Changes in MySQL 5.6.4 (2011-12-20, Milestone 7)

Note

This is a milestone release, for use at your own risk. Upgrades between milestone releases (or from a milestone release to a GA release) are not supported. Significant development changes take place in milestone releases and you may encounter compatibility issues, such as data format changes that require attention in addition to the usual procedure of running mysql_upgrade. For example, you may find it necessary to dump your data with mysqldump before the upgrade and reload it afterward. (Making a backup before the upgrade is a prudent precaution in any case.)

• Condition Handler Changes

• Fractional Seconds Handling

• InnoDB Notes

• Optimizer Notes

• Performance Schema Notes

• Functionality Added or Changed

• Bugs Fixed

Condition Handler Changes

• Incompatible Change: MySQL now supports the GET DIAGNOSTICS statement. GET DIAGNOSTICS provides applications a standardized way to obtain information from the diagnostics area, such as whether the previous SQL statement produced an exception and what it was. See GET DIAGNOSTICS Syntax.

  In addition, several deficiencies in condition handler processing rules were corrected so that MySQL behavior is more like standard SQL:

  • Block scope is used in determining which handler to select. Previously, a stored program was treated as having a single scope for handler selection.

  • Condition precedence is more accurately resolved.

  • Diagnostics area clearing has changed. Bug #55843 caused handled conditions to be cleared from the diagnostics area before activating the handler. This made condition information unavailable within the handler. Now condition information is available within the handler, which can inspect it with the GET DIAGNOSTICS statement. The condition information is cleared when the handler exits, if it has not already been cleared during handler execution.

  • Previously, handlers were activated as soon as a condition occurred. Now they are not activated until the statement in which the condition occurred finishes execution, at which point the most appropriate handler is chosen. This can make a difference for statements that raise multiple conditions, if a condition raised later during statement execution has higher precedence than an earlier condition and there are handlers in the same scope for both conditions. Previously, the handler for the first condition raised would be chosen, even if it had a lower precedence than other handlers. Now the handler for the condition with the highest precedence is chosen, even if it is not the first condition raised by the statement.

  • Issues that caused the server to exit due to incorrect handler call stack processing were fixed.

  The work just described involved several condition-handler bug fixes:

  • The RETURN statement did not clear the diagnostics area as it should have. Now the diagnostics area is cleared before executing RETURN. This prevents a condition in a nested function call from incorrectly propagating to an outer scope. It also means there is no way to return an SQL warning from a stored function. This change is not backward compatible, but the resulting behavior is more like standard SQL.

  • When an SQL HANDLER was activated, the handled condition was immediately removed from the diagnostics area. Consequently, any SQL diagnostic statement executed in the handler was unable to examine the condition that activated the handler. Now condition information is available within the handler.

  • If multiple handlers existed at the same level within a stored program, the wrong one could be chosen. Now the handler for the highest precedence condition is chosen.

  • If an error occurred in a context where different handlers were present at different levels of nesting, an outer handler could be chosen rather than the innermost one.

  For more information, see Scope Rules for Handlers. (Bug #12951117, Bug #38806, Bug #11749343, Bug #55852, Bug #11763171, Bug #61392, Bug #12652873, Bug #11660, Bug #11745196, Bug #48637, Bug #11756690)

Fractional Seconds Handling

• Incompatible Change: MySQL now permits fractional seconds for TIME, DATETIME, and TIMESTAMP values, with up to microseconds (6 digits) precision. To define a column that includes a fractional seconds part, use the syntax type_name(fsp), where type_name is TIME, DATETIME, or TIMESTAMP, and fsp is the fractional seconds precision. For example:

  CREATE TABLE t1 (t TIME(3), dt DATETIME(6));

  The fsp value, if given, must be in the range 0 to 6. A value of 0 signifies that there is no fractional part. If omitted, the default precision is 0. (This differs from the standard SQL default of 6, for compatibility with previous MySQL versions.)

  The following items summarize the implications of this change. See also Fractional Seconds in Time Values.

  • For TIME, DATETIME, and TIMESTAMP columns, the encoding and storage requirements in new tables differ from such columns in tables created previously because these types now include a fractional seconds part. This can affect the output of statements that depend on the row format, such as CHECKSUM TABLE.

