MySQL 6.0.9アルファ版がリリースされました。


MySQLデータベースシステムの新バージョンであるMySQL 6.0.9アルファ版がリリースされました。MySQL 6.0に関する主要なページは、http://www.mysql.com/mysql60/にあります。

結合テーブルと結合バッファの両方へのインデックスアクセスを使用する新たなアルゴリズムが実装されました。Batched Key Access (BKA) Joinアルゴリズムと呼んでいます。当アルゴリズムは、ネストされた外部結合とネストされたsemi-joinを含む、内部結合、外部結合、およびsemi-joinの操作をサポートします。内部結合に対してのみ以前に使用されたBlock Nested Loop Joinアルゴリズムは改良され、ネストされた外部結合とネストされたsemi-joinを含む、外部結合、およびsemi-joinの操作も加わりました。

MySQL 6.0は、トランザクションのFalconエンジンと、クラッシュ・セーフなMariaエンジンの二つの新たなストレージエンジンを含んでいます。

Falconストレージエンジンが初めてで、より多くの情報が必要な場合、Falcon Evaluation Guide

と、Falcon White Paper

Mariaストレージエンジンは、MyISAMのクラッシュ・セーフなバージョンです。MariaはMyISAMエンジンの主な機能のすべてをサポートします。それだけでなく、システムクラッシュが発生した場合のリカバリサポート、CREATE, DROP, RENAME, TRUNCATEを含む十分なロギング機能、全てのMyISAMテーブルのrow format、Maria固有の新たなrow formatもMariaはサポートしています。Mariaについてはhttp://dev.mysql.com/doc/refman/6.0/en/se-maria.htmlに記述されています。

MySQL 6.0.9アルファ版は、http://dev.mysql.com/downloads/mysql/6.0.htmlとミラーサイトのダウンロード・ページから、多くのプラットフォーム用のソースコード及びバイナリで現在利用可能です。



MySQL 6.0の以前のバージョン以来の重要な変更、互換性のない変更、およびセキュリティ変更を以下のセクションに記載します。より多くの修正を含む全ての変更ログはhttp://dev.mysql.com/doc/refman/6.0/en/news-6-0-9.htmlで見ることができます。


   Functionality added or changed:

     * BACKUP DATABASE and RESTORE now indicate in the server's error
       log which databases are being backed up or restored.

     * Performance of SELECT * retrievals from
       INFORMATION_SCHEMA.COLUMNS was improved slightly.

     * Previously, index hints did not work for FULLTEXT searches.
       Now they work as follows:
       For natural language mode searches, index hints are silently
       ignored. For example, IGNORE INDEX(i) is ignored with no
       warning and the index is still used.
       For boolean mode searches, index hints with FOR ORDER BY or
       FOR GROUP BY are silently ignored. Index hints with FOR JOIN
       or no FOR modifier are honored. In contrast to how hints apply
       for non-FULLTEXT searches, the hint is used for all phases of
       query execution (finding rows and retrieval, grouping, and
       ordering). This is true even if the hint is given for a
       non-FULLTEXT index. (Bug#38842:http://bugs.mysql.com/38842)

     * MySQL support for adding collations using LDML specifications
       did not support the <i> identity rule that indicates one
       character sorts identically to another. The <i> rule now is
       supported. (Bug#37129:http://bugs.mysql.com/37129)

     * Previously, RESTORE overwrote any databases with information
       from the backup image. Now, RESTORE aborts with an error if
       the backup image contains any databases that currently exist
       on the server, unless the optional keyword OVERWRITE is given
       following the image filename.

     * A new statement, PURGE BACKUP LOGS, enables the contents of
       the MySQL Backup logs to be culled. See Section,
       "PURGE BACKUP LOGS Syntax."

   Bugs fixed:

     * Security Enhancement: When the DATA DIRECTORY or INDEX
       DIRECTORY clause of a CREATE TABLE statement referred to a
       subdirectory of the data directory via a symlinked component
       of the data directory path, it was accepted, when for security
       reasons it should be rejected.

     * Incompatible Change: CHECK TABLE ... FOR UPGRADE did not check
       for collation changes made in MySQL 6.0.1 to latin2_czech_cs
       or collation changes made in MySQL 6.0.6 to big5_chinese_ci,
       cp866_general_ci, gb2312_chinese_ci, and gbk_chinese_ci. (This
       also affects mysqlcheck and mysql_upgrade, which cause that
       statement to be executed.)

