MySQL 5.7 Release Notes  /  Changes in MySQL 5.7.19 (2017-07-17, General Availability)

Changes in MySQL 5.7.19 (2017-07-17, General Availability)

Account Management Notes

  • During data directory initialization or upgrade, MySQL now creates a 'mysql.session'@'localhost' reserved account. This account is used internally by plugins to access the server. It is locked so that it cannot be used for client connections. (Bug #25642343)

Compilation Notes

  • Solaris: On Solaris, MySQL binary distributions are now linked against libatomic.so, so they are no longer dependent on libstatomic.so. (Bug #25909965)

Deprecation and Removal Notes

Packaging Notes

  • mysqladmin was added to Docker/Minimal packages because it is needed by InnoDB Cluster. (Bug #25998285)

  • Debian/Ubuntu packages now support multiple MySQL instances with systemd. See Managing MySQL Server with systemd. (Bug #24559588, Bug #82785)

Security Notes

  • Security Fix: The linked OpenSSL library for the MySQL Commercial Server has been updated to version 1.0.2l. Issues fixed in the new OpenSSL version are described at http://www.openssl.org/news/vulnerabilities.html.

    This change does not affect the Oracle-produced MySQL Community build of MySQL Server, which uses the yaSSL library instead. (Bug #26160962)

  • MySQL Enterprise Edition now supports LDAP pluggable authentication of MySQL users. This enables MySQL Server to use LDAP (Lightweight Directory Access Protocol) to authenticate MySQL users by accessing directory services such as X.500. For more information, see Pluggable Authentication.

  • MySQL Enterprise Edition now includes a keyring_aws plugin that communicates with the Amazon Web Services Key Management Service as a back end for key generation and uses a local file for key storage. For more information, see The MySQL Keyring.

Spatial Data Support

Test Suite Notes

  • mysql-test-run.pl now has a --test-progress option to cause display of the percentage of tests remaining. (Bug #25601131, Bug #20755059, Bug #76455)

  • These changes were made to the --xml-report option for mysql-test-run.pl:

    • A <failure> tag identifies any test failing on a retry attempt.

    • Separate statistics and fields are included for skipped and disabled tests.

    • The XML report is created in the build directory if no absolute path is given for out-of-source builds.

    • For combination runs, a field named variation is included in the <testcase> tag.

    • Results for all tests belonging to a suite are aggregated within a single <testsuite> tag.

    • Information regarding failures is reported with a brief reason in an attribute named message, along with other details within the tag.

    (Bug #25349924)

  • mysql-test-run.pl now supports a --report-unstable-tests option that has these effects:

    • Reports any test that has passed using at least one retry attempt in a separate category called Unstable tests in the summary.

    • If all failures are due to unstable tests, mysql-test-run.pl produces a warning but exits successfully.

    • Adds a new XML tag to report unstable tests, if the --xml-report option is also specified.

    (Bug #24473420, Bug #25984429)

  • mysqltest now supports a replace_numeric_round command that takes an argument value from 0 to 16 indicating the number of decimals to round numeric values to. This can be used to help prevent result content mismatch errors for tests in which slightly different results are obtained across platforms due to precision differences. Thanks to Daniel Black for the patch. (Bug #23280117, Bug #81399)

X Plugin Notes

  • X Plugin now handles expired SSL certificates correctly. (Bug #25835833)

  • The output of clauses such as GROUP BY and HAVING has been improved by moving the grouping_criteria expression to the derived query. (Bug #25549637, Bug #24497007)

Platform-Specific Notes

  • Linux: The generic Linux build for MySQL 5.7 is now built on Oracle Linux 6 using glibc 2.12. Systems that use the build need to have glibc 2.12 or later installed on them. (Bug #26005558)

  • Linux: The generic Linux build for MySQL 5.7 now supports Non-Uniform Memory Access (NUMA) on its host system. Systems that use the build need to have libnuma installed on them. See Installing MySQL on Unix/Linux Using Generic Binaries for more details. (Bug #26005558)

Functionality Added or Changed

  • Replication: The group_replication_transaction_size_limit variable was added to enable you to protect a group against large transactions causing a failure. (Bug #84785, Bug #25510757)

  • Replication: Group Replication now supports SAVEPOINT SQL transactions.

