On 5/11/10 4:07 PM +0300, Nicolas Barbier wrote:
> 2010/5/11 Marko Tiikkaja<marko.tiikkaja@cs.helsinki.fi>:
>
>> This is getting way off topic, but:
>>
>> On 5/11/10 3:55 PM +0300, Nicolas Barbier wrote:
>>>
>>> T2> SELECT i FROM a WHERE i = 1 FOR SHARE; -- Lock a with i = 1 FOR
>>> SHARE.
>>> i
>>> ---
>>> 1
>>> (1 Zeile)
>>>
>>> T2> SELECT a_id FROM b WHERE a_id = 1; -- Check whether it's got
>>> anything pointing to it.
>>> a_id
>>> ------
>>> (0 Zeilen)
>>>
>>> T2> DELETE FROM a WHERE i = 1; -- Nope, so delete a with i = 1 (this
>>> blocks, because T1 is still holding the lock).
>>
>> Obviously you wouldn't delete anything with a SHARE lock.
>
> So where would you put a SELECT ... FOR SHARE to fix the problem? (Per
> "Will SELECT ... FOR SHARE not help?".) I agree that my second FOR
> SHARE doesn't really make a lot of sense, but that doesn't disprove
> the fact that the first FOR SHARE fails to ensure consistency.
I took the "SELECT ... FOR SHARE" suggestion in a more general way,
suggesting the use of row-level locks. T2 should be holding an
exclusive row-level lock (SELECT ... FOR UPDATE) when checking for
references.
Regards,
Marko Tiikkaja