[Mondrian] RE: Eigenbase perforce change 10142 for review

[Ajit Vasudeo Joglekar]

Thanks Julian,

I have made the test strategy changes

I have added three specific test cases for different 
NullMemberRepresentation values in megatest doAllTests()

The tests that would have failed now recuse themselves whenever the 
NullMemberRepresentation value is other than  default. Other possibility 
was to form the expected result values using the current 
NullMemberRepresentation value atleast in the case of non xml file based 
tests

-Ajit





"Julian Hyde" <jhyde at pentaho.org> 
11/04/2007 07:49 AM
Please respond to
<julian at hydromatic.net>


To
"'Ajit Joglekar'" <ajogleka at thoughtworks.com>
cc
<mondrian at pentaho.org>
Subject
RE: Eigenbase perforce change 10142 for review






Ajit,

The change looks great - but I'd have used a different testing strategy.

With properties which can't readily be changed at runtime, I generally use 
a
different testing strategy. I don't try to change the property at runtime,
but modify megatest to run the test with the property set both ways.
Sometimes a test has to recuse itself if the property setting is
incompatible with the test, but that's OK. The nice thing is that then all
tests can be run from Main, or indeed standalone.

To see this, change the line in bin/megatest from

  doTest jdk1.6 access

to

  doTest jdk1.6 access 
mondrian.olap.NullMemberRepresentation=TheNullMember

and you'll get about 10 failures. Can you fix up the suite so it works
whatever the value of the property. You should also be able to back out 
the
ant changes, such as the junit-nobuild-nomember target.

Julian



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.pentaho.org/pipermail/mondrian/attachments/20071106/ea7879eb/attachment.html