Manuel de référence MySQL 5.0

Fonctionnalités ajoutées ou modifiées :

• Updated version of libedit to 2.9. (Bug #2596)

Bogues corrigés :

• MAX() for an INT UNSIGNED (unsigned 4-byte integer) column could return negative values if the column contained values larger than 2^31. (Bug #9298)
• SHOW CREATE VIEW got confused and could not find the view if there was a temporary table with the same name as the view. (Bug #8921)
• Fixed a deadlock resulting from use of FLUSH TABLES WITH READ LOCK while an INSERT DELAYED statement is in progress. (Bug #7823)
• The optimizer was choosing suboptimal execution plans for certain outer joins where the right table of a left join (or left table of a right join) had both ON and WHERE conditions. (Bug #10162)
• RENAME TABLE for an ARCHIVE table failed if the .arn file was not present. (Bug #9911)
• Invoking a stored function that executed a SHOW statement resulted in a server crash. (Bug #8408)
• Fixed problems with static variables and do not link with libsupc++ to allow building on FreeBSD 5.3. (Bug #9714)
• Fixed some awk script portability problems in cmd-line-utils/libedit/makelist.sh. (Bug #9954)
• Fixed a problem with mishandling of NULL key parts in hash indexes on VARCHAR columns, resulting in incorrect query results. (Bug #9489, Bug #10176)

Fonctionnalités ajoutées ou modifiées :

• Added support for the BIT data type to the MEMORY, InnoDB, and BDB storage engines.
• SHOW VARIABLES no longer displays the deprecated log_update system variable. (Bug #9738)
• --innodb-fast-shutdown is now also settable on the fly (global variable innodb_fast_shutdown). It now accepts values 0, 1 and 2 (except on Netware where 2 is disabled); if set to 2, then when the MySQL server shuts down, InnoDB will just flush its logs and then shut down brutally (and quickly) as if it was a MySQL crash; no committed transaction will be lost, but a crash recovery will be done at next startup.

Bogues corrigés :

• Security fix : If mysqld was started with --user=non_existent_user, it would run using the privileges of the account it was invoked from, even if that was root. (Bug #9833)
• Multiple-table updates could produce spurious data-truncation warnings if they used a join across columns that are indexed using a column prefix. (Bug #9103)
• Fixed a string-length comparison problem that caused mysql to fail loading dump files containing certain '\'-sequences. (Bug #9756)
• Fixed a failure to resolve a column reference properly when an outer join involving a view contained a subquery and the column was used in the subquery and the outer query. (Bug #6106, Bug #6107)
• Use of a subquery that used WITH ROLLUP in the FROM clause of the main query sometimes resulted in a Column cannot be null error. (Bug #9681)
• Fixed a memory leak that occurred when selecting from a view that contained a subquery. (Bug #10107)
• Fixed an optimizer bug in computing the union of two ranges for the OR operator. (Bug #9348)
• Fixed a segmentation fault in mysqlcheck that occurred when the last table checked in --auto-repair mode returned an error (such as the table being a MERGE table). (Bug #9492)
• Incorrect results were returned for queries of the form SELECT ... LEFT JOIN ... WHERE EXISTS (subquery), where the subquery selected rows based on an IS NULL condition. (Bug #9516)
• Executing LOCK TABLES and then calling a stored procedure caused an error and resulting in the server thinking that no stored procedures exist. (Bug #9566)
• Selecting from a view containing a subquery caused the server to hang. (Bug #8490)
• Attempting to execute a multiple-table UPDATE within a stored procedure failed with a Table 'tbl_name' was locked with a READ lock and can't be updated error. (Bug #9486)
• Starting mysqld with the --skip-innodb and --default-storage-engine=innodb (or --default-table-type=innodb caused a server crash. (Bug #9815)
• Queries containing CURRENT_USER() incorrectly were registered in the query cache. (Bug #9796)
• Setting the storage_engine system variable to MEMORY succeeded, but retrieving the variable resulted in a value of HEAP (the old name for the MEMORY storage engine) rather than MEMORY. (Bug #10039)
• mysqlshow displayed an incorrect row count for tables. (Bug #9391)
• The server died with signal 11 if a non-existent location was specified for the location of the binary log. Now the server exits after printing an appropriate error messsage. (Bug #9542)
• Fixed a problem in the client/server protocol where the server closed the connection before sending the final error message. The problem could show up as a Lost connection to MySQL server during query when attempting to connect to access a non-existent database. (Bug #6387, Bug #9455)
• Fixed a readline-related crash in mysql when the user pressed Control-R. (Bug #9568)
• For stored functions that should return a YEAR value, corrected a failure of the value to be in YEAR format. (Bug #8861)
• Fixed a server crash resulting from invocation of a stored function that returned a value having an ENUM or SET data type. (Bug #9775)
• Fixed a server crash resulting from invocation of a stored function that returned a value having a BLOB data type. (Bug #9102)
• Fixed a server crash resulting from invocation of a stored function that returned a value having a BIT data type. (Bug #7648)
• TIMEDIFF() with a negative time first argument and postive time second argument produced incorrect results. (Bug #8068)
• Fixed a problem with OPTIMIZE TABLE for InnoDB tables being written twice to the binary log. (Bug #9149)
• InnoDB : Prevent ALTER TABLE from changing the storage engine if there are foreign key constraints on the table. (Bug #5574, Bug #5670)
• InnoDB : Fixed a bug where next-key locking doesn't allow the insert which does not produce a phantom. (Bug #9354) If the range is of type 'a' <= uniquecolumn, InnoDB lock only the RECORD, if the record with the column value 'a' exists in a CLUSTERED index. This allows inserts before a range.
• InnoDB : When FOREIGN_KEY_CHECKS=0, ALTER TABLE and RENAME TABLE will ignore any type incompatibilities between referencing and referenced columns. Thus, it will be possible to convert the character sets of columns that participate in a foreign key. Be sure to convert all tables before modifying any data! (Bug #9802)
• Provide more informative error messages in clustered setting when a query is issued against a table that has been modified by another mysqld server. (Bug #6762)

Fonctionnalités ajoutées ou modifiées :

• Added ENGINE=MyISAM table option when creating mysql.proc table in mysql_create_system_tables script to make sure the table is created as a MyISAM table even if the default storage engine has been changed. (Bug #9496)
• SHOW CREATE TABLE for an INFORMATION_SCHEMA table no longer prints a MAX_ROWS value because the value has no meaning. (Bug #8941)
• Invalid DEFAULT values for CREATE TABLE now generate errors. (Bug #5902)
• Added --show-table-type option to mysqlshow, to display a column indicating the table type, as in SHOW FULL TABLES. (Bug #5036)
• The way the time zone information is stored into the binary log was changed, so that it's now possible to have a replication master and slave running with different global time zones. A drawback is that replication from 5.0.4 masters to pre-5.0.4 slaves is impossible.
• Added --with-big-tables compilation option to configure. (Previously it was necessary to pass -DBIG_TABLES to the compiler manually in order to enable large table support.) Voir Section 2.4.2, « Options habituelles de configure »2.4.2. Options habituelles de configure for details.
• New configuration directives !include and !includedir implemented for including option files and searching directories for option files. Voir Section 4.3.2, « Fichier d'options my.cnf »4.3.2. Fichier d'options my.cnf for usage.

Bogues corrigés :

• The use of XOR together with NOT ISNULL() erroneously resulted in some outer joins being converted to inner joins by the optimizer. (Bug #9017)
• Fixed an optimizer problem where extraneous comparisons between NULL values in indexed columns were being done for operators such as = that are never true for NULL. (Bug #8877)
• Fixed the client/server protocol for prepared statements so that reconnection works properly when the connection is killed while reconnect is enabled. (Bug #8866)
• A server installed as a Windows service and started with --shared-memory could not be stopped. (Bug #9665)
• Fixed a server crash resulting from multiple executions of a prepared statement involving a join of an INFORMATION_SCHEMA table with another table. (Bug #9383)
• Fixed utf8_spanish2_ci and ucs2_spanish2_ci collations to not consider 'r' equal to 'rr'. If you upgrade to this version from an earlier version, you should rebuild the indexes of affected tables. (Bug #9269)
• mysqldump dumped core when invoked with --tmp and --single-transaction options and a non-existent table name. (Bug #9175)
• Allow extra HKSCS and cp950 characters (big5 extension characters) to be accepted in big5 columns. (Bug #9357)
• mysql.server no longer uses non-portable alias command or LSB functions. (Bug #9852)
• Fixed a server crash resulting from GROUP BY on a decimal expression. (Bug #9210)
• In prepared statements, subqueries containing parameters were erroneously treated as const tables during preparation, resulting in a server crash. (Bug #8807)
• InnoDB : ENUM and SET columns were treated incorrectly as character strings. This bug did not manifest itself with latin1 collations if there were less than about 100 elements in an ENUM, but it caused malfunction with UTF-8. Old tables will continue to work. In new tables, ENUM and SET will be internally stored as unsigned integers. (Bug #9526)
• InnoDB : Avoid test suite failures caused by a locking conflict between two server instances at server shutdown/startup. This conflict on advisory locks appears to be the result of a bug in the operating system; these locks should be released when the files are closed, but somehow that does not always happen immediately in Linux. (Bug #9381)
• InnoDB : True VARCHAR : InnoDB stored the 'position' of a row wrong in a column prefix primary key index; this could cause MySQL to complain ERROR 1032: Can't find record ... in an update of the primary key, and also some ORDER BY or DISTINCT queries. (Bug #9314)
• InnoDB : Fix bug in MySQL/InnoDB 5.0.3 : SQL statements were not rolled back on error. (Bug #8650)
• Fixed a Commands out of sync error when two prepared statements for single-row result sets were open simultaneously. (Bug #8880)
• Fixed a server crash after a call to mysql_stmt_close() for single-row result set. (Bug #9159)
• Fixed server crashes for CREATE TABLE ... SELECT or INSERT INTO ... SELECT when selecting from multiple-table view. (Bug #8703, Bug #9398)
• TRADITIONAL SQL mode should prevent inserts where a column with no default value is omitted or set to a value of DEFAULT. Fixed cases where this restriction was not enforced. (Bug #5986)
• Fixed a server crash when creating a PRIMARY KEY for a table, if the table contained a BIT column. (Bug #9571)
• Warning message from GROUP_CONCAT() did not always indicate correct number of lines. (Bug #8681)
• The commit count cache for NDB was not properly invalidated when deleting a record using a cursor. (Bug #8585)
• Fixed option-parsing code for the embedded server to understand K, M, and G suffixes for the net_buffer_length and max_allowed_packet options. (Bug #9472)
• Selecting a BIT column failed if the binary client/server protocol was used. (Bug #9608)
• Fixed a permissions problem whereby information in INFORMATION_SCHEMA could be exposed to a user with insufficient privileges. (Bug #7214)
• An error now occurs if you try to insert an invalid value via a stored procedure in STRICT mode. (Bug #5907)
• Link with libsupc++ on Fedora Core 3 to get language support functions. (Bug #6554)
• The value of the CHARACTER_MAXIMUM_LENGTH and CHARACTER_OCTET_LENGTH columns of the INFORMATION_SCHEMA.COLUMNS table must be NULL for numeric columns, but were not. (Bug #9344)
• DROP TABLE did not drop triggers that were defined for the table. DROP DATABASE did not drop triggers in the database. (Bug #5859, Bug #6559)
• CREATE OR REPLACE VIEW and ALTER VIEW now require the CREATE VIEW and DROP privileges, not CREATE VIEW and DELETE. (DELETE is a row-level privilege, not a table-level privilege.) (Bug #9260)
• Some user variables were not being handled with « implicit » coercibility. (Bug #9425)
• Setting the max_error_count system variable to 0 resulted in a setting of 1. (Bug #9072)
• Fixed a collation coercibility problem that caused a union between binary and non-binary columns to fail. (Bug #6519)
• Fixed a bug in division of floating point numbers. It could cause nine zeroes (000000000) to be inserted in the middle of the quotient. (Bug #9501)
• INFORMATION_SCHEMA tables had an implicit upper limit for the number of rows. As a result, not all data could be returned for some queries. (Bug #9317)
• Fixed a problem with the tee command in mysql that resulted in mysql crashing. (Bug #8499)
• CAST() now produces warnings when casting a wrong INTEGER and CHAR values. This also applies to implicite string to number casts. (Bug #5912)
• ALTER TABLE now fails in STRICT mode if generates warnings.
• Using CONVERT('0000-00-00',date) or CAST('0000-00-00' as date) in TRADITIONAL mode now produces a warning. (Bug #6145)
• Inserting a zero date in a DATE, DATETIME or TIMESTAMP column during TRADITIONAL mode now produces an error. (Bug #5933)
• Inserting a zero date into a DATETIME column in TRADITIONAL mode now produces an error.
• STR_TO_DATE() now produces errors in strict mode (and warnings otherwise) when given an illegal argument. (Bug #5902)
• Fixed a problem with ORDER BY that sometimes caused incorrect sorting of utf8 data. (Bug #9309)
• Fixed server crash resulting from queries that combined SELECT DISTINCT, SUM(), and ROLLUP. (Bug #8615)
• Incorrect results were returned from queries that combined SELECT DISTINCT, GROUP BY, and ROLLUP. (Bug #8616)
• Too many rows were returned from queries that combined ROLLUP and LIMIT if SQL_CALC_FOUND_ROWS was given. (Bug #8617)
• If on replication master a LOAD DATA INFILE is interrupted in the middle (integrity constraint violation, killed connection...), the slave used to skip this LOAD DATA INFILE entirely, thus missing some changes if this command permanently inserted/updated some table records before being interrupted. This is now fixed. (Bug #3247)

Note : This Beta release, as any other pre-production release, should not be installed on « production » level systems or systems with critical data. It is good practice to back up your data before installing any new version of software. Although MySQL has done its best to ensure a high level of quality, protect your data by making a backup as you would for any software beta release.

Fonctionnalités ajoutées ou modifiées :

• New privilege CREATE USER was added.
• Security improvement : The server creates .frm, .MYD, .MYI, .MRG, .ISD, and .ISM table files only if a file with the same name does not already exist. Thanks to Stefano Di Paola <> for finding and informing us about this issue. (CAN-2005-0711)
• Security improvement : User-defined functions should have at least one symbol defined in addition to the xxx symbol that corresponds to the main xxx() function. These auxiliary symbols correspond to the xxx_init(), xxx_deinit(), xxx_reset(), xxx_clear(), and xxx_add() functions. mysqld by default no longer loads UDFs unless they have at least one auxiliary symbol defined in addition to the main symbol. The --allow-suspicious-udfs option controls whether UDFs that have only an xxx symbol can be loaded. By default, the option is off. mysqld also checks UDF filenames when it reads them from the mysql.func table and rejects those that contain directory pathname separator characters. (It already checked names as given in CREATE FUNCTION statements.) Voir Section 27.2.3.1, « Fonctions utilisateur : appeler des fonctions simples »27.2.3.1. Fonctions utilisateur : appeler des fonctions simples, Section 27.2.3.2, « Appeler des fonctions utilisateurs pour les groupements »27.2.3.2. Appeler des fonctions utilisateurs pour les groupements, and Section 27.2.3.6, « Précautions à prendre avec les fonctions utilisateur »27.2.3.6. Précautions à prendre avec les fonctions utilisateur. Thanks to Stefano Di Paola <> for finding and informing us about this issue. (CAN-2005-0709, CAN-2005-0710)
• Support for the ISAM storage engine has been removed. If you have ISAM tables, you should convert them before upgrading. Voir Section 2.6.1, « Passer en de version 4.1 en version 5.0 »2.6.1. Passer en de version 4.1 en version 5.0.
• Support for RAID options in MyISAM tables has been removed. If you have tables that use these options, you should convert them before upgrading. Voir Section 2.6.1, « Passer en de version 4.1 en version 5.0 »2.6.1. Passer en de version 4.1 en version 5.0.
• Added support for AVG(DISTINCT).
• ONLY_FULL_GROUP_BY no longer is included in the ANSI composite SQL mode. (Bug #8510)
• mysqld_safe will create the directory where the UNIX socket file is to be located if the directory does not exist. This applies only to the last component of the directory pathname. (Bug #8513)
• The coercibility for the return value of functions such as USER() or VERSION() now is « system constant » rather than « implicit. » This makes these functions more coercible than column values so that comparisons of the two do not result in Illegal mix of collations errors. COERCIBILITY() was modified to accommodate this new coercibility value. Voir Section 12.8.3, « Fonctions d'informations »12.8.3. Fonctions d'informations.
• User variable coercibility has been changed from « coercible » to « implicit. » That is, user variables have the same coercibility as column values.
• Boolean full-text phrase searching now requires only that matches contain exactly the same words as the phrase and in the same order. Non-word characters no longer need match exactly.
• CHECKSUM TABLE returns a warning for non-existing tables. The checksum value remains NULL as before. (Bug #8256)
• The server now includes a timestamp in the Ready for connections message that is written to the error log at startup. (Bug #8444)
• Added SQL_NOTES session variable to cause Note-level warnings not to be recorded. (Bug #6662)
• Allowed the service-installation command for Windows servers to specify a single option other than --defaults-file following the service name. This is for compatibility with MySQL 4.1. (Bug #7856)
• InnoDB : Upgrading from 4.1 : The sorting order for end-space in TEXT columns for InnoDB tables has changed. Starting from 5.0.3, InnoDB compares TEXT columns as space-padded at the end. If you have a non-unique index on a TEXT column, you should run CHECK TABLE on it, and run OPTIMIZE TABLE if the check reports errors. If you have a UNIQUE INDEX on a TEXT column, you should rebuild the table with OPTIMIZE TABLE.
• InnoDB : Commit after every 10,000 copied rows when executing ALTER TABLE, CREATE INDEX, DROP INDEX or OPTIMIZE TABLE. This makes it much faster to recover from an aborted operation.
• Added VAR_POP() and STDDEV_POP() as standard SQL aliases for the VARIANCE() and STDDEV() functions that compute population variance and standard deviation. Added new VAR_SAMP() and STDDEV_SAMP() functions to compute sample variance and standard deviation. (Bug #3190)
• Fixed a problem with out-of-order packets being sent (ERROR after OK or EOF) following a KILL QUERY statement. (Bug #6804)
• Retrieving from a view defined as a SELECT that mixed UNION ALL and UNION DISTINCT resulted in a different result than retrieving from the original SELECT. (Bug #6565)
• Fixed a problem with non-optimal index_merge query execution plans being chosen on IRIX. (Bug #8578)
• BIT in column definitions now is a distinct data type; it no longer is treated as a synonym for TINYINT(1).
• Bit-field values can be written using b'value' notation. value is a binary value written using 0s and 1s.
• From the Windows distribution, predefined accounts without passwords for remote users ("root@%", "@%") were removed (other distributions never had them).
• Added mysql_library_init() and mysql_library_end() as synonyms for the mysql_server_init() and mysql_server_end() C API functions. mysql_library_init() and mysql_library_end() are #define symbols, but the names more clearly indicate that they should be called when beginning and ending use of a MySQL C API library no matter whether the application uses libmysqlclient or libmysqld. (Bug #6149)
• SHOW COLUMNS now displays NO rather than blank in the Null output column if the corresponding table column cannot be NULL.
• Changed XML format for mysql from <col_name>col_value</col_name> to <field name="col_name">col_value</field> to allow for proper encoding of column names that are not legal as element names. (Bug #7811)
• Added --innodb-checksums and --innodb-doublewrite options for mysqld.
• Added --large-pages option for mysqld.
• Added multi_read_range system variable.
• SHOW DATABASES, SHOW TABLES, SHOW COLUMNS, and so forth display information about the INFORMATION_SCHEMA database. Also, several SHOW statements now accept a WHERE clause specifying which output rows to display. Voir Chapitre 22, La base de données d'informations INFORMATION_SCHEMAChapitre 22. La base de données d'informations INFORMATION_SCHEMA.
• Added the CREATE ROUTINE and ALTER ROUTINE privileges, and made the EXECUTE privilege operational.
• InnoDB : Corrected a bug in the crash recovery of ROW_FORMAT=COMPACT tables that caused corruption. (Bug #7973) There may still be bugs in the crash recovery, especially in COMPACT tables.
• When the MyISAM storage engine detects corruption of a MyISAM table, a message describing the problem now is written to the error log.
• InnoDB : When MySQL/InnoDB is compiled on Mac OS X 10.2 or earlier, detect the operating system version at run time and use the fcntl() file flush method on Mac OS X versions 10.3 and later. Apple had disabled fsync() in Mac OS X for internal disk drives, which caused corruption at power outages.
• InnoDB : Implemented fast TRUNCATE TABLE. The old approach (deleting rows one by one) may be used if the table is being referenced by foreign keys. (Bug #7150)
• Added cp932 (SJIS for Windows Japanese) and eucjpms (UJIS for Windows Japanese) character sets.
• Added several InnoDB status variables. Voir Section 5.2.4, « Variables de statut du serveur  »5.2.4. Variables de statut du serveur .
• Added the FEDERATED storage engine. Voir Section 14.6, « Le moteur de table FEDERATED »14.6. Le moteur de table FEDERATED.
• SHOW CREATE TABLE now uses USING index_type rather than TYPE index_type to specify an index type. (Bug #7233)
• InnoDB now supports a fast TRUNCATE TABLE. One visible change from this is that auto-increment values for this table are reset on TRUNCATE.
• Added an error member to the MYSQL_BIND data structure that is used in the C API for prepared statements. This member is used for reporting data truncation errors. Truncation reporting is enabled via the new MYSQL_REPORT_DATA_TRUNCATION option for the mysql_options() C API function.
• API change : the reconnect flag in the MYSQL structure is now set to 0 by mysql_real_connect(). Only those client programs which didn't explicitly set this flag to 0 or 1 after mysql_real_connect() experience a change. Having automatic reconnection enabled by default was considered too dangerous (after reconnection, table locks, temporary tables, user and session variables are lost).
• FLUSH TABLES WITH READ LOCK is now killable while it's waiting for running COMMIT statements to finish.
• MEMORY (HEAP) can have VARCHAR() fields.
• VARCHAR columns now remember end space. A VARCHAR() column can now contain up to 65535 bytes. For more details, Voir Section C.1, « Changements de la version 5.0.0 (Développement) »C.1. Changements de la version 5.0.0 (Développement). If the table handler doesn't support the new VARCHAR type, then it's converted to a CHAR column. Currently this happens for NDB tables.
• InnoDB : Introduced a compact record format that does not store the number of columns or the lengths of fixed-size columns. The old format can be requested by specifying ROW_FORMAT=REDUNDANT. The new format (ROW_FORMAT=COMPACT) is the default. The new format typically saves 20 % of disk space and memory.
• InnoDB : Setting the initial AUTO_INCREMENT value for an InnoDB table using CREATE TABLE ... AUTO_INCREMENT = n now works, and ALTER TABLE ... AUTO_INCREMENT = n resets the current value.
• Seconds_Behind_Master is NULL (which means « unknown ») if the slave SQL thread is not running, or if the slave I/O thread is not running or not connected to master. It is zero if the SQL thread has caught up to the I/O thread. It no longer grows indefinitely if the master is idle.
• The MySQL server aborts immediately instead of simply issuing a warning if it is started with the --log-bin option but cannot initialize the binary log at startup (that is, an error occurs when writing to the binary log file or binary log index file).
• The binary log file and binary log index file now are handled the same way as MyISAM tables when there is a « disk full » or « quota exceeded » error. Voir Section A.4.3, « Comment MySQL gère un disque plein »A.4.3. Comment MySQL gère un disque plein.
• The MySQL server now aborts when started with option --log-bin-index and without --log-bin, and when started with --log-slave-updates and without --log-bin.
• If the MySQL server is started without an argument to --log-bin and without --log-bin-index, thus not providing a name for the binary log index file, a warning is issued because MySQL falls back to using the hostname for that name, and this is prone to replication issues if the server's hostname's gets changed later. Voir Section 1.5.7.4, « Bugs connus / limitations de MySQL »1.5.7.4. Bugs connus / limitations de MySQL.
• Added account-specific MAX_USER_CONNECTIONS limit, which allows you to specify the maximum number of concurrent connections for the account. Also, all limited resources now are counted per account (instead of being counted per user + host pair as it was before). Use the --old-style-user-limits option to get the old behavior.
• InnoDB : A shared record lock (LOCK_REC_NOT_GAP) is now taken for a matching record in the foreign key check because inserts can be allowed into gaps.
• InnoDB : Relaxed locking in INSERT...SELECT, single table UPDATE...SELECT and single table DELETE...SELECT clauses when innobase_locks_unsafe_for_binlog is used and isolation level of the transaction is not serializable. InnoDB uses consistent read in these cases for a selected table.
• Added a new global system variable slave_transaction_retries : if the replication slave SQL thread fails to execute a transaction because of an InnoDB deadlock or exceeded InnoDB's innodb_lock_wait_timeout or NDBCluster's TransactionDeadlockDetectionTimeout or TransactionInactiveTimeout, it automatically retries slave_transaction_retries times before stopping with an error. The default is 10. (Bug #8325)
• When a client releases a user-level lock, DO RELEASE_LOCK() will not be written to the binary log anymore (this makes the binary log smaller); as a counterpart, the slave does not actually take the lock when it executes GET_LOCK(). This is mainly an optimization and should not affect existing setups. (Bug #7998)
• The way the character set information is stored into the binary log was changed, so that it's now possible to have a replication master and slave running with different global character sets. A drawback is that replication from 5.0.3 masters to pre-5.0.3 slaves is impossible.
• The LOAD DATA statement was extended to support user variables in the target column list, and an optional SET clause. Now one can perform some transformations on data after they have been read and before they are inserted into the table. For example :

LOAD DATA INFILE 'file.txt'
  INTO TABLE t1
  (column1, @var1)
  SET column2 = @var1/100;

Also, replication of LOAD DATA was changed, so you can't replicate such statements from a 5.0.3 master to pre-5.0.3 slaves.

