JDBC bug: Incorrect number of conflicts is reported by CachedRowSetWriter.writeData

[Frank Ding]

Hi Lance,
   Sorry for late response and thanks for your comment.  You mean I can 
write a jtreg test case that connects to Java DB?  I can do that.

Best regards,
Frank

On 11/13/2012 10:13 PM, Lance Andersen - Oracle wrote:
> Hi Frank,
>
>
> Thank you for the note
> If you could in the future, please provide  a  complete test program to repro the issue as it would save time with the reviews.  Ideally if the issue is not database specific it would be good to leverage Java DB as it is included within Oracle JDK
>
> I will look at this sometime this week
>
> Best
> Lance
> On Nov 12, 2012, at 9:25 PM, Frank Ding wrote:
>
>> Hi Lance
>>   Thanks for your quick response.  Please find the bug info below.
>>
>>   The problem:
>>   When CachedRowSetImpl.acceptChanges() is called, incorrect number of conflicts, if any, is reported.  The number of conflicts is the actual number of existing rows in database, which is the size of variable 'status' defined in CachedRowSetWriter.writeData().  It's not the conflict number that is supposed to be.
>>
>>   Test case:
>>   The bug can be easily manifested in all SQL server environment. Here take PostgreSQL for example.
>>   1. The sql script to create a table
>> CREATE TABLE ressystem.roomdescription
>> (
>>   roomdescription_id serial NOT NULL,
>>   roomdescription character varying NOT NULL,
>>   CONSTRAINT roomdescription_pkey PRIMARY KEY (roomdescription_id)
>> )
>>
>>   2.  Manually insert 3 rows
>> (1, "Test 1")
>> (2, "Test 2")
>> (3, "Test 3")
>>
>>   3. Create a Java class that connects the established database and then execute the following code snippet.
>> String query = "select roomdescription_id, roomdescription from ressystem.roomdescription";
>> Object[] values = {2, "Test2"};
>> rs.setCommand(query);
>> rs.execute(conn);
>> rs.moveToInsertRow();
>> for(int i=0; i<values.length; i++) {
>> rs.updateObject(i+1,values[i]);
>> }
>> rs.insertRow();
>> rs.moveToCurrentRow();
>> rs.acceptChanges(conn);
>>
>>   4. An exception occurs with following output.
>> javax.sql.rowset.spi.SyncProviderException: 4conflicts while synchronizing
>>     at com.sun.rowset.internal.CachedRowSetWriter.writeData(CachedRowSetWriter.java:412)
>>     at com.sun.rowset.CachedRowSetImpl.acceptChanges(CachedRowSetImpl.java:880)
>>
>> 5. In fact, there is only one conflicting row but 4 were reported.
>>
>> Best regards,
>> Frank
>>
>> On 11/9/2012 7:41 PM, Lance Andersen - Oracle wrote:
>>> Frank,
>>>
>>> If you can please post the bug info here, I will take a look at your patch
>>>
>>> Best
>>> Lance
>>> On Nov 8, 2012, at 10:01 PM, Frank Ding wrote:
>>>
>>>> Hi guys,
>>>>   We discovered a bug in CachedRowSetWriter.writeData method where incorrect number of conflicts is reported.  I searched in Oracle bug database and no similar record was found.  So I submitted a new one whose internal review ID is 2376620.  A test case with code is illustrated in the bug submission that leverages PostgreSQL server but the issue is platform independent and easy to reproduce.
>>>>   I provided a patch review, available @
>>>>   http://cr.openjdk.java.net/~dingxmin/2376620/webrev.01/
>>>>   Is there anybody who is interested in patch and can also review bug 2376620?  Your reply is appreciated.
>>>>
>>>> Best regards,
>>>> Frank
>>>>
>>>>
>>>
>>> Lance Andersen| Principal Member of Technical Staff | +1.781.442.2037
>>> Oracle Java Engineering
>>> 1 Network Drive
>>> Burlington, MA 01803
>>> Lance.Andersen at oracle.com
>>>
>
>
> Lance Andersen| Principal Member of Technical Staff | +1.781.442.2037
> Oracle Java Engineering
> 1 Network Drive
> Burlington, MA 01803
> Lance.Andersen at oracle.com
>