[This is part 2 of the announcement] Bugs Fixed, cont.
* Replication: The group replication applier channel does not support DATABASE as the slave_parallel_type; when group replication is started, this is checked for explicitly, and handled correctly. However, it remained possible to change this value indirectly at a later point in time by increasing the value of slave_parallel_workers while the slave SQL thread was stopped, which caused the applier to fail with an error. To fix this problem, the slave_parallel_type for the group_replication_applier is now checked to make sure that it is set to LOGICAL_CLOCK whenever the number of slave_parallel_workers is set greater then 0, and not merely when group replication is first started. (Bug #21798804) * Replication: As binlog_error_action=ABORT_SERVER is the default in MySQL 5.7.7 and later it is being used for more error situations. The behavior has been adjusted to generate a core dump to improve troubleshooting possibilities. (Bug #21486161, Bug #77738) * Replication: At runtime, some Gtid_set objects could be instrumented with a performance schema mutex key equal to 0 (which is invalid), due to its use as the effective default value when the mutex key was not actually supplied. This allowed these objects to be created without a valid key, which led to further issues when using them. (Bug #21485997) * Replication: When running the server with gtid_mode=ON, a DELETE from a MEMORY table following a restart was not written to the binary log correctly. (Bug #21045848) * Replication: The locking behavior of replication administration statements has changed to make SHOW SLAVE STATUS more concurrent. This makes the NONBLOCKING clause redundant for SHOW SLAVE STATUS and it has been removed. (Bug #20593028) * Replication: ER_CANT_USE_AUTO_POSITION_WITH_GTID_MODE_OFF errors were not reported using the correct format. (Bug #20545943) * Replication: When the dump thread was killed while dumping an inactive binary log, some events in this log could be skipped and thus not replicated. (Bug #78337, Bug #21816399) References: See also Bug #74607, Bug #19975697. * Replication: XA transactions could cause an assert condition on XA COMMIT; this was happening because the internal transaction state was not reset between XA PREPARE and XA COMMIT or XA ROLLBACK, due to the fact that these operations constitute separate transactions under XA. In addition, XA ROLLBACK statements were not handled properly in some cases. (Bug #78264, Bug #21755890) * Replication: The interface between the Group Replication plugin and the Performance Schema engine made use of a type of memory allocation which was passed to the server, and was a potential source of problems when passing information between the plugin and performance_schema tables. The implementation for this interface has been reworked so as to avoid performing this type of memory allocation when sharing data. (Bug #78263, Bug #21755699) * Replication: The MTS submode set for each channel was ignored by the worker threads, which continued to read and use the global flag set for all slave channels. This could lead to errors when the coordinator was of one type and its workers of another. (Bug #77763, Bug #21464737) * Replication: Replication slaves could fail for having insufficient privileges when they had been granted only the REPLICATION SLAVE privilege. (Bug #77732, Bug #21455603) * Replication: The status variable Slave_open_temp_tables keeps track of the number of temporary tables that are opened by the replication slave. If multi-source replication is enabled, it is the total number of temporary tables for all channels. This fix addresses the following issues relating to this variable: + RESET SLAVE FOR CHANNEL channel forced the value of Slave_open_temp_tables to 0; in the event that some other replication channel had open temporary tables which were later dropped, the value wrapped around to a large negative value (1 - 2^32). This also caused spurious or missed warnings when issuing a STOP SLAVE or CHANGE MASTER TO statement. + The internal function that modifies Slave_open_temp_tables in such cases relied on two incorrect assumptions: 1. That the variable is updated by only one thread when multi-threaded slaves are not enabled, which is not true in the case of multi-source replication. That non-atomic operations are safe with a single writer and multiple readers, which is not necessarily true for some platforms supported by MySQL. (Bug #77585, Bug #21357008) * Replication: The warning '@@session.gtid_executed' is deprecated and will be removed in a future release. was printed even when the session variable gtid_executed was not included in the result of a query. In addition, the result of SELECT @@session.gtid_executed included a duplicate warning. Both issues occurred because the warning was printed whenever the value of gtid_executed was accessed by a statement, such access occurring as a matter of course, whether or not a given variable is actually included in the result. To