db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Deepa Remesh (JIRA)" <derby-...@db.apache.org>
Subject [jira] Commented: (DERBY-706) Improve testing and increase code coverage for Network Server classes
Date Wed, 07 Dec 2005 23:17:08 GMT
    [ http://issues.apache.org/jira/browse/DERBY-706?page=comments#action_12359665 ] 

Deepa Remesh commented on DERBY-706:

I have attached a second patch 'derby-706_2.diff' which adds some more tests and re-arranges/removes
some internal methods. Details of patch are:

* DDMWriter: Removes min method because Math.min is used. Use doesRequestContainData() method.
* DDMReader: Use readCodePoint() method
* DRDAConnThread: QRYOPTVAL codepoint is not part of DRDA spec. parseCcsidSBC was giving NPE.
* CodePointNameTable: Add codepoints to be used in protocol.tests
* protocol.tests: Added tests to cover methods which were not tested before.
* Session, DRDAStatement, Database, DRDAProtocolException: Remove some methods which are not
needed. The same can be done using other methods.
* TestProto: Use lookup method.
* testProtocol: Added "derby.drda.traceAll=true" to derby.properties file to test methods
in DssTrace
* testProperties: Added test for trace off command. Added "derby.drda.traceAll=true" to derby.properties
file to test memcheck.
* runtimeinfo: Rearranged test because the output was not predictable on different jvms/machines.

Ran derbyall with Sun JDK 1.4.2 on WinXP. No failures. Please review/commit this patch. 

Kathey, I looked into your question:
//this looks highly suspect. Why does timeSlice setSoTimeout?
if (timeSlice != 0)

>From what I understand, setSoTimeout socket option is used so that a thread waits only
for the time specified.  Otherwise for blocking operations, a thread can wait indefinitely.
In case of network server, it can wait indefinitely without considering value specified in
timeslice property. I have not included the test for these in this patch.

I have couple of questions:

1. In one case where agentError is thrown (DRDAConnThread#parseNOCMorNOCS), it seems to me
that a syntax error should be thrown here. For agentErrors, no code point is sent back to
the client. In DDM manual, I saw AGNPRMRM is used for severe agent errors. Is there something
else which can be sent back for non-severe errors?
2. QRYINSID is checked in CNTQRY and CLSQRY but it is not being used. Should this be used?

I would appreciate if someone can provide answers for above. Thanks. 

> Improve testing and increase code coverage for Network Server classes
> ---------------------------------------------------------------------
>          Key: DERBY-706
>          URL: http://issues.apache.org/jira/browse/DERBY-706
>      Project: Derby
>         Type: Improvement
>   Components: Test
>     Versions:
>     Reporter: Deepa Remesh
>     Assignee: Deepa Remesh
>     Priority: Minor
>  Attachments: derby-706.diff, derby-706.status, derby-706_2.diff, derby-706_2.status
> Improve testing of network server by increasing tests to cover 100% of the classes and
increase the method level coverage. The current code coverage(class/method) for network server
based on svn revision 208786 are:
> org.apache.derby.drda.NetworkServercontrol.java - 100/72 
> org.apache.derby.impl.drda - 97/78 
> Details of code coverage from EMMA tool are available from Derby Wiki. Links are:
> http://wiki.apache.org/db-derby/CodeCoverage
> http://people.apache.org/~fuzzylogic/codecoverage/208786/_files/c5.html
> http://people.apache.org/~fuzzylogic/codecoverage/208786/_files/25.html

This message is automatically generated by JIRA.
If you think it was sent incorrectly contact one of the administrators:
For more information on JIRA, see:

View raw message