Bogues corrigés :

• If a MyISAM table on Windows had INDEX DIRECTORY or DATA DIRECTORY table options, mysqldump dumped the directory pathnames with single-backslash pathname separators. This would cause syntax errors when importing the dump file. mysqldump now changes '\' to '/' in the pathnames on Windows. (Bug #6660)
• mysql_fix_privilege_tables now fixes that the mysql privilege tables can be used in MySQL 4.1. This allows one to easily downgrade to 4.1 or run MySQL 5.0 and 4.1 with the same privilege files for testing purposes.
• Fixed bug creating user with GRANT fails with password but works without, (Bug #7905)
• mysqldump misinterpreted '_' and '%' characters in the names of tables to be dumped as wildcard characters. (Bug #9123)
• The definition of the enumeration-valued sql_mode column of the mysql.proc table was missing some of the current allowable SQL modes, so stored routines would not necessarily execute with the SQL mode in effect at the time of routine definition. (Bug #8902)
• REPAIR TABLE did not invalidate query results in the query cache that were generated from the table. (Bug #8480)
• In strict or traditional SQL mode, too-long string values assigned to string columns (CHAR, VARCHAR, BINARY, VARBINARY, TEXT, or BLOB) were correctly truncated, but the server returned an SQLSTATE value of 01000 (should be 22001). (Bug #6999, Bug #9029)
• Stored functions that used cursors could return incorrect results. (Bug #8386)
• AES_DECRYPT(col_name,key) could fail to return NULL for invalid values in col_name, if col_name was declared as NOT NULL. (Bug #8669)
• Ordering by unsigned expression (more complex than a column reference) was treating the value as signed, producing incorrectly sorted results. (Bug #7425)
• HAVING was treating unsigned columns as signed. (Bug #7425)
• Fixed a problem with boolean full-text searches on utf8 columns where a double quote in the search string caused a server crash. (Bug #8351)
• For a query with both GROUP BY and COUNT(DISTINCT) clauses and a FROM clause with a subquery, NULL was returned for any VARCHAR column selected by the subquery. (Bug #8218)
• Fixed a bug in TRUNCATE, which did not work within stored procedures. A workaround has been made so that within stored procedures, TRUNCATE is executed like DELETE. This was necessary because TRUNCATE is implicitly locking tables. (Bug #8850)
• Fixed an optimizer bug that caused incorrectly ordered result from a query that used a FULLTEXT index to retrieve rows and there was another index that was usable for ORDER BY. For such a query, EXPLAIN showed fulltext join type, but regular (not FULLTEXT) index in the Key column. (Bug #6635)
• If SELECT DISTINCT named an index column multiple times in the select list, the server tried to access different key fields for each instance of the column, which could result in a crash. (Bug #8532)
• For a stored function that refers to a given table, invoking the function while selecting from the same table resulted in a server crash. (Bug #8405)
• Comparison of a DECIMAL column containing NULL to a subquery that produced DECIMAL values resulted in a server crash. (Bug #8397)
• The --set-character-set option for myisamchk was changed to --set-collation. The value needed for specifying how to sort indexes is a collation name, not a character set name. (Bug #8349)
• Hostname matching didn't work if a netmask was specified for table-specific privileges. (Bug #3309)
• Corruption of MyISAM table indexes could occur with TRUNCATE TABLE if the table had already been opened. For example, this was possible if the table had been opened implicitly by selecting from a MERGE table that mapped to the MyISAM table. The server now issues an error message for TRUNCATE TABLE under these conditions. (Bug #8306)
• Setting the connection collation to a value different from the server collation followed by a CREATE TABLE statement that included a quoted default value resulted in a server crash. (Bug #8235)
• Fixed handling of table-name matching in mysqlhotcopy to accommodate DBD::mysql 2.9003 and up (which implement identifier quoting). (Bug #8136)
• Selecting from a view defined as a join caused a server crash if the query cache was enabled. (Bug #8054)
• Results in the query cache generated from a view were not properly invalidated after ALTER VIEW or DROP VIEW on that view. (Bug #8050)
• FOUND_ROWS() returned an incorrect value after a SELECT SQL_CALC_FOUND_ROWS DISTINCT statement that selected constants and included GROUP BY and LIMIT clauses. (Bug #7945)
• Selecting from an INFORMATION_SCHEMA table combined with a subselect on an INFORMATION_SCHEMA table caused an error with the message Table tbl_name is corrupted. (Bug #8164)
• Fixed a problem with equality propagation optimization for prepared statements and stored procedures that caused a server crash upon re-execution of the prepared statement or stored procedure. (Bug #8115, Bug #8849)
• LEFT OUTER JOIN between an empty base table and a view on an empty base table caused a server crash. (Bug #7433)
• Use of GROUP_CONCAT() in the select list when selecting from a view caused a server crash. (Bug #7116)
• Use of a view in a correlated subquery that contains HAVING but no GROUP BY caused a server crash. (Bug #6894)
• Handling by mysql_list_fields() of references to stored functions within views was incorrect and could result in a server crash. (Bug #6814)
• mysqldump now avoids writing SET NAMES to the dump output if the server is older than version 4.1 and would not understand that statement. (Bug #7997)
• Fixed problems when selecting from a view that had an EXISTS or NOT EXISTS subquery. Selecting columns by name caused a server crash. With SELECT *, a crash did not occur, but columns in outer query were not resolved properly. (Bug #6394)
• DDL statements for views were not being written to the binary log (and thus not subject to replication). (Bug #4838)
• The CHAR() function was not ignoring NULL arguments, contrary to the documentation. (Bug #6317)
• Creating a table using a name containing a character that is illegal in character_set_client resulted in the character being stripped from the name and no error. The character now is considered an error. (Bug #8041)
• Fixed a problem with the Cyrillic letters I and SHORT I being treated the same by the utf8_general_ci collation. (Bug #8385)
• Some INFORMATION_SCHEMA columns that contained catalog identifiers were of type LONGTEXT. These were changed to VARCHAR(N, where N is the appropriate maximum identifier length. (Bug #7215)
• Some INFORMATION_SCHEMA columns that contained timestamp values were of type VARBINARY. These were changed to TIMESTAMP. (Bug #7217)
• An expression that tested a case-insensitive character column against string constants that differed in lettercase could fail because the constants were treated as having a binary collation. (For example, WHERE city='London' AND city='london' could fail.) (Bug #7098, Bug #8690)
• The output of the STATUS (\s) command in mysql had the values for the server and client character sets reversed. (Bug #7571)
• If the slave was running with --replicate-*-table options which excluded one temporary table and included another, and the two tables were used in a single DROP TEMPORARY TABLE IF EXISTS statement, as the ones the master automatically writes to its binary log upon client's disconnection when client has not explicitly dropped these, the slave could forget to delete the included replicated temporary table. Only the slave needs to be upgraded. (Bug #8055)
• When setting integer system variables to a negative value with SET VARIABLES, the value was treated as a positive value modulo 2^32. (Bug #6958)
• Corrected a problem with references to DUAL where statements such as SELECT 1 AS a FROM DUAL would succeed but statements such as SELECT 1 AS a FROM DUAL LIMIT 1 would fail. (Bug #8023)
• Fixed a server crash caused by DELETE FROM tbl_name ... WHERE ... ORDER BY tbl_name.col_name when the ORDER BY column was qualified with the table name. (Bug #8392)
• Fixed a bug in MATCH ... AGAINST in natural language mode that could cause a server crash if the FULLTEXT index was not used in a join (EXPLAIN did not show fulltext join mode) and the search query matched no rows in the table (Bug #8522).
• InnoDB : Honor the --tmpdir startup option when creating temporary files. Previously, InnoDB temporary files were always created in the temporary directory of the operating system. On Netware, InnoDB will continue to ignore --tmpdir. (Bug #5822)
• Platform and architecture information in version information produced for --version option on Windows was always Win95/Win98 (i32). More accurately determine platform as Win32 or Win64 for 32-bit or 64-bit Windows, and architecture as ia32 for x86, ia64 for Itanium, and axp for Alpha. (Bug #4445)
• If multiple semicolon-separated statements were received in a single packet, they were written to the binary log as a single event rather than as separate per-statement events. For a server serving as a replication master, this caused replication to fail when the event was sent to slave servers. (Bug #8436)
• Fixed LOAD INDEX statement to actually load index in memory. (Bug #8452)
• Fixed a failure of multiple-table updates to replicate properly on slave servers when --replicate-*-table options had been specified. (Bug #7011)
• Fixed failure of CREATE TABLE ... LIKE Windows when the source or destination table was located in a symlinked database directory. (Bug #6607)
• With lower_case_table_names set to 1, mysqldump on Windows could write the same table name in different lettercase for different SQL statements. Fixed so that consistent lettercase is used. (Bug #5185)
• mysqld_safe now understands the --help option. Previously, it ignored the option and attempted to start the server anyway. (Bug #7931)
• Fixed problem in NO_BACKSLASH_ESCAPES SQL mode for strings that contained both the string quoting character and backslash. (Bug #6368)
• Fixed some portability issues with overflow in floating point values.
• Prepared statements now gives warnings on prepare.
• Fixed bug in prepared statements with SUM(DISTINCT...).
• Fixed bug in prepared statements with OUTER JOIN.
• Fixed a bug in CONV() function returning unsigned BIGINT number (third argument is positive, and return value does not fit in 32 bits). (Bug #7751)
• Fixed a failure of the IN() operator to return correct result if all values in the list were constants and some of them were using substring functions, for example, LEFT(), RIGHT(), or MID(). (Bug #7716)
• Fixed a crash in CONVERT_TZ() function when its second or third argument was from a const table (Voir Section 7.2.1, « Syntaxe de EXPLAIN (Obtenir des informations sur les SELECT) »7.2.1. Syntaxe de EXPLAIN (Obtenir des informations sur les SELECT)). (Bug #7705)
• Fixed a problem with calculation of number of columns in row comparison against subquery. (Bug #8020)
• Fixed erroneous output resulting from SELECT DISTINCT combined with a subquery and GROUP BY. (Bug #7946)
• Fixed server crash in comparing a nested row expression (for example row(1,(2,3))) with a subquery. (Bug #8022)
• Fixed server crash resulting from certain correlated subqueries with forward references (references to an alias defined later in the outer query). (Bug #8025)
• Fixed server crash resulting from re-execution of prepared statements containing subqueries. (Bug #8125)
• Fixed a bug where ALTER TABLE improperly would accept an index on a TIMESTAMP column that CREATE TABLE would reject. (Bug #7884)
• SHOW CREATE TABLE now reports ENGINE=MEMORY rather than ENGINE=HEAP for a MEMORY table (unless the MYSQL323 SQL mode is enabled). (Bug #6659)
• Fixed a bug where the use of GROUP_CONCAT() with HAVING caused a server crash. (Bug #7769)
• Fixed a bug where comparing the result of a subquery to a non-existent column caused a server crash on Windows. (Bug #7885)
• Fixed a bug in a combination of -not and trunc* operators of full-text search. Using more than one truncated negative search term, was causing empty result set.
• InnoDB : Corrected the handling of trailing spaces in the ucs2 character set. (Bug #7350, Bug #8771)
• InnoDB : Use native tmpfile() function on Netware. All InnoDB temporary files are created under sys:\tmp. Previously, InnoDB temporary files were never deleted on Netware.
• Fixed a bug in max_heap_table_size handling, that resulted in Table is full error when the table was still smaller than the limit. (Bug #7791).
• Fixed a symlink vulnerability in the mysqlaccess script. Reported by Javier Fernandez-Sanguino Pena and Debian Security Audit Team. (CAN-2005-0004)
• Fixed a bug that caused server crash if some error occured during filling of temporary table created for derived table or view handling. (Bug #7413)
• Fixed a bug which caused server crash if query containing CONVERT_TZ() function with constant arguments was prepared. (Bug #6849)
• Prevent adding CREATE TABLE .. SELECT query to the binary log when the insertion of new records partially failed. (Bug #6682)
• Fixed a bug which caused a crash when only the slave I/O thread was stopped and started. (Bug #6148)
• Giving mysqld a SIGHUP caused it to crash.
• Changed semantics of CREATE/ALTER/DROP DATABASE statements so that replication of CREATE DATABASE is possible when using --binlog-do-db and --binlog-ignore-db. (Bug #6391)
• A sequence of BEGIN (or SET AUTOCOMMIT=0), FLUSH TABLES WITH READ LOCK, transactional update, COMMIT, FLUSH TABLES WITH READ LOCK could hang the connection forever and possibly the MySQL server itself. This happened for example when running the innobackup script several times. (Bug #6732)
• mysqlbinlog did not print SET PSEUDO_THREAD_ID statements in front of LOAD DATA INFILE statements inserting into temporary tables, thus causing potential problems when rolling forward these statements after restoring a backup. (Bug #6671)
• InnoDB : Fixed a bug no error message for ALTER with InnoDB and AUTO_INCREMENT (Bug #7061). InnoDB now supports ALTER TABLE...AUTO_INCREMENT = x query to set auto increment value for a table.
• Made the MySQL server accept executing SHOW CREATE DATABASE even if the connection has an open transaction or locked tables; refusing it made mysqldump --single-transaction sometimes fail to print a complete CREATE DATABASE statement for some dumped databases. (Bug #7358)
• Fixed that, when encountering a « disk full » or « quota exceeded » write error, MyISAM sometimes didn't sleep and retry the write, thus resulting in a corrupted table. (Bug #7714)
• Fixed that --expire-log-days was not honored if using only transactions. (Bug #7236)
• Fixed that a slave could crash after replicating many ANALYZE TABLE, OPTIMIZE TABLE, or REPAIR TABLE statements from the master. (Bug #6461, Bug #7658)
• mysqlbinlog forgot to add backquotes around the collation of user variables (causing later parsing problems as BINARY is a reserved word). (Bug #7793)
• Ensured that mysqldump --single-transaction sets its transaction isolation level to REPEATABLE READ before proceeding (otherwise if the MySQL server was configured to run with a default isolation level lower than REPEATABLE READ it could give an inconsistent dump). (Bug #7850)
• Fixed that when using the RPAD() function (or any function adding spaces to the right) in a query that had to be resolved by using a temporary table, all resulting strings had rightmost spaces removed (i.e. RPAD() did not work) (Bug #4048)
• Fixed that a 5.0.3 slave can connect to a master < 3.23.50 without hanging (the reason for the hang is a bug in these quite old masters -- SELECT @@unknown_var hangs them -- which was fixed in MySQL 3.23.50). (Bug #7965)
• InnoDB : Fixed a deadlock without any locking, simple select and update (Bug #7975). InnoDB now takes an exclusive lock when INSERT ON DUPLICATE KEY UPDATE is checking duplicate keys.
• Fixed a bug where MySQL was allowing concurrent updates (inserts, deletes) to a table if binary logging is enabled. Changed to ensure that all updates are executed in a serialized fashion, because they are executed serialized when binlog is replayed. (Bug #7879)
• Fixed a rare race condition which could lead to FLUSH TABLES WITH READ LOCK hanging. (Bug #8682)
• Fixed a bug that caused the slave to stop on statements that produced an error on the master. (Bug #8412)

Fonctionnalités ajoutées ou modifiées :

• The SCHEMA and SCHEMAS keywords now are accepted as synonyms for DATABASE and DATABASES.
• Added initial support for rudimentary triggers.
• Added basic support for read-only server side cursors.
• Added --start-datetime, --stop-datetime, --start-position, --stop-position options to mysqlbinlog (makes point-in-time recovery easier).
• Made the MySQL server not react to signals SIGHUP and SIGQUIT on Mac OS X 10.3. This is needed because under this OS, the MySQL server receives lots of these signals (reported as Bug #2030).
• New --auto-increment-increment and --auto-increment-offset startup options. These allow you to set up a server to generate auto-increment values that don't conflict with another server.
• MySQL now by default will check dates and only allow fully correct dates. If you want to MySQL to behave as default, you should enable the new ALLOW_INVALID_DATES SQL mode.
• Added STRICT_TRANS_TABLES, STRICT_ALL_TABLES, NO_ZERO_IN_DATE, NO_ZERO_DATE, ERROR_FOR_DIVISION_BY_ZERO, and TRADITIONAL SQL modes. The TRADITIONAL mode is shorthand for all the preceding modes. When using mode TRADITIONAL, MySQL generates an error if you try to insert a wrong value in a column. It does not adjust the value to the closest possible legal value.
• MySQL now remembers which columns were declared to have default values. In STRICT_TRANS_TABLES/STRICT_ALL_TABLES mode, you now get an error if you do an INSERT without specifying all columns that don't have a default value. A side effect of this is that when you do SHOW CREATE for a new table, you will no longer Voir a DEFAULT value for a column for which you didn't specify a default value.
• The compilation flag DONT_USE_DEFAULT_FIELDS was removed because you can get the same behavior by setting the sql_mode system variable to STRICT_TRANS_TABLES.
• We now detect too-large floating point numbers during statement parsing and generate an error messages for them.

• Renamed the sql_updatable_view_key system variable to updatable_views_with_limit. This variable now can have only two values:

  • 1 or YES: Don't issue an error message (warning only) if a VIEW without presence of a key in the underlying table is used in queries with a LIMIT clause for updating. (This is the default value.)
  • 0 or NO: Prohibit update of a VIEW, which does not contain a key in the underlying table and the query uses a LIMIT clause (usually get from GUI tools).
• Reverted output format of SHOW TABLES to old pre-5.0.1 format that did not include a table type column. To get the additional column that lists the table type, use SHOW FULL TABLES now.
• The mysql_fix_privilege_tables script now initializes the global CREATE VIEW and SHOW VIEW privileges in the user table to the value of the CREATE privilege in that table.
• If the server finds that the user table has not been upgraded to include the view-related privilege columns, it treats each account as having view privileges that are the same as its CREATE privilege.

Bogues corrigés :

• Fixed that mysqlbinlog --read-from-remote-server sometimes couldn't accept two binary logfiles on the command line. (Bug #4507)
• Fixed that mysqlbinlog --position --read-from-remote-server had incorrect # at lines. (Bug #4506)
• Fixed that CREATE TABLE ... TYPE=HEAP ... AS SELECT... caused replication slave to stop. (Bug #4971)
• Fixed that mysql_options(...,MYSQL_OPT_LOCAL_INFILE,...) failed to disable LOAD DATA LOCAL INFILE. (Bug #5038)
• Fixed that disable-local-infile option had no effect if client read it from a configuration file using mysql_options(...,MYSQL_READ_DEFAULT,...). (Bug #5073)
• Fixed that SET GLOBAL SYNC_BINLOG did not work on some platforms (Mac OS X). (Bug #5064)
• Fixed that mysql-test-run failed on the rpl_trunc_binlog test if running test from the installed (the target of 'make install') directory. (Bug #5050)
• Fixed that mysql-test-run failed on the grant_cache test when run as Unix user 'root'. (Bug #4678)
• Fixed an unlikely deadlock which could happen when using KILL. (Bug #4810)
• Fixed a crash when one connection got KILLed while it was doing START SLAVE. (Bug #4827)
• Made FLUSH TABLES WITH READ LOCK block COMMIT if server is running with binary logging; this ensures that the binary log position can be trusted when doing a full backup of tables and the binary log. (Bug #4953)
• Fixed that the counter of an auto_increment column was not reset by TRUNCATE TABLE is the table was a temporary one. (Bug #5033)
• Fixed slave SQL thread so that the SET COLLATION_SERVER... statements it replicates don't advance its position (so that if it gets interrupted before the actual update query, it will later redo the SET). (Bug #5705)
• Fixed that if the slave SQL thread found a syntax error in a query (which should be rare, as the master parsed it successfully), it stops. (Bug #5711)
• Fixed that if a write to a MyISAM table fails because of a full disk or an exceeded disk quota, it prints a message to the error log every 10 minutes, and waits until disk becomes free. (Bug #3248)
• Fixed problem introduced in 4.0.21 where a connection starting a transaction, doing updates, then FLUSH TABLES WITH READ LOCK, then COMMIT, would cause replication slaves to stop complaing about error 1223. Bug surfaced when using the InnoDB innobackup script. (Bug #5949)

Fonctionnalité ajoutée ou modifiée :

• For replication of MEMORY (HEAP) tables: Made the master automatically write a DELETE FROM statement to its binary log when a MEMORY table is opened for the first time since master's startup. This is for the case where the slave has replicated a non-empty MEMORY table, then the master is shut down and restarted: the table is now empty on master; the DELETE FROM empties it on slave too. Note that even with this fix, between the master's restart and the first use of the table on master, the slave still has out-of-date data in the table. But if you use the --init-file option to populate the MEMORY table on the master at startup, it ensures that the failing time interval is zero. (Bug #2477)
• When a session having open temporary tables terminates, the statement automatically written to the binary log is now DROP TEMPORARY TABLE IF EXISTS instead of DROP TEMPORARY TABLE, for more robustness.
• The MySQL server now returns an error if SET SQL_LOG_BIN is issued by a user without the SUPER privilege (in previous versions it just silently ignored the statement in this case).
• Changed that when the MySQL server has binary logging disabled (that is, no log-bin option was used) then no transaction binlog cache is allocated for connections (this should save binlog_cache_size bytes of memory (32 kilobytes by default) for every connection).
• Added option --replicate-same-server-id.