  • RPM .spec files now include support for running unit tests. (Bug #25814143, Bug #85743)

  • For Windows, MSI installer packages now include a check for the required Visual Studio redistributable package, and produce a message asking the user to install it if it is missing. (Bug #25658832)

  • The mysql client now supports a --binary-as-hex option that causes display of binary data using hexadecimal notation (0xvalue). Thanks to Daniël van Eeden for the patch. (Bug #25340722, Bug #84391)

  • The audit_log filter parser now produces errors for unexpected JSON elements in filter definitions. (Bug #24360663)

Bugs Fixed

  • Incompatible Change: Plugins such as Group Replication and X Plugin now use the mysql.session account added in this version. If you are upgrading from a previous version which did not include the mysql.session account you must run mysql_upgrade to ensure the account is created. If mysql_upgrade is not run, plugins fail to start with the error message There was an error when trying to access the server with user: mysql.session@localhost. Make sure the user is present in the server and that mysql_upgrade was run after a server update. (Bug #26042764)

    References: See also: Bug #24311527, Bug #25642343, Bug #25750822, Bug #25103980, Bug #83841.

  • InnoDB: An ALTER TABLE operation that rebuilt an encrypted table did not set the encryption attribute properly. (Bug #26243264)

  • InnoDB: The length of a virtual column field in a virtual index record was less than the expected template column length. (Bug #25793677)

  • InnoDB: The server allocated memory unnecessarily for an operation that rebuilt the table. (Bug #25573565, Bug #85043)

  • InnoDB: With foreign_key_checks disabled, InnoDB incorrectly printed messages to the error log when operations were performed on a table that referenced a non-existent foreign key. (Bug #25365223)

  • InnoDB: During a TRUNCATE TABLE operation on a file-per-table tablespace, a dictionary operation lock was released before eviction of dirty pages from the buffer pool, causing a latch order violation. (Bug #25357789)

  • InnoDB: When using an index merge optimizer switch, a SELECT COUNT(*) operation sometimes returned 0. Partitioning code incorrectly performed a memcpy instead of a column copy of columns read by the index, causing the wrong records to be copied. (Bug #25332330, Bug #81031)

  • InnoDB: After a RENAME TABLE operation that moved a table to a different schema, InnoDB returned an error on restart indicating that it could not locate the tablespace data file. InnoDB failed to update INNODB_SYS_DATAFILES data dictionary table during the RENAME TABLE operation. (Bug #25189192, Bug #84038)

  • InnoDB: Loading InnoDB tables required more memory in MySQL 5.7 due primarily to the addition of in-memory structure members introduced with temporary table optimizations. The in-memory structure members, only used for temporary tables, are now only allocated as needed. (Bug #25080442)

  • InnoDB: During an ALTER TABLE operation that rebuilt a table containing a virtual column, InnoDB failed to apply a concurrent insert log record. (Bug #24961167)

  • InnoDB: InnoDB failed to apply the concurrent delete log for an in-place ALTER TABLE operation due to a virtual column validation issue. (Bug #24960450)

  • InnoDB: Changes to the InnoDB recovery process in MySQL 5.7 could require up to three scans of the redo log during recovery. To reduce the number of scans, the first and second scans were merged. With this change, there is only one scan unless the redo log record hash table that is populated by the scan reaches its memory threshold. In this case, a second scan is initiated that performs a simultaneous scan and apply. (Bug #22963951, Bug #80788)

  • Partitioning: When a query performed a reverse-ordered range scan (to satisfy ORDER BY ... DESC) using index condition pushdown on a partitioned InnoDB table, it could take an unreasonably long time to finish. This was due to the fact that the condition pushdown check failed to get the bounds of the range, so that the scan continued to read index tuples until it reached the first value in the index. (Bug #83470, Bug #24929748)

    References: See also: Bug #84107.

