MySQL 5.6.16/5.5.36 发布

完全改进：

Changes in MySQL 5.6.16 (2014-01-31)

A known limitation of this release:

Note

Building MySQL from source on Windows using Visual Studio 2008          fails with an identifier not found error. Later versions of Visual Studio are unaffected. The          workaround is to set the CMake build          option, INNODB_PAGE_ATOMIC_REF_COUNT, toOFF. This option isONby default.

Functionality Added or Changed

• Previously, ALTER TABLE in MySQL        5.6 could alter a table such that the result had temporal        columns in both 5.5 and 5.6 format. Now ALTER TABLE upgrades old temporal        columns to 5.6 format forADD COLUMN,CHANGE COLUMN,MODIFY        COLUMN,ADD INDEX, andFORCEoperations. This conversion cannot be        done using theINPLACEalgorithm, so        specifyingALGORITHM=INPLACEin these cases        results in an error.       (Bug #17246318)

• CMake now supports a -DTMPDIR=dir_name option to specify the default tmpdir value. If unspecified,        the value defaults toP_tmpdirin<stdio.h>. Thanks to Honza Horak for        the patch.       (Bug #68338, Bug #16316074)

Bugs Fixed

• InnoDB; Replication: Using the InnoDB memcachedplugin (see InnoDBIntegration with memcached) with innodb_api_enable_binlog set to        1 caused the server to leak memory.       (Bug #70757, Bug #17675622)

• InnoDB: A boolean mode full-text search query would result in a memory        access violation during parsing.       (Bug #17978763)

• InnoDB: When new indexes are added by an ALTER        TABLE operation, instead of only saving table-level        statistics and statistics for the new indexes,InnoDBwould save statistics for the entire        table, including the table's other indexes. This behavior slowed ALTER TABLE performance.       (Bug #17848838, Bug #16511145)

• InnoDB: Due to a parser error, full-text search queries that include a        sub-expression could return the wrong result.       (Bug #17840768)

• InnoDB: The innochecksum tool did not use a        Windows-specific API to retrieve file size information, which        resulted in an incorrect error message (Error:        ibdata1 cannot be found) when the MySQL 5.6 innochecksum 2GB file size limit was        exceeded. innochecksum now provides support        for files larger than 2GB in both MySQL 5.6 and MySQL 5.7.       (Bug #17810862, Bug #70936)

• InnoDB: Due to a regression introduced by the fix for Bug#17371537,        memory was not allocated for the default memcached engine when        using the default memcached engine as the backstore for data        instead ofInnoDB.       (Bug #17800829)

• InnoDB: InnoDBwould report an incorrect operating        system error code after failing to initialize.       (Bug #17788055, Bug #70867)

• InnoDB: Manipulating a table after discarding its tablespace using ALTER TABLE ...        DISCARD TABLESPACE could result in a serious error.       (Bug #17700280)

• InnoDB: Persistent optimizer statistics would cause stalls due to latch        contention.       (Bug #17699331, Bug #70768)

• InnoDB:MATCH() ... AGAINST queries that        use a long string as an argument forAGAINST()could result in an error when run        on anInnoDBtable with a full-text search        index.       (Bug #17640261)

• InnoDB: AnInnoDBfull-text search failure would        occur due to an “unended” token. The string and        string length should be passed for string comparison.       (Bug #17659310)

• InnoDB: In debug builds, a merge insert buffer during a page read would        cause a memory access violation.       (Bug #17561188)

• InnoDB: Truncating a memcached InnoDBtable while memcached is performing DML operations would        result in a serious error.       (Bug #17468031)

• InnoDB: Insync0rw.ic,rw_lock_x_lock_func_nowaitwould needlessly        callos_thread_get_curr_id.       (Bug #17509710, Bug #70417)

• InnoDB: Attempting to rename a table to a missing database would result        in a serious error.       (Bug #17447500)

• InnoDB: If a tablespace data file path is updated in a.islfile and then a crash recovery is        performed, the updated tablespace data file path is read from        the.islfile but theSYS_DATAFILEStable would not be not updated.        TheSYS_DATAFILEStable is now updated with        the new data file path after crash recovery.       (Bug #17448389)

• InnoDB: If the first page (page 0) of file-per-table tablespace data        file was corrupt, recovery would be halted even though the        doublewrite buffer contained a clean copy of the page.       (Bug #17335427, Bug #70087)

