MySQL 5.1.23-rc has been released (part 2)

[Joerg Bruehe]

Dear MySQL users,

We are proud to present to you the MySQL Server 5.1.23-rc release,
a new "release candidate" version of the popular open source database.

Bear in mind that this is still a "candidate" release, and as with any
other pre-production release, caution should be taken when installing on
production level systems or systems with critical data. For production
level systems using 5.0, we would like to direct your attention to the
product description of MySQL Enterprise at:

             http://mysql.com/products/enterprise/

The MySQL 5.1.23-rc release is now available in source and binary form
for a number of platforms from our download pages at

             http://dev.mysql.com/downloads/

and mirror sites. Note that not all mirror sites may be up to date at
this point in time, so if you can't find this version on some mirror,
please try again later or choose another download site.

Please also note that some of our mirrors are currently experiencing
problems that may result in serving corrupted files. We are working with
the mirror maintainers to resolve this.

We welcome and appreciate your feedback, bug reports, bug fixes,
patches etc.:

             http://forge.mysql.com/wiki/Contributing

The description of the changes from version 5.1.22-rc to this 5.1.23-rc
is some 1,800 lines long, that is about 96 kB. As some mail systems are
bound to truncate long mail at 64 kB, I split the announcement into
three parts - this is part 2 only.

To ensure these important items do not get lost, I repeat the notes
about functionality changes, the security fixes, and other changes
labeled "important". All these were also listed in part 1.
If you want to skip this duplication, search for the text
     "This ends the duplicated part."

Following this, there is the next set of changes from version to version
in the MySQL source code since the latest released version of MySQL 5.1,
the MySQL 5.1.22-rc release. It can also be viewed online at

             http://dev.mysql.com/doc/refman/5.1/en/news-5-1-23.html


Functionality added or changed:
   * Important Change: Partitioning: Security Fix: It was
     possible, by creating a partitioned table using the DATA
     DIRECTORY and INDEX DIRECTORY options to gain privileges
     on other tables having the same name as the partitioned
     table. As a result of this fix, any table-level DATA
     DIRECTORY or INDEX DIRECTORY options are now ignored for
     partitioned tables.
     (Bug#32091: http://bugs.mysql.com/32091, CVE-2007-5970
     (http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2007-5970))
     See also Bug#29325: http://bugs.mysql.com/29325,
     Bug#32111: http://bugs.mysql.com/32111
   * Incompatible Change: In MySQL 5.1.6, when log tables were
     implemented, the default log destination for the general
     query and slow query log was TABLE. This default has been
     changed to FILE, which is compatible with MySQL 5.0, but
     incompatible with earlier releases of MySQL 5.1 from
     5.1.6 to 5.1.20. If you are upgrading from MySQL 5.0 to
     this release, no logging option changes should be
     necessary. However, if you are upgrading from 5.1.6
     through 5.1.20 to this release and were using TABLE
     logging, use the --log-output=TABLE option explicitly to
     preserve your server's table-logging behavior.
     In MySQL 5.1.x, this bug was addressed twice because it
     turned out that the default was set in two places, only
     one of which was fixed the first time.
     (Bug#29993: http://bugs.mysql.com/29993)
   * Incompatible Change
     The parser accepted statements that contained /* ... */
     that were not properly closed with */, such as SELECT 1
     /* + 2. Statements that contain unclosed /*-comments now
     are rejected with a syntax error.
     This fix has the potential to cause incompatibilities.
     Because of Bug#26302: http://bugs.mysql.com/26302, which
     caused the trailing */ to be truncated from comments in
     views, stored routines, triggers, and events, it is
     possible that objects of those types may have been stored
     with definitions that now will be rejected as
     syntactically invalid. Such objects should be dropped and
     re-created so that their definitions do not contain
     truncated comments.
     (Bug#28779: http://bugs.mysql.com/28779)
   * MySQL Cluster: The following improvements have been made
     in the ndb_size.pl utility:
        + The script can now be used with multiple databases;
          lists of databases and tables can also be excluded
          from analysis.
        + Schema name information has been added to index
          table calculations.
        + The database name is now an optional parameter, the
          exclusion of which causes all databases to be