  • Replication: In the case of delayed initialization of the Group Replication plugin, deployed in single-primary mode, secondaries were able to get writes through an asynchronous replication channel, which is not allowed in normal initialization of the Group Replication plugin. (Bug #26314756)

  • Replication: With GTIDs generated for incident log events, MySQL error code 1590 (ER_SLAVE_INCIDENT) could not be skipped using the --slave-skip-errors=1590 startup option on a replication slave. (Bug #26266758)

  • Replication: A USE statement that followed a SET GTID_NEXT statement sometimes had no effect. (Bug #26128931)

  • Replication: Groups can now contain members running different server versions to enable you to do online upgrades of a replication group. The rules for combining members in a group with different versions are:

    • If you have a group with 8.0 members, you cannot add a 5.7 member

    • If you have a group with 5.7 members you can add a 8.0 member, but it remains in read-only mode. Writing to this member is dangerous while the group contains multiple server versions and should be avoided.

    In a single-primary group, if the current primary leaves the group and a new primary must be elected, the primary is first chosen from the lower version members. If no lower version member is found, the primary is chosen from newer version members. (Bug #25876807)

  • Replication: When binlog_checksum=NONE was set on a MySQL server after startup, and then Group Replication was started, if an error occurred, the server remained in RECOVERING state and could not be shut down. (Bug #25793366, Bug #85667)

  • Replication: In a Group Replication setup where circular asynchronous replication was implemented between members of different replication groups, view change log events were repeatedly replicated between the groups with new generated GTIDs each time. The fix ensures that view change log events are ignored outside the named replication group where they occur, and never generate new GTIDs. (Bug #25674926)

    References: See also: Bug #26049695, Bug #25928854, Bug #25721175.

  • Replication: When first starting the MySQL server following an installation from RPM, passwword validation plugin is activated by default (true only for RPM installations). If binary logging was already enabled at this time, the activation was logged, even though plugin activations should not be recorded in the binary log. (Bug #25672750)

  • Replication: In a setup where single-primary Group Replication was combined with asynchronous replication, for example with S1 and S2 forming a group and with S2 and S3 functioning as master and slave, secondaries such as S2 were accepting transactions and these could then enter the group. The fix prevents secondaries creating an asynchronous replication channel when belonging to a single-primary group, and Group Replication cannot be started when asynchronous replication is running. (Bug #25574200, Bug #85047)

    References: See also: Bug #86325, Bug #26078602.

  • Replication: In the event that a member failed to join a group the member was not stopping and continued to accept transactions. To avoid this set your members to have super_read_only=1 in the my.cfg file. Group Replication now checks for this setting upon successful start up and sets super_read_only=0. This ensures that members which do not successfully join a group cannot accept transactions. (Bug #25474736, Bug #84728)

  • Replication: If the binary log on a master server was rotated and a full disk condition occurred on the partition where the binary log file was being stored, the server could stop unexpectedly. The fix adds a check for the existence of the binary log when the dump thread switches to next binary log file. If the binary log is disabled, all binary logs up to the current active log are transmitted to slave and an error is returned to the receiver thread. (Bug #25076007)

  • Replication: Interleaved transactions could sometimes deadlock the slave applier when the transaction isolation level was set to REPEATABLE-READ. (Bug #25040331)

  • Replication: If a relay log index file named relay log files that did not exist, RESET SLAVE ALL sometimes did not fully clean up properly. (Bug #24901077)

  • Replication: The slave_skip_errors system variable did not permit error numbers larger than 3000. Thanks to Tsubasa Tanaka for the patch. (Bug #24748639, Bug #83184)

