The problem was that SHOW CREATE EVENT released all metadata locks
held by the current transaction. This made any exisiting savepoints
invalid, triggering the assert when ROLLBACK TO SAVEPOINT later
was executed.
This patch fixes the problem by making sure SHOW CREATE EVENT only
releases metadata locks acquired by the statement itself.
Test case added to event_trans.test.
The event scheduler was not designed to work in embedded mode. This
patch disables and excludes the event scheduler when the server is
compiled for embedded build.
libmysqld/Makefile.am:
Reduce the amount of event code in an embedded build.
mysql-test/t/events_trans.test:
Disable test if run in embedded mode.
sql/Makefile.am:
Introduce definition HAVE_EVENT_SCHEDULER and one new source file.
sql/event_data_objects.cc:
Refactor Event_parse_data to new file.
sql/event_data_objects.h:
Refactor Event_parse_data to new file.
Move global definitions to new file.
sql/event_queue.cc:
Move all parsed items to Event_parse_data for easier modularization.
sql/events.cc:
Move all parsed items to Event_parse_data for easier modularization.
sql/mysqld.cc:
Disable the event schedular subsystem if the server is compiled in
embedded mode.
sql/set_var.cc:
Disable the event schedular subsystem if the server is compiled in
embedded mode.
sql/set_var.h:
Disable the event schedular subsystem if the server is compiled in
embedded mode.
sql/sql_db.cc:
Disable the event schedular subsystem if the server is compiled in
embedded mode.
sql/sql_parse.cc:
Disable the event schedular subsystem if the server is compiled in
embedded mode.
sql/sql_show.cc:
Disable the event schedular subsystem if the server is compiled in
embedded mode.
sql/sql_test.cc:
Disable the event schedular subsystem if the server is compiled in
embedded mode.
sql/sql_yacc.yy:
Only include event-code needed for parsing to reduce impact on embedded
build.
Move all constants to Event_parse_data class.
mysql-test/r/events_embedded.result:
Add test case to make sure the 'event_scheduler' can't be activated
in embedded mode.
mysql-test/r/is_embedded.require:
Add test case to make sure the 'event_scheduler' can't be activated
in embedded mode.
mysql-test/t/events_embedded.test:
Add test case to make sure the 'event_scheduler' can't be activated
in embedded mode.
sql/event_parse_data.cc:
New file. Extracted Event_parse data into a new file.
sql/event_parse_data.h:
New file. Extracted Event_parse data into a new file.
bug#26338 events_bugs.test fail on Debian
and
bug#28285 Test "events_bugs" has instable results of "select /*1*/ ... from processlist"
mysql-test/r/events_bugs.result:
uppercase
mysql-test/t/events.test:
wait at the end of the script for event which haven't
finished their execution. This should solve
bug#26338 events_bugs.test fail on Debian
and
bug#28285 Test "events_bugs" has instable results of "select /*1*/ ... from processlist"
mysql-test/t/events_bugs.test:
wait at the end of the script for event which haven't
finished their execution. This should solve
bug#26338 events_bugs.test fail on Debian
and
bug#28285 Test "events_bugs" has instable results of "select /*1*/ ... from processlist"
mysql-test/t/events_grant.test:
wait at the end of the script for event which haven't
finished their execution. This should solve
bug#26338 events_bugs.test fail on Debian
and
bug#28285 Test "events_bugs" has instable results of "select /*1*/ ... from processlist"
mysql-test/t/events_logs_tests.test:
wait at the end of the script for event which haven't
finished their execution. This should solve
bug#26338 events_bugs.test fail on Debian
and
bug#28285 Test "events_bugs" has instable results of "select /*1*/ ... from processlist"
mysql-test/t/events_scheduling.test:
wait at the end of the script for event which haven't
finished their execution. This should solve
bug#26338 events_bugs.test fail on Debian
and
bug#28285 Test "events_bugs" has instable results of "select /*1*/ ... from processlist"
mysql-test/t/events_stress.test:
wait at the end of the script for event which haven't
finished their execution. This should solve
bug#26338 events_bugs.test fail on Debian
and
bug#28285 Test "events_bugs" has instable results of "select /*1*/ ... from processlist"
mysql-test/t/events_time_zone.test:
wait at the end of the script for event which haven't
finished their execution. This should solve
bug#26338 events_bugs.test fail on Debian
and
bug#28285 Test "events_bugs" has instable results of "select /*1*/ ... from processlist"
mysql-test/t/events_trans.test:
wait at the end of the script for event which haven't
finished their execution. This should solve
bug#26338 events_bugs.test fail on Debian
and
bug#28285 Test "events_bugs" has instable results of "select /*1*/ ... from processlist"
mysql-test/t/events_trans_notembedded.test:
wait at the end of the script for event which haven't
finished their execution. This should solve
bug#26338 events_bugs.test fail on Debian
and
bug#28285 Test "events_bugs" has instable results of "select /*1*/ ... from processlist"
mysql-test/r/events_trans.result:
result fixed
mysql-test/r/windows.result:
merging fix
mysql-test/t/events_trans.test:
moved to events_trans_notembedded.test
mysql-test/r/events_trans_notembedded.result:
New BitKeeper file ``mysql-test/r/events_trans_notembedded.result''
mysql-test/t/events_trans_notembedded.test:
New BitKeeper file ``mysql-test/t/events_trans_notembedded.test''
when there are no up-to-date system tables to support it:
- initialize the scheduler before reporting "Ready for connections".
