On 07/24/14 17:38, Adriano dos Santos Fernandes wrote:
> On 24/07/2014 07:30, Alex Peshkoff wrote:
>> On 07/23/14 19:59, Adriano dos Santos Fernandes wrote:
>>> Hi!
>>>
>>> After I reported an TCS fail to Alex in January, he moved this method
>>> from IStatement to IResultSet.
>> That move was not directly related with TCS fail. Originally setting
>> cursor name remained in effect only for cursor life-time - i.e. after
>> isc_dsql_free_statement with DSQL_close option cursor name was
>> destroyed. If we set cursor name for statement, semantically it should
>> exist after closing cursor. But this breaks applications (starting with
>> IBO components) that tend to set new cursor name for each statement
>> execution.
> But are you claiming this is currently fixed or not? Because CORE-4489
> describes this situation...
I've missed that issue being at vacations. I was sure it's fixed. May be
same error was introduced again later.
>>> But I fail to understand. Not talking about implementation detail, but
>>> about interface.
>>>
>>> IMO makes no sense to first open the cursor and then name it later.
>>>
>>> We can have two prepared statement, while the second uses the cursor
>>> name of the first. But now we need to open the first before prepare the
>>> second.
>>>
>>> Isn't this just wrong?
>> Slightly changing FB_SQL_2 test
>>
>> diff -u -r1.1 FB_SQL_2.script
>> --- FB_SQL_2.script 27 Dec 2009 16:51:23 -0000 1.1
>> +++ FB_SQL_2.script 24 Jul 2014 10:24:31 -0000
>> @@ -62,8 +62,6 @@
>> }
>> if (isc_dsql_set_cursor_name(userStatus, &stmt1, "C1", 0))
>> isc_print_status(userStatus);
>> - if (isc_dsql_execute(userStatus, &gds__trans, &stmt1, 1, NULL))
>> - isc_print_status(userStatus);
>>
>> stmt2 = 0;
>> if (isc_dsql_allocate_statement(userStatus, &db, &stmt2))
>> @@ -74,6 +72,9 @@
>> isc_print_status(userStatus);
>> }
>>
>> + if (isc_dsql_execute(userStatus, &gds__trans, &stmt1, 1, NULL))
>> + isc_print_status(userStatus);
>> +
>> printf("Original values:\n");
>>
>> while (1)
>>
>> (doing exactly what you talk about - no execute before preparing stmt2)
>> runs fine on 2.5, but fails with 3.0 - it's enough to change client
>> library. I.e. I certainly agree that we have a problem.
>>
>>
>> But what solution can we suggest? Move setCursorName back to IStatement
>> and blank it when cursor is closed?
> It's better than have setCursorName in IResultSet.
Therefore I talk about it...
>> Last is required to make IBO happy.
>> IMHO in old API this is required. But what about new one? What it worse
>> - change in behavior or ugly semantics?
>>
>>
> Maybe a flag in the new API close method to use the old or the good
> behavior?
>
Also ugly.
I would better like to support setCursorName(NULL) in order to emulate
old API.
------------------------------------------------------------------------------
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds
Firebird-Devel mailing list, web interface at
https://lists.sourceforge.net/lists/listinfo/firebird-devel
