BUG#54872 MBR: replication failure caused by using tmp table inside transaction
Changed criteria to classify a statement as unsafe in order to reduce the
number of spurious warnings. So a statement is classified as unsafe when
there is on-going transaction at any point of the execution if:
1. The mixed statement is about to update a transactional table and
a non-transactional table.
2. The mixed statement is about to update a temporary transactional
table and a non-transactional table.
3. The mixed statement is about to update a transactional table and
read from a non-transactional table.
4. The mixed statement is about to update a temporary transactional
table and read from a non-transactional table.
5. The mixed statement is about to update a non-transactional table
and read from a transactional table when the isolation level is
lower than repeatable read.
After updating a transactional table if:
6. The mixed statement is about to update a non-transactional table
and read from a temporary transactional table.
7. The mixed statement is about to update a non-transactional table
and read from a temporary transactional table.
8. The mixed statement is about to update a non-transactionala table
and read from a temporary non-transactional table.
9. The mixed statement is about to update a temporary non-transactional
table and update a non-transactional table.
10. The mixed statement is about to update a temporary non-transactional
table and read from a non-transactional table.
11. A statement is about to update a non-transactional table and the
option variables.binlog_direct_non_trans_update is OFF.
The reason for this is that locks acquired may not protected a concurrent
transaction of interfering in the current execution and by consequence in
the result. So the patch reduced the number of spurious unsafe warnings.
Besides we fixed a regression caused by BUG#51894, which makes temporary
tables to go into the trx-cache if there is an on-going transaction. In
MIXED mode, the patch for BUG#51894 ignores that the trx-cache may have
updates to temporary non-transactional tables that must be written to the
binary log while rolling back the transaction.
So we fix this problem by writing the content of the trx-cache to the
binary log while rolling back a transaction if a non-transactional
temporary table was updated and the binary logging format is MIXED.
The problem is that QUICK_SELECT_DESC behaviour depends
on used_key_parts value which can be bigger than selected
best_key_parts value if an engine supports clustered key.
But used_key_parts is overwritten with best_key_parts
value that prevents from correct selection of index
access method. The fix is to preserve used_key_parts
value for further use in QUICK_SELECT_DESC.
mysql-test/r/innodb_mysql.result:
test case
mysql-test/t/innodb_mysql.test:
test case
sql/sql_select.cc:
preserve used_key_parts value for further use in QUICK_SELECT_DESC
mysqld-debug.exe in 5.5.3 on windows
Fix:
- Do not rename PDB, install mysqld.pdb matching
mysqld-debug.exe into bin\debug subdirectory
- Stack tracing code will now additionally look in
debug subdirectory of the application directory
for debug symbols.
- Small cleanup in stacktracing code: link with
dbghelp rather than load functions dynamically
at runtime, since dbghelp.dll is always present.
- Install debug binaries with WiX
cmake/install_macros.cmake:
Add optional COMPONENT and PDB_DESTINATION
to INSTALL_DEBUG_TARGET
mysys/stacktrace.c:
If binary is build with DBUG, also look in debug subdirectory
of executable directory. Packaging will put some PDBs there
(e.g bin\mysqld-debug.exe will have corresponding pdb in
bin\debug)
Also some cleanup: do not load dbghelp dynamically, instead
link with it. dbghelp is present on all Windows starting with
XP.
packaging/WiX/CPackWixConfig.cmake:
Install debug binaries
sql/CMakeLists.txt:
Do not rename PDB for mysqld-debug.exe, install it in debug subdirectory
starting via mysqld_safe
Rather than hardcode the plugin directory, enhance mysql_config
to fix plugin path when running a relocated install, and use it
to provide the plugin directory to mysqld_safe.
automatic reconnect
A client with automatic reconnect enabled will see the error
message "Lost connection to MySQL server during query" if the
connection is lost between mysql_stmt_prepare() and
mysql_stmt_execute(). The mysql_stmt_errno() number, however,
is 0 -- not the corresponding value 2013.
This patch checks for the case where the prepared statement
has been pruned due to a connection loss (i.e., stmt->mysql
has been set to NULL) during a call to cli_advanced_command(),
and avoids changing the last_errno to the result of the last
reconnect attempt.
switching binlog format to ROW
BUG 52616 fixed the case which the user would switch from STMT to
ROW binlog format, but the server would silently ignore it. After
that fix thd->is_current_stmt_binlog_format_row() reports correct
value at logging time and events are logged in ROW (as expected)
instead of STMT as they were previously and wrongly logged.
