[Insight-developers] How should I handle a trivial regression test failure?

[Luis Ibanez]

Frank,

Great!,

Thanks for committing the fix.

Once we see the extent of the impact in other platforms
we can judge whether we should fine tune parameters in
this test or replace/add a baseline image.


   Thanks


      Luis


--------------------------
frankmiller at jhmi.edu wrote:
> Luis,
> 
> Thanks for the advice. 
> 
> On my system, ImageRegistration14 is the only test that fails so I have
> committed my changes as you suggested.
> 
> Frank
> 
> On Wed, Jun 11, 2008 at 05:07:15PM -0400, Luis Ibanez wrote:
> 
>>
>>Hi Frank,
>>
>>Thanks for addressing this bug.
>>
>>The fist question is:
>>
>>  Is ImageRegistration14 the only failing test ?
>>  or do other tests fail as well ?
>>
>>
>>If this is the only failing test, we could first
>>fiddle with the registration parameters to make
>>sure that the optimizer is working its way to a
>>point very close to the optimum.
>>
>>That is, it may be that by running more iterations
>>or forcing a more stringent stopping criteria the
>>optimizer may work harder and get to the previous
>>answer.
>>
>>If this is the only test failing, I would suggest
>>that you commit the fix, and then we assess the
>>impact of the fix on all the platforms tomorrow.
>>
>>
>>The last resort will be to commit a new baseline
>>image, but before we do that we must really make
>>the case that the previous baseline image was an
>>insatisfactory answer.
>>
>>
>>   Luis
>>
>>
>>
>>------------------------------
>>frankmiller at jhmi.edu wrote:
>>
>>>Greetings,
>>>I have a fix for bug #4547 implemented locally which is a minor correction 
>>>to
>>>the way itk::CenteredTransformInitializer calculates the transform center.
>>>Interestingly, this change causes ImageRegistration14Test to fail.
>>>  http://www.cdash.org/CDash/testDetails.php?test=4638007&build=97344
>>>By visual inspection, it looks like it is performing correctly. Unmodified 
>>>output:
>>>  [ ... snip ... ]
>>>  Initial transform parameters = [0.15, 110.5, 128.5, 10, 14]
>>>  [ ... snip ... ]
>>>  Result =    Angle (radians) 0.174418
>>>   Angle (degrees) 9.99344
>>>   Center X      = 110.5
>>>   Center Y      = 128.5
>>>   Translation X = 12.9975
>>>   Translation Y = 15.9851
>>>   Iterations    = 617
>>>   Metric value  = 1.47025
>>>  [ ... snip ... ]
>>>Modified output:
>>>  [ ... snip ... ]
>>>  Initial transform parameters = [0.15, 110, 128, 10, 14]
>>>  [ ... snip ... ]
>>>  Result =    Angle (radians) 0.174399
>>>   Angle (degrees) 9.99235
>>>   Center X      = 110
>>>   Center Y      = 128
>>>   Translation X = 13.1092
>>>   Translation Y = 15.8775
>>>   Iterations    = 645
>>>   Metric value  = 1.47007
>>>  [ ... snip ... ]
>>>How should I handle this?
>>>Frank
>>>_______________________________________________
>>>Insight-developers mailing list
>>>Insight-developers at itk.org
>>>http://www.itk.org/mailman/listinfo/insight-developers
> 
>