fix this issue, we make handling of @@session.gtid_executed consistent with how the also-deprecated variable @@global.sql_log_bin is treated in such cases, by making the following changes: + gtid_executed is no longer included in the performance_schema.session_variables table. + gtid_executed is still included in the information_schema.session_variables table, but when show_compatibility_56 = ON, the warning is not issued when querying the session_variables table, or when issuing SHOW VARIABLES or SHOW SESSION VARIABLES, even when using a matching LIKE clause with either of the SHOW statements. The warning is still issued by a statement such as SELECT @@session.gtid_executed which accesses the value of the variable directly. (Bug #77574, Bug #21354712) References: See also Bug #75980, Bug #20575529, Bug #76626, Bug #20854952. * Replication: When a transaction consisting of a single statement with a specified GTID failed in autocommit mode, its GTID was not released when rolling it back when binary logging was disabled. (Bug #77521, Bug #21338147) * Replication: The slave group event parser did not properly register an XA_ROLLBACK event as a transaction boundary. (Bug #77392, Bug #21273010) References: See also Bug #20920851. * Replication: mysqlbinlog printed a ROLLBACK statement at the end of the binary log file, which when played back failed with error 1782 @@SESSION.GTID_NEXT cannot be set to ANONYMOUS when @@GLOBAL.GTID_MODE = ON. This occurred when the binary log file did not include any data related events, or when the relay log file included a Format_description_log_event that had been generated on the master at server startup. The fix for this issue causes a relay log's Format_description_log_event to do nothing if it is applied by a BINLOG statement, and stops a ROLLBACK from setting gtid_next to ANONYMOUS when the state of gtid_next has not yet been determined by a subsequent event. (Bug #76887, Bug #20980932) * Replication: SAVEPOINT and ROLLBACK TO SAVEPOINT within a trigger led to an assertion. (Bug #76727, Bug #20901025) * Replication: While a SHOW BINLOG EVENTS statement was executing, any parallel transaction was blocked. The fix ensures that the SHOW BINLOG EVENTS process now only acquires a lock for the duration of calculating the file's end position, therefore parallel transactions are not blocked for long durations. (Bug #76618, Bug #20928790) * Replication: If a CREATE VIEW statement failed, it was being incorrectly written to the binary log even though it did not result in the creation of a partial view. The fix ensures that such statements are not recorded in the binary log. Additionally it was found that when a statement which had failed on a master was received by a slave with an expected error, if the statement was skipped on the slave, for example due to a replication filter, the expected error was being compared with the actual error that happened on the slave. The fix ensures that if a statement with an expected error is received by a slave, if the statement has not been filtered, only then is it compared with the actual error that happened on the slave. (Bug #76493, Bug #20797764) * Replication: The action specified for binlog_error_action was not always honored correctly after a hardware failure occurred during log rotation. (Bug #76379, Bug #20805298) * Replication: When using MySQL 5.7.6 and later with binlog_format=row and gtid_mode=off, if CREATE ... SELECT was killed during execution it could lead to an inconsistent state, breaking replication. The cause was that in MySQL 5.7.6 the way CREATE ... SELECT was logged was changed, so that a commit was introduced between the CREATE TABLE and SELECT steps. The fix ensures that CREATE ... SELECT does not commit in the middle of the transaction when binlog_format=row. (Bug #76320, Bug #77098, Bug #20742519, Bug #21114464) * Replication: Modifying the master_info_repository or relay_log_info_repository inside a transaction and later rolling back that transaction left the repository in an unusable state. We fix this by preventing any modification of these repositories inside a transaction. (Bug #74950, Bug #20074353) * Replication: Transactions added to gtid_executed using SET gtid_purged were not taken into account by WAIT_FOR_EXECUTED_GTID_SET() until a subsequent transaction was committed by a client or slave thread. (Bug #73838, Bug #19579811) * Replication: When the running with --relay-log-info-repository=TABLE, the mysql.slave_relay_log_info table is updated when a transaction is committed or when a flush is performed explicitly, such during relay log rotation. If a transaction that uses any nontransactional tables (such as MyISAM tables) is split across multiple relay logs, it is partially committed on relay log flush. When gtid_mode=ON, this caused the same GTID to be used for the remaining portion of the transaction, which raised an ER_GTID_NEXT_TYPE_UNDEFINED_GROUP error. We fix this issue by postponing in such cases the update of the relay log information repository that