Hi Dag,
I am glad we are making Network Server and Embedded behave the same way in this arena.
One thing that would be good to have is extend the test that you added to show what happens if we are on the last row in the resultset and updateRow/deleteRow is issued and rollback is sent after that.
thanks,
Mamta
On 12/12/05, Bernt M. Johnsen <[EMAIL PROTECTED]> wrote:
>>>>>>>>>>>> Dag H. Wanvik wrote (2005-12-13 02:18:39):
>
> Hi,
>
> I have a made patch for this, can somebody have a look at it,
> please?
1) I think you should add a comment describing the purpose of the
chainAutoCommit variable (why you have to store the
willAutoCommitGenerateFlow) in ResultSet.java
2) There is an unnecessary whitespace change in ResultSet.java
>
> From the JIRA:
>
> "I have uploaded a fix for this problem (611.{stat,diff}), and added a
> new test (look for "Positive Test34") to lang/updatableResultSet.java.
> The network client driver canons are also changed in the case of three
> error messages due to the changed behavior, in addition to the change
> due to new test.
>
> The patch removes the explicit (auto)commit code in the network client
> driver in deleteRow and updateRow, making it behave as the embedded
> driver in this respect (modulo error messages).
>
> I have run lang/updatableResultSet.java successfully with Sun jdk 1.3,
> 1.4 and 1.5 (the two latter have another canon) on Solaris
> 10/x86. Also, there is a separate canon under j9_13 for this test,
> which I have updated, but not been able to test.
>
> I have also run derbyall successfully with jdk 1.4, also on Solaris
> 10/x86, albeit with some errors in unrelated tests (lang/dcl,
> store/OnlineBackupTest1.java)"
>
> Thanks,
>
> Dag
--
Bernt Marius Johnsen, Database Technology Group,
Sun Microsystems, Trondheim, Norway