          examined.
        + If selecting from INFORMATION_SCHEMA fails, the
          script now attempts to fall back to SHOW TABLES.
        + A --real_table_name option has been added; this
          designates a table to handle unique index size
          calculations.
        + The report title has been amended to cover cases
          where more than one database is being analyzed.
     Support for a --socket option was also added.
     For more information, see Section 15.9.15, "ndb_size.pl
     --- NDBCluster Size Requirement Estimator."
     (Bug#28683: http://bugs.mysql.com/28683,
     Bug#28253: http://bugs.mysql.com/28253)
   * MySQL Cluster: Mapping of NDB error codes to MySQL
     storage engine error codes has been improved.
     (Bug#28423: http://bugs.mysql.com/28423)
   * MySQL Cluster: The output from the cluster management
     client showing the progress of data node starts has been
     improved. (Bug#23354: http://bugs.mysql.com/23354)
   * Partitioning: Error messages for partitioning syntax
     errors have been made more descriptive.
     (Bug#29368: http://bugs.mysql.com/29368)
   * Replication: Replication of the following now switches to
     row-based logging in MIXED mode, and generates a warning
     in STATEMENT mode:
        + USER()
        + CURRENT_USER()
        + CURRENT_USER
        + FOUND_ROWS()
        + ROW_COUNT()
     See Section 5.2.4.3, "Mixed Binary Logging (MBL) Format,"
     for more information.
     (Bug#12092: http://bugs.mysql.com/12092,
     Bug#28086: http://bugs.mysql.com/28086,
     Bug#30244: http://bugs.mysql.com/30244)
   * mysqltest now has a change_user command to change the
     user for the current connection. (It invokes the
     mysql_change_user() C API function.)
     (Bug#31608: http://bugs.mysql.com/31608)
   * mysql-test-run.pl now allows a suite name prefix to be
     specified in command-line arguments that name test cases.
     The test name syntax now is
     [suite_name.]test_name[.suffix]. For example,
     mysql-test-run.pl binlog.mytest runs the mytest.test test
     in the binlog test suite.
     (Bug#31400: http://bugs.mysql.com/31400)
   * The --event-scheduler option without a value disabled the
     event scheduler. Now it enables the event scheduler.
     (Bug#31332: http://bugs.mysql.com/31332)
   * mysqldump produces a -- Dump completed on DATE comment at
     the end of the dump if --comments is given. The date
     causes dump files for identical data take at different
     times to appear to be different. The new options
     --dump-date and --skip-dump-date control whether the date
     is added to the comment. --skip-dump-date suppresses date
     printing. The default is --dump-date (include the date in
     the comment). (Bug#31077: http://bugs.mysql.com/31077)
   * Server parser performance was improved for expression
     parsing by lowering the number of state transitions and
     reductions needed.
     (Bug#30625: http://bugs.mysql.com/30625)
   * Server parser performance was improved for identifier
     lists, expression lists, and UDF expression lists.
     (Bug#30333: http://bugs.mysql.com/30333)
   * Server parser performance was improved for boolean
     expressions. (Bug#30237: http://bugs.mysql.com/30237)
   * The LAST_EXECUTED column of the INFORMATION_SCHEMA.EVENTS
     table now indicates when the event started executing
     rather than when it finished executing. As a result, the
     ENDS column is never less than LAST_EXECUTED.
     (Bug#29830: http://bugs.mysql.com/29830)
   * The mysql_odbc_escape_string() C API function has been
     removed. It has multi-byte character escaping issues,
     doesn't honor the NO_BACKSLASH_ESCAPES SQL mode and is
     not needed anymore by Connector/ODBC as of 3.51.17.
     (Bug#29592: http://bugs.mysql.com/29592)
   * If a MyISAM table is created with no DATA DIRECTORY
     option, the .MYD file is created in the database
     directory. By default, if MyISAM finds an existing .MYD
     file in this case, it overwrites it. The same applies to
     .MYI files for tables created with no INDEX DIRECTORY
     option. To suppress this behavior, start the server with
     the new --keep_files_on_create option, in which case
     MyISAM will not overwrite existing files and returns an
     error instead. (Bug#29325: http://bugs.mysql.com/29325)
   * The default value of the connect_timeout system variable
     was increased from 5 to 10 seconds. This might help in
     cases where clients frequently encounter errors of the
     form Lost connection to MySQL server at 'XXX', system
     error: errno. (Bug#28359: http://bugs.mysql.com/28359)
   * MySQL now can be compiled with gcc 4.2.x. There was a
     problem involving a conflict with the min() and max()
     macros in my_global.h.
     (Bug#28184: http://bugs.mysql.com/28184)