     * Partitioning: Replication: Changing the transaction isolation
       level while replicating partitioned InnoDB tables could cause
       statement-based logging to fail.

     * Partitioning: This bug was introduced in MySQL 6.0.5.
       This regression was introduced by

     * Partitioning: With READ COMMITTED transaction isolation level,
       InnoDB uses a semi-consistent read that releases non-matching
       rows after MySQL has evaluated the WHERE clause. However, this
       was not happening if the table used partitions.

     * Partitioning: A SELECT using a range WHERE condition with an
       ORDER BY on a partitioned table caused a server crash.

     * Partitioning: For a partitioned table having an AUTO_INCREMENT
       column: If the first statement following a start of the server
       or a FLUSH TABLES statement was an UPDATE statement, the
       AUTO_INCREMENT column was not incremented correctly.

     * Partitioning: The server attempted to execute the statements
       ... REORGANIZE PARTITION on tables that were not partitioned.
       See alsoBug#20129:http://bugs.mysql.com/20129.

     * Partitioning: The value of the CREATE_COLUMNS column in
       INFORMATION_SCHEMA.TABLES was not partitioned for partitioned
       tables. (Bug#38909:http://bugs.mysql.com/38909)

     * Partitioning: When executing an ORDER BY query on a
       partitioned InnoDB table using an index that was not in the
       partition expression, the results were sorted on a
       per-partition basis rather than for the table as a whole.

     * Partitioning: Partitioned table checking sometimes returned a
       warning with an error code of 0, making proper response to
       errors impossible. The fix also renders the error message
       subject to translation in non-English deployments.

     * Partitioning: When SHOW CREATE TABLE was used on a partitioned
       table, all of the table's PARTITION and SUBPARTITION clauses
       were output on a single line, making it difficult to read or
       parse. (Bug#14326:http://bugs.mysql.com/14326)

     * Replication: Row-based replication failed with non-partitioned
       MyISAM tables having no indexes.

     * An assertion failure occurred for a join query when a small
       size of the join buffer was set and the value of
       record_per_key for the index used for a ref access with this
       join buffer was big enough.

     * Unique indexes on FALCON tables can not be created when the
       the column is NOT NULL.

     * Accessing user variables within triggers could cause a server
       crash. (Bug#40770:http://bugs.mysql.com/40770)

     * For single-table UPDATE statements, an assertion failure
       resulted from a runtime error in a stored function (such as a
       recursive function call or an attempt to update the same table
       as in the UPDATE statement).

     * Date values of 000-00-00 inserted into a FALCON table were
       incorrectly recognized and returned when performing a SELECT
       on a field with an index.

     * Several MySQL Backup-related memory-use issues identified by
       Valgrind were corrected.

     * When executing concurrent CREATE TABLE ... SELECT statements
       on a Maria table, the error Error: Memory allocated at
       trnman.c:129 was underrun, discovered at ma_close.c:65 error
       would be logged in the error file, and the server would
       eventually crash. (Bug#40416:http://bugs.mysql.com/40416)

     * Prepared statements allowed invalid dates to be inserted when
       the ALLOW_INVALID_DATES SQL mode was not enabled.

     * With statement-based binary logging format and a transaction
       isolation level of READ COMMITTED or stricter, InnoDB printed
       an error because statement-based logging might lead to
       inconsistency between master and slave databases. However,
       this error was printed even when binary logging was not
       enabled (in which case, no such inconsistency can occur).

     * A query with an outer join where the ON expression evaluated
       to the constant FALSE could return incorrect results when a
       join buffer was used for the outer join operation.

     * Errors from a BACKUP DATABASE or RESTORE operation were shown
       by SHOW WARNINGS as warnings, not errors.

     * If several errors occurred during a BACKUP DATABASE or RESTORE
       operation, the final error was returned to the client, even
       though the first error is usually more pertinent.

     * Creation of a tablespace file within FALCON could create a
       tablespace entry in the
       data file had not been created.