• InnoDB: TheInnoDBmemcached Readme file (README-innodb_memcached)        incorrectly stated that libevent 1.6.0 is linked statically into        daemon memcached. The bundled version of        libevent is 1.4.12, not 1.6.0.       (Bug #17324419, Bug #70034)

• InnoDB: Attempting to reset a replication slave while innodb_force_recovery is        greater than0would return a cryptic error        message: ERROR(1030) HY000: Got error -1 from storage        engine. The error message has been changed to: ERROR HY000: Operation not allowed when        innodb_force_recovery > 0. Replication options such        as --relay-log-info-repository=TABLE and --master-info-repository=TABLE store information in tables inInnoDB. When innodb_force_recovery is        greater than 0, replication tables cannot be updated which may        cause replication administration commands to fail.       (Bug #17287443, Bug #69907)

• InnoDB: The ALTER TABLE INPLACEalgorithm would fail to decrease the        auto-increment value.       (Bug #17250787, Bug #69882)

• InnoDB: Comments inbtr0cur.ccincorrectly stated        thatbtr_cur_pessimistic_update()andbtr_cur_optimistic_update()would accept a        NULL value.       (Bug #17231743, Bug #69847)

• InnoDB: dict_table_schema_checkwould calldtype_sql_nameneedlessly.       (Bug #17193801, Bug #69802)

• InnoDB: The functionos_file_get_statuswould not        work with raw devices.       (Bug #17023438, Bug #69424)

• InnoDB: During crash recovery, an incorrect transaction active time        would result in rolling back an uncommitted transaction.       (Bug #16936961, Bug #69438)

• InnoDB: Heap block debugging information (file_name,lineno), used for logging diagnostics, would        appear in release builds. This information should only appear in        debug builds.       (Bug #16924719, Bug #69422)

• InnoDB: Renaming a column while also adding or dropping columns in the        same ALTER TABLE operation would        cause an error.       (Bug #16864981)

• InnoDB: An online ALTER TABLE operation        would consume more memory than expected. During an onlineALTER TABLEoperation, an online log buffer        containing a head and tail buffer is created for each index that        is created or rebuilt. The tail buffer is the writer context and        is only required for concurrent write operations on an index        while theALTER TABLEoperation is in        progress. The head buffer is the reader context and is only        required during the log apply phase. To reduce memory        consumption, the tail buffer is now allocated when the first DML        statement is run on the index, and the head buffer is only        allocated in the log apply phase and freed afterwards.       (Bug #16868967, Bug #69325, Bug #17911720)

• InnoDB: On Windows, the full-text search (FTS) object ID was not in the        expected hexadecimal format.       (Bug #16559254)

  References: See also Bug #16559119.

• InnoDB: Fetching and releasing pages from the buffer pool and tracking        the page state are expensive and complex operations. Prior to        the bug fix, these operations were performed using a page mutex.        Using a page mutex to track several things is expensive and does        not scale well. The bug fix separates fetch and release tracking        (in-use state) of a page from page I/O state tracking. Fetch and        release is now tracked using atomics where available.

  For portability, a newCMakebuild option, INNODB_PAGE_ATOMIC_REF_COUNT (defaultON), can be used to disable atomic        page reference counting on platforms where atomics support is        not available. When atomic page reference counting is enabled        (default), “[Note] InnoDB: Using atomics to ref        count buffer pool pages” is printed to the        error log at server startup. If atomic page reference counting        is disabled, “[Note] InnoDB: Using mutexes to ref        count buffer pool pages” is printed instead.       (Bug #16249481, Bug #68079)

• InnoDB: Table renaming errors would appear in theLATEST        FOREIGN KEY ERRORsection of theSHOW ENGINE        INNODB STATUSoutput.       (Bug #12762390, Bug #61746)

• InnoDB: UNIV_SYNC_DEBUG, which was disabled inuniv.iwith the fix for Bug#16720368, is now        enabled.       (Bug #69617, Bug #17033591)

• Partitioning: Queries using theindex_mergeoptimization        (see Index Merge Optimization) could return        invalid results when run against tables that were partitioned byHASH.       (Bug #17588348, Bug #70588)

• Partitioning: When no partition had returned a row since the last HA_ERR_KEY_NOT_FOUND error, the use of        uninitialized memory in the priority queue used for returning        rows in sorted order could lead to a crash of the server.       (Bug #17401628)

• Replication: When the binary log I/O cache grew to exactly 32768 bytes and        the current transaction was preceded by a transaction whose size        was greater than 32768 bytes, events could be corrupted when        written into the binary log.       (Bug #17842137)

• Replication: Creating and dropping large numbers of temporary tables could        lead to increased memory consumption.       (Bug #17806014)

• Replication:mysqlbinlog--verbose failed when it        encountered a corrupt row event in the binary log. Such a row        event could also cause the slave to fail.       (Bug #17632978)

  References: See also Bug #16960133.