Bugs fixed:
   * Security Fix: Replication: It was possible for any
     connected user to issue a BINLOG statement, which could
     be used to escalate privileges.
     Use of the BINLOG statement now requires the SUPER
     privilege. (Bug#31611: http://bugs.mysql.com/31611,
     CVE-2007-6313
     (http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2007-6
     313))
   * Security Fix: Three vulnerabilities in yaSSL versions
     1.7.5 and earlier were discovered that could lead to a
     server crash or execution of unauthorized code. The
     exploit requires a server with yaSSL enabled and TCP/IP
     connections enabled, but does not require valid MySQL
     account credentials. The exploit does not apply to
     OpenSSL.
     Note
        The proof-of-concept exploit is freely available on the
        Internet. Everyone with a vulnerable MySQL configuration
        is advised to upgrade immediately.
        (Bug#33814: http://bugs.mysql.com/33814, CVE-2008-0226
        (http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2008-0226),
        CVE-2008-0227
        (http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2008-0227))
   * Security Fix: Using RENAME TABLE against a table with
     explicit DATA DIRECTORY and INDEX DIRECTORY options can
     be used to overwrite system table information by
     replacing the symbolic link points. the file to which the
     symlink points.
     MySQL will now return an error when the file to which the
     symlink points already exists.
     (Bug#32111: http://bugs.mysql.com/32111, CVE-2007-5969
     (http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2007-5969))
   * Security Fix: ALTER VIEW retained the original DEFINER
     value, even when altered by another user, which could
     allow that user to gain the access rights of the view.
     Now ALTER VIEW is allowed only to the original definer or
     users with the SUPER privilege.
     (Bug#29908: http://bugs.mysql.com/29908)
   * Security Fix: When using a FEDERATED table, the local
     server could be forced to crash if the remote server
     returned a result with fewer columns than expected.
     (Bug#29801: http://bugs.mysql.com/29801)
   * Security Enhancement: It was possible to force an error
     message of excessive length which could lead to a buffer
     overflow. This has been made no longer possible as a
     security precaution.
     (Bug#32707: http://bugs.mysql.com/32707)
   * Important Change: Incompatible Change: A number of
     problems existed in the implementation of MERGE tables
     that could cause problems. The problems are summarized
     below:
        + Bug#26379: http://bugs.mysql.com/26379 - Combination
          of FLUSH TABLE and REPAIR TABLE corrupts a MERGE
          table. This was caused in a number of situations:
            1. A thread trying to lock a MERGE table performs
               busy waiting while REPAIR TABLE or a similar
               table administration task is ongoing on one or
               more of its MyISAM tables.
            2. A thread trying to lock a MERGE table performs
               busy waiting until all threads that did REPAIR
               TABLE or similar table administration tasks on
               one or more of its MyISAM tables in LOCK TABLES
               segments do UNLOCK TABLES. The difference
               against problem #1 is that the busy waiting
               takes place after the administration task. It
               is terminated by UNLOCK TABLES only.
            3. Two FLUSH TABLES within a LOCK TABLES segment
               can invalidate the lock. This does not require
               a MERGE table. The first FLUSH TABLES can be
               replaced by any statement that requires other
               threads to reopen the table. In 5.0 and 5.1 a
               single FLUSH TABLES can provoke the problem.
        + Bug#26867: http://bugs.mysql.com/26867 -
          Simultaneously executing LOCK TABLES and REPAIR
          TABLE on a MERGE table would result in memory/cpu
          hogging.
          Trying DML on a MERGE table, which has a child
          locked and repaired by another thread, made an
          infinite loop in the server.
        + Bug#26377: http://bugs.mysql.com/26377 - Deadlock
          with MERGE and FLUSH TABLE
          Locking a MERGE table and its children in
          parent-child order and flushing the child deadlocked
          the server.
        + Bug#25038: http://bugs.mysql.com/25038 - Waiting
          TRUNCATE
          Truncating a MERGE child, while the MERGE table was
          in use, let the truncate fail instead of waiting for
          the table to become free.
        + Bug#25700: http://bugs.mysql.com/25700 - MERGE base
          tables get corrupted by OPTIMIZE/ANALYZE/REPAIR
          TABLE
          Repairing a child of an open MERGE table corrupted
          the child. It was necessary to FLUSH the child
          first.
        + Bug#30275: http://bugs.mysql.com/30275 - MERGE
          tables: FLUSH TABLES or UNLOCK TABLES causes server
          to crash.
          Flushing and optimizing locked MERGE children
          crashed the server.
        + Bug#19627: http://bugs.mysql.com/19627 - temporary
          merge table locking
          Use of a temporary MERGE table with non-temporary
          children could corrupt the children.
          Temporary tables are never locked. Creation of
          tables with non-temporary children of a temporary
          MERGE table is now prohibited.
        + Bug#27660: http://bugs.mysql.com/27660 - Falcon:
          MERGE table possible
          It was possible to create a MERGE table with
          non-MyISAM children.
        + Bug#30273: http://bugs.mysql.com/30273 - MERGE
          tables: Can't lock file (errno: 155)
          This was a Windows-only bug. Table administration
          statements sometimes failed with "Can't lock file
          (errno: 155)".
     The fix introduces the following changes in behavior:
        + This patch changes the behavior of temporary MERGE
          tables. Temporary MERGE must have temporary
          children. The old behavior was wrong. A temporary
          table is not locked. Hence even non-temporary
          children were not locked. See
          Bug#19627: http://bugs.mysql.com/19627.
        + You cannot change the union list of a non-temporary
