The problem was that creating a DECIMAL column from a decimal
value could lead to a failed assertion as decimal values can
have a higher precision than those attached to a table. The
assert could be triggered by creating a table from a decimal
with a large (> 30) scale. Also, there was a problem in
calculating the number of digits in the integral and fractional
parts if both exceeded the maximum number of digits permitted
by the new decimal type.
The solution is to ensure that truncation procedure is executed
when deducing a DECIMAL column from a decimal value of higher
precision. If the integer part is equal to or bigger than the
maximum precision for the DECIMAL type (65), the integer part
is truncated to fit and the fractional becomes zero. Otherwise,
the fractional part is truncated to fit into the space left
after the integer part is copied.
This patch borrows code and ideas from Martin Hansson's patch.
match against.
Server crashes when executing prepared statement with duplicating
MATCH() function calls in SELECT and ORDER BY expressions, e.g.:
SELECT MATCH(a) AGAINST('test') FROM t1 ORDER BY MATCH(a) AGAINST('test')
This query gets optimized by the server, so the value returned
by MATCH() from the SELECT list is reused for ORDER BY purposes.
To make this optimization server is comparing items from
SELECT and ORDER BY lists. We were getting server crash because
comparision function for MATCH() item is not intended to be called
at this point of execution.
In 5.0 and 5.1 this problem is workarounded by resetting MATCH()
item to the state as it was during PREPARE.
In 6.0 correct comparision function will be implemented and
duplicating MATCH() items from the ORDER BY list will be
optimized.
Using DECIMAL constants with more than 65 digits in CREATE
TABLE ... SELECT led to bogus errors in release builds or
assertion failures in debug builds.
The problem was in inconsistency in how DECIMAL constants and
fields are handled internally. We allow arbitrarily long
DECIMAL constants, whereas DECIMAL(M,D) columns are limited to
M<=65 and D<=30. my_decimal_precision_to_length() was used in
both Item and Field code and truncated precision to
DECIMAL_MAX_PRECISION when calculating value length without
adjusting precision and decimals. As a result, a DECIMAL
constant with more than 65 digits ended up having length less
than precision or decimals which led to assertion failures.
Fixed by modifying my_decimal_precision_to_length() so that
precision is truncated to DECIMAL_MAX_PRECISION only for Field
object which is indicated by the new 'truncate' parameter.
Another inconsistency fixed by this patch is how DECIMAL
constants and expressions are handled for CREATE ... SELECT.
create_tmp_field_from_item() (which is used for constants) was
changed as a part of the bugfix for bug #24907 to handle long
DECIMAL constants gracefully. Item_func::tmp_table_field()
(which is used for expressions) on the other hand was still
using a simplistic approach when creating a Field_new_decimal
from a DECIMAL expression.
assertion .\filesort.cc, line 797
A query with the "ORDER BY @@some_system_variable" clause,
where @@some_system_variable is NULL, causes assertion
failure in the filesort procedures.
The reason of the failure is in the value of
Item_func_get_system_var::maybe_null: it was unconditionally
set to false even if the value of a variable was NULL.
The RAND(N) function where the N is a field of "constant" table
(table of single row) failed with a SIGFPE.
Evaluation of RAND(N) rely on constant status of its argument.
Current server "seeded" random value for each constant argument
only once, in the Item_func_rand::fix_fields method.
Then the server skipped a call to seed_random() in the
Item_func_rand::val_real method for such constant arguments.
However, non-constant state of an argument may be changed
after the call to fix_fields, if an argument is a field of
"constant" table. Thus, pre-initialization of random value
in the fix_fields method is too early.
Initialization of random value by seed_random() has been
removed from Item_func_rand::fix_fields method.
The Item_func_rand::val_real method has been modified to
call seed_random() on the first evaluation of this method
if an argument is a function.
Problem: storing "SELECT ... INTO @var ..." results in variables we used val_xxx()
methods which returned results of the current row.
So, in some cases (e.g. SELECT DISTINCT, GROUP BY or HAVING) we got data
from the first row of a new group (where we evaluate a clause) instead of
data from the last row of the previous group.