• Replication: When log_warnings is greater        than 1, the master prints binary log dump thread        information—containing the slave server ID, binary log        file name, and binary log position—inmysqld.1.err. A slave server ID greater        than 2 billion was printed with a negative value in such cases.       (Bug #17641586, Bug #70685)

• Replication:mysqlbinlog did not properly decode DECIMAL values in a row-based        binary log. This could cause invalid values to be printed out        forDECIMALcolumns.       (Bug #17544169)

  References: See also Bug #14309019.

• Replication: Seconds_Behind_Masterin the output of SHOW SLAVE STATUS could under        some conditions be reported as 0 when it should have had a value        greater than zero.       (Bug #17233214)

  References: See also Bug #16579028.

• Replication: The semisynchronous replication plugin was called twice for a        DDL statement, incrementing Rpl_semi_sync_master_yes_tx by        2 instead of 1 each time such a statement was executed.       (Bug #70410, Bug #17509011)

• Compilation errors occurred on Solaris 10; resolved by includingmy_config.hbefore system header files.       (Bug #18046811)

• FORCE INDEX [FOR ORDER BY]        (index_name)did not work        for joins.

  The fix for this bug also also changes the warning created for EXPLAIN. Instead of printing only{IGNORE|USE|FORCE} INDEXit now also printsFOR {GROUP BY|ORDER BY|JOIN}if that was        specified in the query.       (Bug #17889511)

• With the compressed client/server protocol enabled, Performance        Schema statement instrumentation could raise an assertion.       (Bug #17794846)

• In some cases, UNIX_TIMESTAMP() could returnNULLwhen it should return 0.       (Bug #17728371)

• An assertion could be raised if afilesortfailed to resize its main buffer when record properties changed.       (Bug #17757914)

• The cache used for the Index Merge access method was freed only        after successful retrieval of all rows. Interruption or failure        of the operation led to a file descriptor leak.       (Bug #17708621)

• Using the mysqldump--set-gtid-purged option with        no value caused mysqldump to crash.       (Bug #17650245)

• A race condition between Performance Schema statement event        threads led to a server exit.       (Bug #17637970)

• In a view definition requireing resolution of aggregrate        expressions within a subquery to an outer query, selecting from        the view could cause a server exit.       (Bug #17547804)

  References: This bug is a regression of Bug #16436383.

• An addressing error in accessing the join buffer could produce        invalid results or a server exit.       (Bug #17513341)

• mysql_config incorrectly included some flags        to generate compiler warning output.       (Bug #17400967)

• With semi-join optimization enabled, queries with nested        subqueries could cause a server exit due to incorrect resolution        of references to columns in the MIDdle query block.       (Bug #17398972)

• In some cases, the optimizer wrote fixed-length temporaryMyISAMtables to disk rather than        variable-length temporary tables.       (Bug #17231940)

• Enabling thevalidate_passwordplugin could        result in incorrect password hashes being stored in themysql.usertable.       (Bug #17065383)

• For accounts authenticated using thesha256_passwordplugin, setting the password        after the password had been expired did not clear the        password-expired flag.       (Bug #16872181)

• On Mac OS X 10.7, a race condition involvingvio_shutdown()and the select-based        implementation ofvio_io_wait()could cause a        server exit.       (Bug #16354789, Bug #17733393)

• Host names in example URLs used within the source code were        replaced by names in the example.com domain, the domain intended        by IANA for this purpose.       (Bug #15890092)

• Forutf8andutf8mb4strings, handler functions unnecessarily called a Unicode        conversion function.       (Bug #14057034)

• Several-Wwarning flags were turned off for        compilation in maintainer mode if MySQL was configured with-DWITH_INNODB_MEMCACHED=1.       (Bug #13898319)