          MERGE table when LOCK TABLES is in effect. The
          following does not work:
          CREATE TABLE m1 ... ENGINE=MRG_MYISAM ...;
          LOCK TABLES t1 WRITE, t2 WRITE, m1 WRITE;
          ALTER TABLE m1 ... UNION=(t1,t2) ...;
          However, you can do this with a temporary MERGE
          table.
        + You cannot create a MERGE table with CREATE ...
          SELECT, neither as a temporary MERGE table, nor as a
          non-temporary MERGE table. For example:
          CREATE TABLE m1 ... ENGINE=MRG_MYISAM ... SELECT ...;
          gives error message: table is not BASE TABLE.
     (Bug#19627: http://bugs.mysql.com/19627,
     Bug#25038: http://bugs.mysql.com/25038,
     Bug#25700: http://bugs.mysql.com/25700,
     Bug#26377: http://bugs.mysql.com/26377,
     Bug#26379: http://bugs.mysql.com/26379,
     Bug#26867: http://bugs.mysql.com/26867,
     Bug#27660: http://bugs.mysql.com/27660,
     Bug#30275: http://bugs.mysql.com/30275,
     Bug#30491: http://bugs.mysql.com/30491)
   * Incompatible Change: It is no longer possible to create
     CSV tables with NULL columns. However, for backwards
     compatibility, you can continue to use such tables that
     were created in previous MySQL releases.
     (Bug#32050: http://bugs.mysql.com/32050)
   * Incompatible Change: Inserting a row with a NULL value
     for a DATETIME column results in a CSV file that the
     storage engine cannot read.
     All CSV tables now need to be defined with each column
     marked as NOT NULL. An error is raised if you try to
     create a CSV table with columns that are not defined with
     NOT NULL. (Bug#31473: http://bugs.mysql.com/31473,
     Bug#32817: http://bugs.mysql.com/32817)
   * Incompatible Change: SET PASSWORD statements now cause an
     implicit commit, and thus are prohibited within stored
     functions and triggers.
     (Bug#30904: http://bugs.mysql.com/30904)
   * Incompatible Change: The mysql_install_db script could
     fail to locate some components (including resolveip)
     during execution if the --basedir option was specified on
     the command-line or within the my.cnf file. This was due
     to a conflict when comparing the compiled-in values and
     the supplied values.
     The --source-install command-line option to the script
     has been removed and replaced with the --srcdir option.
     mysql_install_db now locates components either using the
     compiled-in options, the --basedir option or --srcdir
     option. (Bug#30759: http://bugs.mysql.com/30759)
   * Incompatible Change: Within a stored routine, it is no
     longer allowable to declare a cursor for a SHOW
     statement. This happened to work in some instances, but
     is no longer supported.
     (Bug#29223: http://bugs.mysql.com/29223)
   * Incompatible Change: GRANT and REVOKE statements now
     cause an implicit commit, and thus are prohibited within
     stored functions and triggers.
     (Bug#21975: http://bugs.mysql.com/21975,
     Bug#21422: http://bugs.mysql.com/21422,
     Bug#17244: http://bugs.mysql.com/17244)
   * Incompatible Change: It was possible for option files to
     be read twice at program startup, if some of the standard
     option file locations turned out to be the same
     directory. Now duplicates are removed from the list of
     files to be read.
     Also, users could not override system-wide settings using
     ~/.my.cnf because SYSCONFDIR/my.cnf was read last. The
     latter file now is read earlier so that ~/.my.cnf can
     override system-wide settings.
     (Bug#20748: http://bugs.mysql.com/20748)
   * Important Change: MySQL Cluster: AUTO_INCREMENT columns
     had the following problems when used in NDB tables:
        + The AUTO_INCREMENT counter was not updated correctly
          when such a column was updated.
        + AUTO_INCREMENT values were not prefetched beyond
          statement boundaries.
        + AUTO_INCREMENT values were not handled correctly
          with INSERT IGNORE statements.
        + After being set, ndb_autoincrement_prefetch_sz
          showed a value of 1, regardless of the value it had
          actually been set to.
     As part of this fix, the behavior of
     ndb_autoincrement_prefetch_sz has changed. Setting this
     to less than 32 no longer has any effect on prefetching
     within statements (where IDs are now always obtained in
     batches of 32 or more), but only between statements. The
     default value for this variable has also changed, and is
     now 1. (Bug#25176: http://bugs.mysql.com/25176,
     Bug#31956: http://bugs.mysql.com/31956,
     Bug#32055: http://bugs.mysql.com/32055)
   * Partitioning: Important Note: An apostrophe or single
     quote character (') used in the DATA DIRECTORY, INDEX
     DIRECTORY, or COMMENT for a PARTITION clause caused the
     server to crash. When used as part of a CREATE TABLE
     statement, the crash was immediate. When used in an ALTER
     TABLE statement, the crash did not occur until trying to
     perform a SELECT or DML statement on the table. In either
     case, the server could not be completely restarted until
     the .FRM file corresponding to the newly created or
     altered table was deleted.
     Note
        Upgrading to the current (or later) release solves this
        problem only for tables that are newly created or
        altered. Tables created or altered in previous versions
        of the server to include ' characters in PARTITION
        options must still be removed by deleting the
        corresponding .FRM files and re-creating them afterwards.
        (Bug#30695: http://bugs.mysql.com/30695)
   * Important Note: The RENAME DATABASE statement was removed
     and replaced with ALTER DATABASE db_name UPGRADE DATA
     DIRECTORY NAME. The RENAME DATABASE statement was
     intended for upgrading database directory names to the
     encoding format used in 5.1 for representing identifiers
     in the filesystem (see Section 7.2.3, "Mapping of
     Identifiers to Filenames"). However, the statement was
     found to be dangerous because it could result in loss of
     database contents. See Section 11.1.18, "RENAME DATABASE
     Syntax" and Section 11.1.1, "ALTER DATABASE Syntax."
     (Bug#17565: http://bugs.mysql.com/17565,
     Bug#21741: http://bugs.mysql.com/21741,
     Bug#28360: http://bugs.mysql.com/28360)