  • Replication: mysqlbinlog, if invoked with the --raw option, does not flush the output file until the process terminates. But if also invoked with the --stop-never option, the process never terminates, thus nothing is ever written to the output file. Now the output is flushed after each event. (Bug #24609402)

  • Replication: A memory leak in mysqlbinlog was fixed. The leak happened when processing fake rotate events, or when using --raw and the destination log file could not be created. The leak only occurred when processing events from a remote server. Thanks to Laurynas Biveinis for his contribution to fixing this bug. (Bug #24323288, Bug #82283)

  • Replication: A slave server could lose events not yet applied when MASTER_AUTO_POSITION=0, both replication threads were stopped, and the applier delay was changed using CHANGE MASTER TO MASTER_DELAY=N. (Bug #23203678, Bug #81232)

    References: See also: Bug #25340185, Bug #84375.

  • Replication: Transmission of large GCS messages could take so long the sender appeared to have died. (Bug #22671846)

  • Replication: Multi-threaded slaves could not be configured with small queue sizes using slave_pending_jobs_size_max if they ever needed to process transactions larger than that size. Any packet larger than slave_pending_jobs_size_max was rejected with the error ER_MTS_EVENT_BIGGER_PENDING_JOBS_SIZE_MAX, even if the packet was smaller than the limit set by slave_max_allowed_packet.

    With this fix, slave_pending_jobs_size_max becomes a soft limit rather than a hard limit. If the size of a packet exceeds slave_pending_jobs_size_max but is less than slave_max_allowed_packet, the transaction is held until all the slave workers have empty queues, and then processed. All subsequent transactions are held until the large transaction has been completed. The queue size for slave workers can therefore be limited while still allowing occasional larger transactions. (Bug #21280753, Bug #77406)

  • Replication: An incident event that broke replication was not written to the binary log with a GTID, so that it was not possible to skip the event using SET gtid_next=value. Instead, it was necessary to set the relay log file and relay log positions directly; this meant that, when autopositioning was enabled, it was necessary first to disable it, then to set the relay log file and position, and finally to re-enable autopositioning.

    Now in such cases MySQL writes the incident event into the statement cache, so that a GTID is generated and written for it prior to flushing, and that the slave applier works with the change. Then users can skip the event using the SQL statement SET gtid_next=value, followed by BEGIN and COMMIT. (Bug #19594845)

  • Replication: In certain cases, the master could write to the binary log a last_committed value which was smaller than it should have been. This could cause the slave to execute in parallel transactions which should not have been, leading to inconsistencies or other errors. (Bug #84471, Bug #25379659)

  • Replication: When using group_replication_ip_whitelist=AUTOMATIC, IPs in the private network are permitted automatically, but some class C IP addresses were not being permitted correctly. (Bug #84329, Bug #25503458)

  • Replication: When an existing GTID_NEXT transaction was assigned a conflicting GTID by the server, Group Replication generated an assert upon detecting two transactions with same GTID. This was because Group Replication generates the GTID after conflict detection, which is later than with master/slave replication. The fix relaxes some conditions to only be called when commit is done and a message has been added to alert you when a GTID has already been used. (Bug #84153, Bug #25232042)

  • Replication: The replication applier thread returns Error 3002 ER_INCONSISTENT_ERROR when there is a difference between an expected error number and the actual error number. It is now possible to ignore this error by using 3002 with slave_skip_errors. (Bug #83186, Bug #24753281)

  • Replication: MySQL lost its GTID position following a restart when a dump from mysqldump had been used to load data.

    To keep this problem from occurring, the mysql.gtid_executed table is now excluded automatically from dumps made by mysqldump. (Bug #82848, Bug #24590891)

  • Replication: Corruption of relay logs for one channel in multi-source replication caused good channels not to be initalized during a server restart. In addition, when run with --skip-slave-start=false, the server also failed to start slave threads for those channels which were in good condition, despite the fact that it should have started the slave threads for all good channels.