     * Specifying the --log-backup-output option without an argument
       set the destination for the backup logs to FILE rather than to
       the default of TABLE. (Bug#40282:http://bugs.mysql.com/40282)

     * mc.exe is no longer needed to compile MySQL on Windows. This
       makes it possible to build MySQL from source using Visual
       Studio Express 2008. (Bug#40280:http://bugs.mysql.com/40280)

     * The server could generate extra rows in the result set for a
       query with a nested outer join if the inner tables of the
       outer join were joined using join buffers.

     * If BACKUP DATABASE was used to back up an empty database and
       binary logging enabled, the backup image was flagged as
       containing binary log information even though it did not.
       Using RESTORE with the backup image then crashed trying to
       parse the binary log filename.

     * The backup_history_log_file and backup_progress_log_file
       system variables were not settable at server startup. Now they
       are. (Bug#40219:http://bugs.mysql.com/40219)

     * The default value of the backup_history_log and
       backup_progress_log system variables is ON, but explicitly
       setting them to DEFAULT set them to OFF.

     * When an outer join employed a join buffer to join the first
       inner table by the Blocked Nested Loops algorithm, extra
       NULL-complemented rows could be generated if the WHERE clause
       contained conditions that can be pushed down to this table.

     * When the optimizer joined an inner table of an outer join
       using both "not exists" optimization and a join buffer, an
       incorrect result set could be returned.

     * Support for the revision field in .frm files has been removed.
       This addresses the downgrading problem introduced by the fix

     * The MySQL Backup message logger caused an assertion failure.

     * Retrieval speed from the following INFORMATION_SCHEMA tables
       was improved by shortening the VARIABLE_VALUE column to 1024
       As a result of this change, any variable value longer than
       1024 characters will be truncated with a warning. This affects
       only the init_connect system variable.

     * If the operating system is configured to return leap seconds
       from OS time calls or if the MySQL server uses a time zone
       definition that has leap seconds, functions such as NOW()
       could return a value having a time part that ends with :59:60
       or :59:61. If such values are inserted into a table, they
       would be dumped as is by mysqldump but considered invalid when
       reloaded, leading to backup/restore problems.
       Now leap second values are returned with a time part that ends
       with :59:59. This means that a function such as NOW() can
       return the same value for two or three consecutive seconds
       during the leap second. It remains true that literal temporal
       values having a time part that ends with :59:60 or :59:61 are
       considered invalid.
       For additional details about leap-second handling, see Section
       9.7.2, "Time Zone Leap Second Support."

     * The server could crash during a sort-order optimization of a
       dependent subquery. (Bug#39844:http://bugs.mysql.com/39844)

     * Recovery of a tablespace for FALCON tables could fail if the
       tablespace was already in use.

     * Creating a FALCON table while specifying a specific tablespace
       and partition to be used for the table will fail if the
       specified tablespace does not already exist, returning a error
       indicating general table creation failure. The message has
       been updated to indicate that the failure is due to
       non-existent tablespace.

     * With the ONLY_FULL_GROUP_BY SQL mode enabled, the check for
       non-aggregated columns in queries with aggregate functions,
       but without a GROUP BY clause was treating all the parts of
       the query as if they were in the select list. This is fixed by
       ignoring the non-aggregated columns in the WHERE clause.

     * Concurrent execution of BACKUP DATABASE and DML operations on
       MyISAM tables could produce a deadlock.


     * On 64-bit Windows systems, the server accepted key_buffer_size
       values larger than 4GB, but allocated less. (For example,
       specifying a value of 5GB resulted in 1GB being allocated.)

     * Compiling MySQL with FALCON support enabled with a compiler
       that does not support exceptions would fail to complete
       successfully. configure has been updated to switch off FALCON
       support if the specified compiler does not support exceptions.

     * Use of the PACK_KEYS or MAX_ROWS table option in ALTER TABLE
       should have triggered table reconstruction but did not.

     * The server returned a column type of VARBINARY rather than
       DATE as the result from the COALESCE(), IFNULL(), IF(),
       GREATEST(), or LEAST() functions or CASE expression if the
       result was obtained using filesort in an anonymous temporary
       table during the query execution.

     * Starting MySQL with FALCON support when MySQL has not been
       compiled with a compiler supporting exceptions would lead to
       strange errors and results. MySQL will now fail to initialize
       if you have compiled without exceptions enabled with the
       following message: 081116 12:21:12 [ERROR] Falcon must be
       compiled with C++ exceptions enabled to work. Please adjust
       your compile flags.  [Falcon] Error: Falcon exiting process

     * A server built using yaSSL for SSL support would crash if
       configured to use an RSA key and a client sent a cipher list
       containing a non-RSA key as acceptable.

