[
https://issues.apache.org/jira/browse/DERBY-6373?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Knut Anders Hatlen updated DERBY-6373:
--------------------------------------
Attachment: d6373-1a.diff
The attached patch makes SqlWarning.getSQLWarning() follow the warning chain
instead of the exception chain. Since it has already checked that the warning
chain is not null, the NPE is avoided.
The patch also adds a test case which verifies that the NPE is no longer
thrown, and that all the warnings from the batch are received on the client.
(The behaviour seems to be slightly different on embedded. It has only a single
warning on the statement after executing the batch. The client has one warning
for each of the statements in the batch that got a warning.)
All regression tests ran cleanly with the patch.
> NPE in Statement.getWarnings()
> ------------------------------
>
> Key: DERBY-6373
> URL: https://issues.apache.org/jira/browse/DERBY-6373
> Project: Derby
> Issue Type: Bug
> Components: Network Client
> Affects Versions: 10.10.1.1
> Environment: JDK 7
> Reporter: Glenn McGregor
> Assignee: Knut Anders Hatlen
> Priority: Minor
> Attachments: d6373-1a.diff, D6373.java
>
>
> After executing a batch in a prepared statement, I call getWarnings() on that
> statement. A NPE is thrown.
> at
> org.apache.derby.client.am.SqlWarning.getSQLWarning(SqlWarning.java:117)
> at org.apache.derby.client.am.Statement.getWarnings(Statement.java:862)
> ...
> Addtional info:
> In a unit test, which worked for 10.8.2.2, a batch of 4 deletes were issued.
> They were expected to fail, as there were no matching rows. There was a
> SQLWarning chain in the statement, 4 deep, one for each row that failed
> with text:
> No row was found for FETCH, UPDATE or DELETE; or the result of a query is an
> empty table.
> However, each warning had 'nextException_' null. When attempting to chain on
> the exceptions when building the warning, a NPE is thrown.
> Perhaps something like changing line 105 of SqlWarning.java to
> if ( nextWarning_ != null && nextException_ != null )
> Of course perhaps there's always supposed to be a corresponding exception.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