    Now, regardless of any errors on other channels, the server attempts to create and initialize channels that are in good condition, and starts slave threads for the good channels if --skip-slave-start is disabled. As part of this fix, START SLAVE and STOP SLAVE, which are intended to operate on all channels, are also modified such that they continue executing on all good channels even if they find bad channels among them. (Bug #82209, Bug #24285104)

  • Replication: The SQL thread was unable to GTID skip a partial transaction. (Bug #81119, Bug #25800025)

  • Debian client packages were missing information about conflicts with akonadi-backend-mysql packages. (Bug #26002288)

  • mysqldump could write database names in USE statements incorrectly. (Bug #25998635)

  • If the mysql_stmt_close() C API function was called, it freed memory that later could be accessed if mysql_stmt_error(), mysql_stmt_errno(), or mysql_stmt_sqlstate() was called. To obtain error information after a call to mysql_stmt_close(), call mysql_error(), mysql_errno(), or mysql_sqlstate() instead. (Bug #25988681)

  • Queries could be cached incorrectly, leading to incorrect query results, under these circumstances: InnoDB table; rows are being inserted but have not yet been committed; a query uses the table as a base table in a derived table; the optimizer chooses to materialize the derived table. (Bug #25943038, Bug #86047)

  • Man pages for a few utilities were missing from Debian/Ubuntu packages. (Bug #25811814)

  • The field-t unit test failed to run with AddressSanitizer enabled. Thanks to Laurynas Biveinis for the patch. (Bug #25803823, Bug #85678)

  • Debian client packages were missing information about conflicts with native packages. (Bug #25799475)

  • The Perl path in #! lines at the beginning of Perl scripts has been adjusted to /usr/local/bin/perl for FreeBSD 11. (Bug #25719975)

  • For debug builds, the assertion added for Bug#59686 was too strict and could be raised when it should not have been. (Bug #25685958)

    References: See also: Bug #59686.

  • A server exit occurred for downgrades to MySQL 5.7 from a MySQL 8.0 installation for which the optimizer cost tables contained generated columns. (Bug #25650399)

  • The server exited abnormally attempting to access invalid memory. (Bug #25501659)

  • With mysqld secured by TCP wrappers and the hosts.allow and hosts.deny files configured to restrict access from an IP address, connection attempts from that address resulted in too many messages to the error log. (Bug #25476479, Bug #84708)

  • mysqlpump no longer includes the slave_master_info and slave_relay_log_info tables in dumps of the mysql system database. Restoring a dump file containing these tables caused problems by changing the replication state improperly. (Bug #25469190)

  • Changes made by calling mysql_options() to set MYSQL_OPT_SSL_MODE could be affected by later mysql_options() calls. Now setting MYSQL_OPT_SSL_MODE is unaffected by later mysql_options() calls. (Bug #25452210)

  • A race condition could occur for CREATE TABLE statements with DATA DIRECTORY or INDEX DIRECTORY clauses. (Bug #25451091)

  • MySQL compilation in different directories produced different builds to leakage of absolute paths into debug information and __FILE__. (Bug #25436469, Bug #84608)

  • On SELinux in Enforcing mode, the keyring_okv plugin failed to connect to the Oracle Key Vault server. (Bug #25420001)

  • ALTER TABLE ... MODIFY on a DATETIME NOT NULL column using an AFTER clause resulted in an ER_INVALID_USE_OF_NULL error. (Bug #25385334)

  • The range optimizer could create an incorrect query tree, resulting in a server exit. (Bug #25369742, Bug #25586531)

  • mysqld_failed to start the server if the --datadir option was specified with a relative path name. (Bug #25364806)

  • XA PREPARE, XA ROLLBACK, and XA COMMIT for a transaction from a disconnected session did not take a global commit lock and modified the binary log and InnoDB redo log even when FLUSH TABLES WITH READ LOCK was in effect. This could lead to inconsistent backups when backup tools assumed that the server was in a read-only state. (Bug #25364178, Bug #84442)

  • GROUP_CONCAT(DISTINCT) returned nonunique values if the data size was greater than the value of the tmp_table_size system variable. (Bug #25331425, Bug #84320)

  • The fix for Bug #78777 had different effects depending on whether the Performance Schema is enabled. (Bug #25309017, Bug #84305)

    References: This issue is a regression of: Bug #78777.