     * When built with Valgrind, the server failed to access tables
       created with the DATA DIRECTORY or INDEX DIRECTORY table
       option. (Bug#39102:http://bugs.mysql.com/39102)

     * With binary logging enabled CREATE VIEW was subject to
       possible buffer overwrite and a server crash.

     * The fast mutex implementation was subject to excessive lock
       contention. (Bug#38941:http://bugs.mysql.com/38941)

     * Use of InnoDB monitoring (SHOW ENGINE INNODB STATUS or one of
       the InnoDB Monitor tables) could cause a server crash due to
       invalid access to a shared variable in a concurrent
       environment. (Bug#38883:http://bugs.mysql.com/38883)

     * Column names constructed due to wild-card expansion done
       inside a stored procedure could point to freed memory if the
       expansion was performed after the first call to the stored
       procedure. (Bug#38823:http://bugs.mysql.com/38823)

     * If delayed insert failed to upgrade the lock, it did not free
       the temporary memory storage used to keep newly constructed
       BLOB values in memory, resulting in a memory leak.

     * The server unnecessarily acquired a query cache mutex even
       with the query cache disabled, resulting in a small
       performance decrement.

     * On Windows, a five-second delay occurred at shutdown of
       applications that used the embedded server.

     * On Solaris, a scheduling policy applied to the main server
       process could be unintentionally overwritten in
       client-servicing threads.

     * BACKUP DATABASE failed to use the native driver for a Falcon
       table if the table was partitioned.

     * On Windows, the embedded server would crash in
       mysql_library_init() if the language file was missing.

     * The Event Scheduler no longer logs "started in thread" or
       "executed" successfully messages to the error log.

     * Setting the debug system variable and executing a SELECT
       statement resulted in a Valgrind warning.

     * An incorrectly checked XOR subquery optimization resulted in
       an assertion failure. (Bug#37899:http://bugs.mysql.com/37899)

     * A SELECT with a NULL NOT IN condition containing a complex
       subquery from the same table as in the outer select caused an
       assertion failure. (Bug#37894:http://bugs.mysql.com/37894)

     * Use of an uninitialized constant in EXPLAIN evaluation caused
       an assertion failure. (Bug#37870:http://bugs.mysql.com/37870)

     * A query that could use one index to produce the desired
       ordering and another index for range access with index
       condition pushdown could cause a server crash.

     * Renaming an ARCHIVE table to the same name with different
       lettercase and then selecting from it could cause a server
       crash. (Bug#37719:http://bugs.mysql.com/37719)

     * For queries executed with the batched-key access method, an
       incorrect value of an internal parameter caused a server
       crashe if join_buffer_size was less then 256.

     * Compiling MySQL with FALCON support enabled on Solaris 9 using
       the Sun Studio compiler would fail with error: "Interlock.h",
       line 149: Error: #error cas not defined. We need>= Solaris 10.

     * TIMEDIFF() was erroneously treated as always returning a
       positive result. Also, CAST() of TIME values to DECIMAL
       dropped the sign of negative values.

     * mysqlcheck used SHOW FULL TABLES to get the list of tables in
       a database. For some problems, such as an empty .frm file for
       a table, this would fail and mysqlcheck then would neglect to
       check other tables in the database.

     * Updating a view with a subquery in the CHECK option could
       cause an assertion failure.

     * Statements that displayed the value of system variables (for
       example, SHOW VARIABLES) expect variable values to be encoded
       in character_set_system. However, variables set from the
       command line such as basedir or datadir were encoded using
       character_set_filesystem and not converted correctly.

     * On a 32-bit server built without big tables support, the
       offset argument in a LIMIT clause might be truncated due to a
       64-bit to 32-bit cast.

     * Specifying a database name twice to BACKUP DATABASE caused a
       server crash. Now BACKUP DATABASE ignores duplicate names.

     * If a non-directory file f without an extension was created in
       the data directory, the server would allow clients to execute
       a USE f statement even though f could not be a database. The
       server now verifies that that the named database corresponds
       to a directory. (Bug#36897:http://bugs.mysql.com/36897)

     * The FALCON storage would silently recreate missing tablespace
       files if they did not exist. Errors are now written to the
       MySQL error log when the FALCON system tablespace files are
       found to be missing. Warnings are produce in the log file when
       attempting to access data tablespace files that do not exist.

     * Use of CONVERT() with GROUP BY to convert numeric values to
       CHAR could return truncated results.

     * The mysql client, when built with Visual Studio 2005, did not
       display Japanese characters.

     * Setting the slave_compressed_protocol system variable to
       DEFAULT failed in the embedded server.

     * Processing for NULL-complemented rows in the result sets of
       queries with nested outer joins could be incorrect.

     * The columns that store character set and collation names in
       several INFORMATION_SCHEMA tables were lengthened because they
       were not long enough to store some possible values: SCHEMATA,

     * Queries executed using the batched-key access method could
       cause an assertion fail when key expressions for a ref access
       depended on columns not only from the previous join table.

     * Selecting from an INFORMATION_SCHEMA table into an incorrectly
       defined MERGE table caused an assertion failure.

     * perror on Windows did not know about Win32 system error codes.

     * EXPLAIN EXTENDED evaluation of aggregate functions that
       required a temporary table caused a server crash.

     * BACKUP DATABASE produced an incorrect error message when the
       backup image filename contained a non-existent directory.

     * SHOW GLOBAL STATUS shows values that aggregate the session
       status values for all threads. This did not work correctly for
       the embedded server. (Bug#34517:http://bugs.mysql.com/34517)

     * There were spurious warnings about "Truncated incorrect DOUBLE
       value" in queries with MATCH ... AGAINST and > or < with a
       constant (which was reported as an incorrect DOUBLE value) in
       the WHERE condition. (Bug#34374:http://bugs.mysql.com/34374)

     * mysqldumpslow did not aggregate times.

     * mysql_config did not output -ldl (or equivalent) when needed
       for --libmysqld-libs, so its output could be insufficient to
       build applications that use the embedded server.

     * For a stored procedure containing a SELECT * ... RIGHT JOIN
       query, execution failed for the second call.

       REFERENTIAL_CONSTRAINTS.DATA_TYPE columns were declared longer
       than the maximum allowed identifier length.

     * If a TEMPORARY table existed with the same name as a regular
       table, BACKUP DATABASE saved the temporary table, causing a
       subsequent RESTORE to fail.

     * Previously, use of index hints with views (which do not have
       indexes) produced the error ERROR 1221 (HY000): Incorrect
       usage of USE/IGNORE INDEX and VIEW. Now this produces ERROR
       1176 (HY000): Key '...' doesn't exist in table '...', the same
       error as for base tables without an appropriate index.

     * Execution of a prepared statement that referred to a system
       variable caused a server crash.

     * Some division operations produced a result with incorrect
       precision. (Bug#31616:http://bugs.mysql.com/31616)

     * Server variables could not be set to their current values on
       Linux platforms. (These fixes are in addition to those made in
       MySQL 6.0.5.) (Bug#31177:http://bugs.mysql.com/31177)
       See alsoBug#6958:http://bugs.mysql.com/6958.

     * Searching for text values on a column using a character set
       that provides multi-weight characters and sequences on an
       INNODB or FALCON table with an index would fail to find the
       expanded value. (Bug#29246:http://bugs.mysql.com/29246)

     * Some SHOW statements and retrievals from the
       INFORMATION_SCHEMA TRIGGERS and EVENTS tables used a temporary
       table and incremented the Created_tmp_disk_tables status
       variable, due to the way that TEXT columns are handled. The
       columns now are VARCHAR to avoid this problem.

     * XA transaction rollbacks could result in corrupted transaction
       states and a server crash.

     * There were cases where string-to-number conversions would
       produce warnings for CHAR values but not for VARCHAR values.

     * For several read only system variables that were viewable with
       SHOW VARIABLES, attempting to view them with SELECT @@var_name
       or set their values with SET resulted in an unknown system
       variable error. Now they can be viewed with SELECT @@var_name
       and attempting to set their values results in a message
       indicating that they are read only.

     * ALTER TABLE for an ENUM column could change column values.

     * Setting the session value of the max_allowed_packet or
       net_buffer_length system variable was allowed but had no
       effect. The session value of these variables is now read only.

     * A race condition between the mysqld.exe server and the Windows
       service manager could lead to inability to stop the server
       from the service manager.