  • Due to these changes in encoding and storage requirements for MySQL's DATETIME and TIMESTAMP types, importing pre-MySQL 5.6.4 InnoDB tables using ALTER TABLE ... IMPORT TABLESPACE that contain DATETIME and TIMESTAMP types into MySQL 5.6.4 (or later) requires a workaround procedure which is described in the “Server Changes” section of Changes in MySQL 5.6 .

  • Syntax for temporal literals now produces temporal values: DATE 'str', TIME 'str', and TIMESTAMP 'str', and the ODBC-syntax equivalents. The resulting value includes a trailing fractional seconds part if specified. Previously, the temporal type keyword was ignored and these constructs produced the string value. See Standard SQL and ODBC Date and Time Literals

  • Functions that take temporal arguments accept values with fractional seconds. Return values from temporal functions include fractional seconds as appropriate.

  • Three INFORMATION_SCHEMA tables, COLUMNS, PARAMETERS, and ROUTINES, now have a DATETIME_PRECISION column. Its value is the fractional seconds precision for TIME, DATETIME, and TIMESTAMP columns, and NULL for other data types.

  • The C API accommodates fractional seconds as follows:

    • In the MYSQL_FIELD column metadata structure, the decimals member indicates the fractional seconds precision for TIME, DATETIME, and TIMESTAMP columns. Clients can determine whether a result set temporal column has a fractional seconds part by checking for a nonzero decimals value in the corresponding MYSQL_FIELD structure. Previously, the decimals member indicated the precision for numeric columns and was zero otherwise.

    • In the MYSQL_TIME structure used for the binary protocol, the second_part member indicates the microseconds part for TIME, DATETIME, and TIMESTAMP columns. Previously, the second_part member was unused.

  In some cases, previously accepted syntax may produce different results. The following items indicate where existing code may need to be changed to avoid problems:

  • Some expressions produce results that differ from previous results. Examples: The timestamp system variable returns a value that includes a microseconds fractional part rather than an integer value. Functions that return a result that includes the current time (such as CURTIME(), SYSDATE(), or UTC_TIMESTAMP()) interpret an argument as an fsp value and the return value includes a fractional seconds part of that many digits. Previously, these functions permitted an argument but ignored it.

  • TIME values are converted to DATETIME by adding the time to the current date. (This means that the date part of the result differs from the current date if the time value is outside the range from '00:00:00' to '23:59:59'.) Previously, conversion of TIME values to DATETIME was unreliable. See Conversion Between Date and Time Types.

  • TIMESTAMP(N) was permitted in old MySQL versions, but N was a display width rather than fractional seconds precision. Support for this behavior was removed in MySQL 5.5.3, so applications that are reasonably up to date should not be subject to this issue. Otherwise, code must be rewritten.

  Note

  There may be problems replicating from a master server that understands fractional seconds to an older slave that does not:

  • For CREATE TABLE statements containing columns that have an fsp value greater than 0, replication will fail due to parser errors.

  • Statements that use temporal data types with an fsp value of 0 will work for with statement-based logging but not row-based logging. In the latter case, the data types have binary formats and type codes on the master that differ from those on the slave.

  • Some expression results will differ on master and slave. For example, expressions that involve the timestamp system variable or functions that return the current time have different results, as described earlier.

  (Bug #8523, Bug #11745064)

InnoDB Notes

• MySQL now supports FULLTEXT indexes for InnoDB tables. The core syntax is very similar to the FULLTEXT capability from earlier releases, with the CREATE TABLE and CREATE INDEX statements, and MATCH() ... AGAINST() clause in the SELECT statement. The new @ operator allows proximity searches for terms that are near each other in the document. The detailed search processing is controlled by a new set of configuration options: innodb_ft_enable_stopword, innodb_ft_server_stopword_table, innodb_ft_user_stopword_table, innodb_ft_cache_size, innodb_ft_min_token_size, and innodb_ft_max_token_size. You can monitor the workings of the InnoDB full-text search system by querying new INFORMATION_SCHEMA tables: innodb_ft_default_stopword, innodb_ft_index_table, innodb_ft_index_cache, innodb_ft_config, innodb_ft_deleted, and innodb_ft_being_deleted.

