[Insight-users] IJ Test that will fail

[Luis Ibanez]

Hi Kevin,

Thanks for the clarification.


A) From your description, it looks like the code is not going
    to get better by keeping it on your side.

    ...and I mean that in a good way...        :-)

    So, I will suggest  that you go ahead and submit it to the
    Insight Journal, and include in the paper a note about your
    observations related to the different results obtained between
    parallel and serial.

    Once the code is in the Insight Journal, it will be easier for
    other developers to look at the code, and make suggestions
    for improvements, (assuming that this is something caused
    by your code), or help locate a potential bug in ITK.


B) Providing and explicit test that compares the parallel and
     serial results will be great.  Given that we know that there
     is a problem there, we may as well expose it clearly so
     we start working towards fixing it.

     I would think that a "failing" test make more sense than
     a "passing" tests in this case.


C)  Thanks for filing the bug report:
http://public.kitware.com/Bug/view.php?id=9766

       I'll take a look at the problem.


    Thanks


             Luis


------------------------------------------------------------------
On Mon, Oct 26, 2009 at 1:27 PM, Kevin H. Hobbs <hobbsk at ohiou.edu> wrote:
> My comments are below :
>
> On Mon, 2009-10-26 at 12:57 -0400, Luis Ibanez wrote:
>> A) You could add the test yourself by
>>      following the procedure described in:
>> http://www.itk.org/Wiki/ITK_Procedure_for_Adding_a_Test
>>
>> or
>>
>> B) You could send us the test's source code
>>      and we could take care of adding it to the
>>      toolkit.
>>
>
> I'm sorry, my original e-mail was not clear enough.
>
> I was trying to ask what I should do for an Insight Journal submission
> that I'm working on. ( I now have permission to submit. ) I was not
> asking about what I should do if I found a bug in ITK or IJ.
>
> I know that my implementation of MPI Fast Marching will not produce the
> exact same results as the serial algorithm but the MPI version seems to
> work well enough for my requirements.
>
> Should I directly compare the results from a serial run to those from a
> parallel run and produce a failing test? This would make it very clear
> what the implementation does not do.
>
> Or should I produce a passing test that verifies the difference?
>
>>
>> Please let us know what is your preference.
>>
>>
>>  BTW: It would seem that filing a bug report
>> on this issue may also be a good idea.
>
> As to bugs in IJ and ITK: I already filed bug reports (or I have been
> convinced that what I saw was not a bug) on everything I've noticed
> recently.
>
> I filed a feature request for IJ to support MPI.
>
> I filed a bug report for ITK
> http://public.kitware.com/Bug/view.php?id=9766 about how extracting a
> piece from a reader which has had UpdateOutputInformation called but not
> Update produces incorrect results. This bug report has code attached
> which could become a test.
>