[Insight-developers] Need to add images for new tests to Data -- how to do that in a Gerrit topic?

[Bill Lorensen]

I did some experiments. I reported them at gerrit:
http://review.source.kitware.com/#change,301

Here is a summary:
Maybe I don't understand what is going on.

1) I ran ctest -V -R itkImageFileWriterUpdateLargestPossibleRegionTest
Worked great. It reported:
The following tests passed:
itkImageFileWriterUpdateLargestPossibleRegionTest_fetchData
itkImageFileWriterUpdateLargestPossibleRegionTest

2) Then I hand edited cthead1.png.md5 simulating a change image.
Everything still passed. Same results.

3) Then I deleted cthead1.png.md5. Everything still passed. Same results.

4) Then I disconnected my ethernet cable. Everything still passed. Same results.



On Fri, Nov 5, 2010 at 3:22 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
> First, I do not agree that the system is crippled. It could certainly
> be improved.
>
> Adding complexity is not an improvement.
>
> I suggest we split the data into large, small and baselines. Keep the
> small and baselines with the repository (the way it used to be, no
> submodule). Use MIDAS for the large data.
>
> Bill
>
> On Fri, Nov 5, 2010 at 2:50 PM, Luis Ibanez <luis.ibanez at kitware.com> wrote:
>> Gaetan,
>>
>>
>> 2010/11/5 Gaëtan Lehmann <gaetan.lehmann at jouy.inra.fr>
>>>
>>> This method looks interesting for the large images.
>>> IMO, it doesn't look that interesting for the small test images we are
>>> used to have in Testing/Data, compared to a direct inclusion in the ITK
>>> repository.
>>>
>>> The procedure to add a new image on midas looks quite complex also -
>>> nothing close to the "cvs add" we had before ITK 4 and git.
>>>
>>
>>
>> The method we use to have with cvs was certainly
>> simple, but also quite disorganized.  A quick look
>> at the Testing/Data/Input directory will show that
>> we had no systematic organization for the images
>> that get included in that directory.
>>
>> Starting by the fact that tree images,
>> take half of the size of the entire directory.    :-(
>>
>> * GradientRecursiveGaussianImageFilterTest.mha (1.5Mb)
>> * TensorsCorpusCallosum.raw (3.7Mb)
>> *  DwiCorpusCallosum.raw  (4.8Mb)
>>
>> and each one of those giant images are used in a
>> single tests (out of the 1,700+ tests that we have).
>>
>>
>>>
>>> Do we really need a new layer of complexity like this one? I don't think
>>> so.
>>
>>
>> It depends on whether the added complexity
>> is helping to overcome the limitations of the
>> simpler older method.
>>
>>
>>>
>>> Things should be kept simple when possible.
>>>
>>
>> Simplicity is great,
>> when it doesn't implies a crippled system.
>>
>> This is our only chance to create the ITK for the next
>> ten years. If we are going to claim to provide support
>> for large images, we should test with large images.
>>
>> Currently we barely have 3D tests,
>> and we are talking about providing serious
>> support for 4D images.
>>
>> This level of testing infrastructure requires
>> a more scalable method.
>>
>>
>>
>>       Luis
>>
>>
>>
>