normally occurs on relay log rotation until the commit for the transaction in question has been executed. This issue did not affect tables using transactional storage engines such as InnoDB. (Bug #68525, Bug #16418100) * For an index-only scan over an indexed generated column, the server could do random calculations; the random results were not exposed to the user, but Valgrind warnings could occur, and the server could exit when calculations involved functions which did not expect such incorrect data. (Bug #21833760) * Although the use of JSON values with GREATEST() or LEAST() is not currently supported, the server did not handle attempts to do so correctly, leading to an assert (Linux) or exit() call (Windows) in debug builds. Now when you try to use JSON values with either of these functions, the server emits a suitable warning (ER_NOT_SUPPORTED_YET). (Bug #21828321) References: See also Bug #21383497. * For debug builds, a DROP TRIGGER statement could raise an assertion if the trigger was defined on a table that contained a generated column. (Bug #21824519) * A table that included a generated column referencing a JSON column in some cases become corrupted, so that a subsequent access of the table using a different connection caused the server to fail. (Bug #21808680) References: See also Bug #21824519, Bug #78408. * For tables with VIRTUAL generated columns, an INSERT with an empty values list could cause a server exit. (Bug #21807818) * CMake configuration was adjusted to handle new warnings reported by Clang 3.7. (Bug #21803314) * Using a materialized view defined over a table containing generated columns could cause a server exit. (Bug #21797776) * For partitioned InnoDB tables containing a virtual generated column, reads from the table could return random data for the column. (Bug #21779011) * The CMake checks for NUMA availability could cause compilation problems on platforms without NUMA support. (Bug #21774859) * The optimizer did not consider nonfunctional expressions such as (a AND b) = 1 when looking for indexed generated columns to substitute for the (a AND b) expression. Now expressions using the AND and OR logical operators are considered. (Bug #21770798) * For debug builds, when the optimizer tried to clone certain types of keys for a range optimization, an assertion was raised. (Bug #21761867) * For debug builds, the server could exit when the optimizer attempted to estimate the cost for processing unique values when there were no keys. (Bug #21697002) * An INSERT into a view with a subquery could fail if executed as a prepared statement. (Bug #21696206) * mysql-test-run.pl now has an --valgrind-clients option that causes all clients started by .test files to be run with valgrind. This option requires valgrind 3.9 or later. In addition, several client memory leak issues were fixed. (Bug #21672747) * For queries on InnoDB tables for which the optimizer used SPATIAL indexes for full index scans, the result was empty because such indexes do not support a full scan. The optimizer no longer considers SPATIAL indexes as candidates for full index scans. (Bug #21663612) * For some inputs, ST_Intersection() could return an invalid polygon. (Bug #21658453) * Added Microsoft Visual Studio 2015 support. Changes include using the native (added in VS 2015) timespec library if it exists, renamed lfind/lsearch and timezone/tzname to avoid redefinition problems, set TMPDIR to "" by default as P_tmpdir no longer exists, deprecated std::hash_map in favor of std::unordered_map, and added Wix Toolset 3.10 support. (Bug #21657078) * Spatial functions could simplify geometry values in contexts where the value might be used elsewhere in a query, producing incorrect results. (Bug #21652012) * If ST_ConvexHull() or ST_SRID() were used in a view definition, the resulting definition contained ST_Convex_Hull() (misspelled) or SRID() (deprecated). (Bug #21651588) * JSON_TYPE() returned OPAQUE for some binary values that it should have identified as BLOB. (Bug #21649073) * The mysql client parser incorrectly interpreted optimizer hint comments that contained ;, ", ', or ` characters. (Bug #21646026) * These Version Tokens issues were resolved: + version_tokens_delete() now strips whitespace surrounding token names in its argument, similar to version_tokens_set() and version_tokens_edit(). + Passing NULL to version_tokens_delete() caused a server exit. + Passing an argument with an empty token name to to version_tokens_set() or version_tokens_edit() caused a server exit. + Passing NULL as the timeout value to version_tokens_lock_exclusive() or version_tokens_lock_shared() caused a server exit. (Bug #21645001, Bug #21646106, Bug #21645944, Bug #21646017) * For debug builds, enabling the PAD_CHAR_TO_FULL_LENGTH SQL mode could cause SHOW FUNCTION STATUS to raise an assertion. (Bug #21632425) * mysqlpump did not exit with a message for some combinations of incompatible options. (Bug #21628662) * An assertion could be raised if the optimizer tried to create a temporary table