Fix: use val_xxx_result() counterparts to get proper results.
in trigger
Interchangeable calls to the mysql_change_user client function
and invocations of a trigger changing some user variable caused
a memory corruption and a crash.
The mysql_change_user API call forces TDH::cleanup() on a server
that frees user variable entries.
However it didn't reset Item_func_set_user_var::entry to NULL
because Item_func_set_user_var::cleanup() was not overloaded.
So, Item_func_set_user_var::entry held a pointer to freed memory,
that caused a crash.
The Item_func_set_user_var::cleanup method has been overloaded
to cleanup the Item_func_set_user_var::entry field.
IF(..., CAST(longtext AS UNSIGNED), signed_val)
(was: LEFT JOIN on inline view crashes server)
Select from a LONGTEXT column wrapped with an expression
like "IF(..., CAST(longtext_column AS UNSIGNED), smth_signed)"
failed an assertion or crashed the server. IFNULL function was
affected too.
LONGTEXT column item has a maximum length of 32^2-1 bytes,
at the same time this is a maximum possible length of any
MySQL item. CAST(longtext_column AS UNSIGNED) returns some
unsigned numeric result of length 32^2-1, so the result of
IF/IFNULL function of this number and some other signed number
will have text length of (32^2-1)+1=32^2 (one byte for the
minus sign) - there is integer overflow, and the length is
equal to zero. That caused assert/crash.
CAST AS UNSIGNED function has been modified to limit maximal
length of resulting number to 67 (maximal length of DECIMAL
and two characters for minus sign and dot).
IF(..., CAST(longtext AS UNSIGNED), signed_val)
(was: LEFT JOIN on inline view crashes server)
Select from a LONGTEXT column wrapped with an expression
like "IF(..., CAST(longtext_column AS UNSIGNED), smth_signed)"
failed an assertion or crashed the server. IFNULL function was
affected too.
LONGTEXT column item has a maximum length of 32^2-1 bytes,
at the same time this is a maximum possible length of any
MySQL item. CAST(longtext_column AS UNSIGNED) returns some
unsigned numeric result of length 32^2-1, so the result of
IF/IFNULL function of this number and some other signed number
will have text length of (32^2-1)+1=32^2 (one byte for the
minus sign) - there is integer overflow, and the length is
equal to zero. That caused assert/crash.
The bug has been fixed by the same solution as in the CASE
function implementation.
variable settings (rpl_sys)
Problem: under certain conditions (e.g. user variables usage in triggers)
accessing a user defined variable we may use a variables hash table that
belongs to already deleted thread. It happens if
thd= new THD;
has the same address as just deleted thd as we use
if (stored_thd == thd)
to check.
That may lead to unpredictable results, server crash etc.
Fix: use thread_id instead of thd address to distinguish threads.
Note: no simple and repeatable test case.
IS NULL was not checking the correct row in a HAVING context.
At the first row of a new group (where the HAVING clause is evaluated)
the column and SELECT list references in the HAVING clause should
refer to the last row of the previous group and not to the current one.
This was not done for IS NULL, because it was using Item::is_null() doesn't
have a Item_is_null_result() counterpart to access the data from the
last row of the previous group. Note that all the Item::val_xxx() functions
(e.g. Item::val_int()) have their _result counterparts (e.g. Item::val_int_result()).
Fixed by implementing a is_null_result() (similarly to int_result()) and
calling this instead of is_null() column and SELECT list references inside
the HAVING clause.
Fixed the handling of system variable retrieval
in prepared statements : added a cleanup method
that clears up the cache and restores the
original scope of the variable (which is overwritten
at fix_fields()).
The code to get read the value of a system variable was extracting its value
on PREPARE stage and was substituting the value (as a constant) into the parse tree.
Note that this must be a reversible transformation, i.e. it must be reversed before
each re-execution.
Unfortunately this cannot be reliably done using the current code, because there are
other non-reversible source tree transformations that can interfere with this
reversible transformation.
Fixed by not resolving the value at PREPARE, but at EXECUTE (as the rest of the
functions operate). Added a cache of the value (so that it's constant throughout
the execution of the query). Note that the cache also caches NULL values.
