only const tables
The problem was caused by two shortcuts in the optimizer that
are inapplicable in the ROLLUP case.
Normally in a case when only const tables are involved in a
query, DISTINCT clause can be safely optimized away since there
may be only one row produced by the join. Similarly, we don't
need to create a temporary table to resolve DISTINCT/GROUP
BY/ORDER BY. Both of these are inapplicable when the WITH
ROLLUP modifier is present.
Fixed by disabling the said optimizations for the WITH ROLLUP
case.
The SE API requires mysql to notify the storage engine that
it's going to read certain tables at the beginning of the
statement (by calling start_stmt(), store_lock() or
external_lock()).
These are typically called by the lock_tables().
However SHOW CREATE TABLE is not pre-locking the tables
because it's not expected to access the data at all.
But for some view definitions (that include comparing a
date/datetime/timestamp column to a string returning
scalar subquery) the JOIN::prepare may still access data
when materializing the scalar non-correlated subquery
in Arg_comparator::can_compare_as_dates().
Fixed by not materializing the subquery when the function
is called in a SHOW/EXPLAIN/CREATE VIEW
Bug#41756 "Strange error messages about locks from InnoDB".
In JT_EQ_REF (join_read_key()) access method,
don't try to unlock rows in the handler, unless certain that
a) they were locked
b) they are not used.
Unlocking of rows is done by the logic of the nested join loop,
and is unaware of the possible caching that the access method may
have. This could lead to double unlocking, when a row
was unlocked first after reading into the cache, and then
when taken from cache, as well as to unlocking of rows which
were actually used (but taken from cache).
Delegate part of the unlocking logic to the access method,
and in JT_EQ_REF count how many times a record was actually
used in the join. Unlock it only if it's usage count is 0.
Implemented review comments.
Bug#41756 "Strange error messages about locks from InnoDB".
In JT_EQ_REF (join_read_key()) access method,
don't try to unlock rows in the handler, unless certain that
a) they were locked
b) they are not used.
Unlocking of rows is done by the logic of the nested join loop,
and is unaware of the possible caching that the access method may
have. This could lead to double unlocking, when a row
was unlocked first after reading into the cache, and then
when taken from cache, as well as to unlocking of rows which
were actually used (but taken from cache).
Delegate part of the unlocking logic to the access method,
and in JT_EQ_REF count how many times a record was actually
used in the join. Unlock it only if it's usage count is 0.
Implemented review comments.
values return too many records
WHERE clauses with "outer_value_list NOT IN subselect" were
handled incorrectly if the outer value list contained multiple
items where at least one of these could be NULL. The first
outer record with NULL value was handled correctly, but if a
second record with NULL value existed, the optimizer would
choose to reuse the result it got on the last execution of the
subselect. This is incorrect if the outer value list has
multiple items.
The fix is to make Item_in_optimizer::val_int (in
item_cmpfunc.cc) reuse the result of the latest execution
for NULL values only if all values in the outer_value_list
are NULL.
When a sessione is closed, all temporary tables of the session are automatically
dropped and are binlogged. But it will be binlogged with wrong database names when
the length of the temporary tables' database names are greater than the
length of the current database name or the current database is not set.
Query_log_event's db_len is forgot to set when Query_log_event's db is set.
This patch wrote code to set db_len immediately after db has set.
ONLY_FULL_GROUP_BY
Problem was that during checking and preparation of the
partitioining function as a side effect in fix_fields
the full_group_by_flag was changed.
Solution was to set it back to its original value after
calling fix_fields.
Updated patch, to also exclude allow_sum_func from being
affected of fix_fields, as requested by reviewer.
- disabled main.innodb_bug47777.test with InnoDB plugin
until fix for plugin is applied.
- disabled main.innodb-autoinc.test (failing)
- re-enabled main.innodb_bug39438.test
- added error message suppression to innodb_bug39438, as
requested by InnoDB/Oracle
- reverted change to main.innodb_bug34300 as plugin specific.
The reason for the bug is that mysqtest as well as other client tools
running in test suite (mysqlbinlog, mysqldump) will first try to connect
whatever database has created shared memory with default base name
"MySQL" and use this. (Same effect could be seen on Unix if mtr would
not care to calculate "port" and "socket" parameter).
The fix ensures that all client tools and running in mtr use unique
per-database shared memory base parameters, so there is no possibility
to clash with already installed one. We use socket name for shared memory
base (it's known to be unique). This shared-memory-base is written to the
MTR config file to the [client] and [mysqld] sections. Fix made also made
sure all client tools understand and correctly handle --shared-memory-base.