Optimizer Notes

• These query optimizer improvements were implemented:

  • The optimizer detects and optimizes away these useless query parts within IN/ALL/SOME/EXISTS subqueries:

    • DISTINCT

    • GROUP BY, if there is no HAVING clause and no aggregate functions

    • ORDER BY, which has no effect because LIMIT is not supported in these subqueries

Performance Schema Notes

• The Performance Schema has these additions:

  • The Performance Schema now permits instrument and consumer configuration at server startup, which previously was possible only at runtime using UPDATE statements for the setup_instruments and setup_consumers tables. This change was made because configuration at runtime is too late to disable instruments that have already been initialized during server startup. For example, the wait/sync/mutex/sql/LOCK_open mutex is initialized once during server startup, so attempts to disable the corresponding instrument at runtime have no effect.

    To control an instrument at server startup, use an option of this form:

    --performance-schema-instrument='instrument_name=value'

    Here, instrument_name is an instrument name such as wait/sync/mutex/sql/LOCK_open, and value is one of these values:

    • off, false, or 0: Disable the instrument

    • on, true, or 1: Enable and time the instrument

    • counted: Enable and count (rather than time) the instrument

    Each --performance-schema-instrument option can specify only one instrument name, but multiple instances of the option can be given to configure multiple instruments. In addition, patterns are permitted in instrument names to configure instruments that match the pattern. To configure all condition synchronization instruments as enabled and counted, use this option:

    --performance-schema-instrument='wait/synch/cond/%=counted'

    To disable all instruments, use this option:

    --performance-schema-instrument='%=off'

    Longer instrument name strings take precedence over shorter pattern names, regardless of order. For information about specifying patterns to select instruments, see Naming Instruments or Consumers for Filtering Operations.

    An unrecognized instrument name is ignored. It is possible that a plugin installed later may create the instrument, at which time the name is recognized and configured.

    To control a consumer at server startup, use an option of this form:

    --performance-schema-consumer_consumer_name=value

    Here, consumer_name is a consumer name such as events_waits_history, and value is one of these values:

    • off, false, or 0: Do not collect events for the consumer

    • on, true, or 1: Collect events for the consumer

    For example, to enable the events_waits_history consumer, use this option:

    --performance-schema-consumer-events-waits-history=on

    The permitted consumer names can be found by examining the setup_consumers table. Patterns are not permitted.

    Along with the preceding changes to permit configuration at server startup, the default instrument and consumer configuration has changed. Previously, all instruments and consumers were enabled by default. Now, instruments are disabled except the statement, I/O, and idle instruments. Consumers are disabled except the global, thread, and current-statement consumers. These changes produce a default configuration with a low overhead.

  • Tables that have an EVENT_ID column now also have an END_EVENT_ID column to support determination of nested event relationships:

    • events_waits_current, events_waits_history, events_waits_history_long

    • events_stages_current, events_stages_history, events_stages_history_long

    • events_statements_current, events_statements_history, events_statements_history_long

    As before, EVENT_ID is populated with the thread current event counter when an event starts. In addition, END_EVENT_ID is NULL until the event ends, at which point it is set to the new thread current event counter. This permits the relationship “event B is included in event A” to be determined using the following expression, without having to follow each inclusion relationship using NESTING_EVENT_ID:

    A.EVENT_ID <= B.EVENT_ID AND B.END_EVENT_ID <= A.END_EVENT_ID

  • The Performance Schema aggregates file I/O operations in two places, the events_waits_summary_xxx tables and the file_summary_xxx tables. It was possible to join the events_waits_summary_global_by_event_name table to the file_summary_by_event_name by using the EVENT_NAME column. However, it was not possible to do the same with the events_waits_summary_by_instance and file_summary_by_instance tables because the former uses OBJECT_INSTANCE_BEGIN as the instance identifier and the latter uses FILE_NAME. This means that it was possible to obtain both file I/O latency and usage per file, but not to correctly correlate latency to usage when there was more than one form of file (such as multiple redo logs, table files, and so forth).

    To address this issue, the file_summary_by_instance table now has an OBJECT_INSTANCE_BEGIN column. In addition, both file_summary_by_instance and file_summary_by_event_name have additional aggregation columns (such as timer wait information), which in many cases makes it possible to obtain the desired summary information without need for a join at all.

  If you upgrade to this MySQL release from an earlier version, you must run mysql_upgrade (and restart the server) to incorporate these changes into the performance_schema database.

  For more information, see MySQL Performance Schema.

Functionality Added or Changed

• Performance; InnoDB: New optimizations apply to read-only InnoDB transactions. See Optimizing InnoDB Read-Only Transactions for details. The new optimizations make autocommit more applicable to InnoDB queries than before, as a way to signal that a transaction is read-only because it is a single-statement SELECT.

