The problem is that DROP TABLE and other DDL statements failed to
automatically close handlers associated with tables that were marked
for reopen (FLUSH TABLES).
The current implementation fails to properly discard handlers of
dropped tables (that were marked for reopen) because it searches
on the open handler tables list and using the current alias of the
table being dropped. The problem is that it must not use the open
handler tables list to search because the table might have been
closed (marked for reopen) by a flush tables command and also it
must not use the current table alias at all since multiple different
aliases may be associated with a single table. This is specially
visible when a user has two open handlers (using alias) of a same
table and a flush tables command is issued before the table is
dropped (see test case). Scanning the handler table list is also
useless for dropping handlers associated with temporary tables,
because temporary tables are not kept in the THD::handler_tables
list.
The solution is to simple scan the handlers hash table searching
for, and deleting all handlers with matching table names if the
reopen flag is not passed to the flush function, indicating that
the handlers should be deleted. All matching handlers are deleted
even if the associated the table is not open.
different error code depending on platform.
On Mac OS X, KILL statement issued to kill the current
connection would return a different error code and message than on
other platforms ('MySQL server has gone away' instead of 'Shutdown
in progress').
The reason for this difference was that on Mac OS X we have macro
SIGNAL_WITH_VIO_CLOSE defined. This macro forces KILL
implementation to close the communication socket of the thread
that is being killed. SIGNAL_WITH_VIO_CLOSE macro is defined on
platforms where just sending a signal is not a reliable mechanism
to interrupt the thread from sleeping on a blocking system call.
In a nutshell, closing the socket is a hack to work around an
operating system bug and awake the blocked thread no matter what.
However, if the thread that is being killed is the same
thread that issued KILL statement, closing the socket leads to a
prematurely lost connection. At the same time it is not necessary
to close the socket in this case, since the thread in question
is not inside a blocking system call.
The fix, therefore, is to not close the socket if the thread that
is being killed is the same that issued KILL statement, even with
defined SIGNAL_WITH_VIO_CLOSE.
if running as root
Every start of a server in the test suite raised that warning.
The cause was an unconditionla add of the --user option to the
server command line. Only the "root" user (effective user id == 0)
must use that option.
Added check for effective user id == 0 before adding --user.
Thanks to Magnus Svensson for the patch.
wait_condition timeout.
The problem was that the event thread didn't manage to execute
the event in 30 seconds on highly-loaded box. The fix is to
increase timeout.
This is a fix for the test suite.
Merge fix
partition_mgm did not require have_symlink.
Moved the test case to partition_symlink, which
require have_symlink, and should work on both *nix and
Windows
in stored procedure.
The problem was that MySQL used unnecessarily large amounts of
memory if user variables were used as an argument to CONCAT or
CONCAT_WS -- 16M per each user variable used.
Technically, it happened because MySQL used the following
allocation strategy for string functions to avoid multiple
realloc() calls: in the virtual operation fix_length_and_dec()
the attribute max_length was calculated as a sum of max_length
values for each argument.
Although this approach worked well for small (or fixed) data types,
there could be a problem if there as a user variable among
the arguments of a string function -- max_length of the function
would be 16M (as the max_length of a user variable is 16M).
Both CONCAT() and CONCAT_WS() functions suffer from this problem.
The fix is to do not use meta-data for allocating memory.
The following strategy is proposed instead: allocate the exact
length of the result string at the first record, double the amount
of memory allocated when it is required.
No test case for this bug because there is no way to test memory
consumption in a robust way with our test suite.
partitioned table
Post-pushbuild fix
Pushbuild detected a new need for lex initialization in
embedded server.
Fixed test for INSERT DELAYED in partitions_hash.test so that
it works with embedded server.
Problem: there was no standard syntax error when
creating partitions with syntax error in
the partitioning clause.
Solution: added "Syntax error: " to the error message
Problem: the table's INDEX and DATA DIR was taken
directly from the table's first partition.
This allowed rename attack similar to
bug#32111 when ALTER TABLE REMOVE PARTITIONING
Solution: Silently ignore the INDEX/DATA DIR
for the table. (Like some other storage engines
do).
Partitioned tables do not support DATA/INDEX
DIR on the table level, only on its partitions.
the wrong buffer
handler::index_next_same() did not take into account that the
internally called function key_cmp_if_same() uses the fixed
buffer table->record[0] for key comparison instead of the
buffer provided by the caller of handler::index_next_same().
Added code to temporarily redirect table->record[0] and the fields
used for the key to the record buffer provided by the caller of
handler::index_next_same().
The test case is in partition.test already.
RENAME TABLE against a table with DATA/INDEX DIRECTORY overwrites
the file to which the symlink points.
This is security issue, because it is possible to create a table with
some name in some non-system database and set DATA/INDEX DIRECTORY
to mysql system database. Renaming this table to one of mysql system
tables (e.g. user, host) would overwrite the system table.
Return an error when the file to which the symlink points exist.