based on a prepared statement parameter. (Bug #21625929) * Executing a prepared statement with multiple nested subqueries could raise an assertion. (Bug #21624851) * For debug builds, failure of subquery optimization could cause an assertion to be raised due to improper error handling. (Bug #21621313) * Some table and index optimizer hints were lost early in statement processing, so query rewrite plugins did not have access to them. This could cause incorrect matching between incoming statements and statement pattern templates. (Bug #21619780) * Queries containing nested subqueries combining grouping and outer references might cause a server exit. (Bug #21619634) * Passing NULL as the second or third argument to ST_AsGeoJSON() could cause the server to stop responding to the session or (in debug builds) to raise an assertion. Giving input to HANDLER READ that could not be converted to the correct type could cause the server to stop responding to the session or (in debug builds) to raise an assertion. (Bug #21616810, Bug #21650603) * For debug builds, ST_IsValid(NULL) could raise an assertion. (Bug #21616647) * For debug builds, an assertion could be raised for negative zero values when converting time values to decimal. (Bug #21616585) * ST_AsWKB() could cause a server exit if given invalid data. (Bug #21614368) * If an aggregate function was used over a generated column that was itself part of a multiple-column index, the server could exit. (Bug #21613615) * A missing error check during column reference resolution could result in an incorrect error message or (in debug builds) an assertion being raised. (Bug #21613422) * For debug builds, an assertion could be raised in Filesort::make_sortorder() for attempts to sort Item_ref objects. (Bug #21611270) * JSON functions could return incorrect values if a path argument was passed as a user-defined variable that changed values between result set rows. (Bug #21602361) * Columns selected from the right-hand table of a left join, which was also a derived table, might produce incorrect NULL value information when used in an IN subquery. (Bug #21574933) References: This bug was introduced by Bug #14358878. * For debug builds, an assertion was raised for some queries that have a semi-join and use the materialization strategy, if a key length or number of key parts was zero. (Bug #21566735) * Compilation using gcc 4.9 or 5.1 failed on ARM64 platforms. (Bug #21552524) * If a multiple-column UPDATE statement failed to update a JSON column that was then referenced in a later update, the server could exit. (Bug #21547877) * For debug builds, invalid geometry byte strings could cause spatial functions to raise an assertion rather than return an error to the caller. (Bug #21546656) * For debug builds, a too-strict assertion could be raised by invalid characters for LOAD DATA. (Bug #21542698) * The server could exit when InnoDB tried to update a secondary index on a VIRTUAL generated column of type BLOB. (Bug #21530366) * Binary logging of CREATE USER and statements could log the hash of the password hash (rather than the hash itself) when log_backward_compatible_user_definitions was enabled. Binary logging of ALTER USER statements could include attributes not present in the original statements. In consequence of the fix for these issues, log_backward_compatible_user_definitions has been replaced by log_builtin_as_identified_by_password. If this variable is enabled, binary logging for CREATE USER statements involving built-in authentication plugins rewrites the statements to include an IDENTIFIED BY PASSWORD clause, and SET PASSWORD statements are logged as SET PASSWORD statements, rather than being rewritten to ALTER USER statements. (Bug #21516392, Bug #20535561) * For debug builds, incorrect caching of JSON values could cause an assertion to be raised. (Bug #21491442) * An empty string (which is not a valid JSON value) normally is parsed and returned as a JSON null literal, but in some cases could raise an assertion for debug builds. (Bug #21487833) * For temporary tables created to handle UNION statements that selected CHAR or SET columns, the maximum column width could be too long for InnoDB to handle. Now such columns are created as variable-length columns. (Bug #21480999) * Certain JSON functions could return incorrect results when used in prepared statements which had path expression constants. (Bug #21472872) * Valgrind errors could occur during partition pruning for tables containing generated columns. (Bug #21469535) * For builds configured with MAX_INDEXES greater than 64, certain queries for which the server used temporary tables could cause a server exit. (Bug #21466850) * Adding or dropping a VIRTUAL generated column could cause a server exit. (Bug #21465626) * An attempt to use a previously unused time zone with CONVERT_TZ() could produce warnings or errors or (in debug builds) raise an assertion if GTIDs were enabled but the binary log was not enabled. In debug builds, an attempt to use a previously unused