Updated an obsolete related test suite (variables-big) and the code to test the
result type of system variables (as per bug 74).
columns data types
The "SELECT @lastId, @lastId := Id FROM t" query returns
different result sets depending on the type of the Id column
(INT or BIGINT).
Note: this fix doesn't cover the case when a select query
references an user variable and stored function that
updates a value of that variable, in this case a result
is indeterminate.
The server uses incorrect assumption about a constantness of
an user variable value as a select list item:
The server caches a last query number where that variable
was changed and compares this number with a current query
number. If these numbers are different, the server guesses,
that the variable is not updating in the current query, so
a respective select list item is a constant. However, in some
common cases the server updates cached query number too late.
The server has been modified to memorize user variable
assignments during the parse phase to take them into account
on the next (query preparation) phase independently of the
order of user variable references/assignments in a select
item list.
Problem: in mixed and statement mode, a query that refers to a
system variable will use the slave's value when replayed on
slave. So if the value of a system variable is inserted into a
table, the slave will differ from the master.
Fix: mark statements that refer to a system variable as "unsafe",
meaning they will be replicated by row in mixed mode and produce a warning
in statement mode. There are some exceptions: some variables are actually
replicated. Those should *not* be marked as unsafe.
BUG#34732: mysqlbinlog does not print default values for auto_increment variables
Problem: mysqlbinlog does not print default values for some variables,
including auto_increment_increment and others. So if a client executing
the output of mysqlbinlog has different default values, replication will
be wrong.
Fix: Always print default values for all variables that are replicated.
I need to fix the two bugs at the same time, because the test cases would
fail if I only fixed one of them.
NAME_CONST('whatever', -1) * MAX(whatever) bombed since -1 was
not seen as constant, but as FUNCTION_UNARY_MINUS(constant)
while we are at the same time pretending it was a basic const
item. This confused the aggregate handlers in exciting ways.
We now make NAME_CONST() behave more consistently.
between 5.0 and 5.1.
The problem was that in the patch for Bug#11986 it was decided
to store original query in UTF8 encoding for the INFORMATION_SCHEMA.
This approach however turned out to be quite difficult to implement
properly. The main problem is to preserve the same IS-output after
dump/restore.
So, the fix is to rollback to the previous functionality, but also
to fix it to support multi-character-set-queries properly. The idea
is to generate INFORMATION_SCHEMA-query from the item-tree after
parsing view declaration. The IS-query should:
- be completely in UTF8;
- not contain character set introducers.
For more information, see WL4052.
floating point numbers
Some math functions did not check if the result is a valid number
(i.e. neither of +-inf or nan).
Fixed by validating the result where necessary and returning NULL in
case of invalid result.
There's currently no way of knowing the determinicity of an UDF.
And the optimizer and the sequence() UDFs were making wrong
assumptions about what the is_const member means.
Plus there was no implementation of update_system_tables()
causing the optimizer to overwrite the information returned by
the <udf>_init function.
Fixed by equating the assumptions about the semantics of
is_const and providing a implementation of update_used_tables().
Added a TODO item for the UDF API change needed to make a better
implementation.
Problem: setting Item_func_rollup_const::null_value property to argument's null_value
before (without) the argument evaluation may result in a crash due to wrong null_value.
Fix: use is_null() to set Item_func_rollup_const::null_value instead as it evaluates
the argument if necessary and returns a proper value.
There are two problems with ROUND(X, D) on an exact numeric
(DECIMAL, NUMERIC type) field of a table:
1) The implementation of the ROUND function would change the number of decimal
places regardless of the value decided upon in fix_length_and_dec. When the
number of decimal places is not constant, this would cause an inconsistent
state where the number of digits was less than the number of decimal places,
which crashes filesort.
Fixed by not allowing the ROUND operation to add any more decimal places than
was decided in fix_length_and_dec.
2) fix_length_and_dec would allow the number of decimals to be greater than
the maximium configured value for constant values of D. This led to the same
crash as in (1).
Fixed by not allowing the above in fix_length_and_dec.