This ends the duplicated part.
The remaining bug fixes were not listed in part 1 of the announcement:


   * The server crashed on optimizations involving a join of
     INT and MEDIUMINT columns and a system variable in the
     WHERE clause. (Bug#32103: http://bugs.mysql.com/32103)
   * mysql-test-run.pl used the --user option when starting
     mysqld, which produces warnings if the current user is
     not root. Now --user is added only for root.
     (Bug#32078: http://bugs.mysql.com/32078)
   * mysqlslap was missing from the MySQL 5.1.22 Linux RPM
     packages. (Bug#32077: http://bugs.mysql.com/32077)
   * With lower_case_table_names set, CREATE TABLE LIKE was
     treated differently by libmysqld than by the non-embedded
     server. (Bug#32063: http://bugs.mysql.com/32063)
   * Within a subquery, UNION was handled differently than at
     the top level, which could result in incorrect results or
     a server crash. (Bug#32036: http://bugs.mysql.com/32036,
     Bug#32051: http://bugs.mysql.com/32051)
   * On 64-bit platforms, assignments of values to
     enumeration-valued storage engine-specific system
     variables were not validated and could result in
     unexpected values.
     (Bug#32034: http://bugs.mysql.com/32034)
   * A DELETE statement with a subquery in the WHERE clause
     would sometimes ignore an error during subquery
     evaluation and proceed with the delete operation.
     (Bug#32030: http://bugs.mysql.com/32030)
   * Using dates in the range '0000-00-01' to '0000-00-99'
     range in the WHERE clause could result in an incorrect
     result set. (These dates are not in the supported range
     for DATE, but different results for a given query could
     occur depending on position of records containing the
     dates within a table.)
     (Bug#32021: http://bugs.mysql.com/32021)
   * User-defined functions are not loaded if the server is
     started with the --skip-grant-tables option, but the
     server did not properly handle this case and issued an
     Out of memory error message instead.
     (Bug#32020: http://bugs.mysql.com/32020)
   * If a user-defined function was used in a SELECT
     statement, and an error occurred during UDF
     initialization, the error did not terminate execution of
     the SELECT, but rather was converted to a warning.
     (Bug#32007: http://bugs.mysql.com/32007)
   * HOUR(), MINUTE(), and SECOND() could return non-zero
     values for DATE arguments.
     (Bug#31990: http://bugs.mysql.com/31990)
   * Changing the SQL mode to cause dates with "zero" parts to
     be considered invalid (such as '1000-00-00') could result
     in indexed and non-indexed searches returning different
     results for a column that contained such dates.
     (Bug#31928: http://bugs.mysql.com/31928)
   * The server used unnecessarily large amounts of memory
     when user variables were used as an argument to CONCAT()
     or CONCAT_WS(). (Bug#31898: http://bugs.mysql.com/31898)
   * In debug builds, testing the result of an IN subquery
     against NULL caused an assertion failure.
     (Bug#31884: http://bugs.mysql.com/31884)
   * SHOW CREATE TRIGGER caused a server crash.
     (Bug#31866: http://bugs.mysql.com/31866)
   * The server crashed after insertion of a negative value
     into an AUTO_INCREMENT column of an InnoDB table.
     (Bug#31860: http://bugs.mysql.com/31860)
   * For libmysqld applications, handling of
     mysql_change_user() calls left some pointers improperly
     updated, leading to server crashes.
     (Bug#31850: http://bugs.mysql.com/31850)
   * Using ORDER BY led to the wrong result when using the
     ARCHIVE on a table with a BLOB when the table cache was
     full. The table could also be reported as crashed after
     the query had completed, even though the table data was
     intact. (Bug#31833: http://bugs.mysql.com/31833)
   * Comparison results for BETWEEN were different from those
     for operators like < and > for DATETIME-like values with
     trailing extra characters such as '2007-10-01 00:00:00
     GMT-6'. BETWEEN treated the values as DATETIME, whereas
     the other operators performed a binary-string comparison.
     Now they all uniformly use a DATETIME comparison, but
     generate warnings for values with trailing garbage.
     (Bug#31800: http://bugs.mysql.com/31800)
   * Name resolution for correlated subqueries and HAVING
     clauses failed to distinguish which of two was being
     performed when there was a reference to an outer aliased
     field. This could result in error messages about a HAVING
     clause for queries that had no such clause.
     (Bug#31797: http://bugs.mysql.com/31797)
   * With ONLY_FULL_GROUP_BY SQL mode enabled, queries such as
     SELECT a FROM t1 HAVING COUNT(*)>2 were not being
     rejected as they should have been.
     (Bug#31794: http://bugs.mysql.com/31794)
   * The server could crash during filesort for ORDER BY based
     on expressions with INET_NTOA() or OCT() if those
     functions returned NULL.
     (Bug#31758: http://bugs.mysql.com/31758)
   * For a fatal error during filesort in find_all_keys(), the
     error was returned without the necessary handler
     uninitialization, causing an assertion failure. Fixed by
     uninitializing the handler before returning the error.
     (Bug#31742: http://bugs.mysql.com/31742)
   * mysqlslap failed to commit after the final record load.
     (Bug#31704: http://bugs.mysql.com/31704)
   * The examined-rows count was not incremented for const
     queries. (Bug#31700: http://bugs.mysql.com/31700)
   * The server crashed if a thread was killed while locking
     the general_log table at the beginning of statement
     processing. (Bug#31692: http://bugs.mysql.com/31692)
   * The mysql_change_user() C API function was subject to
     buffer overflow. (Bug#31669: http://bugs.mysql.com/31669)
   * For SELECT ... INTO OUTFILE, if the ENCLOSED BY string is
     empty and the FIELDS TERMINATED BY string started with a
     special character (one of n, t, r, b, 0, Z, or N), every
     occurrence of the character within field values would be
     duplicated. (Bug#31663: http://bugs.mysql.com/31663)
   * SHOW COLUMNS and DESCRIBE displayed null as the column
     type for a view with no valid definer. This caused
     mysqldump to produce a non-reloadable dump file for the
     view. (Bug#31662: http://bugs.mysql.com/31662)
   * The mysqlbug script did not include the correct values of
     CFLAGS and CXXFLAGS that were used to configure the