time zone with CONVERT_TZ() or as the value of the time_zone system variable inside a stored program could raise an assertion. (Bug #21459999, Bug #21459795) * For plugins that use the audit plugin API, MYSQL_AUDIT_GLOBAL_VARIABLE_SET events passed to the notification function did not include the new variable value. (Bug #21457699) * Queries on a table containing an indexed generated column could fail if the table name contained special characters. (Bug #21454155) * When a view was the inner table of an outer join, a JSON column could produce a non-NULL value when NULL was expected. (Bug #21448719) * If JSON_CONTAINS_PATH() was called with a one_or_all argument of all and a path argument contained a wildcard, the function found all matches per path, even though in this case one match is sufficient. (Bug #21442775) * JSON_SET() and JSON_REPLACE() sometimes produced an incorrect result if a path expression identified a nonarray value. (Bug #21442624) * Suppression of JSON conversion errors using non-strict SQL mode or INSERT IGNORE could then cause an assertion to be raised if an empty value inserted into a JSON NOT NULL column was copied to another JSON column. (Bug #21437989) * For debug builds, some spatial functions that accept raw byte data for spatial arguments (for example, specified as hex values) could raise an assertion if such an argument contained extra garbage following valid data. (Bug #21397107) * ALTER TABLE could raise an assertion for a table with an indexed virtual column having a column position greater than 64. (Bug #21391781) * Evaluation of virtual generated columns could fail to evaluate all base columns and result in invalid memory reads. (Bug #21390605) * For deeply nested JSON input, ST_GeomFromGeoJSON() or JSON_VALID() could produce stack overflow. (Bug #21389101, Bug #21377136) * Failure to parse a JSON string that contained a floating-point number with a large, negative exponent could cause a server exit. (Bug #21384048) * For debug builds, an incorrect assertion could be raised during subquery execution. (Bug #21383882) * For debug builds, a missing error check in Item_sum_hybrid::fix_fields() caused an assertion to be raised. (Bug #21383714) * For debug builds, invoking ST_AsGeoJSON() within GROUP BY ... WITH ROLLUP could raise an assertion. (Bug #21383497) * JSON_SEARCH() could return incorrect results if an invalid escape expression was specified. (Bug #21383284) * For debug builds, a NULL first argument to JSON_SET() could raise an assertion. (Bug #21381806) * For expressions of the form (subquery) IN (subquery), where a subquery could return a JSON value, failure to handle a row result could cause a server exit.` (Bug #21376088) * Failure of JSON_APPEND() to handle a legal condition could cause a server exit. (Bug #21373874) * Calls to ST_Buffer() could hang or raise an assertion. (Bug #21372946) * The server could exit in unclean fashion if configured to listen on a TCP/IP port number already in use by another server instance. (Bug #21368299) * Certain subqueries as arguments to PROCEDURE ANALYSE() could cause a server exit. (Bug #21350175) * A query with a right outer join inside a derived table might return wrong data. (Bug #21350125) * Starting the server with --skip-grant-tables (or with options such as --initialize for which --skip-grant-tables is implicit) prevented the INSTALL PLUGIN and UNINSTALL PLUGIN statements from working. (Bug #21335821) * mysql_ssl_rsa_setup could create an unwanted .rnd file in the data directory. (The file is actually created by openssl, which mysql_ssl_ras_setup invokes. mysql_ssl_rsa_setup now cleans up the file.) (Bug #21335818) * For spatial functions, input polygons were automatically closed if open. However, the Open Geospatial Consortium guidelines require that input polygons already be closed. Unclosed polygons are now rejected as invalid rather than being closed. (Bug #21327888) * When a VIRTUAL generated column was added to a table, it was not ensured that data being calculated by the generated column expression would not be out of range for the column. This could lead to inconsistent data being returned and unexpectedly failed statements. ALTER TABLE now supports WITHOUT VALIDATION and WITH VALIDATION clauses to control whether ALTER TABLE validates the data for a VIRTUAL generated column: + With WITHOUT VALIDATION (the default if neither clause is specified), an in-place operation is performed (if possible), data integrity is not checked, and the statement finishes more quickly. However, later reads from the table might report warnings or errors for the column if values are out of range. + With WITH VALIDATION, ALTER TABLE copies the table. If an out-of-range or any other error occurs, the statement fails. Because a table copy is performed, the statement takes longer. WITHOUT VALIDATION and WITH VALIDATION are permitted only with ADD COLUMN, CHANGE COLUMN, and MODIFY COLUMN operations. (Bug #21317507) * Some INFORMATION_SCHEMA queries consumed excessive memory due to suboptimal