Kathey Marsden created DERBY-6210:
-------------------------------------

             Summary: Create a mechanism to exclude some testing of internal 
interfaces and likely to change behavior from compatibility testing
                 Key: DERBY-6210
                 URL: https://issues.apache.org/jira/browse/DERBY-6210
             Project: Derby
          Issue Type: Bug
            Reporter: Kathey Marsden


For compatibility and upgrade testing, in order to expose bugs and omitted 
release notes I run the tests from an older branch against new releases.  The 
premise is that the functional tests are really just a JDBC application which 
should continue to pass with the new version.

The JUnit tests for Derby though are not pure functional tests which are 
backward compatible.  Often they make checks on things that are likely to 
change with new versions. This creates a large amount of noise in the testing: 
For example:

1) Checks for the exact contents or number of system tables.
2) Checks for "Not implemented" JDBC API calls which might become implemented.
3) Tests for client specific behavior which we expect might change to match 
embedded.
4) Unit tests which use internal interfaces that are likely to change.
5) Metadata tests which test the exact number of columns when 
6) Diff based tests which are likely to change with message changes.

It would be good to have a flag to omit these types of tests when doing 
compatibility testing.
My thought is to have a system property derby.tests.testCompat=true   and a 
method in BaseTestCase to test the property testCompat().


Then blocks of testing which should not be run or should have a different 
behavior with compatibility testing can be flagged as:
if (! testCompat())  {

// do testing  that we think might change, e.g. system table query.

}

This will allow the more detailed testing under normal circumstances but take 
some of the pain out of the compatibility testing moving forward.  It would 
require folks to think as they add new tests whether their testing as to 
whether they expect the tested behavior  to remain stable moving forward and 
put the block around it if they do not.

I welcome feedback on whether this or some other approach is preferable.


--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to