windows when it attempts to
     rename a temporary file to the original name if the original
     file had been opened, resulting in loss of the .MYD file.
   * GRANT statements were not replicated if the server was started
     with the --replicate-ignore-table or
     --replicate-wild-ignore-table option.
   * MBROverlaps() returned incorrect values in some cases.
   * A problem in handling of aggregate functions in subqueries
     caused predicates containing aggregate functions to be ignored
     during query execution.
   * Improved out-of-memory detection when sending logs from a
     master server to slaves, and log a message when allocation
     fails. (Bug#26837:
   * MBRDisjoint(), MBRequal(), MBRIntersects(), MBROverlaps(),
     MBRTouches(), and MBRWithin() were inadvertently omitted from
     recent versions of MySQL (5.1.14 to 5.1.17).
   * SHOW CREATE VIEW qualified references to stored functions in
     the view definition with the function's database name, even
     when the database was the default database. This affected
     mysqldump (which uses SHOW CREATE VIEW to dump views) because
     the resulting dump file could not be used to reload the
     database into a different database. SHOW CREATE VIEW now
     suppresses the database name for references to functions in
     the default database. (Bug#23491:
   * When MySQL logged slow query information to a CSV table, it
     used an incorrect formula to calculate the query_time and
     lock_time values. (Bug#27638:
   * With innodb_file_per_table enabled, attempting to rename an
     InnoDB table to a non-existent database caused the server to
     exit. (Bug#27381:
   * mysql_install_db could terminate with an error after failing
     to determine that a system table already existed.
   * For InnoDB tables having a clustered index that began with a
     CHAR or VARCHAR column, deleting a record and then inserting
     another before the deleted record was purged could result in
     table corruption. (Bug#26835:
   * make_win_bin_dist neglected to copy some required MyISAM table
     files. (Bug#26922:
   * Fixed a possible buffer overflow in SHOW PROCEDURE CODE.
   * Selecting the result of AVG() within a UNION could produce
     incorrect values. (Bug#24791:
   * Access via my_pread() or my_pwrite() to table files larger
     than 2GB could fail on some systems.
   * An INTO OUTFILE clause is allowed only for the final SELECT of
     a UNION, but this restriction was not being enforced
     correctly. (Bug#23345:
   * Duplicate entries were not assessed correctly in a MEMORY
     table with a BTREE primary key on a utf8 ENUM column.
   * For MyISAM tables, COUNT(*) could return an incorrect value if
     the WHERE clause compared an indexed TEXT column to the empty
     string (''). This happened if the column contained empty
     strings and also strings starting with control characters such
     as tab or newline. (Bug#26231:
   * For DELETE FROM tbl_name ORDER BY col_name (with no WHERE or
     LIMIT clause), the server did not check whether col_name was a
     valid column in the table.
   * ALTER VIEW requires the CREATE VIEW and DROP privileges for
     the view. However, if the view was created by another user,
     the server erroneously required the SUPER privilege.
   * mysqlbinlog --base64-output produced invalid SQL.
   * In a view, a column that was defined using a GEOMETRY function
     was treated as having the LONGBLOB data type rather than the
     GEOMETRY type. (Bug#27300:
   * Some views could not be created even when the user had the
     requisite privileges. (Bug#24040:
   * With the NO_AUTO_VALUE_ON_ZERO SQL mode enabled,
     LAST_INSERT_ID() could return 0 after INSERT ... ON DUPLICATE
     KEY UPDATE. Additionally, the next rows inserted (by the same
     INSERT, or the following INSERT with or without ON DUPLICATE
     KEY UPDATE), would insert 0 for the auto-generated value if
     the value for the AUTO_INCREMENT column was NULL or missing.
   * Executing an INSERT ... SELECT ... FROM
     INFORMATION_SCHEMA.GLOBAL_STATUS statement from within an
     event caused a server crash.
   * Having the EXECUTE privilege for a routine in a database
     should make it possible to USE that database, but the server
     returned an error instead. This has been corrected. As a
     result of the change, SHOW TABLES for a database in which you
     have only the EXECUTE privilege returns an empty set rather
     than an error. (Bug#9504:
   * When RAND() was called multiple times inside a stored
     procedure, the server did not write the correct random seed
     values to the binary log, resulting in incorrect replication.
   * SOUNDEX() returned an invalid string for international
     characters in multi-byte character sets.
   * Row equalities in WHERE clauses could cause memory corruption.
   * GROUP BY on a ucs2 column caused a server crash when there was
     at least one empty string in the column.
   * Evaluation of an IN() predicate containing a decimal-valued
     argument caused a server crash.
   * Storing NULL values in spatial fields caused excessive memory
     allocation and crashes on some systems.
   * mysql_stmt_fetch() did an invalid memory deallocation when
     used with the embedded server.
   * For FEDERATED tables, SHOW CREATE TABLE could fail when the
     table name was longer than the connection name.
   * In a MEMORY table, using a BTREE index to scan for updatable
     rows could lead to an infinite loop.
   * The range optimizer could cause the server to run out of
     memory. (Bug#26625:
     any privileges. Now these statements require the SUPER
     privilege. (Bug#25671:
   * Concurrent CREATE SERVER and ALTER SERVER statements could
     cause a deadlock. (Bug#25721:
   * Difficult repair or optimization operations could cause an
     assertion failure, resulting in a server crash.