query plans and insufficient materialization. (Bug #21299665) * With the server configured to send error messages to syslog or a log file, messages generated prior to error log setup were sent to stderr or stdout. These messages are now buffered until error log setup has completed, then logged to the proper destination. (Bug #21296553) * Statement digests did not include information about optimizer hint comments, causing statements with and without hints to be aggregated. (Bug #21286261) * Executing a prepared statement using a derived table and an aggregate function in a subquery in the SELECT list could cause a server exit. (Bug #21277074) * GRANT created the account for nonexistent accounts even if the NO_AUTO_CREATE_USER SQL mode was enabled. (Bug #21271571) * ST_SymDifference() with multipolygon arguments could return incorrect results. (Bug #21263152) * A query with a NOT IN subquery that had COUNT(DISTINCT) could return incorrect results. (Bug #21243772) * When started using a very old data directory (from MySQL 5.0), the server could exit due to failure to properly read the old grant tables. (Bug #21216433) * A mulitple-table update involving generated columns that updated used a temporary table could cause a server exit or raise an assertion. (Bug #21216067) * Unit testing now uses Google Mock 1.7 rather than 1.6. (Bug #21215389) * For queries containing an expression of the form (x IN (subquery)) IN (subquery2), a combination of semi-join and subquery materialization strategies could cause a server exit. (Bug #21205577) * If a query contained an outer join such as LEFT JOIN (t1,t2,...) and a hint was used to disable join buffering on a right-side table but not on the others, a server exit occurred. (Bug #21205282) * For a cursor type of CURSOR_TYPE_READ_ONLY, retrieving the result set for the first execution of a prepared CALL staement could be missing the first result set row if the data was numeric; raise an assertion for debug builds if the data was string; cause loss of the server connection when calling mysql_stmt_fetch(). (Bug #21199582) * Dangling blob pointers could remain when closing an InnoDB table, resulting in a subsequent read of invalid memory and a server exit. (Bug #21153489) * For some operations where sorting or grouping required a temporary table, the table could have zero columns and raise an assertion. (Bug #21143151) * For debug builds, merging a derived table into an outer query block could raise an assertion. (Bug #21139722) * Queries containing an expression of the form (x IS NULL) IN (subquery) could cause a server exit. (Bug #21139402) * During server SSL file autogeneration, ca.pem briefly had insecure file permissions. (Bug #21138119) * An assertion could be raised due to incorrect error handling if a SELECT ... FOR UPDATE subquery resulted in deadlock and caused a rollback. (Bug #21096444) * Selecting the result of an INSERT() function call to which input was passed as a hexidecimal string could expose more information than was passed to the function. (Bug #21056907) * Subqueries having COUNT() with GROUP BY could yield incorrect results. (Bug #21055139) * The updatable property of a view is set during view creation. If the underlying table was dropped and re-created as a nonupdatable one, the updatable property of the original view was not revised accordingly. This could cause a server exit for attempts to insert or replace into the view is made. (This problem was specific to views with multiple tables/views and did not occur with update statements.) (Bug #21039264) * The locking functions provided by the version_token plugin were renamed: vtoken_get_read_locks(), vtoken_get_write_locks(), and vtoken_release_locks() are now named version_tokens_lock_shared(), version_tokens_lock_exclusive(), and version_tokens_unlock(), respectively. These functions also failed to have any effect because they were implicitly unlocked at the end of the statement in which they were set. (Bug #21034322, Bug #21280801) * Servers linked against yaSSL and compiled with GCC 4.8.2 could fail to respond correctly to connection attempts until several seconds after startup. yaSSL was upgraded to version 2.3.7e. (Bug #21025377) * When upgrading an old data directory (MySQL 5.0 or 5.1), mysql_upgrade could fail to properly read the mysql.proc table. (Bug #20968596) * For tables with subpartitions, the server could exit due to incorrect error handling during partition pruning if the partition could be identified but not the subpartition. (Bug #20909518) * mysql_upgrade could fail to look for checked tables in the wrong database during the repair phase. (Bug #20868496) * DELETE could check privileges for the wrong database when table aliases were used. (Bug #20777016) * mysqldump used incorrect syntax for generated column definitions. (Bug #20769542) * Within a trigger, use of a cursor that accessed OLD or NEW values from a row could cause a server exit. (Bug #20760261) * If a file was specified using an --init-file option, mysqld --initialize