Prior to this patch it was not the case for mysqltest, mysqlbinlog and
mysql_client_test.
All new connections done from mtr scripts via connect() will by default
set shared-memory-base. And finally, there is a possibility to force
shared memory or pipe connection and overwrite shared memory/pipe base name
from within mtr scripts via optional PIPE or SHM modifier. This functionality
was manually backported from 6.0
(original patch http://lists.mysql.com/commits/74749)
Detailed revision comments:
r6127 | vasil | 2009-10-30 11:18:25 +0200 (Fri, 30 Oct 2009) | 18 lines
branches/5.1:
Backport c6121 from branches/zip:
------------------------------------------------------------------------
r6121 | sunny | 2009-10-30 01:42:11 +0200 (Fri, 30 Oct 2009) | 7 lines
Changed paths:
M /branches/zip/mysql-test/innodb-autoinc.result
branches/zip: This test has been problematic for sometime now. The underlying
bug is that the data dictionaries get out of sync. In the AUTOINC code we
try and apply salve to the symptoms. In the past MySQL made some unrelated
change and the dictionaries stopped getting out of sync and this test started
to fail. Now, it seems they have reverted that changed and the test is
passing again. I suspect this is not he last time that this test will change.
------------------------------------------------------------------------
Detailed revision comments:
r6052 | sunny | 2009-10-12 07:09:56 +0300 (Mon, 12 Oct 2009) | 4 lines
branches/5.1: Reset the statement level autoinc counter on ROLLBACK. Fix
the test results too.
rb://164
r6053 | sunny | 2009-10-12 07:37:49 +0300 (Mon, 12 Oct 2009) | 6 lines
branches/5.1: Copy the maximum AUTOINC value from the old table to the new
table when MySQL does a CREATE INDEX ON T. This is required because MySQL
does a table copy, rename and drops the old table.
Fix Bug#47125: auto_increment start value is ignored if an index is created and engine=innodb
rb://168
Detailed revision comments:
r6051 | sunny | 2009-10-12 07:05:00 +0300 (Mon, 12 Oct 2009) | 6 lines
branches/5.1: Ignore negative values supplied by the user when calculating the
next value to store in dict_table_t. Setting autoincrement columns top negative
values is undefined behavior and this change should bring the behavior of
InnoDB closer to what users expect. Added several tests to check.
rb://162
Detailed revision comments:
r6045 | jyang | 2009-10-08 02:27:08 +0300 (Thu, 08 Oct 2009) | 7 lines
branches/5.1: Fix bug #47777. Treat the Geometry data same as
Binary BLOB in ha_innobase::store_key_val_for_row(), since the
Geometry data is stored as Binary BLOB in Innodb.
Review: rb://180 approved by Marko Makela.
When a query was using a DATE or DATETIME value formatted
using any other separator characters beside hyphen '-', a
query with a greater-or-equal '>=' condition matching only
the greatest value in an indexed column, the result was
empty if index range scan was employed.
The range optimizer got a new feature between 5.1.38 and
5.1.39 that changes a greater-or-equal condition to a
greater-than if the value matching that in the query was not
present in the table. But the value comparison function
compared the dates as strings instead of dates.
The bug was fixed by splitting the function
get_date_from_str in two: One part that parses and does
error checking. This function is now visible outside the
module. The old get_date_from_str now calls the new
function.
with temporary tables
There were two problems the test case from this bug was
triggering:
1. JOIN::rollup_init() was supposed to wrap all constant Items
into another object for queries with the WITH ROLLUP modifier
to ensure they are never considered as constants and therefore
are written into temporary tables if the optimizer chooses to
employ them for DISTINCT/GROUP BY handling.
However, JOIN::rollup_init() was called before
make_join_statistics(), so Items corresponding to fields in
const tables could not be handled as intended, which was
causing all kinds of problems later in the query execution. In
particular, create_tmp_table() assumed all constant items
except "hidden" ones to be removed earlier by remove_const()
which led to improperly initialized Field objects for the
temporary table being created. This is what was causing crashes
and valgrind errors in storage engines.
2. Even when the above problem had been fixed, the query from
the test case produced incorrect results due to some
DISTINCT/GROUP BY optimizations being performed by the
optimizer that are inapplicable in the WITH ROLLUP case.
Fixed by disabling inapplicable DISTINCT/GROUP BY optimizations
when the WITH ROLLUP modifier is present, and splitting the
const-wrapping part of JOIN::rollup_init() into a separate
method which is now invoked after make_join_statistics() when
the const tables are already known.
subquery returning multiple rows
Error handling was missing when handling subqueires in WHERE
and when assigning a SELECT result to a @variable.
This caused crash(es).