This ensures that warnings, if any, are printed before "Ready for
connections", and this message is not mangled.
- do not abort the scheduler if there are no system tables
- check the tables once at start up, remember the status and disable
the scheduler if the tables are not up to date.
If one attempts to use the scheduler with bad tables,
issue an error message.
- clean up the behaviour of the module under LOCK TABLES and pre-locking
mode
- make sure implicit commit of Events DDL works as expected.
- add more tests
Collateral clean ups in the events code.
This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work
when mysql.event is damaged
mysql-test/r/events.result:
Update results.
mysql-test/r/events_bugs.result:
Update results.
mysql-test/r/events_restart_phase1.result:
Update results.
mysql-test/r/events_restart_phase2.result:
Update results.
mysql-test/r/events_restart_phase3.result:
Update results.
mysql-test/r/events_scheduling.result:
Update results.
mysql-test/r/events_time_zone.result:
Update results.
mysql-test/t/events.test:
Add new tests for tampering with mysql.event and some more
tests for sub-statements, LOCK TABLES mode and pre-locking.
mysql-test/t/events_bugs.test:
Move the non-concurrent part of test for Bug 16420 to this file.
mysql-test/t/events_restart_phase1.test:
Rewrite events_restart_* tests to take into account that now
we check mysql.event table only once, at server startup.
mysql-test/t/events_restart_phase2.test:
Rewrite events_restart_* tests to take into account that now
we check mysql.event table only once, at server startup.
mysql-test/t/events_restart_phase3.test:
Rewrite events_restart_* tests to take into account that now
we check mysql.event table only once, at server startup.
mysql-test/t/events_scheduling.test:
Add more coverage for event_scheduler global variable.
mysql-test/t/events_time_zone.test:
Move the non-concurrent part of the tests for Bug 16420 to
events_bugs.test
sql/event_data_objects.cc:
Move update_timing_fields functionality to Event_db_repository.
Make loading of events from a table record more robust to tampering
with the table - now we do not check mysql.event on every table open.
sql/event_data_objects.h:
Cleanup.
sql/event_db_repository.cc:
Now Event_db_repository is responsible for table I/O only.
All the logic of events DDL is handled outside, in Events class please
refer to the added test coverage to see how this change affected
the behavior of Event Scheduler.
Dependency on sp_head.h and sp.h removed.
Make this module robust to tweaks with mysql.event table.
Move check_system_tables from events.cc to this file
sql/event_db_repository.h:
Cleanup declarations (remove unused ones, change return type to bool
from int).
sql/event_queue.cc:
Update to adapt to the new start up scheme of the Event Scheduler.
sql/event_queue.h:
Cleanup declarations.
sql/event_scheduler.cc:
Make all the error messages uniform:
[SEVERITY] Event Scheduler: [user][schema.event] message
Using append_identifier for error logging was an overkill - we may
need it only if the system character set may have NUL (null character)
as part of a valid identifier, this is currently never the case,
whereas additional quoting did not look nice in the log.
sql/event_scheduler.h:
Cleanup the headers.
sql/events.cc:
Use a different start up procedure of Event Scheduler:
- at start up, try to check the system tables first.
If they are not up-to-date, disable the scheduler.
- try to load all the active events. In case of a load error, abort
start up.
- do not parse an event on start up. Parsing only gives some information
about event validity, but far not all.
Consolidate the business logic of Events DDL in this module.
Now opt_event_scheduler may change after start up and thus is protected
by LOCK_event_metadata mutex.
sql/events.h:
Use all-static-data-members approach to implement Singleton pattern.
sql/mysqld.cc:
New invocation scheme of Events. Move some logic to events.cc.
Initialize the scheduler before reporting "Ready for connections".
sql/set_var.cc:
Clean up sys_var_thd_sql_mode::symbolic_mode_representation
to work with a LEX_STRING.
Move more logic related to @@events_scheduler global variable to Events
module.
sql/set_var.h:
Update declarations.
sql/share/errmsg.txt:
If someone tampered with mysql.event table after the server has
started we no longer give him/her a complete report what was actually
broken. Do not send the user to look at the error log in such case,
as there is nothing there (check_table_intact is not executed).
sql/sp_head.cc:
Update to a new declaration of
sys_var_thd_sql_mode::symbolic_mode_representation
sql/sql_db.cc:
New invocation scheme of Events module.
sql/sql_parse.cc:
Move more logic to Events module. Make sure that we are consistent
in the way access rights are checked for Events DDL: always
after committing the current transaction and checking the system tables.
sql/sql_show.cc:
Update to the new declarations of
sys_var_thd_sql_mode::symbolic_mode_representation
sql/sql_test.cc:
New invocation scheme of events.
sql/table.cc:
mysql.event is a system table.
Update check_table_intact to be concurrent, more verbose, and less smart.
sql/table.h:
Add a helper method.
mysql-test/r/events_trans.result:
New BitKeeper file ``mysql-test/r/events_trans.result''
mysql-test/t/events_trans.test:
New BitKeeper file ``mysql-test/t/events_trans.test'':
test cases for Event Scheduler that require a transactional
storage engine.