Bogues corrigés :

• Strange results with index (x, y) ... WHERE x=val_1 AND y>=val_2 ORDER BY pk; (Bug #3155)
• Subquery and order by (Bug #3118)
• ALTER DATABASE caused the client to hang if the database did not exist. (Bug #2333)
• SLAVE START (which is a deprecated syntax, START SLAVE should be used instead) could crash the slave. (Bug #2516)
• Multiple-table DELETE statements were never replicated by the slave if there were any replicate-*-table options. (Bug #2527)
• The MySQL server did not report any error if the query (submitted through mysql_real_query() or mysql_prepare()) was terminated by garbage characters (which can happen if you pass a wrong length parameter to mysql_real_query() or mysql_prepare()); the result was that the garbage characters were written into the binary log. (Bug #2703)
• Replication: If a client connects to a slave server and issues an administrative statement for a table (for example, OPTIMIZE TABLE or REPAIR TABLE), this could sometimes stop the slave SQL thread. This does not lead to any corruption, but you must use START SLAVE to get replication going again. (Bug #1858)
• Made clearer the error message which one gets when an update is refused because of the read-only option. (Bug #2757)
• Fixed that replicate-wild-*-table rules apply to ALTER DATABASE when the table pattern is '%', like it is already the case for CREATE DATABASE and DROP DATABASE. (Bug #3000)
• Fixed that when a Rotate event is found by the slave SQL thread in the middle of a transaction, the value of Relay_Log_Pos in SHOW SLAVE STATUS remains correct. (Bug #3017)
• Corrected the master's binary log position that InnoDB reports when it is doing a crash recovery on a slave server. (Bug #3015)
• Changed the column Seconds_Behind_Master in SHOW SLAVE STATUS to never show a value of -1. (Bug #2826)
• Changed that when a DROP TEMPORARY TABLE statement is automatically written to the binlog when a session ends, the statement is recorded with an error code of value zero (this ensures that killing a SELECT on the master does not result in a superfluous error on the slave). (Bug #3063)
• Changed that when a thread handling INSERT DELAYED (also known as a delayed_insert thread) is killed, its statements are recorded with an error code of value zero (killing such a thread does not endanger replication, so we thus avoid a superfluous error on the slave). (Bug #3081)
• Fixed deadlock when two START SLAVE commands were run at the same time. (Bug #2921)
• Fixed that a statement never triggers a superfluous error on the slave, if it must be excluded given the replicate-* options. The bug was that if the statement had been killed on the master, the slave would stop. (Bug #2983)
• The --local-load option of mysqlbinlog now requires an argument.
• Fixed a segmentation fault when running LOAD DATA FROM MASTER after RESET SLAVE. (Bug #2922)
• mysqlbinlog --read-from-remote-server read all binary logs following the one that was requested. It now stops at the end of the requested file, the same was it does when reading a local binary log. (Bug #3204)
• Fixed mysqlbinlog --read-from-remote-server to print the exact positions of events in the "at #" lines. (Bug #3214)
• Fixed a rare error condition that caused the slave SQL thread spuriously to print the message Binlog has bad magic number and stop when it was not necessary to do so. (Bug #3401)
• Fixed mysqlbinlog not to forget to print a USE statement under rare circumstances where the binary log contained a LOAD DATA INFILE statement. (Bug #3415)
• Fixed a memory corruption when replicating a LOAD DATA INFILE when the master had version 3.23. (Bug #3422)
• Multiple-table DELETE statements were always replicated by the slave if there were some replicate-*-ignore-table options and no replicate-*-do-table options. (Bug #3461)
• Fixed a crash of the MySQL slave server when it was built with --with-debug and replicating itself. (BUG #3568)
• Fixed that in some replication error messages, a very long query caused the rest of the message to be invisible (truncated), by putting the query last in the message. (Bug #3357)

Fonctionnalité ajoutée ou modifiée :

• The KILL statement now takes CONNECTION and QUERY modifiers. The first is the same as KILL with no modifier (it kills a given connection thread). The second kills only the statement currently being executed by the connection.
• Added TIMESTAMPADD() and TIMESTAMPDIFF() functions.
• Added WEEK and QUARTER values as INTERVAL arguments for the DATE_ADD() and DATE_SUB() functions.
• New binary log format that enables replication of these session variables: sql_mode, SQL_AUTO_IS_NULL, FOREIGN_KEY_CHECKS (which was already replicated since 4.0.14, but here it's done more efficiently and takes less space in the binary logs), UNIQUE_CHECKS. Other variables (like character sets, SQL_SELECT_LIMIT, ...) will be replicated in upcoming 5.0.x releases.
• Implemented Index Merge optimization for OR clauses. Voir Section 7.2.6, « Optimisation de combinaison d'index »7.2.6. Optimisation de combinaison d'index.
• Basic support for stored procedures (SQL:2003 style). Voir Chapitre 19, Procédures stockées et fonctionsChapitre 19. Procédures stockées et fonctions.
• Added SELECT INTO list_of_vars, which can be of mixed (that is, global and local) types. Voir Section 19.2.9.3, « Syntaxe de SELECT ... INTO »19.2.9.3. Syntaxe de SELECT ... INTO.
• Easier replication upgrade (5.0.0 masters can read older binary logs and 5.0.0 slaves can read older relay logs). Voir Section 6.5, « Compatibilité de la réplication entre les versions de MySQL »6.5. Compatibilité de la réplication entre les versions de MySQL for more details). The format of the binary log and relay log is changed compared to that of MySQL 4.1 and older.
• Important note: If you upgrade to MySQL 4.1.1 or higher, it is difficult to downgrade back to 4.0 or 4.1.0! That is because, for earlier versions, InnoDB is not aware of multiple tablespaces.

Bogues corrigés :

Fonctionnalités ajoutées ou modifiées :

• Updated version of libedit to 2.9. (Bug #2596)
• InnoDB: When FOREIGN_KEY_CHECKS=0, ALTER TABLE and RENAME TABLE will ignore any type incompatibilities between referencing and referenced columns. Thus, it will be possible to convert the character sets of columns that participate in a foreign key. Be sure to convert all tables before modifying any data! (Bug #9802)
• Previously in MySQL 4.1, an Illegal mix of collations error occurred when mixing strings from same character set when one had a non-binary collation and the other a binary collation. Now the binary collation takes precedence, so that both strings are treated as having the binary collation. This restores compatibility with MySQL 4.0 behavior.

Bogues corrigés :

• Security fix: If mysqld was started with --user=non_existent_user, it would run using the privileges of the account it was invoked from, even if that was root. (Bug #9833)
• MAX() for an INT UNSIGNED (unsigned 4-byte integer) column could return negative values if the column contained values larger than 2^31. (Bug #9298)
• Fixed a deadlock resulting from use of FLUSH TABLES WITH READ LOCK while an INSERT DELAYED statement is in progress. (Bug #7823)
• Multiple-table updates could produce spurious data-truncation warnings if they used a join across columns that are indexed using a column prefix. (Bug #9103)
• Use of a subquery that used WITH ROLLUP in the FROM clause of the main query sometimes resulted in a Column cannot be null error. (Bug #9681)
• RENAME TABLE for an ARCHIVE table failed if the .arn file was not present. (Bug #9911)
• Fixed an optimizer problem where extraneous comparisons between NULL values in indexed columns were being done for operators such as = that are never true for NULL. (Bug #8877)
• SELECT ROUND(expr) produced a different result than CREATE TABLE ... SELECT ROUND(expr). (Bug #9837)
• Fixed some awk script portability problems in cmd-line-utils/libedit/makelist.sh. (Bug #9954)
• Changed metadata for result of SHOW KEYS: Data type for Sub_part column now is SMALLINT rather than TINYINT because key part length can be longer than 255. (Bug #9439)
• Fixed some problems with myisampack on 64-bit systems that resulted in segmentation violations. (Bug #9487)
• Fixed an optimizer bug in computing the union of two ranges for the OR operator. (Bug #9348)
• Fixed an index corruption problem for MyISAM tables that resulted from the 4.1 behavior of padding values with blanks for comparison: Dumping a table with mysqldump, reloading it, and then re-running the binary log against it crashed the index and necessitated a repair. (Bug #9188)
• Fixed a segmentation fault in mysqlcheck that occurred when the last table checked in --auto-repair mode returned an error (such as the table being a MERGE table). (Bug #9492)
• Fixed the client/server protocol for prepared statements so that reconnection works properly when the connection is killed while reconnect is enabled. (Bug #8866)
• INSERT ... ON DUPLICATE KEY UPDATE incorrectly updated a TIMESTAMP column to the current timestamp, even if the update list included col_name = col_name for that column to prevent the update. (Bug #7806)
• Starting mysqld with the --skip-innodb and --default-storage-engine=innodb (or --default-table-type=innodb caused a server crash. (Bug #9815)
• Queries containing CURRENT_USER() incorrectly were registered in the query cache. (Bug #9796)
• A server installed as a Windows service and started with --shared-memory could not be stopped. (Bug #9665)
• mysqldump dumped core when invoked with --tmp and --single-transaction options and a non-existent table name. (Bug #9175)
• Additional fix for mysql_server_init() and mysql_server_end() C API functions so that stopping and restarting the embedded server will not cause a crash. (Bug #7344)
• mysql.server no longer uses non-portable alias command or LSB functions. (Bug #9852)
• Fixed a readline-related crash in mysql when the user pressed Control-R. (Bug #9568)
• TIMEDIFF() with a negative time first argument and postive time second argument produced incorrect results. (Bug #8068)
• Fixed a bug that caused concurrent inserts to be allowed into the tables in the SELECT ... UNION ... part of INSERT ... SELECT ... UNION .... This could result in the incorrect order of queries in the binary log. (Bug #9922)
• Warning message from GROUP_CONCAT() did not always indicate correct number of lines. (Bug #8681)
• InnoDB: ENUM and SET columns were treated incorrectly as character strings. This bug did not manifest itself with latin1 collations, but it caused malfunction with utf8. Old tables will continue to work. In new tables, ENUM and SET will be internally stored as unsigned integers. (Bug #9526)
• InnoDB: Avoid test suite failures caused by a locking conflict between two server instances at server shutdown/startup. This conflict on advisory locks appears to be the result of a bug in the operating system; these locks should be released when the files are closed, but somehow that does not always happen immediately in Linux. (Bug #9381)
• InnoDB: Prevent ALTER TABLE from changing the storage engine if there are foreign key constraints on the table. (Bug #5574, Bug #5670)
• InnoDB: Fixed a deadlock without any locking, simple select and update. (Bug #7975) InnoDB now takes an exclusive lock when INSERT ON DUPLICATE KEY UPDATE is checking duplicate keys.

Fonctionnalités ajoutées ou modifiées :

• ONLY_FULL_GROUP_BY no longer is included in the ANSI composite SQL mode. (Bug #8510)
• mysqld_safe will create the directory where the UNIX socket file is to be located if the directory does not exist. This applies only to the last component of the directory pathname. (Bug #8513)
• The coercibility for the return value of functions such as USER() or VERSION() now is « system constant » rather than « implicit. » This makes these functions more coercible than column values so that comparisons of the two do not result in Illegal mix of collations errors. COERCIBILITY() was modified to accommodate this new coercibility value. Voir Section 12.8.3, « Fonctions d'informations »12.8.3. Fonctions d'informations.
• User variable coercibility has been changed from « coercible » to « implicit. » That is, user variables have the same coercibility as column values.
• NULL now is considered more coercible than string constants. This resolves some Illegal mix of collations conflicts.
• Modified the parser to allow SELECT statements following the UNION keyword to be subqueries in parentheses. (Bug #2435)
• For slave replication servers started with --replicate-* options, statements that should not be replicated according those options no longer are written to the slave's general query log. (Bug #8297)
• Added SQL_NOTES session variable to cause Note-level warnings not to be recorded. (Bug #6662)
• InnoDB: Commit after every 10,000 copied rows when executing CREATE INDEX, DROP INDEX or OPTIMIZE TABLE, which are internally implemented as ALTER TABLE. This makes it much faster to recover from an aborted operation.
• Added a new global system variable slave_transaction_retries: If the replication slave SQL thread fails to execute a transaction because of an InnoDB deadlock or exceeded InnoDB's innodb_lock_wait_timeout or NDBCluster's TransactionDeadlockDetectionTimeout or TransactionInactiveTimeout, it automatically retries slave_transaction_retries times before stopping with an error. The default in MySQL 4.1 is 0. You must explicitly set the value greater than 0 to enable the « retry » behavior. (In MySQL 5.0.3 or newer, the default is 10.) (Bug #8325)
• Added --with-big-tables compilation option to configure. (Previously it was necessary to pass -DBIG_TABLES to the compiler manually in order to enable large table support.) Voir Section 2.4.2, « Options habituelles de configure »2.4.2. Options habituelles de configure for details.
• Added configuration directives !include and !includedir for including option files and searching directories for option files. Voir Section 4.3.2, « Fichier d'options my.cnf »4.3.2. Fichier d'options my.cnf for usage.

Bogues corrigés :

• The use of XOR together with NOT ISNULL() erroneously resulted in some outer joins being converted to inner joins by the optimizer. (Bug #9017)
• Fixed utf8_spanish2_ci and ucs2_spanish2_ci collations to not consider 'r' equal to 'rr'. If you upgrade to this version from an earlier version, you should rebuild the indexes of affected tables. (Bug #9269)
• Allow extra HKSCS and cp950 characters (big5 extension characters) to be accepted in big5 columns. (Bug #9357)
• BLOB(M) and TEXT(M) columns, with M less than 256, were being created as BLOB and TEXT columns rather than TINYBLOB or TINYTEXT columns. (Bug #9303)
• Fixed a problem with INSERT ... SELECT ... ON DUPLICATE KEY UPDATE where a column named in the insert list and in the ON DUPLICATE KEY UPDATE clause was erroneously declared to be ambiguous. (Bug #8147)
• In prepared statements, subqueries containing parameters were erroneously treated as const tables during preparation, resulting in a server crash. (Bug #8807)
• Fixed a problem with OPTIMIZE TABLE for InnoDB tables being written twice to the binary log. (Bug #9149)
• Provide more informative error messages in clustered setting when a query is issued against a table that has been modified by another mysqld server. (Bug #6762)
• For MyISAM tables, REPAIR TABLE no longer discard rows that have incorrect checksum. (Bug #9824)
• Depending on index statistics, GROUP BY col1, col2, ... could return incorrect results if the first table processed for a join had several indexes that cover the grouped columns. (Bug #9213)
• Fixed incorrect evaluation of ALL/ANY subqueries that contain a HAVING clause. (Bug #9350)
• Fixed server crash when left expression of IN/ALL/ANY comparision was a subquery. (Bug #8888)
• Fixed option-parsing code for the embedded server to understand K, M, and G suffixes for the net_buffer_length and max_allowed_packet options. (Bug #9472)
• Fixed a crash when using TIMESTAMP columns with no minute or second parts in GROUP BY with the new system variable set to 1. (Bug #9401)
• If a MyISAM table on Windows had INDEX DIRECTORY or DATA DIRECTORY table options, mysqldump dumped the directory pathnames with single-backslash pathname separators. This would cause syntax errors when importing the dump file. mysqldump now changes '\' to '/' in the pathnames on Windows. (Bug #6660)
• Fixed a server crash caused by use of NOW() is a subquery. (Bug #8824)
• Fixed problems with static variables to allow building on Fedora Core 3. (Bug #6554)
• Some user variables were not being handled with « implicit » coercibility. (Bug #9425)
• Setting the max_error_count system variable to 0 resulted in a setting of 1. (Bug #9072)
• Fixed a collation coercibility problem that caused a union between binary and non-binary columns to fail. (Bug #6519)
• Fixed a problem with the tee command in mysql that resulted in mysql crashing. (Bug #8499)
• On Windows, create shared memory objects with the proper access rights to make them usable when the client and server are running under different accounts. (Bug #8226)
• Bundled zlib in the source distribution was upgraded to 1.2.2. (Bug #9118)
• Fixed server crash resulting from queries that combined SELECT DISTINCT, SUM(), and ROLLUP. (Bug #8615)
• Incorrect results were returned from queries that combined SELECT DISTINCT, GROUP BY, and ROLLUP. (Bug #8616)
• Fixed a bug that under certain circumstances could allow a privilege escalation via database wildcards in GRANT. (CAN-2004-0957)
• Too many rows were returned from queries that combined ROLLUP and LIMIT if SQL_CALC_FOUND_ROWS was given. (Bug #8617)
• mysqldump misinterpreted '_' and '%' characters in the names of tables to be dumped as wildcard characters. (Bug #9123)
• Made the relay_log_space_limit system variable show up in the output of SHOW VARIABLES. (Bug #7100)
• Use of GROUP_CONCAT(x) in a subquery, where x was an alias to a column in the outer query, resulted in a server crash. (Bug #8656)
• The CHARSET(), COLLATION(), and COERCIBILITY() functions sometimes returned NULL. CHARSET() and COLLATION() returned NULL when given any of these arguments that evaluated to NULL: A system function such as DATABASE(); a column value; and a user variable. Now CHARSET() and COLLATION() return the system character set and collation; the column character set and collation; and binary. COERCIBILITY(NULL) now returns « ignorable » coercibility rather than NULL. (Bug #9129)
• Expressions involving nested CONCAT() calls and character set conversion of string constants could return an incorrect result. (Bug #8785)
• The MEMORY storage engine did not properly increment an AUTO_INCREMENT column if there was a second composite index that included the column. (Bug #8489)
• Fixed a bug in the filesort routine such that killing a filesort could cause an assertion failure. (Bug #8799)
• REPAIR TABLE did not invalidate query results in the query cache that were generated from the table. (Bug #8480)
• If max_join_size was set, a query containing a subquery that exceeded the examined-rows limit could hang. (Bug #8726)
• Mixed-case database and table names in the grant tables were ignored for authentication if the lower_case_table_names system variable was set. GRANT will not create such privileges when lower_case_table_names is set, but it is possible to create them by direct manipulation of the grant tables, or that old grant records were present before setting the variable. (Bug #7989)
• AES_DECRYPT(col_name,key) could fail to return NULL for invalid values in col_name, if col_name was declared as NOT NULL. (Bug #8669)
• Ordering by unsigned expression (more complex than a column reference) was treating the value as signed, producing incorrectly sorted results. (Bug #7425)
• HAVING was treating unsigned columns as signed. (Bug #7425)
• Fixed a problem with boolean full-text searches on utf8 columns where a double quote in the search string caused a server crash. (Bug #8351)
• MIN(col_name) and MAX(col_name) could fail to produce the correct result if col_name was contained in multiple indexes and the optimizer did not choose the first index that contained the column. (Bug #8893)
• Table creation for a MyISAM table failed if DATA DIRECTORY or INDEX DIRECTORY options were given that specified the pathname to the database directory where the table files would be created by default. (Bug #8707)
• Fixed a problem with LIKE pattern-matching for strings with the cp1251_bin binary collation. (Bug #8560)
• A join on two tables failed when each contained a BIGINT UNSIGNED column that were compared when their values exceeded 2^63 - 1. The match failed and the join returned no rows. (Bug #8562)
• For a query with both GROUP BY and COUNT(DISTINCT) clauses and a FROM clause with a subquery, NULL was returned for any VARCHAR column selected by the subquery. (Bug #8218)
• Fixed an optimizer bug that caused incorrectly ordered result from a query that used a FULLTEXT index to retrieve rows and there was another index that was usable for ORDER BY. For such a query, EXPLAIN showed fulltext join type, but regular (not FULLTEXT) index in the Key column. (Bug #6635)
• For a statement string that contained multiple slow queries, only the last one would be written to the slow query log. (Bug #8475)
• When the server was started with --skip-name-resolve, specifying hostname values that included netmasks in GRANT statements did not work. (Bug #8471)
• The --set-character-set option for myisamchk was changed to --set-collation. The value needed for specifying how to sort indexes is a collation name, not a character set name. (Bug #8349)
• Hostname matching didn't work if a netmask was specified for table-specific privileges. (Bug #3309)
• Binary data stored in BLOB or BINARY columns would be erroneously dumped if mysqldump was invoked with --hex-blob and --skip-extended-insert arguments. This happened if data contained characters larger then 0x7F (Bug #8830).
• Corruption of MyISAM table indexes could occur with TRUNCATE TABLE if the table had already been opened. For example, this was possible if the table had been opened implicitly by selecting from a MERGE table that mapped to the MyISAM table. The server now issues an error message for TRUNCATE TABLE under these conditions. (Bug #8306)
• Fixed handling of table-name matching in mysqlhotcopy to accommodate DBD::mysql 2.9003 and up (which implement identifier quoting). (Bug #8136)
• In the mysql_real_escape_string() C API function, when a multi-byte character is encountered that is illegal in the current character set, escape only the first byte, not each byte. This avoids creating a valid character from an invalid one. (Bug #8378)
• Fixed a problem with the cp1250_czech_cs collation that caused empty literal strings not to compare equal to empty character columns. (Bug #8840)
• Fixed a problem in index cost calculation that caused a USE INDEX or FORCE INDEX hint not to be used properly for a LEFT JOIN across indexed BLOB columns. (Bug #7520)
• The column type for MAX(datetime_col) was returned as VARCHAR rather than DATETIME if the query included a GROUP BY clause. (Bug #5615)
• FOUND_ROWS() returned an incorrect value for preceding SELECT statements that used no table or view. (Bug #6089)
• In string literals with an escape character ('\') followed by a multi-byte character that has a second byte of '\', the literal was not interpreted correctly. The next character now is escaped, not just the next byte. (Bug #8303)
• InnoDB: Work around a problem in AIX 5.1 patched with ML7 security patch: InnoDB would refuse to open its ibdata files, complaining about an operating system error 0.
• InnoDB: Fixed a memory corruption bug if one created a table with a primary key that contained at least two column prefixes. An example: CREATE TABLE t(a char(100), b tinyblob, PRIMARY KEY(a(5), b(10))).
• InnoDB: Do not try to space-pad BLOB columns containing ucs2 characters. This avoids an assertion failure that was introduced when fixing Bug #7350. (Bug #8771)
• InnoDB: Fixed a bug : MySQL-4.1.8 - 4.1.10 could complain that an InnoDB table created with MySQL-3.23.49 or earlier was in the new compact InnoDB table format of 5.0.3 or later, and InnoDB would refuse to use that table. There is nothing wrong with the table, it is mysqld that is in error. Workaround: wait that 4.1.11 is released before doing an upgrade, or dump the table and recreate it with any MySQL version >= 3.23.50 before upgrading.
• InnoDB: Honor the --tmpdir startup option when creating temporary files. Previously, InnoDB temporary files were always created in the temporary directory of the operating system. On Netware, InnoDB will continue to ignore --tmpdir. (Bug #5822)
• InnoDB: If MySQL wrote to its binlog, but for some reason trx->update_undo and trx->insert_undo were NULL in InnoDB, then trx->commit_lsn was garbage, and InnoDB could assert in the log flush of trx_commit_complete_for_mysql(). (Bug #9277)
• InnoDB: If InnoDB cannot allocate memory, keep retrying for 60 seconds before we intentionally crash mysqld; maybe the memory shortage is just temporary.
• InnoDB: If one used LOCK TABLES, created an InnoDB temp table, and did a multi-table update where a MyISAM table was the update table and the temp table was a read table, then InnoDB asserted in row0sel.c because n_mysql_tables_in_use was 0. Also, we remove the assertion altogether and just print an error to the .err log if this important consistency check fails. (Bug #8677)
• mysqldump now avoids writing SET NAMES to the dump output if the server is older than version 4.1 and would not understand that statement. (Bug #7997)
• Fixed a bug in my_print_defaults that made it ignore the --defaults-extra-file and --defaults-file options.
• Retrieving from a view defined as a SELECT that mixed UNION ALL and UNION DISTINCT resulted in a different result than retrieving from the original SELECT. (Bug #6565)
• Worked around a bug in support for NSS support in glibc when static linking is used and LDAP is one of the NSS sources. The workaround is to detect when the bug causes a segfault and issue a diagnostic message with information about the problem. (Bug #3037, Bug #4872)
• If the mysql prompt was configured to display the default database name, and that database was dropped, mysql did not update the prompt. (Bug #4802)
• perror was printing a spurious extra line of output ("Error code ###: Unknown error ###" printed directly before the correct line with the error message). (Bug #8517)
• The CHAR() function was not ignoring NULL arguments, contrary to the documentation. (Bug #6317)
• Neither SHOW ERRORS nor SHOW WARNINGS were displaying Error-level messages. (Bug #6572)
• Creating a table using a name containing a character that is illegal in character_set_client resulted in the character being stripped from the name and no error. The character now is considered an error. (Bug #8041)
• Fixed a problem with the Cyrillic letters I and SHORT I being treated the same by the utf8_general_ci collation. (Bug #8385)
• The MAX_CONNECTIONS_PER_HOUR resource limit was not being reset hourly and thus imposed an absolute limit on number of connections per account until the server is restarted or the limits flushed. (Bug #8350)
• With a database was dropped with lower_case_table_names=2, tables in the database also were dropped but not being flushed properly from the table cache. If the database was recreated, the tables also would appear to have been recreated. (Bug #8355)
• Changed mysql_server_end() C API function to restore more variables to their initial state so that a subsequent call to mysql_server_init() would not cause a client program crash. (Bug #7344)
• Fixed a problem with accented letters improperly being treated as distinct with the utf_general_ci collation. (Bug #7878)
• ENUM and SET columns in privilege tables incorrectly had a case-sensitive collation, resulting in failure of assignments of values that did not have the same lettercase as given in the column definitions. The collation was changed to be case insensitive. (Bug #7617)
• An expression that tested a case-insensitive character column against string constants that differed in lettercase could fail because the constants were treated as having a binary collation. (For example, WHERE city='London' AND city='london' could fail.) (Bug #7098, Bug #8690)
• The output of the STATUS (\s) command in mysql had the values for the server and client character sets reversed. (Bug #7571)
• If the slave was running with --replicate-*-table options which excluded one temporary table and included another, and the two tables were used in a single DROP TEMPORARY TABLE IF EXISTS statement, as the ones the master automatically writes to its binary log upon client's disconnection when client has not explicitly dropped these, the slave could forget to delete the included replicated temporary table. Only the slave needs to be upgraded. (Bug #8055)
• Treat user variables as having IMPLICIT derivation (coercibility) to avoid « Illegal mix of collations » errors when replicating user variables. (Bug #6676)
• When setting integer system variables to a negative value with SET VARIABLES, the value was treated as a positive value modulo 2^32. (Bug #6958)
• Fixed a bug in bundled readline library that caused segmentation fault in mysql when user entered Shift+Enter. (Bug #5672)
• Fix conversion of strings -> double to get higher accuracy for floating point values that are integers, like: 123.45E+02 (Bug #7840).
• Fixed a bug in MATCH ... AGAINST in natural language mode that could cause a server crash if the FULLTEXT index was not used in a join (EXPLAIN did not show fulltext join mode) and the search query matched no rows in the table (Bug #8522).
• Platform and architecture information in version information produced for --version option on Windows was always Win95/Win98 (i32). More accurately determine platform as Win32 or Win64 for 32-bit or 64-bit Windows, and architecture as ia32 for x86, ia64 for Itanium, and axp for Alpha. (Bug #4445)
• Fixed a rare race condition which could lead to FLUSH TABLES WITH READ LOCK hanging. (Bug #8682)
• Fixed a bug that caused the slave to stop on statements that produced an error on the master. (Bug #8412)
• If multiple semicolon-separated statements were received in a single packet, they were written to the binary log as a single event rather than as separate per-statement events. For a server serving as a replication master, this caused replication to fail when the event was sent to slave servers. (Bug #8436)

Note: The security improvements related to creation of table files and to user-defined functions were made after MySQL 4.1.10 was released and are present in MySQL 4.1.10a. We would like to thank Stefano Di Paola <> for making us aware of these.