  • An aggregate function in some nested queries could cause a server exit. (Bug #25303711)

  • Virtual generated column expressions that used the CONV() or HEX() functions could cause problems if the connection character set was changed. In this context, the table character set is now used for these functions regardless of connection character set. (Bug #25287633)

  • The Rewriter plugin did not perform locking properly if the read_only system variable was enabled. (Bug #25264253)

  • With read_only enabled, creation of non-TEMPORARY tables by non-SUPER users was permitted under certain conditions. (Bug #25250768)

  • For a table having a TIMESTAMP or DATETIME column having a default of CURRENT_TIMESTAMP, the column could be initialized to 0000-00-00 00:00:00' if the table had a BEFORE INSERT trigger. (Bug #25209512, Bug #84077)

  • Certain stored functions, if used in a query WHERE clause, could be handled using Index Condition Pushdown (which should not happen), resulting in a server exit. (Bug #25196653, Bug #25174454)

  • Certain queries against InnoDB tables that used a primary key and a subquery could return incorrect results if the index_merge_intersection flag of the optimizer_switch system variable was enabled. (Bug #24829050, Bug #79675)

  • Initialization of the keyring_okv plugin failed if the STANDBY_SERVER setting was missing from the okvclient.ora configuration file, effectively making this a mandatory setting. STANDBY_SERVER is now optional. (Bug #24816271)

  • On x86 machines, the uint3korr() macro read 4 bytes of data instead of the intended 3 bytes. (Bug #24807826, Bug #83264)

  • An assertion was raised during a fetch operation by the memcached plugin. (Bug #24605783)

  • Queries that contained UNION in a subquery and GROUP BY could return incorrect results. (Bug #24595639)

  • The server could dereference a null pointer when a deterministic function returning LONGTEXT was used in a subquery. (Bug #24595581)

  • Incorrect behavior could occur for INSERT statements executed in stored-program or prepared-statement context, if the VALUES part of an ON DUPLICATE KEY UPDATE clause referred to a BLOB value in the INSERT column list. (Bug #24538207, Bug #25361251, Bug #25530880, Bug #25684790)

  • systemd support scripts in Debian packages contained hardcoded references to the data directory, making it difficult to change the data directory using --datadir. (Bug #24398446, Bug #82417)

  • MySQL failed to compile under macOS 10.10.5 using Clang. (Bug #24352163, Bug #82340)

  • If a REPLACE statement tried to update a row in a table containing a virtual generated column of type BLOB, subsequent DML statements could behave incorrectly. (Bug #23573575)

  • Some PROXY grants were not replicated to slaves, causing incorrect replication. (Bug #23289541, Bug #81424)

  • The help output from mysqlxtest has been improved. (Bug #23107137, Bug #81086)

  • Compiler flags were adjusted to eliminate numerous warnings that occurred when compiling the keyring_file plugin using Clang. (Bug #22834591, Bug #80524)

  • If enabling the Event Scheduler caused an event defined as ON COMPLETION NOT PRESERVE to be dropped because its execution time had passed, the drop event was not written to the binary log, causing slaves not to replicate it and replication failure if an event of the same name was created later. (Bug #22150112)

  • LOAD XML INFILE performance became noticeably slower when the XML file being read contained a great many spaces, such as those introduced by indenting or pretty-printing. Now all leading whitespace is trimmed from each such value before reading it into memory. (Bug #16212207)