Fixed by adding error handling code to both the WHERE
condition evaluation and to assignment to an @variable.
having clause...
The fix for bug 46184 was not very complete. It was not covering
views using temporary tables and multiple tables in a FROM clause.
Fixed by reverting the fix for 46184 and making a more general
check that is checking at the right execution stage and for all
of the non-supported cases.
Now PROCEDURE ANALYZE on non-top level SELECT is also forbidden.
Updated the analyse.test and subselect.test accordingly.
CURRENT_USER() in GRANT ... TO CURRENT_USER() only gave us a definer,
not a full user (i.e., password-element was not initiliazed). Hence
dereferencing the password led to a crash.
Properly initializes definers now, just so there are no misunderstandings.
Also does some magic so IDENTIFIED BY ... works with CURRENT_USER().
If an outer query is broken, a subquery might not even get set up.
EXPLAIN EXTENDED did not expect this and merrily tried to de-ref all
of the half-setup info.
We now catch this case and print as much as we have, as it doesn't cost us
anything (doesn't make regular execution slower).
Queries with nested outer joins may lead to crashes or
bad results because an internal data structure is not handled
correctly.
The optimizer uses bitmaps of nested JOINs to determine
if certain table can be placed at a certain place in the
JOIN order.
It does maintain a bitmap describing in which JOINs
last placed table is nested.
When it puts a table it makes sure the bit of every JOIN that
contains the table in question is set (because JOINs can be nested).
It does that by recursively setting the bit for the next enclosing
JOIN when this is the first table in the JOIN and recursively
resetting the bit if it's the last table in the JOIN.
When it removes a table from the join order it should do the
opposite : recursively unset the bit if it's the only remaining
table in this join and and recursively set the bit if it's removing
the last table of a JOIN.
There was an error in how the bits was set for the upper levels :
when removing a table it was setting the bit for all the enclosing
nested JOINs even if there were more tables left in the current JOIN
(which practically means that the upper nested JOINs were not affected).
Fixed by stopping the recursion at the relevant level.
XA START may cause assertion failure/server crash when it is called
after unilateral roll back issued by the Resource Manager (both
in regular transaction and after XA transaction).
The problem was that rm_error variable wasn't set/reset properly.
Bug#46539 Various crashes on INSERT IGNORE SELECT + SELECT FOR UPDATE.
If a transaction was rolled back inside InnoDB due to a deadlock
or lock wait timeout, and the statement had IGNORE clause,
the server could crash at the end of the statement or on shutdown.
This was caused by the error handling infrastructure's attempt to
ignore a non-ignorable error.
When a transaction rollback request is raised, switch off
current_select->no_error flag, so that the following error
won't be ignored.
Instead, we could add !thd->is_fatal_sub_stmt_error to
my_message_sql(), but since in write_record() we switch
off no_error, the same approach is used in
thd_mark_transaction_to_rollback().
@todo: call thd_mark_transaction_to_rollback() from
handler::print_error(), then we can easily make sure
that the error reported by print_error is not ignored.
BUG#47073 - valgrind errs, corruption,failed repair of partition,
low myisam_sort_buffer_size
Fixed race conditions discovered with the provided test case and
stabilized test case.
inside subquery
Re-setting a fulltext index was a no-operation if not all
the matches of a search were consumed by reading them.
This was preventing a joined table using a fulltext index
in a subquery that requires only 1 row of output (e.g. EXISTS)
from working correctly because the second execution of the
sub-query has the fulltext index cursor in a wrong state and
was not finding results.
Fixed by making the re-init code _ftb_init_index_search()
to re-set open cursors in addition to depleted ones.
Problem 1:
column_priv_hash uses utf8_general_ci collation
for the key comparison. The key consists of user name,
db name and table name. Thus user with privileges on table t1
is able to perform the same operation on T1
(the similar situation with user name & db name, see acl_cache).
So collation which is used for column_priv_hash and acl_cache
should be case sensitive.
The fix:
replace system_charset_info with my_charset_utf8_bin for
column_priv_hash and acl_cache
Problem 2:
The same situation with proc_priv_hash, func_priv_hash,
the only difference is that Routine name is case insensitive.
So the fix is to use my_charset_utf8_bin for
proc_priv_hash & func_priv_hash and convert routine name into lower
case before writing the element into the hash and
before looking up the key.
Additional fix: mysql.procs_priv Routine_name field collation
is changed to utf8_general_ci.
It's necessary for REVOKE command
(to find a field by routine hash element values).
Note:
It's safe for lower-case-table-names mode too because
db name & table name are converted into lower case
(see GRANT_NAME::GRANT_NAME).
