[ http://issues.apache.org/jira/browse/DERBY-1014?page=all ]
Knut Anders Hatlen updated DERBY-1014:
--------------------------------------
Attachment: DERBY-1014-v1.diff
DERBY-1014-v1.stat
Attached a patch (DERBY-1014-v1.diff) which modifies tests in
preparation for DERBY-822. There are two types of changes:
1) Invoke ResultSet.next() to expose error even when the driver
isn't pre-fetching data.
2) Filter out some of the run-time statistics before comparing with
the canon.
Please note that one 'svn add' and two 'svn delete' commands are
required (see DERBY-1014-v1.stat), and probably 'ant clobber' as well.
I have run derbyall successfully (failures also seen in nightly
regression tests) on Sun JVM 1.4.2/Solaris 10 x86.
lang/wisconsin.java:
* Added patterns to the sed file. Cursor names and some of the
run-time statistics are filtered out.
* Canons for DerbyNet and DerbyNetClient were deleted since they are
no longer needed. Each of these files is 1.5M, so the diff is
huge.
lang/scrollCursors1.sql:
* Added patterns to the sed file. Number of rows seen and number of
reads from hash table will be filtered out.
jdbcapi/parameterMetaDataJdbc30.java:
* Added a call to ResultSet.next() to provoke an error in a test
case for DERBY-44 regardless of framework/pre-fetching.
jdbcapi/resultset.java:
* Don't print column headers until one row is successfully retrieved
(achieved by calling ResultSet.next() earlier).
jdbcapi/setTransactionIsolation.java:
* Added a call to ResultSet.next() to provoke a timeout regardless
of framework/pre-fetching.
* Added patterns to the sed file. Filtered out number of pages
visited, number of rows qualified/visited etc.
> Make tests less sensitive to pre-fetching
> -----------------------------------------
>
> Key: DERBY-1014
> URL: http://issues.apache.org/jira/browse/DERBY-1014
> Project: Derby
> Type: Sub-task
> Components: Test
> Reporter: Knut Anders Hatlen
> Assignee: Knut Anders Hatlen
> Priority: Minor
> Attachments: DERBY-1014-v1.diff, DERBY-1014-v1.stat
>
> Some tests are very sensitive to changes in pre-fetching, either
> because they call SYSCS_UTIL.SYSCS_GET_RUNTIMESTATISTICS() which
> displays the number of rows seen, or because a failure that normally
> is exposed when calling ResultSet.next() is exposed when calling
> Statement.execute().
> Most of these tests are not testing when or how much data is
> pre-fetched. If we make the tests less sensitive to pre-fetching, it
> is easier to catch the real regressions when changing the way Derby is
> pre-fetching data. Also, patches in that area of the code will be
> smaller (fewer test changes) and easier to review.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira
