MySQL Community Server 5.0.81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管理システムの新バージョンであるMySQL Community Server 5.0.81がリリースされました。MySQL Community Serverの本リリースは、MySQL Enterprise Server 5.0.80コードベースに基づいています。MySQL Community Server 5.0リリースシリーズにおける今後のリリースはMySQL Enterprise Serverのカウンターパートに対応するバージョン番号を共有する予定です。
MySQL 5.0.81は、http://dev.mysql.com/downloads/およびミラーサイトから、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。




MySQL 5.0.81は、AIX 5.2向けのMySQL 5.0としての最終リリースとなります。AIX 5.2向けのEOL(End of Life) を2009年4月30日に迎える予定です。

以下のセクションは、以前のリリースであるMySQL Community Server 5.0.77以来のMySQLソースコードにおけるバージョン間の変更を記載しています。より多くの修正ログ修正を含む変更ログは以下で確認できます。


MySQL 5.0.81は、現在の製品リリースに対するバグ修正版であり、MySQL 5.0.77に代わるものです。


Functionality added or changed:

      * Security Enhancement: To enable stricter control over the
        location from which user-defined functions can be loaded,
        the plugin_dir system variable has been backported from
        MySQL 5.1. If the value is non-empty, user-defined
        function object files can be loaded only from the
        directory named by this variable. If the value is empty,
        the behavior that is used prior to the inclusion of
        plugin_dir applies: The UDF object files must be located
        in a directory that is searched by your system's dynamic
        linker. (Bug#37428:http://bugs.mysql.com/37428)

      * A new status variable, Queries, indicates the number of
        statements executed by the server. This includes
        statements executed within stored programs, unlike the
        Questions variable which includes only statements sent to
        the server by clients.

      * 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 are honored.

    Bugs fixed:

      * Important Change: Security Fix: Additional corrections
        were made for the symlink-related privilege problem
        originally addressed in MySQL 5.0.60. The original fix
        did not correctly handle the data directory path name if
        it contained symlinked directories in its path, and the
        check was made only at table-creation time, not at
        table-opening time later.
        See alsoBug#39277:http://bugs.mysql.com/39277.

      * Security Enhancement: The server consumed excess memory
        while parsing statements with hundreds or thousands of
        nested boolean conditions (such as OR (OR ... (OR ...
        ))). This could lead to a server crash or incorrect
        statement execution, or cause other client statements to
        fail due to lack of memory. The latter result constitutes
        a denial of service.

      * Incompatible Change: There were some problems using
        DllMain() hook functions on Windows that automatically do
        global and per-thread initialization for libmysqld.dll:

           + Per-thread initialization: MySQL internally counts
             the number of active threads, which causes a delay
             in my_end() if not all threads have exited. But
             there are threads that can be started either by
             Windows internally (often in TCP/IP scenarios) or by
             users. Those threads do not necessarily use
             libmysql.dll functionality but still contribute to
             the open-thread count. (One symptom is a five-second
             delay in times for PHP scripts to finish.)

           + Process-initialization: my_init() calls WSAStartup
             that itself loads DLLs and can lead to a deadlock in
             the Windows loader.
        To correct these problems, DLL initialization code now is
        not invoked from libmysql.dll by default. To obtain the
        previous behavior (DLL initialization code will be
        called), set the LIBMYSQL_DLLINIT environment variable to
        any value. This variable exists only to prevent breakage
        of existing Windows-only applications that do not call
        mysql_thread_init() and work okay today. Use of
        LIBMYSQL_DLLINIT is discouraged and is removed in MySQL
        6.0. (Bug#37226:http://bugs.mysql.com/37226,

      * Incompatible Change: SHOW STATUS took a lot of CPU time
        for calculating the value of the
        Innodb_buffer_pool_pages_latched status variable. Now
        this variable is calculated and included in the output of
        SHOW STATUS only if the UNIV_DEBUG symbol is defined at
        MySQL build time.

      * Incompatible Change: In connection with view creation,
        the server created arc directories inside database
        directories and maintained useless copies of .frm files
        there. Creation and renaming procedures of those copies
        as well as creation of arc directories has been
        This change does cause a problem when downgrading to
        older server versions which manifests itself under these

          1. Create a view v_orig in MySQL 5.0.72 or higher.

          2. Rename the view to v_new and then back to v_orig.

          3. Downgrade to an older 5.0.x server and run

          4. Try to rename v_orig to v_new again. This operation
        As a workaround to avoid this problem, use either of
        these approaches:

           + Dump your data using mysqldump before downgrading
             and reload the dump file after downgrading.

           + Instead of renaming a view after the downgrade, drop
             it and recreate it.
        The downgrade problem introduced by the fix for this bug
        has been addressed as

      * Replication: When rotating relay log files, the slave
        deletes relay log files and then edits the relay log
        index file. Formerly, if the slave shut down unexpectedly
        between these two events, the relay log index file could
        then reference relay logs that no longer existed.
        Depending on the circumstances, this could when
        restarting the slave cause either a race condition or the
        failure of replication.

      * In example option files provided in MySQL distributions,
        the thread_stack value was increased from 64K to 128K.

      * SET PASSWORD caused a server crash if the account name
        was given as CURRENT_USER().

        limited to 7680 rows.

      * In debug builds, obsolete debug code could be used to
        crash the server.

      * CHECK TABLE ... FOR UPGRADE did not check for
        incompatible collation changes made in MySQL 5.0.48
       Bug#29499:http://bugs.mysql.com/29499). This also
        affects mysqlcheck and mysql_upgrade, which cause that
        statement to be executed. See Section 2.18.3, "Checking
        Whether Table Indexes Must Be Rebuilt."
        See alsoBug#39585:http://bugs.mysql.com/39585.

      * Some queries that used a "range checked for each record"
        scan could return incorrect results.

      * Certain SELECT queries could fail with a Duplicate entry
        error. (Bug#40953:http://bugs.mysql.com/40953)

      * The FEDERATED handler had a memory leak.

      * IF(..., CAST(longtext_val AS UNSIGNED), signed_val) as an
        argument to an aggregate function could cause an
        assertion failure.

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

      * 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.

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

      * 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.

      * 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.

      * The server crashed if an integer field in a CSV file did
        not have delimiting quotes.

      * CHECK TABLE failed for MyISAM INFORMATION_SCHEMA tables.

      * InnoDB could hang trying to open an adaptive hash index.

      * For a TIMESTAMP column in an InnoDB table, testing the
        column with multiple conditions in the WHERE clause
        caused a server crash.

      * 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.

      * References to local variables in stored procedures are
        replaced with NAME_CONST(name, value) when written to the
        binary log. However, an "illegal mix of collation" error
        might occur when executing the log contents if the
        value's collation differed from that of the variable. Now
        information about the variable collation is written as
        well. (Bug#39182:http://bugs.mysql.com/39182)

      * Some recent releases for Solaris 10 were built on Solaris
        10 U5, which included a new version of libnsl.so that
        does not work on U4 or earlier. To correct this, Solaris
        10 builds now are created on machines that do not have
        that upgraded libnsl.so, so that they will work on
        Solaris 10 installations both with and without the
        upgraded libnsl.so.

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

      * Queries of the form SELECT ... REGEXP BINARY NULL could
        lead to a hung or crashed server.

      * Statements of the form INSERT ... SELECT .. ON DUPLICATE
        KEY UPDATE col_name = DEFAULT could result in a server
        crash. (Bug#39002:http://bugs.mysql.com/39002)

      * 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.

      * Repeated CREATE TABLE ... SELECT statements, where the
        created table contained an AUTO_INCREMENT column, could
        lead to an assertion failure.

      * 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. (Bug#38693:http://bugs.mysql.com/38693)

      * A server crash resulted from concurrent execution of a
        multiple-table UPDATE that used a NATURAL or USING join
        for the table being updated.

      * On ActiveState Perl, mysql-test-run.pl --start-and-exit
        started but did not exit.

      * Server-side cursors were not initialized properly, which
        could cause a server crash.

      * Stored procedures involving substrings could crash the
        server on certain platforms due to invalid memory reads.