However, the fix was only partially complete, because on
disconnect, at THD cleanup, the implicit logging of temporary
tables is conditionally performed. If the binlog_format==ROW and
thd->is_current_stmt_binlog_format_row() is true then DROPs are
not logged. Given that the user can switch from STMT to ROW, this
is wrong because the server cannot tell, just by relying on the
ROW binlog format, that the tables have been dropped before. This
is effectively similar to the MIXED scenario when a switch from
STMT to ROW is triggered.
We fix this by removing this condition from
close_temporary_tables.
mysql-test/extra/binlog_tests/drop_temp_table.test:
Added binlog test case.
mysql-test/suite/binlog/r/binlog_row_drop_tmp_tbl.result:
Result changes because:
- there is a missing drop on three temporary tables
- it now contains results for the test added
mysql-test/suite/binlog/r/binlog_row_mix_innodb_myisam.result:
Result now contains the implicit drop for the temporary table.
mysql-test/suite/binlog/r/binlog_stm_drop_tmp_tbl.result:
Result file changed because it now contains results for added
test case.
mysql-test/suite/rpl/r/rpl_drop_temp.result:
Result file changed because it now contains results for added
test case.
mysql-test/suite/rpl/t/rpl_drop_temp.test:
Added replication test case.
sql/sql_base.cc:
Removed the condition that would make the server to skip
logging implicit drops when ROW binary log format mode was
in use.
Additionally, deployed DBUG_ENTER/RETURN macros.
The default value of the myisam_max_extra_sort_file_size could be
higher than the maximum accepted value, leading to warnings upon
the server start.
The solution is to simply set the value to the maximum value in a
32-bit built (2147483647, one less than the current). This should
be harmless as the option is currently unused in 5.1.
include/myisam.h:
Remove now-unused macro.
sql/mysqld.cc:
Set max value to INT_MAX32.
The problem was that a user could supply supply data in chunks
via the COM_STMT_SEND_LONG_DATA command to prepared statement
parameter other than of type TEXT or BLOB. This posed a problem
since other parameter types aren't setup to handle long data,
which would lead to a crash when attempting to use the supplied
data.
Given that long data can be supplied at any stage of a prepared
statement, coupled with the fact that the type of a parameter
marker might change between consecutive executions, the solution
is to validate at execution time each parameter marker for which
a data stream was provided. If the parameter type is not TEXT or
BLOB (that is, if the type is not able to handle a data stream),
a error is returned.
sql/sql_prepare.cc:
Before converting the parameter data stream, check the type
compatibility.
tests/mysql_client_test.c:
Add test case.
This bug is a consequence of WL#5349, as the
default storage engine was changed.
The fix was to explicitly add an ENGINE
clause to a CREATE TABLE statement, to
ensure that we test case preservement on
MyISAM.
DROP USER
RENAME USER CURRENT_USER() ...
GRANT ... TO CURRENT_USER()
REVOKE ... FROM CURRENT_USER()
ALTER DEFINER = CURRENT_USER() EVENTbut, When these statements are binlogged, CURRENT_USER() just is binlogged
as 'CURRENT_USER()', it is not expanded to the real user name. When slave
executes the log event, 'CURRENT_USER()' is expand to the user of slave
SQL thread, but SQL thread's user name always NULL. This breaks the replication.
After this patch, session's user will be written into query log events
if these statements call CURREN_USER() or 'ALTER EVENT' does not assign a definer.
mysql-test/include/diff_tables.inc:
Expend its abilities.
Now it can diff not only in sessions of 'master' and 'slave', but
other sessions as well.
mysql-test/include/rpl_diff_tables.inc:
Diff the same table between master and slaves.
sql/log_event.cc:
session's user will be written into Query_log_event, if is_current_user_used() is TRUE.
On slave SQL thread, Only thd->variables.current_user is written into Query_log_event,
if it exists.
sql/sql_acl.cc:
On slave SQL thread, grantor should copy from thd->variables.current_user, if it exists
sql/sql_class.h:
On slave SQL thread, thd->variables.current_user is used to store the applying event's
invoker.
This deadlock happened if DROP DATABASE was blocked due to an open
HANDLER table from a different connection. While DROP DATABASE
is blocked, it holds the LOCK_mysql_create_db mutex. This results
in a deadlock if the connection with the open HANDLER table tries
to execute a CREATE/ALTER/DROP DATABASE statement as they all
try to acquire LOCK_mysql_create_db.