     distribution. (Bug#31644: http://bugs.mysql.com/31644)
   * Queries that include a comparison of an
     INFORMATION_SCHEMA table column to NULL caused a server
     crash. (Bug#31633: http://bugs.mysql.com/31633)
   * EXPLAIN EXTENDED for SELECT from INFORMATION_SCHEMA
     tables caused an assertion failure.
     (Bug#31630: http://bugs.mysql.com/31630)
   * ucs2 does not work as a client character set, but
     attempts to use it as such were not rejected. Now
     character_set_client cannot be set to ucs2. This also
     affects statements such as SET NAMES and SET CHARACTER
     SET. (Bug#31615: http://bugs.mysql.com/31615)
   * A buffer used when setting variables was not dimensioned
     to accommodate the trailing '\0' byte, so a single-byte
     buffer overrun was possible.
     (Bug#31588: http://bugs.mysql.com/31588)
   * HAVING could treat lettercase of table aliases
     incorrectly if lower_case_table_names was enabled.
     (Bug#31562: http://bugs.mysql.com/31562)
   * Spurious duplicate-key errors could occur for
     multiple-row inserts into an InnoDB table that activate a
     trigger. (Bug#31540: http://bugs.mysql.com/31540)
   * Using ALTER EVENT to rename a disabled event caused it to
     become enabled. (Bug#31539: http://bugs.mysql.com/31539)
   * The fix for Bug#24989: http://bugs.mysql.com/24989
     introduced a problem such that a NULL thread handler
     could be used during a rollback operation. This problem
     is unlikely to be seen in practice.
     (Bug#31517: http://bugs.mysql.com/31517)
   * The length of the result from IFNULL() could be
     calculated incorrectly because the sign of the result was
     not taken into account.
     (Bug#31471: http://bugs.mysql.com/31471)
   * Queries that used the ref access method or index-based
     subquery execution over indexes that have DECIMAL columns
     could fail with an error Column col_name cannot be null.
     (Bug#31450: http://bugs.mysql.com/31450)
   * InnoDB now tracks locking and use of tables by MySQL only
     after a table has been successfully locked on behalf of a
     transaction. Previously, the locked flag was set and the
     table in-use counter was updated before checking whether
     the lock on the table succeeded. A subsequent failure in
     obtaining a lock on the table led to an inconsistent
     state as the table was neither locked nor in use.
     (Bug#31444: http://bugs.mysql.com/31444)
   * SELECT 1 REGEX NULL caused an assertion failure for debug
     servers. (Bug#31440: http://bugs.mysql.com/31440)
   * The UpdateXML() function did not check for the validity
     of all its arguments; in some cases, this could lead to a
     crash of the server.
     (Bug#31438: http://bugs.mysql.com/31438)
   * The mysql_change_user() C API function caused advisory
     locks (obtained with GET_LOCK()) to malfunction.
     (Bug#31418: http://bugs.mysql.com/31418)
   * NDB libraries and include files were missing from some
     binary tar file distributions.
     (Bug#31414: http://bugs.mysql.com/31414)
   * Executing RENAME while tables were open for use with
     HANDLER statements could cause a server crash.
     (Bug#31409: http://bugs.mysql.com/31409)
   * mysql-test-run.pl tried to create files in a directory
     where it could not be expected to have write permission.
     mysqltest created .reject files in a directory other than
     the one where test results go.
     (Bug#31398: http://bugs.mysql.com/31398)
   * For a table that had been opened with HANDLER and marked
     for reopening after being closed with FLUSH TABLES, DROP
     TABLE did not properly discard the handler.
     (Bug#31397: http://bugs.mysql.com/31397)
   * Automatically allocated memory for string options
     associated with a plugin was not freed if the plugin did
     not get installed.
     (Bug#31382: http://bugs.mysql.com/31382)
   * INFORMATION_SCHEMA.TABLES was returning incorrect
     information. (Bug#31381: http://bugs.mysql.com/31381)
   * DROP USER caused an increase in memory usage.
     (Bug#31347: http://bugs.mysql.com/31347)
   * For InnoDB tables with READ COMMITTED isolation level,
     UPDATE statements skipped rows locked by another
     transaction, rather than waiting for the locks to be
     released. (Bug#31310: http://bugs.mysql.com/31310)
   * For an almost-full MyISAM table, an insert that failed
     could leave the table in a corrupt state.
     (Bug#31305: http://bugs.mysql.com/31305)
   * myisamchk --unpack could corrupt a table that when
     unpacked has static (fixed-length) row format.
     (Bug#31277: http://bugs.mysql.com/31277)
   * CONVERT(val, DATETIME) would fail on invalid input, but
     processing was not aborted for the WHERE clause, leading
     to a server crash.
     (Bug#31253: http://bugs.mysql.com/31253)
   * Allocation of an insufficiently large group-by buffer
     following creation of a temporary table could lead to a
     server crash. (Bug#31249: http://bugs.mysql.com/31249)
   * Use of DECIMAL(n, n) ZEROFILL in GROUP_CONCAT() could
     cause a server crash.
     (Bug#31227: http://bugs.mysql.com/31227)
   * When a TIMESTAMP with a non-zero time part was converted
     to a DATE value, no warning was generated. This caused
     index lookups to assume that this is a valid conversion
     and was returning rows that match a comparison between a
     TIMESTAMP value and a DATE keypart. Now a warning is
     generated so that TIMESTAMP with a non-zero time part
     will not match DATE values.
     (Bug#31221: http://bugs.mysql.com/31221)
   * Server variables could not be set to their current values
     on Linux platforms.
     (Bug#31177: http://bugs.mysql.com/31177)
     See also Bug#6958: http://bugs.mysql.com/6958
   * WIth small values of myisam_sort_buffer_size, REPAIR
     TABLE for MyISAM tables could cause a server crash.
     (Bug#31174: http://bugs.mysql.com/31174)
   * Use of the @@hostname system variable in inserts in
     mysql_system_tables_data.sql did not replicate. The
     workaround is to select its value into a user variable
     (which does replicate) and insert that.
     (Bug#31167: http://bugs.mysql.com/31167)
   * If MAKETIME() returned NULL when used in an ORDER BY that
     was evaluated using filesort, a server crash could
     result. (Bug#31160: http://bugs.mysql.com/31160)
   * Data in BLOB or GEOMETRY columns could be cropped when
     performing a UNION query.
     (Bug#31158: http://bugs.mysql.com/31158)