• Performance; InnoDB: You can now set the InnoDB page size for uncompressed tables to 8KB or 4KB, as an alternative to the default 16KB. This setting is controlled by the innodb_page_size configuration option. You specify the size when creating the MySQL instance. All InnoDB tablespaces within an instance share the same page size. Smaller page sizes can help to avoid redundant or inefficient I/O for certain combinations of workload and storage devices, particularly SSD devices with small block sizes.

• Replication: Previously, replication slaves could connect to the master server only through master accounts that use native authentication. Now replication slaves can also connect through master accounts that use nonnative authentication if the required client-side plugin is installed on the slave side in the directory named by the slave plugin_dir system variable. (Bug #12897501)

• The optimizer trace capability now tracks temporary tables created by the server during statement execution. (Bug #13400713)

• Performance of metadata locking operations on Windows XP systems was improved by instituting a cache for metadata lock objects. This permits the server to avoid expensive operations for creation and destruction of synchronization objects on XP. A new system variable, metadata_locks_cache_size, permits control over the size of the cache. The default size is 1024. (Bug #12695572)

• Upgrading from an Advanced GPL RPM package to an Advanced RPM package did not work. Now on Linux it is possible to use rpm -U to replace any installed MySQL product by any other of the same release family. It is not necessary to remove the old produce with rpm -e first. (Bug #11886309)

• The make_win_bin_dist script is no longer used and has been removed from MySQL distributions and the manual. (Bug #58241)

• MEMORY table creation time is now available in the CREATE_TIME column of the INFORMATION_SCHEMA.TABLES table and the Create_time column of SHOW TABLE STATUS output. (Bug #51655, Bug #11759349)

• Previously, MySQL servers from 5.1 and up refused to open ARCHIVE tables created in 5.0 because opening them caused a server crash. The server now can open 5.0 ARCHIVE tables, and REPAIR TABLE updates them to the format used in 5.6. However, the recommended upgrade procedure is still to dump 5.0 ARCHIVE tables before upgrading and reload them after upgrading. (Bug #48633, Bug #11756687)

• Error messages that referred only to an error code now also include the corresponding error description. (Bug #48348, Bug #11756433)

• The MySQL code base was changed to permit use of the C++ Standard Library and to enable exceptions and runtime type information (RTTI). This change has the following implications:

  • Libraries and executables depend on some C++ standard library. On Linux, this has not been the case previously. On Solaris, the default dependency has changed from the default library to libstlport, which is now included with binary distributions for users whose system does not have it.

  • The -fno-rtti and -fno-exceptions options are no longer used to build plugins, such as storage engines. Users who write their own plugins should omit these options if they were using tem.

  • C++ users who compile from source should set CXX to a C++ compiler rather than a C compiler. For example, use g++ rather than gcc. This includes the server as well as client programs.

  • mysql_config now has a --cxxflags option. This is like the --cflags option, but produces flags appropriate for a C++ compiler rather than a C compiler.

  • User-defined functions can be written in C++ using standard library features.

Bugs Fixed

• Security Enhancement; Replication: The START SLAVE statement now accepts USER and PASSWORD options. By default, MySQL native authentication is used, and the user name and password are stored in the master.info repository. This behavior can be overridden by additionally specifying the name (DEFAULT_AUTH) and location (PLUGIN_DIR) of an authentication plugin when issuing START SLAVE. .

  As part of this change, warnings are now issued in the following cases:

  1. If START SLAVE USER="..." PASSWORD="..." or CHANGE MASTER TO MASTER_USER="..." MASTER_PASSWORD="..." is executed using an unencrypted connection, the warning message Sending passwords in plain text without SSL/TLS is extremely insecure is generated (ER_INSECURE_PLAIN_TEXT).

  2. If the user name and password are stored in or read from the master.info repository in the course of executing CHANGE MASTER TO, a warning message is printed out to the error log: Storing MySQL user name or password information in the master.info repository is not secure and is therefore not recommended (ER_INSECURE_CHANGE_MASTER).

  Note

  The text of a running START SLAVE statement, including values for USER and PASSWORD, can be seen in the output of a concurrent SHOW PROCESSLIST statement. The complete text of a CHANGE MASTER TO statement is also visible to SHOW PROCESSLIST.

