Hi Serg,
Sergei Golubchik wrote:
> Hi!
>
> On Feb 14, Alexey Kopytov wrote:
>> Below is the list of changes that have just been committed into a local
>> 5.1 repository of kaa. When kaa does a push these changes
>> will be propagated to the main repository and, within 24 hours after the
>> push, to the public repository.
>> For information on how to access the public repository
>> see http://dev.mysql.com/doc/mysql/en/installing-source-tree.html
>>
>> ChangeSet@stripped, 2008-02-14 17:46:39+03:00, kaa@kaamos.(none) +9 -0
>> Fix for bug #31236: Inconsistent division by zero behavior for
>> 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.
>>
>> configure.in@stripped, 2008-02-14 17:46:36+03:00, kaa@kaamos.(none) +2 -2
>> Removed DONT_USE_FINITE, it is not used anywhere.
>
> it's in my_global.h in my 5.1 tree. how comes ?
>
Oops, because I forgot to remove it! The only use of it was:
#ifdef DONT_USE_FINITE /* HPUX 11.x has is_finite() */
#undef HAVE_FINITE
#endif
I.e. "claim that we don't have finite() at all, without providing any
replacements" (like is_finite() which is available on HPUX as the
comments says).
Since the patch changes that to always define isfinite(), either as a
native macro, or an alias to finite(), or a simple replacement, we don't
need that define anymore. We also get rid of checks for HAVE_FINITE in
the code.
Removed in the new patch.
>> diff -Nrup a/sql/item_func.cc b/sql/item_func.cc
>> --- a/sql/item_func.cc 2008-01-19 20:59:07 +03:00
>> +++ b/sql/item_func.cc 2008-02-14 17:46:36 +03:00
>> @@ -1104,7 +1104,7 @@ double Item_func_plus::real_op()
>> double value= args[0]->val_real() + args[1]->val_real();
>> if ((null_value=args[0]->null_value || args[1]->null_value))
>> return 0.0;
>> - return value;
>> + return fix_result(value);
>> }
>
> Is it the best approach ? I wonder perhaps the value could be fixed not
> in every item that does the calculations, but only when we need to do
> something with a result ? that is in double->string conversion. in
> isnull checks, in Field::store().
>
Never passing an invalid value to another part of code looks more robust
to me. Setting it to NULL right away means we won't even have to start a
conversion, or any calculations in another function/operation.
Currently, we assume everywhere that input floating point values are
"sane", and only check if the output (the computed value) is finite.
Best regards,
--
Alexey Kopytov, Software Developer
MySQL AB, www.mysql.com
Are you MySQL certified? www.mysql.com/certification