Fonctionnalités ajoutées ou modifiées :

• Added back faster subquery execution from 4.1.8. This adds also back a bug from 4.1.8 in comparing NULL to the value of a subquery. Voir Section 1.5.7.4, « Bugs connus / limitations de MySQL »1.5.7.4. Bugs connus / limitations de MySQL.
• Security improvement: The server creates .frm, .MYD, .MYI, .MRG, .ISD, and .ISM table files only if a file with the same name does not already exist. Thanks to Stefano Di Paola <> for finding and informing us about this issue. (CAN-2005-0711)
• Security improvement: User-defined functions should have at least one symbol defined in addition to the xxx symbol that corresponds to the main xxx() function. These auxiliary symbols correspond to the xxx_init(), xxx_deinit(), xxx_reset(), xxx_clear(), and xxx_add() functions. mysqld by default no longer loads UDFs unless they have at least one auxiliary symbol defined in addition to the main symbol. The --allow-suspicious-udfs option controls whether UDFs that have only an xxx symbol can be loaded. By default, the option is off. mysqld also checks UDF filenames when it reads them from the mysql.func table and rejects those that contain directory pathname separator characters. (It already checked names as given in CREATE FUNCTION statements.) Voir Section 27.2.3.1, « Fonctions utilisateur : appeler des fonctions simples »27.2.3.1. Fonctions utilisateur : appeler des fonctions simples, Section 27.2.3.2, « Appeler des fonctions utilisateurs pour les groupements »27.2.3.2. Appeler des fonctions utilisateurs pour les groupements, and Section 27.2.3.6, « Précautions à prendre avec les fonctions utilisateur »27.2.3.6. Précautions à prendre avec les fonctions utilisateur. Thanks to Stefano Di Paola <> for finding and informing us about this issue. (CAN-2005-0709, CAN-2005-0710)
• Setting the connection collation to a value different from the server collation followed by a CREATE TABLE statement that included a quoted default value resulted in a server crash. (Bug #8235)
• Thread stack size was increased from 192KB to 256KB on Linux/IA-64 (too small stack size was causing server crashes on some queries). (Bug #8391)
• From the Windows distribution, predefined accounts without passwords for remote users ("root@%", "@%") were removed (other distributions never had them).
• Added mysql_library_init() and mysql_library_end() as synonyms for the mysql_server_init() and mysql_server_end() C API functions. mysql_library_init() and mysql_library_end() are #define symbols, but the names more clearly indicate that they should be called when beginning and ending use of a MySQL C API library no matter whether the application uses libmysqlclient or libmysqld. (Bug #6149)
• The server now issues a warning when lower_case_table_names=2 and the data directory is on a case-sensitive filesystem, just as when lower_case_table_names=0 on a case-insensitive filesystem. (Bug #7887)
• The server now issues a warning to the error log when it encounters older tables that contain character columns that might be interpreted by newer servers to have a different column length. (Bug #6913) Voir Section 2.6.2, « Passer de la version 4.0 à la version 4.1 »2.6.2. Passer de la version 4.0 à la version 4.1 for a discussion of this problem and what to do about it.
• InnoDB: When MySQL/InnoDB is compiled on Mac OS X 10.2 or earlier, detect the operating system version at run time and use the fcntl() file flush method on Mac OS X versions 10.3 and later. Apple had disabled fsync() in Mac OS X for internal disk drives, which caused corruption at power outages.
• InnoDB: A shared record lock (LOCK_REC_NOT_GAP) is now taken for a matching record in the foreign key check because inserts can be allowed into gaps.
• InnoDB: Relaxed locking in INSERT...SELECT, single table UPDATE...SELECT and single table DELETE...SELECT clauses when innobase_locks_unsafe_for_binlog is used and isolation level of the transaction is not serializable. InnoDB uses consistent read in these cases for a selected table.

Bogues corrigés :

• FOUND_ROWS() returned an incorrect value after a SELECT SQL_CALC_FOUND_ROWS DISTINCT statement that selected constants and included GROUP BY and LIMIT clauses. (Bug #7945)
• Fixed a bug in cardinality estimations for HASH indexes of TEMPORARY tables created using MEMORY storage engine. As a result queries that were using this index (as shown by EXPLAIN) could have returned incorrect results. (Bug #8371)
• Corrected a problem with references to DUAL where statements such as SELECT 1 AS a FROM DUAL would succeed but statements such as SELECT 1 AS a FROM DUAL LIMIT 1 would fail. (Bug #8023)
• Fixed a server crash caused by DELETE FROM tbl_name ... WHERE ... ORDER BY tbl_name.col_name when the ORDER BY column was qualified with the table name. (Bug #8392)
• mysqld had problems finding its language files if the basedir value was specified as a very long pathname. (Bug #8015)
• Updates were being written to the binary log when there were binlog-do-db or binlog-ignore-db options even when there was no current database, contrary to Section 5.9.4, « Le log binaire »5.9.4. Le log binaire. (Bug #6749)
• Fixed conversion of floating-point values to character fields when the absolute value of the float was less than 1, and also fixed calculation of length for negative values. (Bug #7774)
• Column headers in query results retrieved from the query cache could be corrupted when a non-4.1 client was served a result originally generated for a 4.1 client. The query cache was not keeping track of which client/server protocol was being used. (Bug #6511)
• Fixed LOAD INDEX statement to actually load index in memory. (Bug #8452)
• If multiple prepared statements were executed without retrieving their results, executing one of them again would cause the client program to crash. (Bug #8330)
• Non-numeric values inserted into a YEAR column were being stored as 2000 rather than as 0000. (Bug #6067)
• Fixed a failure of multiple-table updates to replicate properly on slave servers when --replicate-*-table options had been specified. (Bug #7011)
• mysql_stmt_close() C API function was not clearing an error indicator when a previous prepare call failed, causing subsequent invocations of error-retrieving calls to indicate spurious error values. (Bug #7990)
• Fixed failure of CREATE TABLE ... LIKE Windows when the source or destination table was located in a symlinked database directory. (Bug #6607)
• With lower_case_table_names set to 1, mysqldump on Windows could write the same table name in different lettercase for different SQL statements. Fixed so that consistent lettercase is used. (Bug #5185) HAVING that referred to RAND() or a user-defined function in the SELECT part through an alias could cause a crash or wrong value. (Bug #8216)
• If one used CONVERT_TZ() function in SELECT, which in its turn was used in CREATE TABLE statements, then system time zone tables were added to list of tables joined in SELECT and thus erroneous result was produced. (Bug #7899)
• Fixed a bug in CONV() function returning unsigned BIGINT number (third argument is positive, and return value does not fit in 32 bits). (Bug #7751)
• Fixed a failure of the IN() operator to return correct result if all values in the list were constants and some of them were using substring functions, for example, LEFT(), RIGHT(), or MID().
• Fixed problem with SHOW INDEX reporting Sub_part values in bytes rather than characters for columns with a multi-byte character set. (Bug #7943)
• Fixed a crash in CONVERT_TZ() function when its second or third argument was from a const table (Voir Section 7.2.1, « Syntaxe de EXPLAIN (Obtenir des informations sur les SELECT) »7.2.1. Syntaxe de EXPLAIN (Obtenir des informations sur les SELECT)). (Bug #7705)
• Correct a problem with mysql_config, which was failing to produce proper zlib option for linking under some circumstances. (Bug #6273)
• Fixed a problem with calculation of number of columns in row comparison against a subquery. (Bug #8020)
• Fixed erroneous output resulting from SELECT DISTINCT combined with a subquery and GROUP BY. (Bug #7946)
• Fixed server crash in comparing a nested row expression (for example row(1,(2,3))) with a subquery. (Bug #8022)
• Fixed server crash resulting from certain correlated subqueries with forward references (referring to an alias defined later in the outer query). (Bug #8025)
• Fixed server crash resulting from re-execution of prepared statements containing subqueries. (Bug #8125)
• Removed a dependence of boolean full-text search on --default-character-set option. (Bug #8159)
• Fixed a crash in a boolean full-text search in certain joins. (Bug #8234)
• Fixed erroneous comparison where strings that began with CHAR(31) were considered equal to the empty string. (Bug #8134)
• Add description of debug command to mysqladmin help output. (Bug #8207)
• perror.exe was always returning « Unknown error » on Windows. Voir Section 8.12, « perror, expliquer les codes d'erreurs »8.12. perror, expliquer les codes d'erreurs. (Bug #7390)
• Modify SET statements produced by mysqldump to write quoted strings using single quotes rather than double quotes. This avoids problems if the dump file is reloaded while the ANSI_QUOTES SQL mode is in effect. (Bug #8148)
• Fixed a bug where ALTER TABLE improperly would accept an index on a TIMESTAMP column that CREATE TABLE would reject. (Bug #7884)
• Fixed a bug in multiple-table UPDATE statements that could cause spurious Table '#sql_....' is full errors if the number of rows to update is big enough. (Bug #7788)
• Fixed a problem where SHOW INDEX on a MERGE table could crash a debugging version of the server. (Bug #7377)
• Fixed a problem where adding an ORDER BY clause for an indexed column would cause a SELECT to return an empty result. (Bug #7331)
• Fixed a problem where ALTER TABLE on a TEMPORARY table with a mixed-lettercase name could cause the table to disappear when lower_case_table_names was set to 2. (Bug #7261)
• Fixed a problem with key cache statistics being reported incorrectly by the server after receipt of a SIGHUP signal. (Bug #4285)
• Fixed a problem that caused mysql_stmt_prepare() to be very slow when used in client programs on Windows. (Bug #5787)
• For indexes, SHOW CREATE TABLE now displays the index type even if it is the default, for storage engines that support multiple index types. (Bug #7235)
• Fixed a bug where the use of GROUP_CONCAT() with HAVING caused a server crash. (Bug #7769)
• Fixed a bug where comparing the result of a subquery to a non-existent column caused a server crash on Windows. (Bug #7885)
• Fixed a bug which caused TIMEDIFF() function to return wrong results if one of its arguments had non-zero microsecond part (Bug #7586).
• Fixed a bug which caused TIMESTAMP columns with display width specified to be not identical to DATETIME columns when server was run in MAXDB mode (Bug #7418).
• Fixed a bug in UNION statements that resulted in the wrong number of the examined rows reported in the slow query log.
• Fixed a bug in a combination of -not and trunc* operators of full-text search. Using more than one truncated negative search term, was causing empty result set.
• InnoDB: Fixed a bug introduced in 4.1.9 to the Windows version if you used innodb_file_per_table. mysqld would stop and complain about Windows error number 87 in a file operation. (Voir the Bugs database or the 4.1.9 change notes about a workaround for that bug in 4.1.9). (Bug #8021)
• InnoDB: Corrected the handling of trailing spaces in the ucs2 character set. (Bug #7350)
• InnoDB: Use native tmpfile() function on Netware. All InnoDB temporary files are created under sys:\tmp. Previously, InnoDB temporary files were never deleted on Netware.
• InnoDB: Fix a race condition that could cause the assertion space->n_pending_flushes == 0 to fail in fil0fil.c, in fil_space_free(), in DROP TABLE or in ALTER TABLE.
• InnoDB: ALTER TABLE ... ADD CONSTRAINT PRIMARY KEY ... complained about bad foreign key definition. (Bug #7831)
• InnoDB: Fix a theoretical hang over the adaptive hash latch in InnoDB if one runs INSERT ... SELECT ... (binlog not enabled), or a multi-table UPDATE or DELETE, and only the read tables are InnoDB type, the rest are MyISAM. (Bug #7879)
• Fixed a bug in max_heap_table_size handling, that resulted in Table is full error when the table was still smaller than the limit. (Bug #7791).
• Fixed a symlink vulnerability in the mysqlaccess script. Reported by Javier Fernandez-Sanguino Pena and Debian Security Audit Team. (CAN-2005-0004)
• mysqlbinlog forgot to add backquotes around the collation of user variables (causing later parsing problems as BINARY is a reserved word). (Bug #7793)
• Ensured that mysqldump --single-transaction sets its transaction isolation level to REPEATABLE READ before proceeding (otherwise if the MySQL server was configured to run with a default isolation level lower than REPEATABLE READ it could give an inconsistent dump). (Bug #7850)
• Changed mysql client so that including \p as part of a prompt command uses the name of the shared memory connection when the connection is using shared memory. (Bug #7922)
• Fixed a problem in the server where executing a multi-statement query more than once with the query cache active could yield incorrect result sets. (Bug #7966)
• Fixed that a 4.1.10 slave can connect to a master < 3.23.50 without hanging (the reason for the hang is a bug in these quite old masters -- SELECT @@unknown_var hangs them -- which was fixed in MySQL 3.23.50). (Bug #7965)
• Fixed a bug where MySQL was allowing concurrent updates (inserts, deletes) to a table if binary logging is enabled. Changed to ensure that all updates are executed in a serialized fashion, because they are executed serialized when binlog is replayed. (Bug #7879)

Fonctionnalités ajoutées ou modifiées :

• mysqld_safe no longer tests for the presence of the data directory when using a relatively located server binary. It just assumes the directory is there, and fails to start up if it is not. This allows the data directory location to be specified on the command line, and avoids running a server binary that was not intended. (Bug #7249)

• The naming scheme of the Windows installation packages has changed slightly:
  Examples: mysql-essential-4.1.9-win32.msi, mysql-noinstall-4.1.9-win32.zip Voir Section 2.2.1, « Installer MySQL sous Windows »2.2.1. Installer MySQL sous Windows.

  • The platform suffix was changed from -win to -win32
  • The product descriptions -noinstall and -essential have been moved in front of the version number
• The Mac OS X 10.3 installation disk images now include a MySQL Preference Pane for the Mac OS X Control Panel that enables the user to start and stop the MySQL server via the GUI and activate and deactivate the automatic MySQL server startup on bootup.
• The MySQL-shared-compat Linux RPM now includes the 3.23 as well as the 4.0 libysqlclient.so shared libraries. (Bug #6342)
• Seconds_Behind_Master is NULL (which means « unknown ») if the slave SQL thread is not running, or if the slave I/O thread is not running or not connected to master. It is zero if the SQL thread has caught up with the I/O thread. It no longer grows indefinitely if the master is idle.
• InnoDB: Do not acquire an internal InnoDB table lock in LOCK TABLES if AUTOCOMMIT=1. This helps in porting old MyISAM applications to InnoDB. InnoDB table locks in that case caused deadlocks very easily.
• InnoDB: Print a more descriptive error and refuse to start InnoDB if the size of ibdata files is smaller than what is stored in the tablespace header; innodb_force_recovery overrides this.
• The MySQL server aborts immediately instead of simply issuing a warning if it is started with the --log-bin option but cannot initialize the binary log at startup (that is, an error occurs when writing to the binary log file or binary log index file).
• The binary log file and binary log index file now behave like MyISAM when there is a "disk full" or "quota exceeded" error. Voir Section A.4.3, « Comment MySQL gère un disque plein »A.4.3. Comment MySQL gère un disque plein.

Bogues corrigés :

• Fixed problem where running mysql_fix_privilege_tables could result in grant table columns with too-short lengths if the server character set had been set to a multi-byte character set first. (Bug #7539)
• InnoDB: Fixed the critical bug if you enabled innodb_file_per_table in my.cnf. If you shut down mysqld, records could disappear from the secondary indexes of a table. Unfortunately, on Windows a new Bug #8021 was introduced. Windows users of innodb_file_per_table should put a line innodb_flush_method=unbuffered to their my.cnf or my.ini to work around 8021. (Bug #7496)
• InnoDB: Fixed a bug : 32-bit mysqld binaries built on HP-UX-11 did not work with InnoDB files greater than 2 GB in size. (Bug #6189)
• InnoDB: Return a sensible error code from DISCARD TABLESPACE if it fails because the table is referenced by a FOREIGN KEY.
• InnoDB: Fixed a bug : InnoDB failed to drop a table in the background drop queue if the table was referenced by a FOREIGN KEY constraint.
• InnoDB: Fixed a bug : if we dropped a table where an INSERT was waiting for a lock to check a FOREIGN KEY constraint, then an assertion would fail in lock_reset_all_on_table().
• InnoDB: Fix a little bug: we looked at the physical size of a stored SQL NULL value from a wrong field in the index; this has probably caused no bugs visible to the user. It caused only some extra space to be used in some rare cases.
• InnoDB: Use the fcntl() file flush method on Mac OS X versions 10.3 and up. Apple had disabled fsync() in Mac OS X for internal disk drives, which caused corruption at power outages.
• mysqladmin password now checks whether the server has --old-passwords turned on or predates 4.1 and uses the old-format password if so. (Bug #7451)
• Added a --default-character-set option to mysqladmin to avoid problems when the default character set is not latin1. (Bug #7524)
• Fix a problem with truncation of FLOAT values. (Bug #7361)
• Fixed a bug in PROCEDURE ANALYSE(), which did not quote some ENUM values properly. (Bug #2813)
• Fixed a bug that caused incorrect results for complex datetime expressions containing casts of datetime values to TIME or DATE values. (Bug #6914)
• Include compression library flags in the output from mysql_config --lib_r. (Bug #7021)
• Corrected a problem with mysql_config not producing all relevant flags from CFLAGS. (Bug #6964)
• Corrected a problem with mysqld_safe not properly capturing output from ps. (Bug #5878)
• Fixed a bug that caused a linking failure when linking both the MySQL client library and IMAP library. (Bug #7428)
• Fixed table corruption bug when using INSERT DELAYED with prepared statements.
• Fixed a bug that caused microseconds to be gobbled from the string result of the STR_TO_DATE function, if there is some other specifier in the format string following %f. (Bug #7458)
• Made the MySQL server accept executing SHOW CREATE DATABASE even if the connection has an open transaction or locked tables. Refusing it made mysqldump --single-transaction sometimes fail to print a complete CREATE DATABASE statement for some dumped databases. (Bug #7358)
• Fixed that, when encountering a « disk full » or « quota exceeded » write error, MyISAM sometimes didn't sleep and retry the write, thus resulting in a corrupted table. (Bug #7714)
• Fixed that --expire-log-days was not honored if using only transactions. (Bug #7236)
• Fixed that a slave could crash after replicating many ANALYZE TABLE, OPTIMIZE TABLE, or REPAIR TABLE statements from the master. (Bug #6461, Bug #7658)

Fonctionnalités ajoutées ou modifiées :

Bogues corrigés :

Fonctionnalité ajoutée ou modifiée :

• InnoDB: Made LOCK TABLES behave by default like it did before MySQL 4.0.20 or 4.1.2: no InnoDB lock will be taken. Added a startup option and settable system variable innodb_table_locks for making LOCK TABLE acquire also InnoDB locks. Voir Section 15.17, « Restrictions sur les tables InnoDB »15.17. Restrictions sur les tables InnoDB. (Bug #3299, Bug #5998)

Bogues corrigés :

• Fixed a bug with FOUND_ROWS() used together with LIMIT clause in prepared statements. (Bug#6088)
• Fixed a bug with NATURAL JOIN in prepared statements. (Bug#6046).
• Fixed a bug in join of tables from different databases having columns with identical names (prepared statements). (Bug#6050)
• Now implicit access to system time zone description tables (which happens when you set time_zone variable or use CONVERT_TZ() function) does not require any privileges. (Bug #6116)
• Fixed a bug which caused server crash when deprecated libmysqlclient functilon mysql_create_db() was called. (Bug #6081)
• Fixed REVOKE ALL PRIVILEGES, GRANT OPTION FROM user so that all privileges are revoked correctly. (Bug #5831). This corrects a case that the fix in 4.1.6 could miss.
• Fixed a bug that could cause MyISAM index corruption when key values start with character codes below BLANK. This was caused by the new key sort order in 4.1. (Bug #6151)
• Fixed a bug in the prepared statements protocol when wrong metadata was sent for SELECT statements not returning a result set (such as SELECT ... INTO OUTFILE). (Bug #6059)
• Fixed bug which allowed to circumvent missing UPDATE privilege if one had INSERT and SELECT privileges for table with primary key. (Bug #6173)
• Fixed a bug in libmysqlclient with wrong conversion of negative time values to strings. (Bug #6049).
• Fixed a bug in libmysqlclient with wrong conversion of zero date values (0000-00-00) to strings. (Bug #6058)
• Fixed a bug that caused the server to crash on attempt to prepare a statement with RAND(?). (Bug #5985)
• Fixed a bug with handling of DATE, TIME, and DATETIME columns in the binary protocol. The problem is compiler-specific and could have been observed on HP-UX, AIX, Solaris9, when compiling with native compiler. (Bug #6025)
• Fixed a bug with handling of TINYINT columns in the binary protocol. The problem is specific to platforms where the C compiler has the char data type unsigned by default. (Bug #6024)
• Fixed problem introduced in MySQL 4.0.21 where a connection starting a transaction, doing updates, then FLUSH TABLES WITH READ LOCK, then COMMIT, would cause replication slaves to stop complaing about error 1223. Bug surfaced when using the InnoDB innobackup script. (Bug #5949)

Fonctionnalité ajoutée ou modifiée :

• Added option --sigint-ignore to the mysql command line client to make it ignore SIGINT signals (typically the result of the user pressing Control-C).
• InnoDB: Added the startup option and settable global variable innodb_max_purge_lag for delaying INSERT, UPDATE and DELETE operations when the purge operations are lagging. The default value of this parameter is zero, meaning that there are no delays. Voir Section 15.13, « Implémentation du multi-versionnage »15.13. Implémentation du multi-versionnage.
• InnoDB: The innodb_autoextend_increment startup option that was introduced in release 4.1.5 was made a settable global variable. (Bug #5736)
• InnoDB: If DROP TABLE is invoked on an InnoDB table for which the .ibd file is missing, print to error log that the table was removed from the InnoDB data dictionary, and allow MySQL to delete the .frm file. Maybe DROP TABLE should issue a warning in this case.
• TIMESTAMP columns now can store NULL values. To create such a column, you must explicitly specify the NULL attribute in the column specification. (Unlike all other column types, TIMESTAMP columns are NOT NULL by default.)
• Now if ALTER TABLE converts one AUTO_INCREMENT column to another AUTO_INCREMENT column it preserves zero values (this includes the case that we don't change such column at all).
• Now if ALTER TABLE converts some column to TIMESTAMP NOT NULL column it converts NULL values to current timestamp value (One can still get old behavior by setting system TIMESTAMP variable to zero).
• On Windows, the MySQL configuration files included in the package now use .ini instead of .cnf as the file name suffix.