  See also Pluggable Authentication. (Bug #13083642)

• Incompatible Change; Replication: The statements in the following list are now marked as unsafe for statement-based replication. This is due to the fact that each of these statements depends on the results of a SELECT statement whose order cannot always be determined. When using STATEMENT logging mode, a warning is issued in the binary log for any of these statements; when using MIXED logging mode, the statement is logged using the row-based format.

  • INSERT ... SELECT ... ON DUPLICATE KEY UPDATE

  • REPLACE ... SELECT

  • CREATE TABLE ... IGNORE SELECT

  • CREATE TABLE ... REPLACE SELECT

  • INSERT IGNORE ... SELECT

  • UPDATE IGNORE

  When upgrading, you should note the use of these statements in your applications, keeping in mind that a statement that inserts or replaces rows obtained from a SELECT can take up many times as much space in the binary log when logged using row-based format than when only the statement itself is logged. Depending on the number and size of the rows selected and inserted (or replaced) by any such statements, the difference in size of the binary log after the logging of these statements is switched from statement-based to row-based can potentially be several orders of magnitude. See Advantages and Disadvantages of Statement-Based and Row-Based Replication. (Bug #11758262, Bug #50439)

• Incompatible Change: Previously, “Aborted connection” errors were written to the error log based on the session value of log_warnings, which permitted users with minimal privileges to cause many messages to be written to the log unless restricted by the MAX_CONNECTIONS_PER_HOUR resource limit. Now this logging is based on the global log_warnings variable. There are no remaining uses of the session log_warnings variable, so it has been removed and the variable now has only a global value. (Bug #53466, Bug #11761014)

• Important Change; InnoDB: If an ALTER TABLE statement failed for an InnoDB table due to an error code from an underlying file-renaming system call, InnoDB could lose track of the .ibd file for the table. This issue only occurred when the innodb_file_per_table configuration option was enabled, and when the low-level error persisted through thousands of retry attempts. In MySQL 5.1, this issue applied to the InnoDB Plugin but not the built-in InnoDB storage engine.

  For example, if you encounter an error like the following:

  mysql> ALTER TABLE sb2 ADD COLUMN d2 INT;
  ERROR 1025 (HY000): Error on rename of './sbtest/#sql-1eb9_1' to
  './sbtest/sb2' (errno: -1)

  you might be able to access the #sql* table by copying an .frm file from a table with an identical schema. The table name to use for the .frm filewould be `sbtest.#mysql50##sql-1eb9_1` in the preceding example. (Bug #12884631, Bug #62146)

• Important Change; Replication: Setting an empty user in a CHANGE MASTER TO statement caused an invalid internal result and is no longer permitted. Trying to use MASTER_USER='' or setting MASTER_PASSWORD while leaving MASTER_USER unset causes the statement to fail with an error. (Bug #13427949)

• Performance; InnoDB: The process of deallocating the InnoDB Adaptive Hash Index was made faster, during shutdown or when turning off the AHI with the statement:

  SET GLOBAL innodb_adaptive_hash_index=OFF;

  (Bug #13006367, Bug #62487)

• Performance; InnoDB: This fix improves the performance of instrumentation code for InnoDB buffer pool operations. (Bug #12950803, Bug #62294)

• Performance; InnoDB: This fix improved the efficiency and concurrency of freeing pages in the InnoDB buffer pool when performing a DROP TABLE for an InnoDB table when the innodb_file_per_table option is enabled.

  This change is most noticeable for systems with large buffer pools. During the drop operation, one traversal of the buffer pool memory structure is changed from the LRU list (the entire buffer pool) to the flush list (a much smaller structure). The LRU scanning is reduced, but not entirely eliminated. The buffer pool mutex is also released periodically, so that if the drop operation takes significant time, other threads can proceed concurrently. (Bug #11759044, Bug #51325)

• InnoDB: An internal deadlock could occur within InnoDB, on a server doing a substantial amount of change buffering for DML operations, particularly DELETE statements. (Bug #13340047)

• InnoDB: Fixed a compilation problem that affected the InnoDB source code with gcc 4.6.1. The affected InnoDB source file was btr/btr0cur.c. (Bug #13116045)

• InnoDB: Querying the INFORMATION_SCHEMA.INNODB_SYS_TABLESTATS table could cause the server to halt with an assertion error, in debug builds only. (Bug #12960058)

• InnoDB: Valgrind errors when building with the settings innodb_checksum_algorithm=innodb and innodb_checksum_algorithm=crc32 were fixed. (Bug #12939557)