produced errors for statements in the file such as GRANT that affect user accounts. (Bug #20546898) * Failure during execution of an XA PREPARE statement could result in an invalid XA transaction state. Subsequent attempts to start another XA transaction led to an ER_XAER_OUTSIDE error. (Bug #20538956) * The audit log plugin could audit accounts named in the audit_log_exclude_accounts system variable. (Bug #20408206) * If a generated foreign key index was renamed by the same ALTER TABLE statement that added a new foreign key with the same name, the server could exit. (Bug #20146455) * ALTER TABLE operations that dropped and added the same FULLTEXT index were not performed as in-place (fast) operations that avoid using a temporary copy of the table. (Bug #20106837) * When the number of days calculated by DATE_FORMAT() function was negative, the server could exit. (Bug #19985318) * ALTER TABLE operations to add or modify columns could create geometry columns containing invalid data due to missing validation. (Bug #19880316) * If range optimization was attempted on an index with a string column as its first part and values used for comparison were fully truncated, comparisons would be incorrect and produce incorrect results. (Bug #19333852) References: This bug was introduced by Bug #16407965. * The server could hang due to incorrect cleanup of aggregate functions used in a query. (Bug #18979515) * The server could exit while checking for appropriate indexes to use for certain queries that used aggregate function in the WHERE clause. (Bug #18706592) * On Windows, the validate_password plugin could cause a server exit during the dictionary check. (Bug #18636874) * Invoking a stored program without qualifying it with the database name could lead to stored program compilation errors. (Bug #18599181) * EXPLAIN of statements containing GROUP_CONCAT() could cause a server exit. (Bug #17865675) * The value of the FOUND_ROWS() function that returns the number of rows found in the previous query could change during execution of the next query. Now the value of FOUND_ROWS() for the previous query remains constant during execution of the next query. (Bug #17846246) * Failure to check for error conditions could cause some updates or deletes to result in a server exit. (Bug #17763238) * On Windows, heap corruption in the audit log plugin caused server startup failure. (Bug #14700102) * If the UPDATE part of INSERT ... ON DUPLICATE KEY UPDATE swapped two column values, the server could read incorrect data and exit. (Bug #13901905) * mysqlpump generated incorrect ALTER TABLE statements for adding foreign keys. (Bug #78593, Bug #21907297) * Some stress test files in the mysql-test/suite/innodb_stress directory had the executable file mode set although they were not script files. (Bug #78403, Bug #21822413) * Subqueries that used a derived table and contained a set function referring to a column from that derived table might be aggregated in the wrong query block. (Bug #78250, Bug #21753180) * For some inputs, ST_Union() could return an invalid geometry collection. (Bug #78206, Bug #21689998) * On non-Windows systems, setting range_alloc_block_size or query_alloc_block_size to a value larger than 32 bits at startup could cause a server exit. The maximum value for these system variables (on all platforms) is now limited to 2^32 â 1, rounded down to the nearest multiple of 1024. (Bug #78188, Bug #21682231) * On Windows, if the MySQL server was started as a service, logging to the error log file was disabled. (Bug #77977, Bug #21574096) References: This bug was introduced by Bug #21328041. * In the Performance Schema setup_instruments table, it was possible to set memory instrument to TIMED='YES', although memory operations are never timed. Now such attempts are ignored. It was possible to set built-in memory instruments (with names of the form memory/performance_schema/%) to ENABLED='YES', although built-in memory instruments cannot be disabled. Now such attempts are ignored. (Bug #77944, Bug #21562212) * mysqldump and mysqlpump output included sys schema stored programs even when the sys schema was not dumped. (Bug #77926, Bug #21549860) * RPM installation scripts if configuration files contained multiple datadir lines. Now the last datadir line is used. (Bug #77878, Bug #21527467) * An unnecessary memset() call invoked during Performance Schema digest operations has been removed, which improves performance by reducing overhead. (Bug #77863, Bug #21528683) * A potential race condition for the safe mutex implementation was corrected. This implementation is enabled by default only for debug builds. (Bug #77862, Bug #21522888) * mysqld --initialize produced warnings about missing SSL files, which is unnecessary because initialization does not require SSL. (Bug #77825, Bug #21498544) * When mysqlpump was invoked with the --defer-table-indexes option, it could generate incorrect CREATE TABLE statements for tables with a foreign key and a primary key but not a secondary index. (Bug #77759, Bug #21462732) * The server initialization script used for the service mysql status command on Linux sometimes incorrectly reported that the server was stopped. (Bug #77696, Bug #21768876) * For statements of the form CREATE TABLE ... SELECT, where the table was defined to contain a DECIMAL UNSIGNED, calculation of the row size was incorrect, leading to incorrect values in the table. (Bug #77636, Bug #21383896) * A WHERE predicate containing both TRIM(LEADING ..) and TRIM(TRAILING ...) could be incorrectly optimized away. (Bug #77631, Bug #21447969) * For wait events, the Performance Schema uses the CYCLE timer by default, but failed to fall back to a different timer if CYCLE was unavailable. (Bug #77577, Bug #21374104) * A disk-full condition during execution of a CREATE TABLESPACE statement caused a server exit. (Bug #77556, Bug #21347001) * A privilege precheck for derived tables could fail and cause a server exit. (Bug #77525, Bug #21338077) * For some string functions, data was truncated when evaluated in subqueries due to incorrect space calculations when creating temporary tables to hold intermediate results. (Bug #77473, Bug #21317406) * A predicate of the form WHERE ROUND(X,Y) > 0, where X is a column name and Y is a program local variable, could return false when it should return true. (Bug #77391, Bug #21279005) * Updating VARCHAR and TEXT columns in the same UPDATE statement could produce incorrect results. When a VARCHAR column was assigned to a TEXT column and the VARCHAR column was then set to a different value, the TEXT column's result contained the VARCHAR column's new value. (Bug #77135, Bug #21143080) * A subquery in a HAVING clause that returned more than 1 row could cause a server exit. Additional to the bug fix, EXPLAIN now displays Zero limit rather than Impossible WHERE when optimizing a query with LIMIT 0. (Bug #76998, Bug #21067109) * If an INFORMATION_SCHEMA query that performed a table-open operation encountered a corrupt table and attempted to repair it, a deadlock could occur, resulting in an aborted transaction without an appropriate error being reported. Such queries now do not attempt table repair. (Bug #76912, Bug #21021848) * mysqladmin -u root -p could exit with a segmentation fault. (Bug #76538, Bug #20802751) * The optimizer sometimes generates an index for a derived table (subquery in the FROM clause). If this occurred for a statement executed within a stored program, a memory leak could occur. (Bug #76349, Bug #20728894) * Optimizer estimates for filtering conditions could lead to suboptimal execution plans if the expected number of rows selected from a table was between 0 and 1. The estimate is now made to be at least 1. (Bug #76314, Bug #20701585) * The optimizer could incorrectly assume an out-of-memory condition while optimizing a range scan for the OR operator, resulting in overestimation of the number of qualifying rows. (Bug #75248, Bug #20229614) * The Performance Schema events_statements_history table could have an ERRORS column value of 0 when other columns indicated there were errors. (Bug #74614, Bug #19929832) * View creation from a UNION failed with a duplicate-column error if a SELECT statement in the UNION other than the first used the same column name multiple times. (Bug #74539, Bug #19886430) * Timestamp values written to the slow query log could be incorrect. (Bug #73974, Bug #19646918) * When the mysql client was used to connect to the server in batch mode using an account with an expired password, the error message was not meaningful. mysql now reports "Please use --connect-expired-password option or invoke mysql in interactive mode" in this case. (Bug #72696, Bug #21464621) * For UPDATE statements with ORDER BY, the optimizer could perform an unnecessary filesort on a key that was used for scanning as well as being updated. (Bug #72518, Bug #18698556) * For a query with many range conditions, the optimizer could estimate that too much memory would be required for a range scan and fall back to a less efficient full table scan. A new range_optimizer_max_mem_size system variable now controls the limit on memory consumption for the range optimizer. A value of 0 means "no limit." If an execution plan considered by the optimizer uses the range access method but the optimizer estimates that the amount of memory needed for this method would exceed the limit, it abandons the plan and considers other plans. (Bug #70247, Bug #17413040, Bug #17769777) * Empty XML elements having the form <element/> were not handled correctly by the LOAD XML statement. (Bug #67542, Bug #16171518) * As the number of open MyISAM tables increased, lookups to check whether a table was open became expensive, particularly when the table was not open. Lookup performance has been improved, with the overhead reduction especially beneficial for selects on large number of tables with large values of table_open_cache and table_definition_cache. (Bug #49177, Bug #11757169) -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/mysql