• Calling the ExtractValue() function with an invalid XPath expression could in some cases        lead to a failure of the server.       (Bug #12428404, Bug #61065)

• Use of a nonmulti-byte algorithm for skipping leading spaces in        multi-byte strings could cause a server exit.       (Bug #12368495)

• WithONLY_FULL_GROUP_BYSQL mode enabled, a        query that usesGROUP BYon a column derived        from a subquery in theFROMclause failed        with acolumn isn't in GROUP BYerror, if the        query was in a view.       (Bug #11923239)

• For theutf8_bincollation,ORDER BY        LOWER(col_name)could        produce incorrect ordering.       (Bug #69005, Bug #16691598)

• Several issues identified by the Coverity static analysis tool        were fixed. Thanks to Honza Horak for the patch.       (Bug #70830, Bug #17760511)

• On Windows, the --local-service server option did not work, and was not displayed in the --help message.       (Bug #69637, Bug #17049656)

• It was not possible to query a view with anORDER        BYclause that referenced an alias in theSELECTclause of the view definition, unless        all columns in the view were named in the select list.

  To handle this problem, the server now writes a view differently        into the.frmfile that stores the view        definition. If you experience view-evaluation errors such as        just described, drop and recreate the view so that the.frmfile contains the updated view        representation.       (Bug #69678, Bug #17077305)

• The prototype of the Performance Schema instrumentation APImysql_cond_timedwait()call was fixed to be        drop-in compatible withpthread_cond_timedwait(). This fix affects        only implementers of third-party plugins.       (Bug #70628, Bug #17702677)

• Themake_atomic_cas_body64implementation on        IA32 with gcc but without gcc builtins could be miscompiled due to an        incorrect constraint. The patch also causes MySQL to use builtin        atomics when compiled using Clang.       (Bug #63451, Bug #17242996)

• Complex updates of Performance Schema tables involving joins or        subqueries failed to update every row.       (Bug #70025, Bug #17309657)

• For the path specified with the --basedir option, mysql_plugin attempted to unlink the path        rather than free the memory in which the path was stored.       (Bug #69752, Bug #17168602)

• COUNT(DISTINCT) sometimes produced an incorrect result when the        last read row contained aNULLvalue.       (Bug #68749, Bug #16539979, Bug #71028, Bug #17867117)

• sql_resolver.ccreferred to partitioning        code that should have been protected by an#ifdef, even when MySQL was configured with-DWITH_PARTITION_STORAGE_ENGINE=OFF.       (Bug #71010, Bug #17876794)

• In incorrect result could be returned for a query with an IF() predicate in theWHEREclause combined withOUTER        JOINin a subquery that is transformed to a semi-join.        (A workaround is to disable semi-join usingSET        optimizer_switch='semijoin=off';)       (Bug #70608, Bug #17600176)

• A full-text search combined with derived tables (subqueries in        theFROMclause) caused a server exit.

  Now if a full-text operation depends on a derived table, the        server produces an error indicating that a full-text search        cannot be done on a materialized table.       (Bug #68751, Bug #16539903)

• Some scripts displayed out-of-date information regarding where        to report bugs.       (Bug #68742, Bug #16530527)

• Some files in the Performance Schema file_instances table were not being        removed because the file-removal operation was not instrumented.       (Bug #69782, Bug #17209750)

• mysqldump --single-transaction acquired        metadata locks for each dumped table but did not release them        until the dump operation finished. Consequently, other DDL        operations on a dumped table blocked even after the table itself        had been dumped. mysqldump now attempts to        release metadata locks earlier.       (Bug #71017, Bug #17862905)

• Updating aFEDERATEDtable withUPDATE... JOINcaused a server exit when the        local table contained a single row and that row could be joined        to a row in theFEDERATEDtable.       (Bug #68354, Bug #16324629)

• mysql_install_dbreferred to the obsolete mysqlbug script for reporting problems. It        now refers to http://bugs.mysql.com/ instead.       (Bug #29716, Bug #11746921)

http://dev.mysql.com/doc/relnotes/mysql/5.5/en/news-5-5-36.html

下载：ftp://mirror.switch.ch/mirror/mysql/Downloads/MySQL-5.6/mysql-5.6.16.tar.gz

ftp://mirror.switch.ch/mirror/mysql/Downloads/MySQL-5.5/mysql-5.5.36.tar.gz