• InnoDB: Unused functions were removed from the internal InnoDB code related to mini-transactions, to clarify the logic. (Bug #12626794, Bug #61240)

• InnoDB: Lookups using secondary indexes could give incorrect matches under a specific set of conditions. The conditions involve an index defined on a column prefix, for a BLOB or other long column stored outside the index page, with a table using the Barracuda file format. (Bug #12601439, Bug #12543666)

• InnoDB: An UPDATE statement for an InnoDB table could hang. The issue affects tables using the Barracuda file format and having multiple indexes on column prefixes. The size of an undo log record could exceed the page size, even though the total size of the column prefixes was less than the page size (usually 16KB). In MySQL 5.5 and higher, this error is now reported using the new code ER_UNDO_RECORD_TOO_BIG. In MySQL 5.1 with the InnoDB Plugin, this error is reported using the existing code ER_TOO_BIG_ROWSIZE. (Bug #12547647)

• InnoDB: This fix corrects cases where the MySQL server could hang or abort with a long semaphore wait message. (This is a different issue than when these symptoms occurred during a CHECK TABLE statement.) (Bug #11766591, Bug #59733)

• InnoDB: Issuing INSERT...ON DUPLICATE KEY statements for InnoDB tables from concurrent threads could cause a deadlock, particularly with the INSERT...ON DUPLICATE KEY UPDATE form. The problem could also be triggered by issuing multiple INSERT IGNORE statements. The fix avoids deadlocks caused by the same row being accessed by more than one transaction. Deadlocks could still occur when multiple rows are inserted and updated simultaneously by different transactions in inconsistent order; those types of deadlocks require the standard error handling on the application side, of re-trying the transaction. (Bug #11759688, Bug #52020, Bug #12842206)

• Partitioning: CHECKSUM TABLE returned 0 for a partitioned table unless the statement was used with the EXTENDED option. (Bug #11933226, Bug #60681)

• Partitioning: Error 1214 (ER_TABLE_CANT_HANDLE_FT), given when trying to use a FULLTEXT index with a partitioned table, displayed the misleading text The used table type doesn't support FULLTEXT indexes was misleading and has been replaced with Error 1752 (ER_FULLTEXT_NOT_SUPPORTED_WITH_PARTITIONING) which shows the more accurate FULLTEXT index is not supported for partitioned tables. (Bug #11763825, Bug #56590)

• Partitioning: Using ALTER TABLE to remove partitioning from a valid MyISAM table could corrupt it. (Bug #52599, Bug #11760213)

• Replication: The value set for the slave_parallel_workers system variable (or the corresponding --slave-parallel-workers server option) was not always honored correctly; in such cases a random value was used. (Bug #13334470)

• Replication: Execution of LOAD DATA on a MyISAM table having an after-insert trigger which wrote into an InnoDB table caused multithreaded statement-based replication to abort with error 1742 (Cannot execute the current event group in the parallel mode). (Bug #12982188)

• Replication: Several warnings and informational messages were revised for typographic errors and clarity. (Bug #12947248, Bug #12978113)

• Replication: When a statement containing a large number of rows to be applied on a slave table that does not contain a primary key, a considerable amount of time can be needed to find and change all the rows that are to be changed. The current fix helps diagnose this issue by printing a message to the error log if the execution time for a given statement replicated using row-based replication takes more than 60 seconds. log_warnings must be greater than 1 for this message to be printed to the error log. (Bug #11760927, Bug #53375)

• Replication: mysqlbinlog --hexdump printed the last row of the hex dump incorrectly, in two ways:

  1. If the length of the last row was eight bytes, the end of the previous row was copied to the end of the last row, padding the last row to full length.

  2. If the length of the last row was less than sixteen bytes, its textual representation was not aligned with that of previous rows.

  (Bug #11747887, Bug #34386)

• Replication: A replication master could send damaged events to slaves after the binary log disk on the master became full. To correct this issue, only complete events are now pushed by the master dump thread to the slave I/O thread. In addition, the error text that the master sends to the slave when an incomplete event is found now states that the incomplete event may have been caused by running out of disk space on the master, and provides coordinates of the first and the last event bytes read. (Bug #11747416, Bug #32228)

  References: See also: Bug #64035, Bug #13593869.