Bogues corrigés :

• Fixed a bug that caused the server to crash on attempt to execute a prepared statement with a subselect inside a boolean expression. (Bug #5987)
• Fixed a bug that caused the server to sometimes choose non-optimal execution plan for a prepared statement executed with changed placeholder values. (Bug #6042)
• InnoDB: Make the check for excessive semaphore waits tolerate glitches in the system clock (do not crash the server if the system time is adjusted while InnoDB is under load.). (Bug #5898)
• InnoDB: Fixed a bug in the InnoDB FOREIGN KEY parser that prevented ALTER TABLE of tables containing '#' in their names. (Bug #5856)
• InnoDB: Fixed a bug that prevented ALTER TABLE t DISCARD TABLESPACE from working. (Bug #5851)
• InnoDB: SHOW CREATE TABLE now obeys the SET SQL_MODE=ANSI and SET SQL_QUOTE_SHOW_CREATE=0 settings. (Bug #5292)
• InnoDB: Fixed a bug that caused CREATE TEMPORARY TABLE ... ENGINE=InnoDB to terminate mysqld when running in innodb_file_per_table mode. Per-table tablespaces for temporary tables from now on are created in the temporary directory of mysqld. (Bug #5137)
• InnoDB: Fixed some (not all) UTF-8 bugs in column prefix indexes. (Bug #5975)
• InnoDB: If one updated a column so that its size changed, or updated it to an externally stored (TEXT or BLOB) value, then ANOTHER externally stored column would show up as 512 bytes of good data + 20 bytes of garbage in a consistent read that fetched the old version of the row. (Bug #5960)
• InnoDB: Change error code to HA_ERR_ROW_IS_REFERENCED if we cannot DROP a parent table referenced by a FOREIGN KEY constraint; this error number is less misleading than the previous number HA_ERR_CANNOT_ADD_FOREIGN, but misleading still. (Bug #6202)
• Fixed REVOKE ALL PRIVILEGES, GRANT OPTION FROM user so that all privileges are revoked correctly. (Bug #5831)
• Fixed a bug that caused the server to crash when character set conversion was implicitly used in prepared mode; for example, as in 'abc' LIKE CONVERT('abc' as utf8). (Bug #5688)
• The mysql_change_user() C API function now frees all prepared statements associated with the connection. (Bug #5315)
• Fixed a bug when inserting NULL into an AUTO_INCREMENT column failed, when using prepared statements. (Bug #5510)
• Fixed slave SQL thread so that the SET COLLATION_SERVER... statements it replicates don't advance its position (so that if it gets interrupted before the actual update query, it later redoes the SET). (Bug #5705)
• Fixed that if the slave SQL thread found a syntax error in a query (which should be rare, as the master parsed it successfully), it stops. (Bug #5711)
• Fixed that if a write to a MyISAM table fails because of a full disk or an exceeded disk quota, it prints a message to the error log every 10 minutes, and waits until disk space becomes available. (Bug #3248)
• Now MySQL does not prefer columns, which are mentioned in select list but are renamed, over columns from other tables participating in FROM clause when it resolves GROUP BY clause (e.g. SELECT t1.a AS c FROM t1, t2 ORDER BY a produces an error if both t1 and t2 tables contain a column). (Bug #4302)
• Behavior of ALTER TABLE converting column containing NULL values to AUTO_INCREMENT column is no longer affected by NO_AUTO_VALUE_ON_ZERO mode. (Bug #5915).

Fonctionnalité ajoutée ou modifiée :

• InnoDB: Added configuration option innodb_autoextend_increment for setting the size in megabytes by which InnoDB tablespaces are extended when they become full. The default value is 8, corresponding to the fixed increment of 8MB in previous versions of MySQL.

Bogues corrigés :

• Fixed a bug which caused the server to crash on attempt to execute a prepared statement with BETWEEN ? AND ? and a datetime column. (Bug #5748)
• Fixed name resolving of external fields of subqueries if subquery placed in select list of query with grouping. (Bug #5326)
• Fixed detection of using same table for updating and selecting in multi-update queries. (Bug #5455)
• The values of the max_sort_length, sql_mode, and group_concat_max_len system variables now are stored in the query cache with other query information to avoid returning an incorrect result from the query cache. (Bug #5394) (Bug #5515)
• Fixed syntax analyzer with sql_mode=IGNORE_SPACE. It happened to take phrases like default .07 as identifier.identifier. (Bug #5318)
• Fixed illegal internal field length of user variables of integer type. This showed up when creating a table as SELECT @var_name. (Bug #4788)
• Fixed a buffer overflow in prepared statements API (libmysqlclient) when a statement containing thousands of placeholders was executed. (Bug #5194)
• Fixed a bug in the server when after reaching a certain limit of prepared statements per connection (97), statement ids began to overlap, so occasionally wrong statements were chosen for execution. (Bug #5399)
• Fixed a bug in prepared statements when LIKE used with arguments in different character sets crashed server on first execute. (Bug #4368)
• Fixed a bug in prepared statements when providing '0000-00-00' date to a parameter lead to server crash. (Bug #4231, Bug #4562)
• Fixed a bug in OPTIMIZE TABLE that could cause table corruption on FULLTEXT indexes. (Bug #5327)
• InnoDB: Fixed a bug that InnoDB only allowed a maximum of 1000 connections inside InnoDB at the same time. A higher number could cause an assertion failure in sync0arr.c, line 384. Now we allow 1000, 10000, or 50000, depending on the buffer pool size. (Bug #5414)

Note: To fix a compile problem on systems that do not have automake 1.7 installed, an updated 4.1.4a source tarball has been published. In addition to resolving this automake dependency (Bug #5319), it also fixes some reported libedit compile errors when using a non-gcc compiler (Bug #5353).

Fonctionnalité ajoutée ou modifiée :

• Made internal representation of TIMESTAMP values in InnoDB in 4.1 to be the same as in 4.0. This difference resulted in incorrect datetime values in TIMESTAMP columns in InnoDB tables after an upgrade from 4.0 to 4.1. (Bug #4492) Warning: extra steps during upgrade required! Unfortunately this means that if you are upgrading from 4.1.x, where x <= 3, to 4.1.4 you should use mysqldump for saving and then restoring your InnoDB tables with TIMESTAMP columns.
• The mysqld-opt Windows server was renamed to mysqld. This completes the Windows server renaming begun in MySQL 4.1.2. Voir Section 2.2.8.1, « Choisir un serveur MySQL sur Windows »2.2.8.1. Choisir un serveur MySQL sur Windows.
• Added Latin language collations for the ucs2 and utf8 Unicode character sets. These are called ucs2_roman_ci and utf8_roman_ci.
• Corrected the name of the Mac OS X StartupItem script (it must match the name of the subdirectory, which was renamed to MySQLCOM in MySQL 4.1.2). Thanks to Bryan McCormack for reporting this.
• Added --start-datetime, --stop-datetime, --start-position, and --stop-position options to mysqlbinlog. These make point-in-time recovery easier.
• Killing a CHECK TABLE statement does not result in the table being marked as « corrupted » any more; the table remains as if CHECK TABLE had not even started. Voir Section 13.5.4.3, « Syntaxe de KILL »13.5.4.3. Syntaxe de KILL.
• Made the MySQL server ignore SIGHUP and SIGQUIT on Mac OS X 10.3. This is needed because under this OS, the MySQL server receives lots of these signals (reported as Bug #2030).

Bogues corrigés :

• Fixed a bug that caused libmysql to crash when attempting to fetch a value of MEDIUMINT column. (Bug #5126)
• Fixed a bug that caused the MySQL server to crash when attempting to execute a prepared statement with SELECT ... INTO @var for a second time. (Bug #5034)
• Fixed execution of optimized IN subqueries that use compound indexes. (Bug #4435)
• Prohibited resolving of table fields in inner queries if fields do not take part in grouping for queries with grouping (inside aggregate function arguments, all table fields are still allowed). (Bug #4814)
• Fixed a crash after SLAVE STOP if the IO thread was in a special state. (Bug #4629)
• Fixed an old bug in concurrent accesses to MERGE tables (even one MERGE table and MyISAM tables), that could have resulted in a crash or hang of the server. (Bug #2408)
• Fixed a bug that caused server crash on attempt to execute for a second time a prepared statement with NOT in WHERE or ON clauses. (Bug #4912)
• MATCH ... AGAINST now works in a subquery. (Bug #4769)
• Fixed a bug that omitted the .err extension of the error log file (--log-error) when the hostname contained a domain name. The domain name is now replaced by the extension. (Bug #4997)
• Fixed a crash in myisamchk. (Bug #4901)
• Fixed a bug which caused server crash if one used the CONVERT_TZ() function with time zone described in database as parameter and this time zone was not used before. (Bug #4508)
• Support for %T, %r, %V, %v and %X, %x format specifiers was added to STR_TO_DATE() function. (Bug #4756)
• Fixed a bug (hang) in NATURAL JOIN where joined table had no common column. (Bug #4807)
• Fixed a crash caused by UNHEX(NULL). (Bug #4441)
• mysql_fix_privilege_tables didn't correctly handle the argument of its --password=# option. (Bug #4240, Bug #4543)
• Fixed that mysqlbinlog --read-from-remote-server sometimes couldn't accept 2 binary logs on command line. (Bug #4507)
• Fixed that mysqlbinlog --position --read-from-remote-server had wrong # at lines. (Bug #4506)
• If CREATE TEMPORARY TABLE t SELECT failed while loading the data, the temporary table was not dropped. (Bug #4551)
• Fixed that when a multiple-table DROP TABLE failed to drop a table on the master server, the error code was not written to the binary log. (Bug #4553)
• When the slave SQL thread was replicating a LOAD DATA INFILE statement, it didn't show the statement in the output of SHOW PROCESSLIST. (Bug #4326)
• Fixed an assertion failure when reading the grant tables (Bug #4407)
• Fixed that CREATE TABLE ... TYPE=HEAP ... AS SELECT... caused replication slave to stop. (Bug #4971)
• Fixed that mysql_options(...,MYSQL_OPT_LOCAL_INFILE,...) failed to disable LOAD DATA LOCAL INFILE. (Bug #5038)
• Fixed that disable-local-infile option had no effect if client read it from a configuration file using mysql_options(...,MYSQL_READ_DEFAULT,...). (Bug #5073)
• Fixed that SET GLOBAL SYNC_BINLOG did not work on some platforms (Mac OS X). (Bug #5064)
• Fixed that mysql-test-run failed on the rpl_trunc_binlog test if running test from the installed (the target of 'make install') directory. (Bug #5050)
• Fixed that mysql-test-run failed on the grant_cache test when run as Unix user 'root'. (Bug #4678)
• Fixed an unlikely deadlock which could happen when using KILL. (Bug #4810)
• Fixed a crash when one connection got KILLed while it was doing START SLAVE. (Bug #4827)
• Made FLUSH TABLES WITH READ LOCK block COMMIT if server is running with binary logging; this ensures that the binary log position is trustable when doing a full backup of tables and the binary log. (Bug #4953)
• Fixed that the counter of an auto_increment column was not reset by TRUNCATE TABLE if the table was a temporary table. (Bug #5033)
• Fixed bug which caused error to be reported when column from ORDER BY clause was present in two tables participating in SELECT even if the second instance of column in select list was renamed. (Bug #4302)

Fonctionnalité ajoutée ou modifiée :

Bogues corrigés :

• Fixed a crash of mysqld that was started with binary logging disabled, but with non-zero expire_logs_days variable. (Bug #3807)

Fonctionnalité ajoutée ou modifiée :

• Added explanation of hidden SELECT of UNION in output of EXPLAIN SELECT statement.
• mysql command-line client now supports multiple -e options. (Bug #591)
• New myisam_data_pointer_size system variable. Voir Section 5.2.3, « Variables serveur système »5.2.3. Variables serveur système.
• The --log-warnings server option now is enabled by default. Disable with --skip-log-warnings.
• The --defaults-file=file_name option now requires that the filename must exist (safety fix). (Bug #3413)
• mysqld_multi now creates the log in datadir (from [mysqld] section in my.cnf or compiled in), not in /tmp - vulnerability id CAN-2004-0388. Thanks to Christian Hammers from Debian Security Team for reporting this!
• Warning: Incompatible change! String comparison now works according to the SQL standard. Because we have that 'a' = 'a ' then from it must follow that 'a' > 'a\t'. (The latter was not the case before 4.1.2.) To implement it, we had to change how storage engines compare strings internally. As a side effect, if you have a table where a CHAR or VARCHAR column in some row has a value with the last character less than ASCII(32), you will have to repair this table. CHECK TABLES will tell you if this problem exists. (Bug #3152)
• Added support for DEFAULT CURRENT_TIMESTAMP and for ON UPDATE CURRENT_TIMESTAMP specifications for TIMESTAMP columns. Now you can explicitly say that a TIMESTAMP column should be set automatically to the current timestamp for INSERT and/or UPDATE statements, or even prevent the column from updating automatically. Only one column with such an auto-set feature per table is supported. TIMESTAMP columns created with earlier versions of MySQL behave as before. Behavior of TIMESTAMP columns that were created without explicit specification of default/on as earlier depends on its position in table: If it is the first TIMESTAMP column, it will be treated as having been specified as TIMESTAMP DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP. In other cases, it would be treated as a TIMESTAMP DEFAULT 0 column. NOW is supported as an alias for CURRENT_TIMESTAMP. Warning: Incompatible change! Unlike in previous versions, explicit specification of default values for TIMESTAMP column is never ignored and turns off the auto-set feature (unless you have CURRENT_TIMESTAMP as the default).

• Warning: Incompatible change! Renamed prepared statements C API functions:

  Old Name	New Name
  mysql_bind_param()	mysql_stmt_bind_param()
  mysql_bind_result()	mysql_stmt_bind_result()
  mysql_prepare()	mysql_stmt_prepare()
  mysql_execute()	mysql_stmt_execute()
  mysql_fetch()	mysql_stmt_fetch()
  mysql_fetch_column()	mysql_stmt_fetch_column()
  mysql_param_count()	mysql_stmt_param_count()
  mysql_param_result()	mysql_stmt_param_metadata()
  mysql_get_metadata()	mysql_stmt_result_metadata()
  mysql_send_long_data()	mysql_stmt_send_long_data()

  Now all functions that operate with a MYSQL_STMT structure begin with the prefix mysql_stmt_.

• Warning: Incompatible change! The signature of the mysql_stmt_prepare() function was changed to int mysql_stmt_prepare(MYSQL_STMT *stmt, const char *query, unsigned long length). To create a MYSQL_STMT handle, you should use the mysql_stmt_init() function.

• SHOW GRANTS with no FOR clause or with FOR CURRENT_USER() shows the privileges for the current session.

• The improved character set support introduced in MySQL 4.1.0 for the MyISAM and HEAP storage engines is now available for InnoDB as well.

• A name of « Primary » no longer can be specified as an index name. (That name is reserved for the PRIMARY KEY if the table has one.) (Bug #856)

• MySQL now issues a warning when a SET or ENUM column with duplicate values in the list is created. (Bug #1427)

• Now SQL_SELECT_LIMIT variable has no influence on subqueries. (Bug #2600)

• UNHEX() function implemented. Voir Section 12.3, « Fonctions de chaînes de caractères »12.3. Fonctions de chaînes de caractères.

• History in command line client does not store multiple copies of identical queries that are run consecutively.

• Multi-line queries in the command line client now are stored as a single line.

• UUID() function implemented. Note that it does not work with replication yet. Voir Section 12.8.4, « Fonctions diverses »12.8.4. Fonctions diverses.

• Prepared statements with all types of subqueries fixed.

• MySQL now supports up to 64 keys per table.

• MyISAM tables now support keys up to 1000 bytes long.

• MyISAM and InnoDB tables now support index prefix lengths up to 1000 bytes long.

• If you try to create a key with a key part that is too long, and it is safe to auto-truncate it to a smaller length, MySQL now does so. A warning is generated, rather than an error.

• The ft_boolean_syntax variable now can be changed while the server is running. Voir Section 5.2.3, « Variables serveur système »5.2.3. Variables serveur système.

• REVOKE ALL PRIVILEGES, GRANT FROM user_list is changed to a more consistent REVOKE ALL PRIVILEGES, GRANT OPTION FROM user_list. (Bug #2642)

• Internal string-to-number conversion now supports only SQL:2003 compatible syntax for numbers. In particular, '0x10'+0 will not work anymore. (Actually, it worked only on some systems before, such as Linux. It did not work on others, such as FreeBSD or Solaris. Making these queries OS-independent was the goal of this change). Use CONV() to convert hexadecimal numbers to decimal. E.g. CONV(MID('0x10',3),16,10)+0.

• mysqlhotcopy now works on NetWare.

• ALTER TABLE DROP PRIMARY KEY no longer drops the first UNIQUE index if there is no primary index. (Bug #2361)

• Added latin1_spanish_ci (Modern Spanish) collation for the latin1 character set.

• Added the ENGINE table option as a synonym for the TYPE option for CREATE TABLE and ALTER TABLE.

• Added the --default-storage-engine server option as a synonym for --default-table-type.

• Added the storage_engine system variable as a synonym for table_type.

• Added init_connect and init_slave server variables. The values should be SQL statements to be executed when each client connects or each time a slave's SQL thread starts, respectively.

• C API enhancement: SERVER_QUERY_NO_INDEX_USED and SERVER_QUERY_NO_GOOD_INDEX_USED flags are now set in the server_status field of the MYSQL structure. It is these flags that make the query to be logged as slow if mysqld was started with --log-slow-queries --log-queries-not-using-indexes.

• For replication of MEMORY (HEAP) tables: Made the master automatically write a DELETE FROM statement to its binary log when a MEMORY table is opened for the first time since master's startup. This is for the case where the slave has replicated a non-empty MEMORY table, then the master is shut down and restarted: the table is now empty on master; the DELETE FROM empties it on slave too. Note that even with this fix, between the master's restart and the first use of the table on master, the slave still has out-of-date data in the table. But if you use the init-file option to populate the MEMORY table on the master at startup, it ensures that the failing time interval is zero. (Bug #2477)

• When a session having open temporary tables terminates, the statement automatically written to the binary log is now DROP TEMPORARY TABLE IF EXISTS instead of DROP TEMPORARY TABLE, for more robustness.

• The MySQL server now returns an error if SET SQL_LOG_BIN or SET SQL_LOG_UPDATE is issued by a user without the SUPER privilege (in previous versions it just silently ignored the statement in this case).

• Changed that when the MySQL server has binary logging disabled (that is, no log-bin option was used) then no transaction binlog cache is allocated for connections (this should save binlog_cache_size bytes of memory (32 kilobytes by default) for every connection).

• Added Binlog_cache_use and Binlog_cache_disk_use status variables that count the number of transactions that used transaction binary log and that had to flush this temporary binary log to disk instead of using only buffer in memory. They can be used for tuning the binlog_cache_size system variable.

• Added option --replicate-same-server-id.

• The Mac OS X Startup Item has been moved from the directory /Library/StartupItems/MySQL to /Library/StartupItems/MySQLCOM to avoid a file name collision with the MySQL Startup Item installed with Mac OS X Server. Voir Section 2.8.2, « Notes relatives à Mac OS X »2.8.2. Notes relatives à Mac OS X.

Bogues corrigés :