      * A server crash or Valgrind warnings could result when a
        stored procedure selected from a view that referenced a
        function. (Bug#38291:http://bugs.mysql.com/38291)

      * Incorrect handling of aggregate functions when loose
        index scan was used caused a server crash.

      * Queries containing a subquery with DISTINCT and ORDER BY
        could cause a server crash.

      * Queries with a HAVING clause could return a spurious row.

      * Use of spatial data types in prepared statements could
        cause memory leaks or server crashes.

      * The server crashed if an argument to a stored procedure
        was a subquery that returned more than one row.

      * When analyzing the possible index use cases, the server
        was incorrectly reusing an internal structure, leading to
        a server crash. (Bug#37943:http://bugs.mysql.com/37943)

      * 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.

      * For InnoDB tables, ORDER BY ... DESC sometimes returned
        results in ascending order.

      * If a table has a BIT NOT NULL column c1 with a length
        shorter than 8 bits and some additional NOT NULL columns
        c2, ..., and a SELECT query has a WHERE clause of the
        form (c1 = constant) AND c2 ..., the query could return
        an unexpected result set.

      * Nesting of IF() inside of SUM() could cause an extreme
        server slowdown. (Bug#37662:http://bugs.mysql.com/37662)

      * The MONTHNAME() and DAYNAME() functions returned a binary
        string, so that using LOWER() or UPPER() had no effect.
        Now MONTHNAME() and DAYNAME() return a value in
        character_set_connection character set.

      * TIMEDIFF() was erroneously treated as always returning a
        positive result. Also, CAST() of TIME values to DECIMAL
        dropped the sign of negative values.
        See alsoBug#42525:http://bugs.mysql.com/42525.

      * 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.

      * The <=> operator could return incorrect results when
        comparing NULL to DATE, TIME, or DATETIME values.

      * 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. (Bug#37339:http://bugs.mysql.com/37339)

      * For a MyISAM table with CHECKSUM = 1 and ROW_FORMAT =
        DYNAMIC table options, a data consistency check (maximum
        record length) could fail and cause the table to be
        marked as corrupted.

      * The max_length result set metadata value was calculated
        incorrectly under some circumstances.

      * CREATE INDEX could crash with InnoDB plugin 1.0.1.

      * Certain boolean-mode FULLTEXT searches that used the
        truncation operator did not return matching records and
        calculated relevance incorrectly.

      * The NO_BACKSLASH_ESCAPES SQL mode was ignored for LOAD
        DATA INFILE and SELECT INTO ... OUTFILE. The setting is
        taken into account now.

      * 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.

      * If the server failed to expire binary log files at
        startup, it could crash.

      * The code for the ut_usectime() function in InnoDB did not
        handle errors from the gettimeofday() system call. Now it
        retries gettimeofday() several times and updates the
        value of the Innodb_row_lock_time_max status variable
        only if ut_usectime() was successful.

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

      * A query which had an ORDER BY DESC clause that is
        satisfied with a reverse range scan could cause a server
        crash for some specific CPU/compiler combinations.

      * Dumping information about locks in use by sending a
        SIGHUP signal to the server or by invoking the mysqladmin
        debug command could lead to a server crash in debug
        builds or to undefined behavior in production builds.

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

      * When the fractional part in a multiplication of DECIMAL
        values overflowed, the server truncated the first operand
        rather than the longest. Now the server truncates so as
        to produce more precise multiplications.

      * A read past the end of the string could occur while
        parsing the value of the --innodb-data-file-path option.

      * Host name values in SQL statements were not being checked
        for '@', which is illegal according to RFC952.

      * The UUID() function returned UUIDs with the wrong time;
        this was because the offset for the time part in UUIDs
        was miscalculated.

      * SHOW CREATE TABLE did not display a printable value for
        the default value of BIT columns.

      * mysql_install_db failed on machines that had the host
        name set to localhost.

      * Dynamic plugins failed to load on i5/OS.

      * Freeing of an internal parser stack during parsing of
        complex stored programs caused a server crash.

      * The max_length metadata value was calculated incorrectly
        for the FORMAT() function, which could cause incorrect
        result set metadata to be sent to clients.