This patch makes this deadlock scenario very unlikely by closing and
marking for re-open all HANDLER tables for which there are pending
conflicing locks, before LOCK_mysql_create_db is acquired.
However, there is still a very slight possibility that a connection
could access one of these HANDLER tables between closing/marking for
re-open and the acquisition of LOCK_mysql_create_db.
This patch is for 5.1 only, a separate and complete fix will be
made for 5.5+.
Test case added to schema.test.
Bug#47633 - assert in ha_myisammrg::info during OPTIMIZE
The server crashed on an attempt to optimize a MERGE table with
non-existent child table.
mysql_admin_table() relied on the table to be successfully open
if a table object had been allocated.
Changed code to check return value of the open function before
calling a handler:: function on it.
mysql-test/r/merge.result:
Backport of revid:ingo.struewing@sun.com-20091223200354-r2uzbdkj2v6yv111
Bug#47633 - assert in ha_myisammrg::info during OPTIMIZE
Updated result file.
mysql-test/t/merge.test:
Backport of revid:ingo.struewing@sun.com-20091223200354-r2uzbdkj2v6yv111
Bug#47633 - assert in ha_myisammrg::info during OPTIMIZE
Changed tests to respect changed TEMPORARY MERGE locking (unrelated).
Changed tests to respect changed CREATE TABLE ... LIKE (unrelated).
Changed tests to respect that no new tables can be created
under LOCK TABLE (unrelated).
Added test for Bug#47633.
Changed error numbers to symbolic names.
Added test for child locking for ALTER under LOCK TABLE.
Since Bug 36171 is not pushed yet, not the whole patch has been backported.
mysys/my_delete.c:
Backport of revid:ingo.struewing@sun.com-20091223200354-r2uzbdkj2v6yv111
Bug#47633 - assert in ha_myisammrg::info during OPTIMIZE
Fixed error reporting.
Fixed indentation.
mysys/my_mmap.c:
Backport of revid:ingo.struewing@sun.com-20091223200354-r2uzbdkj2v6yv111
Bug#47633 - assert in ha_myisammrg::info during OPTIMIZE
Added DBUG.
sql/item_func.cc:
Backport of revid:ingo.struewing@sun.com-20091223200354-r2uzbdkj2v6yv111
Added Debug Sync point, required by merge_sync.test.
sql/sql_table.cc:
Backport of revid:ingo.struewing@sun.com-20091223200354-r2uzbdkj2v6yv111
Bug#47633 - assert in ha_myisammrg::info during OPTIMIZE
Do not call handler:: functions if the table was not opened
successfully.
Added Debug Sync point, required by merge_sync.test.
storage/myisam/mi_check.c:
Backport of revid:ingo.struewing@sun.com-20091223200354-r2uzbdkj2v6yv111
Bug#47633 - assert in ha_myisammrg::info during OPTIMIZE
Unmap memory before exchanging data files. Needed on Windows.
storage/myisammrg/ha_myisammrg.cc:
Backport of revid:ingo.struewing@sun.com-20091223200354-r2uzbdkj2v6yv111
Added Debug Sync point, required by merge_sync.test.
merge_sync.test will be introduced by a patch for Bug 36171,
which is not pushed yet.
returns nothing
When looking for table or database names inside INFORMATION_SCHEMA
we must convert the table and database names to lowercase (just as it's
done in the rest of the server) when lowercase_table_names is non-zero.
This will allow us to find the same tables that we would find if there
is no condition.
Fixed by converting to lower case when extracting the database and
table name conditions.
Test case added.
and clarifies the invariant in dict_table_get_on_id().
In Mar 2007 Marko observed a crash during recovery, the crash resulted from
an UNDO operation on a system table. His solution was to acquire an X lock on
the data dictionary, this in hindsight was an overkill. It is unclear what
caused the crash, current hypothesis is that it was a memory corruption.
The X lock results in performance issues by when undoing changes due to
rollback during normal operation on regular tables.
Why the change is safe:
======================
The InnoDB code has changed since the original X lock change was made. In the
new code we always lock the data dictionary in X mode during startup when
UNDOing operations on the system tables (this is a given). This ensures that
the crash Marko observed cannot happen as long as all transactions that update
the system tables follow the standard rules by setting the appropriate DICT_OP
flag when writing the log records when they make the changes.
If transactions violate the above mentioned rule then during recovery (at
startup) the rollback code (see trx0roll.c) will not acquire the X lock
and we will see the crash again. This will however be a different bug.