• Fixed check of EXPLAIN of UNION. (Bug #3639)
• Fixed a bug in a query that used DISTINCT and ORDER BY by column's real name, while the column had an alias, specified in SELECT clause. (Bug #3681)
• Fixed crash of group_concat on expression with ORDER BY and external ORDER BY in a query. (Bug #3752)
• Fixed a bug in ALL/SOME subqueries in case of optimisation (key field present in subquery). (Bug #3646)
• Fixed a bug in SHOW GRANTS and EXPLAIN SELECT character set conversion. (Bug #3403)
• Prepare statements parameter do not cause error message as fields used in select list but not included in ORDER BY list.
• UNION statements did not consult SQL_SELECT_LIMIT value when set. This is now fixed properly, which means that this limit is applied to the top level query, unless LIMIT for entire UNION is used.
• Fixed a bug in multiple-table UPDATE statements that resulted in an error when one of the tables was not updated but was used in the nested query, contained therein.
• Fixed mysql_stmt_send_long_data() behavior on second execution of prepared statement and in case when long data had zero length. (Bug #1664)
• Fixed crash on second execution of prepared statement with UNION. (Bug #3577)
• Fixed incorrect results of aggregate functions in subquery with empty result set. (Bug #3505)
• You can now call mysql_stmt_attr_set(..., STMT_ATTR_UPDATE_MAX_LENGTH) to tell the client library to update MYSQL_FIELD->max_length when doing mysql_stmt_store_result(). (Bug #1647).
• Added support for unsigned integer types to prepared statement API (Bug #3035).
• Fixed crash in prepared statements when subquery in the FROM clause with parameter used. (Bug #3020)
• Fixed unknown error when negative value bind to unsigned. (Bug #3223)
• Fixed aggregate function in prepared statements. (Bug #3360)
• Incorrect error message when wrong table used in multiple-table DELETE statement in prepared statements. (Bug #3411)
• Requiring UPDATE privilege for tables which will not be updated in multiple-table UPDATE statement in prepared statements.
• Fixed prepared statement support for INSERT, REPLACE, CREATE, DELETE, SELECT, DO, SET and SHOW. All other commands are prohibited via prepared statement interface. (Bug #3398, Bug #3406, Bug #2811)
• Fixed a lot of bugs in GROUP_CONCAT(). (Bug #2695, Bug #3381, Bug #3319)
• Added optimization that allows for prepared statements using a large number of tables or tables with a large number of columns to be re-executed significantly faster. (Bug #2050)
• Fixed bug that caused execution of prepared statements to fail then table that this statement were using left table cache. This bug showed up as if this prepared statement used random garbage as column names or as server crashes. (Bug #3307)
• Fixed a problem resulting from setting the character_set_results variable to NULL. (Bug #3296)
• Fixed query cache statistics.
• Fixed bug in ANALYZE TABLE on a BDB table inside a transaction that hangs server thread. (Bug #2342)
• Fixed a symlink vulnerability in mysqlbug script. (Bug #3284)
• Fixed a bug in parallel repair (myisamchk -p, myisam_repair_threads); sometimes the repair process failed to repair a table. (Bug #1334)
• A query that uses both UNION [DISTINCT] and UNION ALL now works correctly. (Bug #1428)
• Table default character set affects LONGBLOB columns. (Bug #2821)
• CONCAT_WS() makes the server die in case of illegal mix of collations. (Bug #3087)
• UTF8 charset breaks joins with mixed column/string constant. (Bug #2959)
• Fixed DROP DATABASE to report number of tables deleted.
• Fixed memory leak in the client library when statement handle was freed on closed connection (call to mysql_stmt_close after mysql_close). (Bug #3073)
• Fixed server segfaults when processing malformed prepared statements commands. (Bug #2795, Bug #2274)
• Fixed using subqueries with OR and AND functions. (Bug #2838)
• Fixed comparison of tables/database names with --lower_case_table_names option. (Bug #2880)
• Removed try to check NULL if index built on column where NULL is impossible in IN subquery optimization. (Bug #2393)
• Fixed incorrect parsing of subqueries in the FROM clause. (Bug #2421)
• Fixed processing of RAND() in subqueries with static tables. (bug #2645)
• Fixed bug with quoting of table names in mysqldump for various values of sql_mode of server. (Bug #2591)
• Fixed bug with storing values that are out of range for DOUBLE and FLOAT columns. (Bug #2082)
• Fixed bug with compiling --with-pstack with binutils 2.13.90. (Bug #1661)
• Fixed a bug in the GRANT system. When a password was assigned to an account at the global level and then privileges were granted at the database level (without specifying any password), the existing password was replaced temporarily in memory until the next FLUSH PRIVILEGES operation or the server was restarted. (Bug #2953)
• Fixed a bug in full-text search on multi-byte character set (such as UTF8) that appeared when a search word was shorter than a matching word from the index (for example, searching for « Uppsala » when table data contain « Uppsa*la »). (Bug #3011)
• Fixed a bug that made Max_used_connections to be less than the actual maximum number of connections in use simultaneously.
• Fixed calculation of Index_length in HEAP table status for BTREE indexes. (Bug #2719)
• Fixed mysql_stmt_affected_rows() call to always return number of rows affected by given statement. (Bug #2247)
• Fixed crash in MATCH ... AGAINST() on a phrase search operator with a missing closing double quote. (Bug #2708)
• Fixed output of mysqldump --tab. (Bug #2705)
• Fix for a bug in UNION operations that prevented proper handling of NULL columns. This happened only if a column in the first SELECT node was NOT NULL. (Bug #2508)
• Fix for a bug in UNION operations with InnoDB storage engine, when some columns from one table where used in one SELECT statement and some were used in another SELECT statement. (Bug #2552)
• Fixed a few years old bug in the range optimizer that caused a segmentation fault on some very rare queries. (Bug #2698)
• Fixed bug with SHOW CREATE TABLE ... which didn't properly double quotes. (Bug #2593)
• Queries with subqueries in FROM clause locks all tables at once for now. This also fixed bugs in EXPLAIN of subqueries in FROM output. (Bug #2120)
• Fixed bug with mysqldump not quoting « tricky » names correctly. (Bug #2592)
• Fix for a bug that prevented table / column privileges from being loaded on startup. (Bug #2546)
• Fixed bug in replication with CREATE TABLE ... LIKE ... that resulted in a statement not being written to the binary log. (Bug #2557)
• Fixed memory leak in INSERT ... ON DUPLICATE KEY UPDATE .... (Bug #2438)
• Fixed bug in the parser, making the syntax CONVERT(expr,type) legal again.
• Fixed parsing of short-form IP addresses in INET_ATON(). (Bug #2310)
• Fixed a bug in CREATE ... SELECT that sometimes caused a string column with a multi-byte character set (such as utf8) to have insufficient length to hold the data.
• Fixed a rare table corruption on adding data (INSERT, REPLACE, UPDATE, etc. but not DELETE) to a FULLTEXT index. (Bug #2417)
• Compile the MySQL-client RPM package against libreadline instead of libedit. (Bug #2289)
• Fix for a crashing bug that was caused by not setting vio_timeout() virtual function for all protocols. This bug occurred on Windows. (Bug #2025)
• Fix for a bug that caused mysql client program to erroneously cache the value of the current database. (Bug #2025)
• Fix for a bug that caused client/server communication to be broken when mysql_set_server_option() or mysql_get_server_option() were invoked. (Bug #2207)
• Fix for a bug that caused wong results when CAST() was applied on NULL to signed or unsigned integer column. (Bug #2219)
• Fix for a crashing bug that occurred in the mysql client program when database name was longer then expected. (Bug #2221)
• Fixed a bug in CHECK TABLE that sometimes resulted in a spurious error Found key at page ... that points to record outside datafile for a table with a FULLTEXT index. (Bug #2190)
• Fixed bug in GRANT with table-level privilege handling. (Bug #2178)
• Fixed bug in ORDER BY on a small column. (Bug #2147)
• Fixed a bug with the INTERVAL() function when 8 or more comparison arguments are provided. (Bug #1561)
• Packaging: Fixed a bug in the Mac OS PKG postinstall script (mysql_install_db was called with an obsolete argument).
• Packaging: Added missing file mysql_create_system_tables to the server RPM package. This bug was fixed for the 4.1.1 RPMs by updating the MySQL-server RPM from MySQL-server-4.1.1-0 to MySQL-server-4.1.1-1. The other RPMs were not affected by this change.
• Fixed a bug in myisamchk and CHECK TABLE that sometimes resulted in a spurious error Found key at page ... that points to record outside datafile for a table with a FULLTEXT index. (Bug #1977)
• Fixed a hang in full-text indexing of strings in multi-byte (all besides utf8) charsets. (Bug #2065)
• Fixed a crash in full-text indexing of UTF8 data. (Bug #2033)
• Replication: a rare race condition in the slave SQL thread that could lead to an incorrect complaint that the relay log is corrupted. (Bug #2011)
• Replication: If a client connects to a slave server and issues an administrative statement for a table (for example, OPTIMIZE TABLE or REPAIR TABLE), this could sometimes stop the slave SQL thread. This does not lead to any corruption, but you must use START SLAVE to get replication going again. (Bug #1858)
• Replication: in the slave SQL thread, a multiple-table UPDATE could produce an incorrect complaint that some record was not found in one table, if the UPDATE was preceded by a INSERT ... SELECT. (Bug #1701)
• Replication: sometimes the master gets a non-fatal error during the execution of a statement but finally the statements succeeds (for example, a write to a MyISAM table first receives "no space left on device" but is able to finally complete, Voir Section A.4.3, « Comment MySQL gère un disque plein »A.4.3. Comment MySQL gère un disque plein); the bug was that the master forgot to reset the error code to 0 after success, so the error code got into its binary log, thus making the slave giving false alarms like "did not get the same error as on master". (Bug #2083)
• Removed a misleading "check permissions on master.info" from a replication error message, because the cause of the problem could be different from permissions. (Bug #2121)
• Fixed a crash when the replication slave was unable to create the first relay log. (Bug #2145)
• ALTER DATABASE caused the client to hang if the database did not exist. (Bug #2333)
• Multiple-table DELETE statements were never replicated by the slave if there were any replicate-*-table options. (Bug #2527)
• Fixed bug in ALTER TABLE RENAME, when rename to the table with the same name in another database silently dropped destination table if it existed. (Bug #2628)
• The MySQL server did not report any error if the query (submitted through mysql_real_query() or mysql_prepare()) was terminated by garbage characters (which can happen if you pass a wrong length parameter to mysql_real_query() or mysql_prepare()); the result was that the garbage characters were written into the binary log. (Bug #2703)
• Fixed bug in client library which caused mysql_fetch and mysql_stmt_store_result() to hang if they were called without prior call of mysql_execute(). Now they give an error instead. (Bug #2248)
• Made clearer the error message which one gets when an update is refused because of the read-only option. (Bug #2757)
• Fixed that replicate-wild-*-table rules apply to ALTER DATABASE when the table pattern is '%', like it is already the case for CREATE DATABASE and DROP DATABASE. (Bug #3000)
• Fixed that when a Rotate event is found by the slave SQL thread in the middle of a transaction, the value of Relay_Log_Pos in SHOW SLAVE STATUS remains correct. (Bug #3017)
• Corrected the master's binary log position that InnoDB reports when it is doing a crash recovery on a slave server. (Bug #3015)
• Changed the column Seconds_Behind_Master in SHOW SLAVE STATUS to never show a value of -1. (Bug #2826)
• Changed that when a DROP TEMPORARY TABLE statement is automatically written to the binary log when a session ends, the statement is recorded with an error code of value zero (this ensures that killing a SELECT on the master does not result in a superfluous error on the slave). (Bug #3063)
• Changed that when a thread handling INSERT DELAYED (also known as a delayed_insert thread) is killed, its statements are recorded with an error code of value zero (killing such a thread does not endanger replication, so we thus avoid a superfluous error on the slave). (Bug #3081)
• Fixed deadlock when two START SLAVE commands were run at the same time. (Bug #2921)
• Fixed that a statement never triggers a superfluous error on the slave, if it must be excluded given the replicate-* options. The bug was that if the statement had been killed on the master, the slave would stop. (Bug #2983)
• The --local-load option of mysqlbinlog now requires an argument.
• Fixed a segmentation fault when running LOAD DATA FROM MASTER after RESET SLAVE. (Bug #2922)
• mysqlbinlog --read-from-remote-server read all binary logs following the one that was requested. It now stops at the end of the requested file, the same was it does when reading a local binary log. (Bug #3204)
• Fixed mysqlbinlog --read-from-remote-server to print the exact positions of events in the "at #" lines. (Bug #3214)
• Fixed a rare error condition that caused the slave SQL thread spuriously to print the message Binlog has bad magic number and stop when it was not necessary to do so. (Bug #3401)
• Fixed the Exec_master_log_pos column and its disk image in the relay-log.info file to be correct if the master had version 3.23. (The value was too big by six bytes.) This bug does not exist in MySQL 5.0. (Bug #3400)
• Fixed mysqlbinlog not to forget to print a USE statement under rare circumstances where the binary log contained a LOAD DATA INFILE statement. (Bug #3415)
• Fixed a memory corruption when replicating a LOAD DATA INFILE when the master had version 3.23. Some smaller problems remain in this setup, Voir Section 6.7, « Fonctionnalités de la réplication et problèmes connus »6.7. Fonctionnalités de la réplication et problèmes connus. (Bug #3422)
• Multiple-table DELETE statements were always replicated by the slave if there were some replicate-*-ignore-table options and no replicate-*-do-table options. (Bug #3461)
• Fixed a crash of the MySQL slave server when it was built with --with-debug and replicating itself. (BUG #3568)
• Fixed that in some replication error messages, a very long query caused the rest of the message to be invisible (truncated), by putting the query last in the message. (Bug #3357)

This release includes all fixes in MySQL 4.0.16 and most of the fixes in MySQL 4.0.17.

Fonctionnalité ajoutée ou modifiée :

• New CHECKSUM TABLE statement for reporting table checksum values.
• Added character_set_client, character_set_connection, character_set_database, character_set_results, character_set_server, character_set_system, collation_connection, collation_database, and collation_server system variables to provide information about character sets and collations.
• It is now possible to create multiple key caches, assign table indexes to particular caches, and to preload indexes into caches. Voir Section 13.5.4.1, « Syntaxe de CACHE INDEX »13.5.4.1. Syntaxe de CACHE INDEX. Voir Section 13.5.4.4, « Syntaxe de LOAD INDEX INTO CACHE »13.5.4.4. Syntaxe de LOAD INDEX INTO CACHE. Structured system variables are introduced as a means of grouping related key cache parameters. Voir Section 9.4.1, « Variables système structurées »9.4.1. Variables système structurées.
• New COERCIBILITY() function to return the collation coercibility of a string.
• The --quote-names option for mysqldump now is enabled by default.
• mysqldump now includes a statement in the dump output to set FOREIGN_KEY_CHECKS to 0 to avoid problems with tables having to be reloaded in a particular order when the dump is reloaded. The existing FOREIGN_KEY_CHECKS value is saved and restored.
• Important note: If you upgrade to InnoDB-4.1.1 or higher, you cannot downgrade to a version lower than 4.1.1 any more! That is because earlier versions of InnoDB are not aware of multiple tablespaces.
• One can revoke all privileges from a user with REVOKE ALL PRIVILEGES, GRANT FROM user_list.
• Added IGNORE option for DELETE statement.
• The MySQL source distribution now also includes the MySQL Internals Manual internals.texi.
• Added mysql_set_server_option() C API client function to allow multiple statement handling in the server to be enabled or disabled.
• The mysql_next_result() C API function now returns -1 if there are no more result sets.
• Renamed CLIENT_MULTI_QUERIES connect option flag to CLIENT_MULTI_STATEMENTS. To allow for a transition period, the old option will continue to be recognized for a while.
• Require DEFAULT before table and database default character set. This enables us to use ALTER TABLE tbl_name ... CHARACTER SET=... to change the character set for all CHAR, VARCHAR, and TEXT columns in a table.
• Added MATCH ... AGAINST( ... WITH QUERY EXPANSION) and the ft_query_expansion_limit server variable.
• Removed unused ft_max_word_len_for_sort system variable.
• Removed unused ft_max_word_len_for_sort variable from myisamchk.
• Full-text search now supports multi-byte character sets and the Unicode utf8 character set. (The Unicode ucs2 character set is not yet supported.)
• Phrase search in MATCH ... AGAINST ( ... IN BOOLEAN MODE) no longer matches partial words.
• Added aggregate function BIT_XOR() for bitwise XOR operations.
• Replication over SSL now works.
• The START SLAVE statement now supports an UNTIL clause for specifying that the slave SQL thread should be started but run only until it reaches a given position in the master's binary logs or in the slave's relay logs.
• Produce warnings even for single-row INSERT statements, not just for multiple-row INSERT statements. Previously, it was necessary to set SQL_WARNINGS=1 to generate warnings for single-row statements.
• Added delimiter (\d) command to the mysql command-line client for changing the statement delimiter (terminator). The default delimiter is semicolon.
• CHAR, VARCHAR, and TEXT columns now have lengths measured in characters rather than in bytes. The character size depends on the column's character set. This means, for example, that a CHAR(n) column for a multi-byte character set will take more storage than before. Similarly, index values on such columns are measured in characters, not bytes.
• LIMIT no longer accepts negative arguments (they used to be treated as very big positive numbers before).
• The DATABASE() function now returns NULL rather than the empty string if there is no database selected.
• Added --sql-mode=NO_AUTO_VALUE_ON_ZERO option to suppress the usual behavior of generating the next sequence number when zero is stored in an AUTO_INCREMENT column. With this mode enabled, zero is stored as zero; only storing NULL generates a sequence number.
• Warning: Incompatible change! Client authentication now is based on 41-byte passwords in the user table, not 45-byte passwords as in 4.1.0. Any 45-byte passwords created for 4.1.0 must be reset after running the mysql_fix_privilege_tables script.
• Added secure_auth global server system variable and --secure-auth server option that disallow authentication for accounts that have old (pre-4.1.1) passwords.
• Added --secure-auth option to mysql command-line client. If this option is set, the client refuses to send passwords in old (pre-4.1.1) format.
• Warning: Incompatible change! Renamed the C API mysql_prepare_result() function to mysql_get_metadata() as the old name was confusing.
• Added DROP USER 'user_name'@'host_name' statement to drop an account that has no privileges.
• The interface to aggregated UDF functions has changed a bit. You must now declare a xxx_clear() function for each aggregate function XXX().
• Added new ADDTIME(), DATE(), DATEDIFF(), LAST_DAY(), MAKEDATE(), MAKETIME(), MICROSECOND(), SUBTIME(), TIME(), TIMEDIFF(), TIMESTAMP(), UTC_DATE(), UTC_TIME(), UTC_TIMESTAMP(), and WEEKOFYEAR() functions.
• Added new syntax for ADDDATE() and SUBDATE(). The second argument now may be a number representing the number of days to be added to or subtracted from the first date argument.
• Added new type values DAY_MICROSECOND, HOUR_MICROSECOND, MINUTE_MICROSECOND, SECOND_MICROSECOND, and MICROSECOND for DATE_ADD(), DATE_SUB(), and EXTRACT().
• Added new %f microseconds format specifier for DATE_FORMAT() and TIME_FORMAT().
• All queries in which at least one SELECT does not use indexes properly now are written to the slow query log when long log format is used.
• It is now possible to create a MERGE table from MyISAM tables in different databases. Formerly, all the MyISAM tables had to be in the same database, and the MERGE table had to be created in that database as well.
• Added new COMPRESS(), UNCOMPRESS(), and UNCOMPRESSED_LENGTH() functions.
• When using SET sql_mode='mode' for a complex mode (like ANSI), we now update the sql_mode variable to include all the individual options implied by the complex mode.
• Added the OLAP (On-Line Analytical Processing) function ROLLUP, which provides summary rows for each GROUP BY level.
• Added SQLSTATE codes for all server errors.
• Added mysql_sqlstate() and mysql_stmt_sqlstate() C API client functions that return the SQLSTATE error code for the last error.
• TIME columns with hour values greater than 24 were returned incorrectly to the client.
• ANALYZE TABLE, OPTIMIZE TABLE, REPAIR TABLE, and FLUSH statements are now stored in the binary log and thus replicated to slaves. This logging does not occur if the optional NO_WRITE_TO_BINLOG keyword (or its alias LOCAL) is given. Exceptions are that FLUSH LOGS, FLUSH MASTER, FLUSH SLAVE, and FLUSH TABLES WITH READ LOCK are not logged in any case. For a syntax example, Voir Section 13.5.4.2, « Syntaxe de FLUSH »13.5.4.2. Syntaxe de FLUSH.
• New global system variable relay_log_purge to enable or disable automatic relay log purging.
• LOAD DATA now produces warnings that can be fetched with SHOW WARNINGS.
• Added support for syntax CREATE TABLE table2 (LIKE table1) that creates an empty table table2 with a definition that is exactly the same as table1, including any indexes.
• CREATE TABLE tbl_name (...) TYPE=storage_engine now generates a warning if the named storage engine is not available. The table is still created as a MyISAM table, as before.
• Most subqueries are now much faster than before.
• Added PURGE BINARY LOGS as an alias for PURGE MASTER LOGS.
• Disabled the PURGE LOGS statement that was added in version 4.1.0. The statement now should be issued as PURGE MASTER LOGS or PURGE BINARY LOGS.
• Added SHOW BDB LOGS as an alias for SHOW LOGS.
• Added SHOW MASTER LOGS (which had been deleted in version 4.1.0) as an alias for SHOW BINARY LOGS.
• Added Slave_IO_State and Seconds_Behind_Master columns to the output of SHOW SLAVE STATUS. Slave_IO_State indicates the state of the slave I/O thread, and Seconds_Behind_Master indicates the number of seconds by which the slave is late compared to the master.
• The --lower-case-table-names=1 server option now also makes aliases case insensitive. (Bug #534)
• Changed that the relay log is flushed to disk by the slave I/O thread every time it reads a relay log event. This reduces the risk of losing some part of the relay log in case of brutal crash.

Bogues corrigés :

• Fixed mysql parser not to erroneously interpret ';' character within /* ... */ comment as statement terminator.
• Fixed merging types and length of result set columns for UNION operations. The types and lengths now are determined taking into account values for all SELECT statements in the UNION, not just the first SELECT.
• Fixed a bug in privilege handling that caused connections from certain IP addresses to be assigned incorrect database-level privileges. A connection could be assigned the database privileges of the previous successful authentication from one of those IP addresses, even if the IP address username and database name were different. (Bug #1636)
• Error-handling functions were not called properly when an error resulted from [CREATE | REPLACE| INSERT] ... SELECT statements.
• HASH, BTREE, RTREE, ERRORS, and WARNINGS no longer are reserved words. (Bug #724)
• Fix for bug in ROLLUP when all tables were const tables. (Bug #714)
• Fixed a bug in UNION that prohibited NULL values from being inserted into result set columns where the first SELECT of the UNION retrieved NOT NULL columns. The type and max_length of the result column is now defined based on all UNION parts.
• Fixed name resolution of columns of reduced subqueries in unions. (Bug #745)
• Fixed memory overrun in subqueries in select list with WHERE clause bigger than outer query WHERE clause. (Bug #726)
• Fixed a bug that caused MyISAM tables with FULLTEXT indexes created in 4.0.x to be unreadable in 4.1.x.
• Fixed a data loss bug in REPAIR TABLE ... USE_FRM when used with tables that contained TIMESTAMP columns and were created in 4.0.x.
• Fixed reduced subquery processing in ORDER BY/GROUP BY clauses. (Bug #442)
• Fixed name resolution of outer columns of subquery in INSERT/REPLACE statements. (Bug #446)
• Fixed bug in marking columns of reduced subqueries. (Bug #679)
• Fixed a bug that made CREATE FULLTEXT INDEX syntax illegal.
• Fixed a crash when a SELECT that required a temporary table (marked by Using temporary in EXPLAIN output) was used as a derived table in EXPLAIN command. (Bug #251)
• Fixed a rare table corruption bug in DELETE from a big table with a new (created by MySQL-4.1) full-text index.
• LAST_INSERT_ID() now returns 0 if the last INSERT statement didn't insert any rows.
• Fixed missing last character in function output. (Bug #447)
• Fixed a rare replication bug when a transaction spanned two or more relay logs, and the slave was stopped while executing the part of the transaction that was in the second or later relay log. Then replication would resume at the beginning of the second or later relay log, which was incorrect. (It should resume at BEGIN, in the first relay log.) (Bug #53)
• CONNECTION_ID() now is properly replicated. (Bug #177)
• The new PASSWORD() function in 4.1 is now properly replicated. (Bug #344)
• Fixed a bug with double freed memory.
• Fixed a crashing bug in UNION operations that involved temporary tables.
• Fixed a crashing bug in DERIVED TABLES when EXPLAIN is used on a DERIVED TABLES with a join.
• Fixed a crashing bug in DELETE with ORDER BY and LIMIT caused by an uninitialized array of reference pointers.
• Fixed a bug in the USER() function caused by an error in the size of the allocated string.
• Fixed a crashing bug when attempting to create a table containing a spatial (GIS) column with a storage engine that does not support spatial types.
• Fixed a crashing bug in UNION caused by the empty select list and a non-existent column being used in some of the individual SELECT statements.
• Fixed a replication bug with a 3.23 master and a 4.0 slave: The slave lost the replicated temporary tables if FLUSH LOGS was issued on the master. (Bug #254)
• Fixed a security bug: A server compiled without SSL support still allowed connections by users who had the REQUIRE SSL option specified for their accounts.
• When an undefined user variable was used in a updating query on the master (such as INSERT INTO t VALUES(@a), where @a had never been set by this connection before), the slave could replicate the query incorrectly if a previous transaction on the master used a user variable of the same name. (Bug #1331)
• Fixed bug with prepared statements: Using the ? prepared statement parameter as the argument to certain functions or statement clauses caused a server crash when mysql_prepare() was invoked. (Bug #1500)
• Fixed bug with prepared statements: after call to mysql_prepare placeholders became allowed in all consequent statements, even if they are not prepared (Bug #1946)
• SLAVE START (which is a deprecated syntax, START SLAVE should be used instead) could crash the slave. (Bug #2516)
• Fixed bug in ALTER TABLE RENAME, when rename to the table with the same name in another database silently dropped destination table if it existed. (Bug #2628)

Fonctionnalité ajoutée ou modifiée :

• Added --compatible option to mysqldump for producing output that is compatible with other database systems or with older MySQL servers.
• The --opt option for mysqldump now is enabled by default, as are all the options implied by --opt.
• New CHARSET() and COLLATION() functions to return the character set and collation of a string.
• Allow index type to be specified explicitly for some storage engines via USING type_name syntax in index definition.
• New function IS_USED_LOCK() for determining the connection identifier of the client that holds a given advisory lock.
• New more secure client authentication based on 45-byte passwords in the user table.
• New CRC32() function to compute cyclic redundancy check value.
• On Windows, we are now using shared memory to communicate between server and client when they are running on the same machine and you are connecting to localhost.
• REPAIR TABLE of MyISAM tables now uses less temporary disk space when sorting char columns.
• DATE/DATETIME checking is now a bit stricter to support the ability to automatically distinguish between date, datetime, and time with microseconds. For example, dates of type YYYYMMDD HHMMDD are no longer supported; you must either have separators between each DATE/TIME part or not at all.
• Server side help for all MySQL functions. One can now type help week in the mysql client and get help for the week() function.
• Added new mysql_get_server_version() C API client function.
• Fixed bug in libmysqlclient that fetched column defaults.
• Fixed bug in mysql command-line client in interpreting quotes within comments. (Bug #539)
• Added record_in_range() method to MERGE tables to be able to choose the right index when there are many to choose from.
• Replication now works with RAND() and user variables @var.
• Allow one to change mode for ANSI_QUOTES on the fly.
• EXPLAIN SELECT now can be killed. Voir Section 13.5.4.3, « Syntaxe de KILL »13.5.4.3. Syntaxe de KILL.
• REPAIR TABLE now can be killed. Voir Section 13.5.4.3, « Syntaxe de KILL »13.5.4.3. Syntaxe de KILL.
• Allow empty index lists to be specified for USE INDEX, IGNORE INDEX, and FORCE INDEX.
• DROP TEMPORARY TABLE now drops only temporary tables and doesn't end transactions.
• Added support for UNION in derived tables.
• Warning: Incompatible change! TIMESTAMP is now returned as a string of type 'YYYY-MM-DD HH:MM:SS' and different timestamp lengths are not supported.
  This change was necessary for SQL standards compliance. In a future version, a further change will be made (backward compatible with this change), allowing the timestamp length to indicate the desired number of digits of fractions of a second.
• New faster client/server protocol that supports prepared statements, bound parameters, and bound result columns, binary transfer of data, warnings.
• Added database and real table name (in case of alias) to the MYSQL_FIELD structure.
• Multi-line queries: You can now issue several queries at once and then read the results in one go.
• In CREATE TABLE foo (a INT not null primary key) the PRIMARY word is now optional.
• In CREATE TABLE the attribute SERIAL is now an alias for BIGINT UNSIGNED NOT NULL AUTO_INCREMENT UNIQUE.
• SELECT ... FROM DUAL is an alias for SELECT .... (To be compatible with some other databases).
• If one creates a too long CHAR/VARCHAR it's now automatically changed to TEXT or BLOB; One will get a warning in this case.
• One can specify the different BLOB/TEXT types with the syntax BLOB(length) and TEXT(length). MySQL will automatically change it to one of the internal BLOB/TEXT types.
• CHAR BYTE is an alias for CHAR BINARY.
• VARCHARACTER is an alias for VARCHAR.
• New operators integer MOD integer and integer DIV integer.
• SERIAL DEFAULT VALUE added as an alias for AUTO_INCREMENT.
• TRUE and FALSE added as alias for 1 and 0, respectively.
• Aliases are now forced in derived tables, as per standard SQL.
• Fixed SELECT .. LIMIT 0 to return proper row count for SQL_CALC_FOUND_ROWS.
• One can specify many temporary directories to be used in a round-robin fashion with: --tmpdir=dirname1:dirname2:dirname3.
• Subqueries: SELECT * from t1 where t1.a=(SELECT t2.b FROM t2).