• Replication: --replicate-rewrite-db=from_name->to_name did not work correctly when the name of the source database (from_name) consisted of only a single character. (Bug #34332, Bug #11747866)

• An incorrect InnoDB assertion could cause the server to halt. This issue only affected debug builds. The assertion referenced the source file btr0pcur.ic and the variable cursor->pos_state. (Bug #13358468)

• A derived table with more than 64 columns caused a server crash. (Bug #13354889)

• With InnoDB change buffering enabled and innodb_page_size set to an 8K or 4K page size, an UPDATE statement could fail if a column being updated contained a value longer than 1/8th of the page size. (Bug #13336585)

• Writes to the slow log involved a call to thd->current_utime() even if no log entries ended up being written, unnecessarily reducing performance. (Bug #13326965)

• Rounding DBL_MAX returned DBL_MAX, not 'inf'. (Bug #13261955)

• For materialized temporary tables, a missing key length check could cause incorrect query results. (Bug #13261277)

• Access privileges were checked for each stored program instruction, even if the instruction used no tables, resulting in reduced performance. (Bug #13251277)

• The error message for ER_EVENT_CANNOT_ALTER_IN_THE_PAST was incorrect. (Bug #13247871)

• During the table-opening process, memory was allocated and later freed that was needed view loading, even for statements that did not use views. These unnecessary allocation and free operations are no longer done. (Bug #13116518)

• Subqueries with OUTER JOIN could return incorrect results if the subquery referred to a column from another SELECT. (Bug #13068506)

• Writes to MyISAM temporary tables could include uninitialized data, which could contain sensitive information. Now only bytes containing initialized data are copied, which also improves performance. (Bug #12997905)

• The Performance Schema nested some network I/O events within the wrong statement. (Bug #12981100)

• mysql_plugin mishandled the --plugin-ini, --mysqld, and --my-print-defaults options under some circumstances. (Bug #12968815)

• mysql_plugin returned the wrong error code from failed server bootstrap execution. (Bug #12968567)

• Internal conversion of zero to binary and back could yield a result with incorrect precision. (Bug #12911710)

• Valgrind warnings generated by filesort operations were fixed. (Bug #12856915)

• An IN-to-EXISTS subquery transformation could yield incorrect results if the outer value list contained NULL. (Bug #12838171)

• With index condition pushdown enabled, STRAIGHT_JOIN queries could produce incorrect results. (Bug #12822678, Bug #12724899)

• The result of ROUND() was incorrect for certain numbers. (Bug #12744991)

• A warning resulting from use of SPACE() referred to REPEAT() in the warning message. (Bug #12735829)

• IN and EXISTS subqueries with DISTINCT and ORDER BY could return incorrect results. (Bug #12709738)

• A memory leak occurred due to failure to clean up after QUICK_INDEX_MERGE_SELECT/Unique. (Bug #12694872, Bug #14542543)

• Several improvements were made to the libedit library bundled with MySQL distributions, and that is available for all platforms that MySQL supports except Windows.

  • Navigation keys did not work for UTF-8 input.

  • Word navigation and delete operations did not work for UTF-8 input with Cyrillic characters.

  • Nonlatin characters were corrupted in overwrite mode for UTF-8 input.

  • Long queries caused the statement history file to become corrupted.

  • The Alt key caused history operations to fail.

  (Bug #12605400, Bug #12613725, Bug #12618092, Bug #12624155, Bug #12617651, Bug #12605388)

• SELECT SQL_BUFFER_RESULT query results included too many rows if a GROUP BY clause was optimized away. (Bug #12578908)

• decimal_round() could cause a server exit when processing long numeric strings. (Bug #12563865)

• mysqldump --all-databases did not dump the replication log tables. (They could be dumped only by naming them explicitly when invoking mysqldump, and using the --master-data option.)

  As a result of the fix for this problem, it is now possible to execute statements requiring read locks on the replication log tables at any time, while any statements requiring a write lock on either or both of these tables are disallowed whenever replication is in progress. For more information, see Replication Relay and Status Logs. (Bug #12402875, Bug #60902)

• The client-server protocol now has the client send authentication data as length-encoded strings so that data longer than 256 bytes can be sent. This is done using the CLIENT_PLUGIN_AUTH_LENENC_CLIENT_DATA client capability. (Bug #11878962)

• mysqld_safe did not properly check for an already running instance of mysqld. (Bug #11878394)

• SUM(DISTINCT) incorrectly converted DATE, DATETIME, TIME, and TIMESTAMP arguments to YEAR. (Bug #73543, Bug #19427648)