      * Index scans performed with the sort_union() access method
        returned wrong results, caused memory to be leaked, and
        caused temporary files to be deleted when the limit set
        by sort_buffer_size was reached.

      * If the server crashed with an InnoDB error due to
        unavailability of undo slots, errors could persist during
        rollback when the server was restarted: There are two
        UNDO slot caches (for INSERT and UPDATE). If all slots
        end up in one of the slot caches, a request for a slot
        from the other slot cache would fail. This can happen if
        the request is for an UPDATE slot and all slots are in
        the INSERT slot cache, or vice versa.

      * For InnoDB tables, ALTER TABLE DROP failed if the name of
        the column to be dropped began with "foreign".

      * perror on Windows did not know about Win32 system error
        codes. (Bug#34825:http://bugs.mysql.com/34825)

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

      * Queries of the form SELECT ... WHERE string = ANY(...)
        failed when the server used a single-byte character set
        and the client used a multi-byte character set.
        See alsoBug#20835:http://bugs.mysql.com/20835.

      * Using OPTIMIZE TABLE as the first statement on an InnoDB
        table with an AUTO_INCREMENT column could cause a server
        crash. (Bug#34286:http://bugs.mysql.com/34286)

      * mysql_install_db failed if the server was running with an
        SQL mode of TRADITIONAL. This program now resets the SQL
        mode internally to avoid this problem.

      * Changes to build files were made to enable the MySQL
        distribution to compile on Microsoft Visual C++ Express
        2008. (Bug#33907:http://bugs.mysql.com/33907)

      * The mysql client incorrectly parsed statements containing
        the word "delimiter" in mid-statement.
        This fix is different from the one applied for this bug
        in MySQL 5.0.66. (Bug#33812:http://bugs.mysql.com/33812)
        See alsoBug#38158:http://bugs.mysql.com/38158.

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

      * 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.

      * Cached queries that used 256 or more tables were not
        properly cached, so that later query invalidation due to
        a TRUNCATE TABLE for one of the tables caused the server
        to hang. (Bug#33362:http://bugs.mysql.com/33362)

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

      * mysql_upgrade attempted to use the /proc file system even
        on systems that do not have it.

      * mysqldump could fail to dump views containing a large
        number of columns.

      * Queries executed using join buffering of BIT columns
        could produce incorrect results.

        TINYTEXT or MEDIUMTEXT columns to a longer text type if
        necessary when converting the column to a different
        character set. (Bug#31291:http://bugs.mysql.com/31291)

      * For installation on Solaris using pkgadd packages, the
        mysql_install_db script was generated in the scripts
        directory, but the temporary files used during the
        process were left there and not deleted.

      * Several MySQL programs could fail if the HOME environment
        variable had an empty value.

      * On NetWare, mysql_install_db could appear to execute
        normally even if it failed to create the initial
        databases. (Bug#30129:http://bugs.mysql.com/30129)

      * The Serbian translation for the
        ER_INCORRECT_GLOBAL_LOCAL_VAR error was corrected.

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

      * The BUILD/check-cpu build script failed if gcc had a
        different name (such as gcc.real on Debian).

      * On Windows, Visual Studio does not take into account some
        x86 hardware limitations, which led to incorrect results
        converting large DOUBLE values to unsigned BIGINT values.

      * SSL support was not included in some "generic" RPM
        packages. (Bug#26760:http://bugs.mysql.com/26760)

      * In some cases, the parser interpreted the ; character as
        the end of input and misinterpreted stored program
        definitions. (Bug#26030:http://bugs.mysql.com/26030)

      * The Questions status variable is intended as a count of
        statements sent by clients to the server, but was also
        counting statements executed within stored routines.

      * For access to the INFORMATION_SCHEMA.VIEWS table, the
        server did not check the SHOW VIEW and SELECT privileges,
        leading to inconsistency between output from that table
        and the SHOW CREATE VIEW statement.

      * The FLUSH PRIVILEGES statement did not produce an error
        when it failed. (Bug#21226:http://bugs.mysql.com/21226)

      * 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.

      * mysqld_safe would sometimes fail to remove the pid file
        for the old mysql process after a crash. As a result, the
        server would fail to start due to a false A mysqld
        process already exists... error.