• Derived tables:

   

  Sélectionnez

  SELECT a.col1, b.col2
         FROM (SELECT MAX(col1) AS col1 FROM root_table) a,
         other_table b
         WHERE a.col1=b.col1;

• Character sets to be defined per column, table and database.

• Unicode (UTF8) support.

• New CONVERT(... USING ...) syntax for converting string values between character sets.

• BTREE index on MEMORY (HEAP) tables.

• Faster embedded server (new internal communication protocol).

• One can add a comment per column in CREATE TABLE.

• SHOW FULL COLUMNS FROM tbl_name shows column comments.

• ALTER DATABASE.

• Support for GIS (Geometrical data). Voir Chapitre 18, Données spatiales avec MySQLChapitre 18. Données spatiales avec MySQL.

• SHOW [COUNT(*)] WARNINGS shows warnings from the last command.

• One can specify a column type for a column in CREATE TABLE ... SELECT by defining the column in the CREATE part.

   

  Sélectionnez

  CREATE TABLE foo (a TINYINT NOT NULL) SELECT b+1 AS a FROM bar;

• expr SOUNDS LIKE expr same as SOUNDEX(expr)=SOUNDEX(expr).

• Added new VARIANCE(expr) function returns the variance of expr

• One can create a table from the existing table using CREATE [TEMPORARY] TABLE [IF NOT EXISTS] table (LIKE table). The table can be either normal or temporary.

• New options --reconnect and --skip-reconnect for the mysql client, to reconnect automatically or not if the connection is lost.

• START SLAVE (STOP SLAVE) no longer returns an error if the slave is already started (stopped); it returns a warning instead.

• SLAVE START and SLAVE STOP are no longer accepted by the query parser; use START SLAVE and STOP SLAVE instead.

Fonctionnalité ajoutée ou modifiée :

• Added --with-big-tables compilation option to configure. (Previously it was necessary to pass -DBIG_TABLES to the compiler manually in order to enable large table support.) Section 2.4.2, « Options habituelles de configure »2.4.2. Options habituelles de configure for details.

Bogues corrigés :

• Fixed a deadlock resulting from use of FLUSH TABLES WITH READ LOCK while an INSERT DELAYED statement is in progress. (Bug #7823)
• Fixed a segmentation fault in mysqlcheck that occurred when the last table checked in --auto-repair mode returned an error (such as the table being a MERGE table). (Bug #9492)
• Fixed faulty display of TIMESTAMP columns retrieved as col_name+0 while the new system variable is set to 1. (Bug #8894)
• Queries containing CURRENT_USER() incorrectly were registered in the query cache. (Bug #9796)
• Fixed problems with static variables to allow building on Fedora Core 3. (Bug #6554)
• An UPDATE that updated only some of the columns in a multiple-column index could result in a loop. (Bug #8942)
• REPAIR TABLE did not invalidate query results in the query cache that were generated from the table. (Bug #8480)
• Fixed a bug that caused concurrent inserts to be allowed into the tables in the SELECT ... UNION ... part of INSERT ... SELECT ... UNION .... This could result in the incorrect order of queries in the binary log. (Bug #9922)
• Fixed a bug that under certain circumstances could allow a privilege escalation via database wildcards in GRANT. (Bug #3924, CAN-2004-0957)
• <=> was not properly comparing NULL values in the WHERE clause of outer joins. (Bug #8711)
• InnoDB: Fixed a bug : MySQL-4.0.23 and 4.0.24 could complain that an InnoDB table created with MySQL-3.23.49 or earlier was in the new compact InnoDB table format of 5.0.3 or later, and InnoDB would refuse to use that table. (The same bug exists in 4.1.8 - 4.1.10.) There is nothing wrong with the table, it is mysqld that is in error. Workaround: wait that 4.0.25 or 4.1.11 is released before doing an upgrade, or dump the table and recreate it with any MySQL version >= 3.23.50 before upgrading to 4.0.23 or 4.0.24.

Fonctionnalité ajoutée ou modifiée :

• Security improvement: The server creates .frm, .MYD, .MYI, .MRG, .ISD, and .ISM table files only if a file with the same name does not already exist. Thanks to Stefano Di Paola <> for finding and informing us about this issue. (CAN-2005-0711)
• Security improvement: User-defined functions should have at least one symbol defined in addition to the xxx symbol that corresponds to the main xxx() function. These auxiliary symbols correspond to the xxx_init(), xxx_deinit(), xxx_reset(), xxx_clear(), and xxx_add() functions. mysqld by default no longer loads UDFs unless they have at least one auxiliary symbol defined in addition to the main symbol. The --allow-suspicious-udfs option controls whether UDFs that have only an xxx symbol can be loaded. By default, the option is off. mysqld also checks UDF filenames when it reads them from the mysql.func table and rejects those that contain directory pathname separator characters. (It already checked names as given in CREATE FUNCTION statements.) Voir Section 27.2.3.1, « Fonctions utilisateur : appeler des fonctions simples »27.2.3.1. Fonctions utilisateur : appeler des fonctions simples, Section 27.2.3.2, « Appeler des fonctions utilisateurs pour les groupements »27.2.3.2. Appeler des fonctions utilisateurs pour les groupements, and Section 27.2.3.6, « Précautions à prendre avec les fonctions utilisateur »27.2.3.6. Précautions à prendre avec les fonctions utilisateur. Thanks to Stefano Di Paola <> for finding and informing us about this issue. (CAN-2005-0709, CAN-2005-0710)
• InnoDB: Added configuration option and settable global variable innodb_autoextend_increment for setting the size in megabytes by which InnoDB tablespaces are extended when they become full. The default value is 8, corresponding to the fixed increment of 8MB in previous versions of MySQL.
• InnoDB: Do not acquire an internal InnoDB table lock in LOCK TABLES if AUTOCOMMIT=1. This helps in porting old MyISAM applications to InnoDB. InnoDB table locks in that case caused deadlocks very easily.

Bogues corrigés :

• AES_DECRYPT(col_name,key) could fail to return NULL for invalid values in col_name, if col_name was declared as NOT NULL. (Bug #8669)
• FOUND_ROWS() returned an incorrect value after a SELECT SQL_CALC_FOUND_ROWS DISTINCT statement that selected constants and included GROUP BY and LIMIT clauses. (Bug #7945)
• Queries of the form (SELECT ...) ORDER BY ... were being treated as a UNION. This improperly resulted in only distinct values being returned (because UNION by default eliminates.) (Bug #7672)
• Index cardinality was not being updated properly for TEMPORARY tables under some circumstances, such as CREATE TABLE ... SELECT followed by ANALYZE TABLE. (Bug #7519)
• Fixed a server crash caused by DELETE FROM tbl_name ... WHERE ... ORDER BY tbl_name.col_name when the ORDER BY column was qualified with the table name. (Bug #8392)
• Fixed a bug in MATCH ... AGAINST in natural language mode that could cause a server crash if the FULLTEXT index was not used in a join (EXPLAIN did not show fulltext join mode) and the search query matched no rows in the table (Bug #8522).
• Platform and architecture information in version information produced for --version option on Windows was always Win95/Win98 (i32). More accurately determine platform as Win32 or Win64 for 32-bit or 64-bit Windows, and architecture as ia32 for x86, ia64 for Itanium, and axp for Alpha. (Bug #4445)
• Fixed an optimization problem that allowed a negative number to be stored in a DOUBLE UNSIGNED column when it was assigned a value from a signed DOUBLE column. (Bug #7700)
• Fixed a failure of multiple-table updates to replicate properly on slave servers when --replicate-*-table options had been specified. (Bug #7011)
• Renamed set_bit() and clear_bit() functions in source code to avoid a conflict with functions of the same names in Linux kernel header files. (Bug #7971)
• Part of the information being used to cache access-permission lookups was not always reinitialized properly, particularly for connections from localhost on Windows. The result was connection failures that appeared to occur randomly. (Bug #5569)
• Corrected a problem with the QUOTE() function returning bad results. (Bug #8248)
• Fixed a problem where INSERT INTO ...SELECT failed when the source and target table were the same. (Bug #6034)
• Fixed a problem where RPM installation on Linux as a non-privileged user would result in incomplete installation. (Bug #7347)
• Change thread stack size used for building Linux RPM distributions to avoid warnings about stack size during server startup. (Bug #6226)
• Fixed a symlink vulnerability in the mysqlaccess script. Reported by Javier Fernandez-Sanguino Pena and Debian Security Audit Team. (CAN-2005-0004)
• Fixed support for C API function mysql_list_fields(), which was accidentally broken in 4.0.22 (Bug#6761)
• Make query_cache_wlock_invalidate system variable visible in SHOW VARIABLES output. (Bug #7594)
• Fixed a bug which caused FROM_UNIXTIME() function to return NULL for zero argument instead of the Epoch. (Bug #7515)
• Now in datetime values two digit year is interpreted as year in 20th or 21st century even with zero month and day. (Bug #7297)
• Fixed a bug in QUOTE function when used in conjunction with some other string functions. This lead to severe buffer overflow and server crashing. (Bug #7495)
• InnoDB: Work around a problem in AIX 5.1 patched with ML7 security patch: InnoDB would refuse to open its ibdata files, complaining about an operating system error 0.
• InnoDB: Fixed a memory corruption bug if one created a table with a primary key that contained at least two column prefixes. An example: CREATE TABLE t(a char(100), b tinyblob, PRIMARY KEY(a(5), b(10))).
• InnoDB: Use native tmpfile() function on Netware. All InnoDB temporary files are created under sys:\tmp. Previously, InnoDB temporary files were never deleted on Netware.
• InnoDB: Honor the --tmpdir startup option when creating temporary files. Previously, InnoDB temporary files were always created in the temporary directory of the operating system. On Netware, InnoDB will continue to ignore --tmpdir. (Bug #5822)
• InnoDB: Fix a theoretical hang over the adaptive hash latch in InnoDB if one runs INSERT ... SELECT ... (binlog not enabled), or a multi-table UPDATE or DELETE, and only the read tables are InnoDB type, the rest are MyISAM; this also fixes bug #7879Bug #7879 for InnoDB type tables. ()
• InnoDB: Fixed a bug : 32-bit mysqld binaries built on HP-UX-11 did not work with InnoDB files greater than 2 GB in size. (Bug #6189)
• InnoDB: Fixed a bug : InnoDB failed to drop a table in the background drop queue if the table was referenced by a foreign key constraint.
• InnoDB: Fixed a bug : if we dropped a table where an INSERT was waiting for a lock to check a FOREIGN KEY constraint, then an assertion would fail in lock_reset_all_on_table(), since that operation assumes no waiting locks on the table or its records.
• Fixed that, when encountering a « disk full » or « quota exceeded » write error, MyISAM sometimes didn't sleep and retry the write, thus resulting in a corrupted table. (Bug #7714)
• Fixed that a slave could crash after replicating many ANALYZE TABLE, OPTIMIZE TABLE, or REPAIR TABLE statements from the master. (Bug #6461, Bug #7658)
• Fixed a bug where MySQL was allowing concurrent updates (inserts, deletes) to a table if binary logging is enabled. Changed to ensure that all updates are executed in a serialized fashion, because they are executed serialized when binlog is replayed. (Bug #7879)
• Fixed a bug that caused the slave to stop on statements that produced an error on the master. (Bug #8412)
• Documented problem with using mysqldump in 4.0.x to dump TIMESTAMP(2) and TIMESTAMP(4) column types. (Bug #6530)

Note: Due to a libtool-related bug in the source distribution, the creation of shared libmysqlclient libraries was not possible (the resulting files were missing the .so file name extension). The file ltmain.sh was updated to fix this problem and the resulting source distribution was released as mysql-4.0.23a.tar.gz. This modification did not affect the binary packages. (Bug #7401)

Fonctionnalité ajoutée ou modifiée :

• Added --hex-blob option to mysqldump for dumping binary string columns using hexadecimal notation.
• Added mysql_hex_string() C API function that hex-encodes a string.
• InnoDB: Do not periodically write SHOW INNODB STATUS information to a temporary file unless the configuration option innodb_status_file=1 is set.
• InnoDB: Made the foreign key parser better aware of quotes. (Bug #6340)
• mysqlbinlog now prints an informative commented line (thread id, timestamp, server id, etc) before each LOAD DATA INFILE, like it does for other queries; unless --short-form is used.

Bogues corrigés :

• Corrected accounts in the mysql.user table in Windows distributions that had been created with a Host value of build rather than %. (Bug #6000)
• Prevent adding CREATE TABLE .. SELECT query to the binary log when the insertion of new records partially failed. (Bug #6682)
• Fixed bug which caused FROM_UNIXTIME() function to return wrong result if the argument was too big. (Bug #6439)
• Fixed bug which caused MySQL server to store wrong values in TIMESTAMP columns and give wrong results for UNIX_TIMESTAMP() function if it was run in time zone with leap seconds. (Bug #6387)
• InnoDB: Fixed a bug in LOAD DATA INFILE...REPLACE printing duplicate key error when executing the same load query several times. (Bug #5835)
• InnoDB: Refuse to open new-style tables created with MySQL 5.0.3 or later. (Bug #7089)
• InnoDB: Do not call rewind() when displaying SHOW INNODB STATUS information on stderr.
• InnoDB: If one used INSERT IGNORE to insert several rows at a time, and the first inserts were ignored because of a duplicate key collision, then InnoDB in a replication slave assigned AUTO_INCREMENT values 1 bigger than in the master. This broke the MySQL replication. (Bug #6287)
• InnoDB: Fix two hangs: FOREIGN KEY constraints treated table and database names as case-insensitive. RENAME TABLE t TO T would hang in an endless loop if t had a foreign key constraint defined on it. Fix also a hang over the dictionary mutex that would occur if one tried in ALTER TABLE or RENAME TABLE to create a foreign key constraint name that collided with another existing name. (Bug #3478)
• InnoDB: Treat character 0xA0 as space in InnoDB's FOREIGN KEY parser if MySQL treats it as space in the default charset. EMS MySQL Manager inserts character 0xA0 after the table name in an ALTER, which confused InnoDB's parser.
• Fixed a bug which caused a crash when only the slave I/O thread was stopped and restarted. (Bug #6148)
• If a connection had an open transaction but had done no updates to transactional tables (for example if had just done a SELECT FOR UPDATE then executed a non-transactional update, that update automatically committed the transaction (thus releasing InnoDB's row-level locks etc). (Bug #5714)
• If a connection was interrupted by a network error and did a rollback, the network error code got stored into the BEGIN and ROLLBACK binary log events; that caused superfluous slave stops. (Bug #6522)
• A sequence of BEGIN (or SET AUTOCOMMIT=0), FLUSH TABLES WITH READ LOCK, transactional update, COMMIT, FLUSH TABLES WITH READ LOCK could hang the connection forever and possibly the MySQL server itself. This happened for example when running the innobackup script several times. (Bug #6732)

Fonctionnalités ajoutées ou modifiées :

• InnoDB: Made LOCK TABLES behave by default like it did before MySQL 4.0.20 or 4.1.2: no InnoDB lock will be taken. Added a startup option and settable system variable innodb_table_locks for making LOCK TABLE acquire also InnoDB locks. Voir Section 15.17, « Restrictions sur les tables InnoDB »15.17. Restrictions sur les tables InnoDB. (Bug #3299, Bug #5998)
• The --with-openssl option for configure now accepts a path prefix as an argument. --with-openssl-includes and --with-openssl-libs still are supported, but are needed only to override the default values. (Bug #5494)
• Added new --without-man option to configure to suppress building/installing the manual pages. (Bug #5379)

Bogues corrigés :

• Fixed bug in server which caused connection stall when one of deprecated libmysqlclient functions mysql_create_db(), mysql_rm_db() were called and were going to return error. (Bug #6081)
• Fixed returning wrong query result from query cache if temporary table haded real tables after putting results to query cache. (Bug #6084)
• Fixed ENABLE KEYS, which failed if tmpdir ran out of space. Now, a full repair is done in this case. (Bug #5625)
• Fixed an improper error message when trying to drop a table which is referenced by a FOREIGN KEY constraint. (Bug #5784)
• Fixed a bug that allowed FLUSH TABLE(S) to close HANDLER tables. HANDLER tables now are re-opened after a FLUSH TABLE(S) when they are next used. However, they lose their file position if this happens. (Bug #4286)
• Fixed a bug that allowed HANDLER tables with the same alias to be multiple opened. HANDLER aliases must now be unique, even though it is syntactically correct in versions below 4.1, to qualify them with their base table's database name (e.g. test_db.handler_tbl, but this will now conflict with e.g. another_db.handler_tbl). (Bug #4335)
• Fixed crash when using MySQL 4.0 with privilege tables from MySQL 5.0.
• InnoDB: Make the check for excessive semaphore waits tolerate glitches in the system clock (do not crash the server if the system time is adjusted while InnoDB is under load.). (Bug #5898)
• mysqlimport now reads input files locally from the client host only if the --local option is given. Previously, it assumed incorrectly in some cases that files were local even without --local. (Bug #5829)
• InnoDB: Fixed a bug in the InnoDB FOREIGN KEY parser that prevented ALTER TABLE of tables containing '#' in their names. (Bug #5856)
• Fixed a bug which resulted in erronously calculated number of examined rows in UNION's. This value is printed in the slow query log. (Bug #5879)
• Fixed bug with crash of server on some values of read_rnd_buffer_size (Bug #5492)
• Fixed bug which caused truncation of values read from or into TIMESTAMP fields if --new mode was enabled. (Bug #4131)
• mysqladmin now returns a status of 0 even when the server denies access; such an error means theserver is running. (Bug #3120)
• InnoDB: Fixed a bug introduced in 4.0.21. An assertion failed if one used mysqldump with the option -l or --opt, or if one used LOCK TABLES ... LOCAL. (Workaround in 4.0.21: use --quick and --single-transaction. (Bug #5538)
• Fixed that if the slave SQL thread found a syntax error in a query (which should be rare, as the master parsed it successfully), it stops. (Bug #5711)
• Fixed that if a write to a MyISAM table fails because of a full disk or an exceeded disk quota, it prints a message to the error log every 10 minutes, and waits until disk becomes free. (Bug #3248)
• Fixed problem with symlinked databases on Windows being shown with SHOW DATABASES even if the database name doesn't match the given wildcard (Bug #5539)
• Fixed problem introduced in 4.0.21 where a connection starting a transaction, doing updates, then FLUSH TABLES WITH READ LOCK, then COMMIT, would cause replication slaves to stop complaing about error 1223. Bug surfaced when using the InnoDB innobackup script. (Bug #5949)

Fonctionnalité ajoutée ou modifiée :

Bogues corrigés :

• Fixed a bug with truncation of big values (> 4294967295) of 64-bit system variables. (Bug #3754)

Fonctionnalité ajoutée ou modifiée :

• Phrase search in MATCH ... AGAINST ( ... IN BOOLEAN MODE) no longer matches partial words.

Bogues corrigés :

• Fixed a bug in division / reporting incorrect metadata (number of digits after the decimal point). It can be Voirn, e.g. in CREATE TABLE t1 SELECT "0.01"/"3". (Bug #3612)
• Fixed a problem with non-working DROP DATABASE on some configurations (in particular, Linux 2.6.5 with ext3 are known to expose this bug). (Bug #3594)
• Fixed that in some replication error messages, a very long query caused the rest of the message to be invisible (truncated), by putting the query last in the message. (Bug #3357)

Note: The MySQL 4.0.19 binaries were uploaded to the download mirrors on May, 10th. However, a potential crashing bug was found just before the 4.0.19 release was publicly announced and published from the 4.0 download pages at http://dev.mysql.com/.

Voir (Bug #3596) for details (it was reported against MySQL-4.1, but was confirmed to affect 4.0.19 as well).

A fix for this bug was pushed into the MySQL source tree shortly after it could be reproduced and will be included in the upcoming MySQL 4.0.20, to be released shortly. We recommend users to stick to MySQL 4.0.18 for the time being, until MySQL 4.0.20 has been released (this specific bug was introduced after MySQL 4.0.18 was released, so older versions were not affected by it). We apologize for the inconvenience!

Fonctionnalité ajoutée ou modifiée :

• If length of a timestamp field is defined as 19, the timestamp will be displayed as "YYYY-MM-DD HH:MM:SS". This is done to make it easier to use tables created in MySQL 4.1 to be used in MySQL 4.0.
• If you use RAID_CHUNKS with a value > 255 it will be set to 255. This was made to ensure that all raid directories are always 2 hex bytes. (Bug #3182)
• Changed that the optimizer will now consider the index specified in FORCE INDEX clause as a candidate to resolve ORDER BY as well.
• Non-standard behavior of UNION statements has changed to the standard ones. So far, a table name in the ORDER BY clause was tolerated. From now on a proper error message is issued (Bug #3064).
• Added max_insert_delayed_threads system variable as a synonym for max_delayed_threads.
• Added query_cache_wlock_invalidate system variable. It allow emulation of MyISAM table write-locking behavior, even for queries in the query cache. (Bug #2693)
• The keyword MASTER_SERVER_ID is not reserved anymore.
• The following is mainly relevant for Mac OS X users who use a case-insensitive filesystem. This is not relevant for Windows users as InnoDB in this case always stores file names in lower case:
  One can now force lower_case_table_names to 0 from the command line or a configuration file. This is useful with case-insensitive filesystems when you have previously not used lower_case_table_names=1 or lower_case_table_names=2 and your have already created InnoDB tables. With lower_case_table_names=0, InnoDB tables were stored in mixed case while setting lower_case_table_names <> 0 will now force it to lower case (to make the table names case insensitive).
  Because it's possible to crash MyISAM tables by referring to them with different case on a case-insensitive filesystem, we recommend that you use lower_case_table_names or lower_case_table_names=2 on such filesystems.
  The easiest way to convert to use lower_case_table_names=2 is to dump all your InnoDB tables with mysqldump, drop them and then restore them.
• Non-standard behavior of UNION statements has changed to the standard ones. So far, a table name in the ORDER BY clause was tolerated. From now on a proper error message is issued (Bug #3064).
• Added max_insert_delayed_threads system variable as a synonym for max_delayed_threads.
• Added query_cache_wlock_invalidate system variable. It allow emulation of MyISAM table write-locking behavior, even for queries in the query cache. (Bug #2693)
• Changed that the relay log is flushed to disk by the slave I/O thread every time it reads a relay log event. This reduces the risk of losing some part of the relay log in case of brutal crash.
• When a session having open temporary tables terminates, the statement automatically written to the binary log is now DROP TEMPORARY TABLE IF EXISTS instead of DROP TEMPORARY TABLE, for more robustness.
• Added option --replicate-same-server-id.

Bogues corrigés :

• Added missing full-text variable ft_stopword_file to myisamchk.
• Don't allow stray ',' at the end of field specifications. (Bug #3481)
• INTERVAL now can handle big values for seconds, minutes and hours. (Bug #3498)
• Blank hostname did not work as documented for table and column privileges. Now it's works the same way as '%'. (Bug #3473)
• Fixed a harmless buffer overflow in replace utility. (Bug# 3541)
• Fixed SOUNDEX() to ignore non-alphabetic characters also in the beginning of the string. (Bug #3556)
• Fixed a bug in MATCH ... AGAINST() searches when another thread was doing concurrent inserts into the MyISAM table in question. The first --- full-text search --- query could return incorrect results in this case (e.g. « phantom » rows or not all matching rows, even an empty result set). The easiest way to check whether you are affected is to start mysqld with --skip-concurrent-insert switch and Voir if it helps.
• Fixed bug when doing DROP DATABASE on a directory containing non- MySQL files. Now a proper error message is returned.
• Fixed bug in ANALYZE TABLE on a BDB table inside a transaction that hangs server thread. (Bug #2342)
• Fixed a symlink vulnerability in mysqlbug script. (Bug #3284)
• Fixed core dump bug in SELECT DISTINCT where all selected parts where constants and there were hidden columns in the created temporary table. (Bug #3203)
• Fixed core dump bug in COUNT(DISTINCT) when there was a lot of values and one had a big value for max_heap_table_size.
• Fixed problem with multi-table-update and BDB tables. (Bug: #3098)
• Fixed memory leak when dropping database with RAID tables. (Bug #2882)
• Fixed core dump crash in replication during relay-log switch when the relay log went over max_relay_log_size and the slave thread did a flush_io_cache() at the same time.
• Fixed hangup bug when issuing multiple SLAVE START from different threads at the same time. (Bug #2921)
• Fixed bug when using DROP DATABASE with lower_case_table_names=2.
• Fixed wrong result in UNION when using lower_case_table_names=2. (Bug #2858)
• One can now kill threads that is 'stuck' in the join optimizer (can happen when there is MANY tables in the join in which case the optimizer can take really long time). (Bug #2825)
• Rollback DELETE and UPDATE statements if thread is killed. (Bug #2422)
• Ensure that all rows in an INSERT DELAYED statement is written at once if binary logging is enabled. (Bug #2491).
• Fixed bug in query cache statistic, more accurate formula linked statistic variables mentioned in the manual.
• Fixed a bug in parallel repair (myisamchk -p, myisam_repair_threads) - sometimes repair process failed to repair a table. (Bug #1334)
• Fixed bugs with names of tables, databases and columns that end to space (Bug #2985)
• Fixed a bug in multiple-table UPDATE statements involving at least one constant table. Bug was exhibited in allowing non matching row to be updated. (Bug #2996).
• Fixed all bugs in scripts for creating/upgrading system database (Bug #2874) Added tests which guarantee against such bugs in the future.
• Fixed bug in mysql command-line client in interpreting quotes within comments. (Bug #539)
• --set-character-set and --character-sets-dir options in myisamchk now work.
• Fixed a bug in mysqlbinlog that caused one pointer to be free'd twice in some cases.
• Fixed a bug in boolean full-text search, that sometimes could lead to false matches in queries with several levels of subexpressions using + operator (for example, MATCH ... AGAINST('+(+(word1 word2)) +word3*' IN BOOLEAN MODE).
• Fixed Windows-specific portability bugs in myisam_ftdump.
• Fixed a bug in multiple-table DELETE that was caused by foreign key constraints. If the order of the tables established by MySQL optimizer did not match parent-child order, no rows were deleted and no error message was provided. (Bug #2799)
• Fixed a few years old bug in the range optimizer that caused a segmentation fault on some very rare queries. (Bug #2698)
• Replication: If a client connects to a slave server and issues an administrative statement for a table (for example, OPTIMIZE TABLE or REPAIR TABLE), this could sometimes stop the slave SQL thread. This does not lead to any corruption, but you must use START SLAVE to get replication going again. (Bug #1858) The bug was accidentally not fixed in 4.0.17 as it was unfortunately earlier said.
• Fixed that when a Rotate event is found by the slave SQL thread in the middle of a transaction, the value of Relay_Log_Pos in SHOW SLAVE STATUS remains correct. (Bug #3017)
• Corrected the master's binary log position that InnoDB reports when it is doing a crash recovery on a slave server. (Bug #3015)
• Changed that when a DROP TEMPORARY TABLE statement is automatically written to the binary log when a session ends, the statement is recorded with an error code of value zero (this ensures that killing a SELECT on the master does not result in a superfluous error on the slave). (Bug #3063)
• Changed that when a thread handling INSERT DELAYED (also known as a delayed_insert thread) is killed, its statements are recorded with an error code of value zero (killing such a thread does not endanger replication, so we thus avoid a superfluous error on the slave). (Bug #3081)
• Fixed deadlock when two START SLAVE commands were run at the same time. (Bug #2921)
• Fixed that a statement never triggers a superfluous error on the slave, if it must be excluded given the replicate-* options. The bug was that if the statement had been killed on the master, the slave would stop. (Bug #2983)
• The --local-load option of mysqlbinlog now requires an argument.
• Fixed a segmentation fault when running LOAD DATA FROM MASTER after RESET SLAVE. (Bug #2922)
• Fixed a rare error condition that caused the slave SQL thread spuriously to print the message Binlog has bad magic number and stop when it was not necessary to do so. (Bug #3401)
• Fixed the column Exec_master_log_pos (and its disk image in the relay-log.info file) to be correct if the master had version 3.23 (it was too big by 6 bytes). This bug does not exist in the 5.0 version. (Bug #3400)
• Fixed that mysqlbinlog does not forget to print a USE command under rare circumstances where the binary log contained a LOAD DATA INFILE command. (Bug #3415)
• Fixed a memory corruption when replicating a LOAD DATA INFILE when the master had version 3.23. Some smaller problems remain in this setup, Voir Section 6.7, « Fonctionnalités de la réplication et problèmes connus »6.7. Fonctionnalités de la réplication et problèmes connus. (Bug #3422)
• Multiple-table DELETE statements were always replicated by the slave if there were some replicate-*-ignore-table options and no replicate-*-do-table options. (Bug #3461)
• Fixed a crash of the MySQL slave server when it was built with --with-debug and replicating itself. (BUG #3568)

Fonctionnalité ajoutée ou modifiée :

• Fixed processing of LOAD DATA by mysqlbinlog in remote mode. (Bug #1378)
• New utility program myisam_ftdump was added to binary distributions.
• ENGINE is now a synonym for the TYPE option for CREATE TABLE and ALTER TABLE.
• lower_case_table_names system variable now can take a value of 2, to store table names in mixed case on case-insensitive filesystems. It's forced to 2 if the database directory is located on a case-insensitive filesystem.
• For replication of MEMORY (HEAP) tables: Made the master automatically write a DELETE FROM statement to its binary log when a MEMORY table is opened for the first time since master's startup. This is for the case where the slave has replicated a non-empty MEMORY table, then the master is shut down and restarted: the table is now empty on master; the DELETE FROM empties it on slave too. Note that even with this fix, between the master's restart and the first use of the table on master, the slave still has out-of-date data in the table. But if you use the init-file option to populate the MEMORY table on the master at startup, it ensures that the failing time interval is zero. (Bug #2477)
• Optimizer is now better tuned for the case where the first used key part (of many) is a constant. (Bug #1679)
• Removed old non-working --old-rpl-compat server option, which was a holdover from the very first 4.0.x versions. (Bug #2428)

Bogues corrigés :

• mysqlhotcopy now works on NetWare.
• DROP DATABASE could not drop databases with RAID tables that had more than nine RAID_CHUNKS. (Bug #2627)
• Fixed bug in range optimizer when using overlapping ranges. (Bug #2448)
• Limit wait_timeout to 2147483 on Windows (OS limit). (Bug #2400)
• Fixed bug when --init-file crashes MySQL if it contains a large SELECT. (Bug #2526)
• SHOW KEYS now shows NULL in the Sub_part column for FULLTEXT indexes.
• The signal thread's stack size was increased to enable mysqld to run on Debian/IA-64 with a TLS-enabled glibc. (Bug #2599)
• Now only the SELECT privilege is needed for tables that are only read in multiple-table UPDATE statements. (Bug #2377)
• Give proper error message if one uses LOCK TABLES ... ; INSERT ... SELECT and one used the same table in the INSERT and SELECT part. (Bug #2296)
• SELECT INTO ... DUMPFILE now deletes the generated file on error.
• Fixed foreign key reference handling to allow references to column names that contain spaces. (Bug #1725)
• Fixed problem with index reads on character columns with BDB tables. The symptom was that data could be returned in the wrong lettercase. (Bug #2509)
• Fixed a spurious table corruption problem that could sometimes appear on tables with indexed TEXT columns if these columns happened to contain values having trailing spaces. This bug was introduced in 4.0.17.
• Fixed a problem where some queries could hang if a condition like indexed_TEXT_column =expr was present and the column contained values having trailing spaces. This bug was introduced in 4.0.17.
• Fixed a bug that could cause incorrect results from a query that involved range conditions on indexed TEXT columns that happened to contain values having trailing spaces. This bug was introduced in 4.0.17. (Bug #2295)
• Fixed incorrect path names in some of the manual pages. (Bug #2270)
• Fixed spurious « table corrupted » errors in parallel repair operations. Voir Section 5.2.3, « Variables serveur système »5.2.3. Variables serveur système.
• Fixed a crashing bug in parallel repair operations. Voir Section 5.2.3, « Variables serveur système »5.2.3. Variables serveur système.
• Fixed bug in updating MyISAM tables for BLOB values longer than 16MB. (Bug #2159)
• Fixed bug in mysqld_safe when running multiple instances of MySQL. (Bug #2114)
• Fixed a bug in using HANDLER statement with tables not from a current database. (Bug #2304)
• Fixed a crashing bug that occurred due to the fact that multiple-table UPDATE statements did not check that there was only one table to be updated. (Bug #2103)
• Fixed a crashing bug that occurred due to BLOB column type index size being calculated incorrectly in MIN() and MAX() optimizations. (Bug #2189)
• Fixed a bug with incorrect syntax for LOCK TABLES in mysqldump. (Bug #2242)
• Fixed a bug in mysqld_safe that caused mysqld to generate a warning about duplicate user=xxx options if this option was specified in the [mysqld] or [server] sections of my.cnf. (Bug #2163)
• INSERT DELAYED ... SELECT ... could cause table corruption because tables were not locked properly. This is now fixed by ignoring DELAYED in this context. (Bug #1983)
• Replication: Sometimes the master gets a non-fatal error during the execution of a statement that does not immediately succeed. (For example, a write to a MyISAM table may first receive « no space left on device, » but later complete when disk space becomes available. Voir Section A.4.3, « Comment MySQL gère un disque plein »A.4.3. Comment MySQL gère un disque plein.) The bug was that the master forgot to reset the error code to 0 after success, so the error code got into its binary log, thus causing the slave to issue false alarms such as « did not get the same error as on master. » (Bug #2083)
• Removed a misleading « check permissions on master.info » from a replication error message, because the cause of the problem could be something other than permissions. (Bug #2121)
• Fixed a crash when the replication slave was unable to create the first relay log. (Bug #2145)
• Replication of LOAD DATA INFILE for an empty file from a 3.23 master to a 4.0 slave caused the slave to print an error. (Bug #2452)
• When automatically forcing lower_case_table_names to 1 if the file system was case insensitive, mysqld could crash. This bug existed only in MySQL 4.0.17. (Bug #2481)
• Restored ability to specify default values for TIMESTAMP columns that was erroneously disabled in previous release. (Bug #2539) Fixed SHOW CREATE TABLE to reflect these values. (Bug #1885) Note that because of the auto-update feature for the first TIMESTAMP column in a table, it makes no sense to specify a default value for the column. Any such default will be silently ignored (unless another TIMESTAMP column is added before this one). Also fixed the meaning of the DEFAULT keyword when it is used to specify the value to be inserted into a TIMESTAMP column other than the first. (Bug #2464)
• Fixed bug for out-of-range arguments on QNX platform that caused UNIX_TIMESTAMP() to produce incorrect results or that caused non-zero values to be inserted into TIMESTAMP columns. (Bug #2523) Also, current time zone now is taken into account when checking if datetime values satisfy both range boundaries for TIMESTAMP columns. The range allowed for a TIMESTAMP column is time zone-dependent and equivalent to a range of 1970-01-01 00:00:01 UTC to 2037-12-31 23:59:59 UTC.
• Multiple-table DELETE statements were never replicated by the slave if there were any replicate-*-table options. (Bug #2527)
• Changes to session counterparts of variables query_prealloc_size, query_alloc_block_size, trans_prealloc_size, trans_alloc_block_size now have an effect. (Bug #1948)
• Fixed bug in ALTER TABLE RENAME, when rename to the table with the same name in another database silently dropped destination table if it existed. (Bug #2628)

Fonctionnalité ajoutée ou modifiée :

• mysqldump no longer dumps data for MERGE tables. (Bug #1846)
• lower_case_table_names is now forced to 1 if the database directory is located on a case-insensitive file system. (Bug #1812)
• Symlink creation is now disabled on systems where realpath() doesn't work. (Before one could use CREATE TABLE .. DATA DIRECTORY=.. even if HAVE_BROKEN_REALPATH was defined. This is now disabled to avoid problems when running ALTER TABLE).
• Inserting a negative AUTO_INCREMENT value in a MyISAM table no longer updates the AUTO_INCREMENT counter to a big unsigned value. (Bug #1366)
• Added four new modes to WEEK(..., mode) function. Voir WEEK(date: (mode)). (Bug #1178)
• Allow UNION DISTINCT syntax.
• mysql_server_init() now returns 1 if it can't initialize the environment. (Previously mysql_server_init() called exit(1) if it could not create a key with pthread_key_create(). (Bug #2062)
• Allow spaces in Windows service names.
• Changed the default Windows service name for mysqld from MySql to MySQL. This should not affect usage, because service names are not case sensitive.
• When you install mysqld as a service on Windows systems, mysqld will read startup options in option files from the option group with the same name as the service name. (Except when the service name is MySQL).

Bogues corrigés :

• One can now configure MySQL as a Windows service as a normal user. (Bug #1802). Thanks to Richard Hansen for fixing this.
• Database names are now compared in lowercase in ON clauses when lower_case_table_names is set. (Bug #1736)
• IGNORE ... LINES option to LOAD DATA INFILE didn't work when used with fixed length rows. (Bug #1704)
• Fixed problem with UNIX_TIMESTAMP() for timestamps close to 0. (Bug #1998)
• Fixed problem with character values greater than 128 in the QUOTE() function. (Bug #1868)
• Fixed searching of TEXT with end space. (Bug #1651)
• Fixed caching bug in multiple-table updates where same table was used twice. (Bug #1711)
• Fixed directory permissions for the MySQL-server RPM documentation directory. (Bug #1672)
• Fixed server crash when updating an ENUM column that is set to the empty string (for example, with REPLACE()). (Bug #2023)
• mysql client program now correctly prints connection identifier returned by mysql_thread_id() as unsigned integer rather than as signed integer. (Bug #1951)
• FOUND_ROWS() could return incorrect number of rows after a query with an impossible WHERE condition. (Bug #1468)
• SHOW DATABASES no longer shows .sym files (on Windows) that do not point to a valid directory. (Bug #1385)
• Fixed a possible memory leak on Mac OS X when using the shared libmysql.so library. (from pthread_key_create()). (Bug #2061)
• Fixed bug in UNION statement with alias *. (Bug #1249)
• Fixed a bug in DELETE ... ORDER BY ... LIMIT where the rows where not deleted in the proper order. (Bug #1024, Bug #1697).
• Fixed serious problem with multi-threaded programs on Windows that used the embedded MySQL libraries. (Locks of tables were not handled correctly between different threads).
• Code cleanup: Fixed a few code defects (potential memory leaks, null pointer dereferences, uninitialized variables). Thanks to Reasoning Inc. for informing us about these findings.
• Fixed a buffer overflow error that occurred with prepended '0' characters in some columns of type DECIMAL. (Bug #2128)
• Filesort was never shown in EXPLAIN if query contained an ORDER BY NULL clause. (Bug #1335)
• Fixed invalidation of whole query cache on DROP DATABASE. (Bug #1898)
• Fixed bug in range optimizer that caused wrong results for some unlikely AND/OR queries. (Bug #1828)
• Fixed a crash in ORDER BY when ordering by expression and identifier. (Bug #1945)
• Fixed a crash in an open HANDLER when an ALTER TABLE was executed in a different connection. (Bug #1826)
• Fixed a bug in trunc* operator of full-text search which sometimes caused MySQL not to find all matched rows.
• Fixed bug in prepending '0' characters to DECIMAL column values.
• Fixed optimizer bug, introduced in 4.0.16, when REF access plan was preferred to more efficient RANGE on another column.
• Fixed problem when installing a MySQL server as a Windows service using a command of the form mysqld --install mysql --defaults-file=path-to-file. (Bug #1643)
• Fixed an incorrect result from a query that uses only const tables (such as one-row tables) and non-constant expression (such as RAND()). (Bug #1271)
• Fixed bug when the optimizer did not take SQL_CALC_FOUND_ROWS into account if LIMIT clause was present. (Bug #1274)
• mysqlbinlog now asks for a password at the console when the -p or --password option is used with no argument. This is consistent with the way that other clients such mysqladmin and mysqldump already behave. Note: A consequence of this change is that it is no longer possible to invoke mysqlbinlog as mysqlbinlog -p pass_val (with a space between the -p option and the following password value). (Bug #1595)
• Fixed bug accidentally introduced in 4.0.16 where the slave SQL thread deleted its replicated temporary tables when STOP SLAVE was issued.
• In a « chain » replication setup A->B->C, if 2 sessions on A updated temporary tables of the same name at the same time, the binary log of B became incorrect, resulting in C becoming confused. (Bug #1686)
• In a « chain » replication setup A->B->C, if STOP SLAVE was issued on B while it was replicating a temporary table from A, then when START SLAVE was issued on B, the binary log of B became incorrect, resulting in C becoming confused. (Bug #1240)
• When MASTER_LOG_FILE and MASTER_LOG_POS were not specified, CHANGE MASTER used the coordinates of the slave I/O thread to set up replication, which broke replication if the slave SQL thread lagged behind the slave I/O thread. This caused the slave SQL thread to lose some events. The new behavior is to use the coordinates of the slave SQL thread instead. Voir Section 13.6.2.1, « CHANGE MASTER TO »13.6.2.1. CHANGE MASTER TO. (Bug #1870)
• Now if integer is stored or converted to TIMESTAMP or DATETIME value checks of year, month, day, hour, minute and second ranges are performed and numbers representing illegal timestamps are converted to 0 value. This behavior is consistent with manual and with behavior of string to TIMESTAMP/DATETIME conversion. (Bug #1448)
• Fixed bug when BIT_AND() and BIT_OR() group functions returned incorrect value if SELECT used a temporary table and no rows were found. (Bug #1790).
• BIT_AND() is now unsigned in all contexts. This means that it will now return 18446744073709551615 (= 0xffffffffffffffff) instead of -1 if there were no rows in the result.
• Fixed bug with BIT_AND() still returning signed value for an empty set in some cases. (Bug #1972)
• Fixed bug with ^ (XOR) and >> (bit shift) still returning signed value in some cases. (Bug #1993)
• Replication: a rare race condition in the slave SQL thread, which could lead to a wrong complain that the relay log is corrupted. (Bug #2011)
• Replication: if an administrative command on a table (OPTIMIZE TABLE, REPAIR TABLE etc) was run on the slave, this could sometimes stop the slave SQL thread (this did not led to any corruption; one just had to type START SLAVE to get replication going again). (Bug #1858)
• Replication: in the slave SQL thread, a multi-table UPDATE could produce a wrong complain that some record was not found in one table, if the UPDATE was preceded by a INSERT ... SELECT. (Bug #1701)
• Fixed deficiency in MySQL code which is responsible for scanning directories. This deficiency caused SHOW TABLE STATUS to be very slow for big number of tables in database even if single particular table were specified. (Bug #1952)

Fonctionnalité ajoutée ou modifiée :

• Option values in option files now may be quoted. This is useful for values that contain whitespace or comment characters.
• Write memory allocation information to error log when doing mysqladmin debug. This works only on systems that support the mallinfo() call (like newer Linux systems).
• Added the following new server variables to allow more precise memory allocation: range_alloc_block_size, query_alloc_block_size, query_prealloc_size, transaction_alloc_block_size, and transaction_prealloc_size.
• mysqlbinlog now reads option files. To make this work, you must now specify --read-from-remote-server when reading binary logs from a MySQL server. (Note that using a remote server is deprecated and may disappear in future mysqlbinlog versions).
• Block SIGPIPE signals also for non-threaded programs. The blocking is moved from mysql_init() to mysql_server_init(), which is automatically called on the first call to mysql_init().
• Added --libs_r and --include options to mysql_config.
• New `> prompt for mysql. This prompt is similar to the '> and "> prompts, but indicates that an identifier quoted with backticks was begun on an earlier line and the closing backtick has not yet been Voirn.
• Updated mysql_install_db to be able to use the local machine's IP address instead of the hostname when building the initial grant tables if skip-name-resolve has been specified. This option can be helpful on FreeBSD to avoid thread-safety problems with the FreeBSD resolver libraries. (Thanks to Jeremy Zawodny for the patch.)
• A documentation change: Added a note that when backing up a slave, it is necessary also to back up the master.info and relay-log.info files, as well as any SQL_LOAD-* files located in the directory specified by the --slave-load-tmpdir option. All these files are needed when the slave resumes replication after you restore the slave's data.

Bogues corrigés :

• Fixed a spurious error ERROR 14: Can't change size of file (Errcode: 2) on Windows in DELETE FROM tbl_name without a WHERE clause or TRUNCATE TABLE tbl_name, when tbl_name is a MyISAM table. (Bug #1397)
• Fixed a bug that resulted in thr_alarm queue is full warnings after increasing the max_connections variable with SET GLOBAL. (Bug #1435)
• Made LOCK TABLES to work when Lock_tables_priv is granted on the database level and Select_priv is granted on the table level.
• Fixed crash of FLUSH QUERY CACHE on queries that use same table several times (Bug #988).
• Fixed core dump bug when setting an enum system variable (such as SQL_WARNINGS) to NULL.
• Extended the default timeout value for Windows clients from 30 seconds to 1 year. (The timeout that was added in MySQL 4.0.15 was way too short). This fixes a bug that caused ERROR 2013: Lost connection to MySQL server during query for queries that lasted longer than 30 seconds, if the client didn't specify a limit with mysql_options(). Users of 4.0.15 on Windows should upgrade to avoid this problem.
• More « out of memory » checking in range optimizer.
• Fixed and documented a problem when setting and using a user variable within the same SELECT statement. (Bug #1194).
• Fixed bug in overrun check for BLOB values with compressed tables. This was a bug introduced in 4.0.14. It caused MySQL to regard some correct tables containing BLOB values as corrupted. (Bug #770, Bug #1304, and maybe Bug #1295)
• SHOW GRANTS showed USAGE instead of the real column-level privileges when no table-level privileges were given.
• When copying a database from the master, LOAD DATA FROM MASTER dropped the corresponding database on the slave, thus erroneously dropping tables that had no counterpart on the master and tables that may have been excluded from replication using replicate-*-table rules. Now LOAD DATA FROM MASTER no longer drops the database. Instead, it drops only the tables that have a counterpart on the master and that match the replicate-*-table rules. replicate-*-db rules can still be used to include or exclude a database as a whole from LOAD DATA FROM MASTER. A database will also be included or excluded as a whole if there are some rules like replicate-wild-do-table=db1.% or replicate-wild-ignore-table=db1.%, as is already the case for CREATE DATABASE and DROP DATABASE in replication. (Bug #1248)
• Fixed a bug where mysqlbinlog crashed with a segmentation fault when used with the -h or --host option. (Bug #1258)
• Fixed a bug where mysqlbinlog crashed with a segmentation fault when used on a binary log containing only final events for LOAD DATA. (Bug #1340)
• mysqlbinlog will not reuse temporary file names from previous runs. Previously mysqlbinlog failed if was used several times on the same binary log file that contained a LOAD DATA command.
• Fixed compilation problem when compiling with OpenSSL 0.9.7 with disabled old DES support (If OPENSSL_DISABLE_OLD_DES_SUPPORT option was enabled).
• Fixed a bug when two (or more) MySQL servers were running on the same machine, and they were both slaves, and at least one of them was replicating some LOAD DATA INFILE command from its master. The bug was that one slave MySQL server sometimes deleted the SQL_LOAD-* files (used for replication of LOAD DATA INFILE and located in the slave-load-tmpdir directory, which defaults to tmpdir) belonging to the other slave MySQL server of this machine, if these slaves had the same slave-load-tmpdir directory. When that happened, the other slave could not replicate LOAD DATA INFILE and complained about not being able to open some SQL_LOAD-* file. (Bug #1357)
• If LOAD DATA INFILE failed for a small file, the master forgot to write a marker (a Delete_file event) in its binary log, so the slave could not delete 2 files (SQL_LOAD-*.info and SQL_LOAD-*.data from its tmpdir. (Bug #1391)
• On Windows, the slave forgot to delete a SQL_LOAD-*.info file from tmpdir after successfully replicating a LOAD DATA INFILE command. (Bug #1392)
• When a connection terminates, MySQL writes DROP TEMPORARY TABLE statements to the binary log for all temporary tables which the connection had not explicitly dropped. MySQL forgot to use backticks to quote the database and table names in the statement. (Bug #1345)
• On some 64-bit machines (some HP-UX and Solaris machines), a slave installed with the 64-bit MySQL binary could not connect to its master (it connected to itself instead). (Bug #1256, Bug #1381)
• Code was introduced in MySQL 4.0.15 for the slave to detect that the master had died while writing a transaction to its binary log. This code reported an error in a legal situation: When the slave I/O thread was stopped while copying a transaction to the relay log, the slave SQL thread would later pretend that it found an unfinished transaction. (Bug #1475)