• If index condition pushdown access was chosen and then abandoned, some variables were not cleared, leading to incorrect query results. (Bug #62533)

• The CMake configuration checks did not properly test whether the C compiler supports the inline keyword. (Bug #61708, Bug #12711108)

• For a lower_case_table_names value of 1 or 2 and a database having a mixed-case name, calling a stored function using a fully qualified name including the database name failed. (Bug #60347, Bug #11840395)

• mysql_upgrade did not upgrade the system tables or create the mysql_upgrade_info file when run with the --write-binlog or --skip-write-binlog option. (Bug #60223, Bug #11827359)

• A multiple-table UPDATE statement required the UPDATE privilege on a view which was only read if the view was processed using the merge algorithm. (Bug #59957, Bug #11766767)

• When a join operation contained a view, the optimizer sometimes failed to associate the view's WHERE clause with the first table or view in a join when it was possible to do so, resulting in a less efficient query. (Bug #59696, Bug #11766559)

• An assertion was raised when selecting from a view that selects from a view that used a user-defined function that had been deleted. (Bug #59546, Bug #11766440)

• The help message for mysql_install_db did not indicate that it supports the --defaults-file, --defaults-extra-file and --no-defaults options. (Bug #58898, Bug #11765888)

• mysql_install_db printed the --skip-grant-tables server option as --skip-grant in one of its error messages. (Bug #58534, Bug #11765553)

• An assertion designed to detect zero-length sort keys also was raised when the entire key set fit in memory. (Bug #58200, Bug #11765254)

• During optimization, ZEROFILL values may be converted to string constants. However, CASE expressions did not handle switching data types after the planning stage, leading to CASE finding a null pointer instead of its argument. (Bug #57135, Bug #11764313)

• If a plugin was uninstalled, thread local variables for plugin variables of string type with wth PLUGIN_VAR_MEMALLOC flag were not freed. (Bug #56652, Bug #11763882)

• Deadlock could occur when these four things happened at the same time: 1) An old dump thread was waiting for the binary log to grow. 2) The slave server that replicates from the old dump thread tried to reconnect. During reconnection, the new dump thread tried to kill the old dump thread. 3) A KILL statement tried to kill the old dump thread. 4) An INSERT statement caused a binary log rotation. (Bug #56299, Bug #11763573)

• myisampack could create corrupt FULLTEXT indexes when compressing tables. (Bug #53646, Bug #11761180)

• The SQL_BIG_RESULT modifier could change the results for queries that included a GROUP BY clause. (Bug #53534, Bug #11761078)

• ARCHIVE tables with NULL columns could cause server crashes or become corrupt under concurrent load. (Bug #51252, Bug #11758979)

• InnoDB used incorrect identifier quoting style in an error message that resulted in an error if a user followed the suggestion in the message. (Bug #49556, Bug #11757503)

• OPTIMIZE TABLE could corrupt MyISAM tables if myisam_use_mmap was enabled. (Bug #49030, Bug #11757032)

• Concurrent access to ARCHIVE tables could cause corruption. (Bug #42784, Bug #11751793)

• A query that selected a GROUP_CONCAT() function result could return different values depending on whether an ORDER BY of the function result was present. (Bug #41090, Bug #11750518)

• A linking problem prevented the FEDERATED storage engine plugin from loading. (Bug #40942, Bug #11750417)

• Subqueries could return incorrect results when materialization was enabled. (Bug #40037, Bug #11749901, Bug #12705660, Bug #12908058)

• For debug builds, an assertion could be raised for ALTER statements that performed a RENAME operation. This occurred for storage engine handlertons that exposed the HTON_FLUSH_AFTER_RENAME flag. (Bug #38028, Bug #11749050)

• The estimate of space required for filesort operations could be too high, resulting in inefficient initialization. (Bug #37359, Bug #11748783)

• An ALTER TABLE that included an ADD ... AFTER operation to add a new column after a column that had been modified earlier in the statement failed to find the existing column. (Bug #34972, Bug #11748057)

• For FEDERATED tables, loss of connection to the remote table during some insert operations could cause a server crash. (Bug #34660, Bug #11747970)

• For some queries, the index_merge access method was used even when more expensive then ref access. (Bug #32254, Bug #11747423)

• Collation for the SPACE() function was determined by the parse time value of the collation_connection system variable (instead of the runtime value), which could give unexpected results from prepared statements, triggers, and stored procedures. (Bug #23637, Bug #11746123)