[Insight-developers] itkImageFileWriterPastingTest2_02 failures

[Bill Lorensen]

Brad,

We don't like to see failing tests linger. They interfere with newly
failing tests. People stop looking at the failing tests when there are
tests that always fail.

Can we either fix the problem or remove the test until someone fixes it?

Thanks,

Bill

On Sat, Jan 24, 2009 at 10:59 AM, Bradley Lowekamp
<blowekamp at mail.nih.gov> wrote:
> Hello Bill,
> This test is expected to fail. The output of the difference image does look
> right. However at the end of the output is the following:
>
> MetaImage: M_ReadElements: data not read completely
>    ideal = 124992 : actual = 82256
> MetaImage: M_ReadElements: data not read completely
>    ideal = 124992 : actual = 82256
>
> The file is actually an invalid MetaIO file.
> What the test is trying to do is paste a region to the middle of a image
> file when there is no file. The expected behavior should be to create a
> valid image file where just the pixels of the paste region are defined. The
> test in fact will need to be modified to check just this
> subregion. Unfortunately, metaio does not currently conform to this
> behavior. The way metaio currently works is to just write the ioregion at
> the appropriate spot of the file, and does not make the file the full valid
> size. It only extends the file to where the ioregion is. That is the file
> output of this test is truncated and not a valid file. This works when
> streaming the entire file, like itkImageFileWriterStreamingTest1 or
> itkImageFileWriterPasting1.
> Currently, no changes have been made it itkMetaImageIO, or the
> MetaImage utility library. This was does on purpose to verify backwards
> compatibility. So incase some one had there old streaming IO it would still
> work with out modifications.
> The next step is to modify MetaIO to improve it's robustness in cases like
> this and a couple other, so that it will either create a valid output or
> throw and exception.
> -Brad
> On Jan 24, 2009, at 8:29 AM, Bill Lorensen wrote:
>
> Brad,
>
> I think the failing test itkImageFileWriterPastingTest2_02
> http://www.cdash.org/CDash/testDetails.php?test=15954282&build=256618
>
> may be producing correct results. This test should paste an io region
> into a new image. Seems like it is doing that. I think that it just
> needs a valid regression image. It's companion test,
> itkImageFileWriterPastingTest2_01 pastes into an existing image. It
> passes.
>
> Comments?
>
> Bill
>
>
>
> ========================================================
>
> Bradley Lowekamp
>
> Lockheed Martin Contractor for
>
> Office of High Performance Computing and Communications
>
> National Library of Medicine
>
> blowekamp at mail.nih.gov
>