[qmtest] supporting unsupported tests

[Mark Mitchell]

Stefan Seefeld wrote:

>> What method is typically used to mark a test as not to be run for one
>> platform?  Should I be thinking about having the test run but always
>> pass (perhaps by replacing all my use of ExecTest and ShellCommandTest
>> with subclasses that check a "platform type" property before either
>> forwarding to the base class or always returning success), or is their
>> a concept of a "UNTESTABLE" state separate from "PASS" and "FAIL".
> 
> I think the UNTESTED pseudo result is most suitable to indicate that.
> However, in order for the  test database to generate such outcomes,
> you either need a custom test database, or you need to make all your
> tests depend on a common 'resource' that will fail on some platforms,
> and thus cause dependent tests not to get run.

I don't think that's true; I think that if the tests are marked as being
for a particular target, and then run on another target, QMTest will
automatically skip them.  I might be wrong about that, but that's my
recollection.

I think that the right thing to do is to create a target file containing
only the current target (i.e., the one for the machine on which you're
actually running), and use "qmtest run -T that_file".

-- 
Mark Mitchell
CodeSourcery
mark at codesourcery.com
(650) 331-3385 x713