asserts.
The patch for Bug#21074 replaces acquisition of the global LOCK_open lock
with exclusive locks on table names in such operations ad DROP TABLE
and RENAME TABLE.
Unfortunately, NDB internally assumes that LOCK_open is acquired and
tries to release it.
This dependency should be fixed by a separate (and significant in size)
patch. For now we just satisfy it - after all, the original
goal of the patch for Bug#21074 was to move query_cache_invalidate
outside of the scope of LOCK_open, and we still can do that.
This fixes some failing NDB tests in the runtime tree.
results.
When executing a CREATE EVENT statement with ON COMPLETION NOT PRESERVE
clause (explicit or implicit) and completion date in the past, we do not
create the event. Or, put it differently, we create it and then drop
immediately.
A warning is issued in this case, not an error -- we want to load
successfully old database dumps, and such dumps may contain events
that are no longer valid.
Update the warning text to not imply an erroneous condition.
pseudo_thread_id was reset to zero via mysql_change_user() handling
whereas there is no reason to do that. Moreover, having two
concurrent threads that change user and create a namesake temp tables
leads to recording the dup pair of queries:
set @@session.pseudo_thread_id = 0;
CREATE temporary table `the namesake`;
which will stall the slave as the second instance can not be created.
And that is the bug case.
Fixed by correcting pseudo_thread_id value after mysql_change_user().
we now have exclusive name lock on the table name in mysql_rm_table_part2,
we still should keep LOCK_open - some storage engines are not
ready for locking scope change and assume that LOCK_open is kept.
Still, the binary logging and query cache invalidation calls
moved out of LOCK_open scope.
Fixes some of the broken 5.1-runtime tests (tests break on asserts).
incomplete in 5.0 (and review fixes).
When in 5.0.13 I introduced class Prepared_statement and methods
::prepare and ::execute, general logging was left out of this class.
This was good for stored procedures, since in stored procedures
we do not log sub-statements, but introduced a regression in case of SQL
syntax for prepared statements, as previously we would log the actual
statements to the log, and after the change we would log only
COM_QUERY text.
Restore the old behavior, but still suppress logging if inside a stored
procedure.
Based on a community contributed patch from Vladimir Shebordaev.
No test case since we do not have a mechanism to test output
of the general log.
causes full table lock on innodb table.
Also fixes Bug#28502 Triggers that update another innodb table
will block on X lock unnecessarily (duplciate).
Code review fixes.
Both bugs' synopses are misleading: InnoDB table is
not X locked. The statements, however, cannot proceed concurrently,
but this happens due to lock conflicts for tables used in triggers,
not for the InnoDB table.
If a user had an InnoDB table, and two triggers, AFTER UPDATE and
AFTER INSERT, competing for different resources (e.g. two distinct
MyISAM tables), then these two triggers would not be able to execute
concurrently. Moreover, INSERTS/UPDATES of the InnoDB table would
not be able to run concurrently.
The problem had other side-effects (see respective bug reports).
This behavior was a consequence of a shortcoming of the pre-locking
algorithm, which would not distinguish between different DML operations
(e.g. INSERT and DELETE) and pre-lock all the tables
that are used by any trigger defined on the subject table.
The idea of the fix is to extend the pre-locking algorithm to keep track,
for each table, what DML operation it is used for and not
load triggers that are known to never be fired.
causing wrong error message in Falcon
An error message about a duplicate key could show a wrong key
value when not all columns of the key were used to select the
rows for update.
Some storage engines return a record with only the selected
columns filled.
This is fixed by re-reading the record with a read_set which
includes all columns of the duplicate key after a duplicate key
error happens and before the error message is printed.
A race condition in the integration between MyISAM and the query cache code
caused the query cache to fail to invalidate itself on concurrently inserted
data.
This patch fix this problem by using the existing handler interface which, upon
each statement cache attempt, compare the size of the table as viewed from the
cache writing thread and with any snap shot of the global table state. If the
two sizes are different the global table size is unknown and the current
statement can't be cached.
SHOW CREATE TABLE or SELECT FROM I_S.
This is the last patch for this bug, which depends on the big
CS patch and was pending.
The problem was that SHOW CREATE statements returned original
queries in the binary character set. That could cause the query
to be unreadable.
The fix is to use original character_set_client when sending
the original query to the client. In order to preserve the query
in mysqldump, 'binary' character set results should be set when
issuing SHOW CREATE statement. If either source or destination
character set is 'binary' , no conversion is performed.
The idea is that since the source character set is no longer
'binary', we fix the destination character set to still produce
valid dumps.
An assertion abort could occur for some grouping queries that employed
decimal user variables with assignments to them.
The problem appeared the constructors of the class Field_new_decimal
because the function my_decimal_length_to_precision did not guarantee
returning decimal precision not greater than DECIMAL_MAX_PRECISION.
The cast operation ignored the cases when the precision and/or the scale exceeded
the limits, 65 and 30 respectively. No errors were reported in these cases.
For some queries this may lead to an assertion abort.
Fixed by throwing errors for such cases.
For GCov builds, if the server crashes, the normal exit handler for writing
coverage information is not executed due to the abnormal termination.
Fix this by explicitly calling the __gcov_flush function in our crash handler.
The need arose when working on Bug 26141, where it became
necessary to replace TABLE_LIST with its forward declaration in a few
headers, and this involved a lot of s/TABLE_LIST/st_table_list/.
Although other workarounds exist, this patch is in line
with our general strategy of moving away from typedef-ed names.
Sometime in future we might also rename TABLE_LIST to follow the
coding style, but this is a huge change.