   * LAST_INSERT_ID() execution could be handled improperly in
     subqueries. (Bug#31157: http://bugs.mysql.com/31157)
   * An assertion designed to detect a bug in the ROLLUP
     implementation would incorrectly be triggered when used
     in a subquery context with non-cacheable statements.
     (Bug#31156: http://bugs.mysql.com/31156)
   * Selecting spatial types in a UNION could cause a server
     crash. (Bug#31155: http://bugs.mysql.com/31155)
   * Use of GROUP_CONCAT(DISTINCT bit_column) caused an
     assertion failure.
     (Bug#31154: http://bugs.mysql.com/31154)
   * The server crashed in the parser when running out of
     memory. Memory handling in the parser has been improved
     to gracefully return an error when out-of-memory
     conditions occur in the parser.
     (Bug#31153: http://bugs.mysql.com/31153)
   * MySQL declares a UNIQUE key as a PRIMARY key if it
     doesn't have NULL columns and is not a partial key, and
     the PRIMARY key must alway be the first key. However, in
     some cases, a non-first key could be reported as PRIMARY,
     leading to an assert failure by InnoDB. This is fixed by
     correcting the key sort order.
     (Bug#31137: http://bugs.mysql.com/31137)
   * mysqldump failed to handle databases containing a `-'
     character in the name.
     (Bug#31113: http://bugs.mysql.com/31113)
   * Starting the server using --read-only and with the Event
     Scheduler enabled caused it to crash.
     Note
        This issue occurred only when the server had been built
        with certain nonstandard combinations of configure
        options.
     (Bug#31111: http://bugs.mysql.com/31111)
   * A rule to prefer filesort over an indexed ORDER BY when
     accessing all rows of a table was being used even if a
     LIMIT clause was present.
     (Bug#31094: http://bugs.mysql.com/31094)
   * REGEXP operations could cause a server crash for
     character sets such as ucs2. Now the arguments are
     converted to utf8 if possible, to allow correct results
     to be produced if the resulting strings contain only
     8-bit characters.
     (Bug#31081: http://bugs.mysql.com/31081)
   * Expressions of the form WHERE col NOT IN (col, ...),
     where the same column was named both times, could cause a
     server crash in the optimizer.
     (Bug#31075: http://bugs.mysql.com/31075)
   * Internal conversion routines could fail for several
     multi-byte character sets (big5, cp932, euckr, gb2312,
     sjis) for empty strings or during evaluation of SOUNDS
     LIKE. (Bug#31069: http://bugs.mysql.com/31069,
     Bug#31070: http://bugs.mysql.com/31070)
   * Many nested subqueries in a single query could led to
     excessive memory consumption and possibly a crash of the
     server. (Bug#31048: http://bugs.mysql.com/31048)
   * Using ORDER BY with ARCHIVE tables caused a server crash.
     (Bug#31036: http://bugs.mysql.com/31036)
   * A server crash could occur when a non-DETERMINISTIC
     stored function was used in a GROUP BY clause.
     (Bug#31035: http://bugs.mysql.com/31035)
   * The MOD() function and the % operator crashed the server
     for a divisor less than 1 with a very long fractional
     part. (Bug#31019: http://bugs.mysql.com/31019)
   * Transactions were committed prematurely when LOCK TABLE
     and SET AUTOCOMMIT=OFF were used together.
     (Bug#30996: http://bugs.mysql.com/30996)
   * On Windows, the pthread_mutex_trylock() implementation
     was incorrect. (Bug#30992: http://bugs.mysql.com/30992)
   * A character set introducer followed by a hexadecimal or
     bit-value literal did not check its argument and could
     return an ill-formed result for invalid input.
     (Bug#30986: http://bugs.mysql.com/30986)
   * CHAR(str USING charset) did not check its argument and
     could return an ill-formed result for invalid input.
     (Bug#30982: http://bugs.mysql.com/30982)
   * The result from CHAR(str USING ucs2) did not add a
     leading 0x00 byte for input strings with an odd number of
     bytes. (Bug#30981: http://bugs.mysql.com/30981)
   * A cluster restart could sometimes fail due to an issue
     with table IDs. (Bug#30975: http://bugs.mysql.com/30975)
   * The GeomFromText() function could cause a server crash if
     the first argument was NULL or the empty string.
     (Bug#30955: http://bugs.mysql.com/30955)
   * MAKEDATE() incorrectly moved year values in the 100-200
     range into the 1970-2069 range. (This is legitimate for
     00-99, but three-digit years should be used unchanged.)
     (Bug#30951: http://bugs.mysql.com/30951)
   * When invoked with constant arguments, STR_TO_DATE() could
     use a cached value for the format string and return
     incorrect results.
     (Bug#30942: http://bugs.mysql.com/30942)
   * GROUP_CONCAT() returned ',' rather than an empty string
     when the argument column contained only empty strings.
     (Bug#30897: http://bugs.mysql.com/30897)
   * For MEMORY tables, lookups for NULL values in BTREE
     indexes could return incorrect results.
     (Bug#30885: http://bugs.mysql.com/30885)
   * A server crash could occur if a stored function that
     contained a DROP TEMPORARY TABLE statement was invoked by
     a CREATE TEMPORARY TABLE statement that created a table
     of the same name.
     (Bug#30882: http://bugs.mysql.com/30882)
   * Calling NAME_CONST() with non-constant arguments
     triggered an assertion failure. Non-constant arguments
     are now disallowed.
     (Bug#30832: http://bugs.mysql.com/30832)
   * For a spatial column with a regular (non-SPATIAL) index,
     queries failed if the optimizer tried to use the index.
     (Bug#30825: http://bugs.mysql.com/30825)
   * Values for the --tc-heuristic-recover option incorrectly
     were treated as values for the --myisam-stats-method
     option. (Bug#30821: http://bugs.mysql.com/30821)
   * INFORMATION_SCHEMA.SCHEMATA was returning incorrect
     information. (Bug#30795: http://bugs.mysql.com/30795)
   * The optimizer incorrectly optimized conditions out of the
     WHERE clause in some queries involving subqueries and
     indexed columns. (Bug#30788: http://bugs.mysql.com/30788)
   * Improper calculation of CASE expression results could
     lead to value truncation.
     (Bug#30782: http://bugs.mysql.com/30782)
   * On Windows, the pthread_mutex_trylock() implementation
     was incorrect. One symptom was that invalidating the
     query cache could cause a server crash.
     (Bug#30768: http://bugs.mysql.com/30768)
   * User-supplied names foreign key names might not be set to
     the right key, leading to foreign keys with no name.
     (Bug#30747: http://bugs.mysql.com/30747)
   * Under some circumstances, CREATE TABLE ... SELECT could
     crash the server or incorrectly report that the table row
     size was too large.
     (Bug#30736: http://bugs.mysql.com/30736)
   * Using the MIN() or MAX() function to select one part of a
     multi-part key could cause a crash when the function
     result was NULL. (Bug#30715: http://bugs.mysql.com/30715)
   * The embedded server did not properly check column-level
     privileges. (Bug#30710: http://bugs.mysql.com/30710)
   * INFORMATION_SCHEMA.VIEWS.VIEW_DEFINITION was incorrect
     for views that were defined to select from other
     INFORMATION_SCHEMA tables.
     (Bug#30689: http://bugs.mysql.com/30689)
   * Issuing an ALTER SERVER statement to update the settings
     for a FEDERATED server would cause the mysqld to crash.
     (Bug#30671: http://bugs.mysql.com/30671)
   * The optimizer could ignore ORDER BY in cases when the
     result set is ordered by filesort, resulting in rows
     being returned in incorrect order.
     (Bug#30666: http://bugs.mysql.com/30666)
   * A different execution plan was displayed for EXPLAIN than
     would actually have been used for the SELECT because the
     test of sort keys for ORDER BY did not consider keys
     mentioned in IGNORE KEYS FOR ORDER BY.
     (Bug#30665: http://bugs.mysql.com/30665)
   * On Windows, LIMIT arguments greater than 2^32 did not
     work correctly. (Bug#30639: http://bugs.mysql.com/30639)
   * MyISAM tables could not exceed 4294967295 (2^32 - 1) rows
     on Windows. (Bug#30638: http://bugs.mysql.com/30638)
   * A failed HANDLER ... READ operation could leave the table
     in a locked state.
     (Bug#30632: http://bugs.mysql.com/30632)
   * mysql-test-run.pl could not run mysqld with root
     privileges. (Bug#30630: http://bugs.mysql.com/30630)
   * The mysqld_safe script contained a syntax error.
     (Bug#30624: http://bugs.mysql.com/30624)
   * The optimization that uses a unique index to remove GROUP
     BY did not ensure that the index was actually used, thus
     violating the ORDER BY that is implied by GROUP BY.
     (Bug#30596: http://bugs.mysql.com/30596)
   * SHOW STATUS LIKE 'Ssl_cipher_list' from a MySQL client
     connected via SSL returned an empty string rather than a
     list of available ciphers.
     (Bug#30593: http://bugs.mysql.com/30593)
   * For MEMORY tables, DELETE statements that remove rows
     based on an index read could fail to remove all matching
     rows. (Bug#30590: http://bugs.mysql.com/30590)
   * Using GROUP BY on an expression of the form timestamp_col
     DIV number caused a server crash due to incorrect
     calculation of number of decimals.
     (Bug#30587: http://bugs.mysql.com/30587)
   * Executing a SELECT COUNT(*) query on an InnoDB table
     partitioned by KEY that used a DOUBLE column as the
     partitioning key caused the server to crash.
     (Bug#30583: http://bugs.mysql.com/30583)
   * The options available to the CHECK TABLE statement were
     also allowed in OPTIMIZE TABLE and ANALYZE TABLE
     statements, but caused corruption during their execution.
     These options were never supported for the these
     statements, and an error is now raised if you try to
     apply these options to these statements.
     (Bug#30495: http://bugs.mysql.com/30495)
   * A self-referencing trigger on a partitioned table caused
     the server to crash instead of failing with an error.
     (Bug#30484: http://bugs.mysql.com/30484)
   * The mysql_change_user() C API function did not correctly
     reset the character set variables to the values they had
     just after initially connecting.
     (Bug#30472: http://bugs.mysql.com/30472)
   * When expanding a * in a USING or NATURAL join, the check
     for table access for both tables in the join was done
     using only the grant information of the first table.
     (Bug#30468: http://bugs.mysql.com/30468)
   * When casting a string value to an integer, cases where
     the input string contained a decimal point and was long
     enough to overrun the unsigned long long type were not
     handled correctly. The position of the decimal point was
     not taken into account which resulted in miscalculated
     numbers and incorrect truncation to appropriate SQL data
     type limits. (Bug#30453: http://bugs.mysql.com/30453)
   * Versions of mysqldump from MySQL 4.1 or higher tried to
     use START TRANSACTION WITH CONSISTENT SNAPSHOT if the
     --single-transaction and --master-data options were
     given, even with servers older than 4.1 that do not
     support consistent snapshots.
     (Bug#30444: http://bugs.mysql.com/30444)
   * With libmysqld, use of prepared statements and the query
     cache at the same time caused problems.
     (Bug#30430: http://bugs.mysql.com/30430)
   * Issuing a DELETE statement having both an ORDER BY clause
     and a LIMIT clause could cause mysqld to crash.
     (Bug#30385: http://bugs.mysql.com/30385)
   * For CREATE ... SELECT ... FROM, where the resulting table
     contained indexes, adding SQL_BUFFER_RESULT to the SELECT
     part caused index corruption in the table.
     (Bug#30384: http://bugs.mysql.com/30384)
   * The Last_query_cost status variable value can be computed
     accurately only for simple "flat" queries, not complex
     queries such as those with subqueries or UNION. However,
     the value was not consistently being set to 0 for complex
     queries. (Bug#30377: http://bugs.mysql.com/30377)
   * The optimizer made incorrect assumptions about the value
     of the is_member value for user-defined functions,
     sometimes resulting in incorrect ordering of UDF results.
     (Bug#30355: http://bugs.mysql.com/30355)
   * Queries that had a GROUP BY clause and selected
     COUNT(DISTINCT bit_column) returned incorrect results.
     (Bug#30324: http://bugs.mysql.com/30324)
   * Some valid euc-kr characters having the second byte in
     the ranges [0x41..0x5A] and [0x61..0x7A] were rejected.
     (Bug#30315: http://bugs.mysql.com/30315)


Please also check parts 1 and 3 for any item you are interested in.


Enjoy !
Jörg

--
Joerg Bruehe, Senior Production Engineer
MySQL AB, www.mysql.com













--
MySQL Packagers Mailing List
For list archives: http://lists.mysql.com/packagers
To unsubscribe:    http://lists.mysql.com/packagers?unsub=lists@...