Re: itk and tiff
On Wed, 2013-03-20 at 08:50 +0100, Mathieu Malaterre wrote:
> >> Anyway I should probably summarize what I have been trying to do lately.
> >> Basically, I have been trying to get all the tests to pass on a sparc
> >> machine. From what I can tell, most of them have to do with gdcm, and it
> >> not working on a big-endian machine. I noticed that gdcm itself is built
> >> with testing off, and when you turn it back on, all sorts of tests fail,
> >> so that isn't good.
> >
> > No, that's not good. Mathieu: can you comment on why testing is disabled?
>
> Technically most of the tests needs gdcmData (a set of DICOM files)
> which is not shipped. Also gdcm when testing is ON compiles a few more
> classes, which I did not want to be present in release dist.
>
> Having said that, test suite is much more important than those, I'll
> turn the test suite back on and check why this is failing on sparc.
I never checked if the failing tests had to do with missing gdcmData. So
that might be the issue.
Note, itk gives test failures with both gdcm installed from the
repository, and with the gdcm included with itk source
(ITK_USE_SYSTEM_GDCM ON and OFF). I am pretty sure it is gdcm, I have
stepped through the itkGDCMImageIOTest1 and it went into gdcm. It seemed
to have a hard time with the file, which I think might be a curve ball
from itk, as the dicom header in the test file looked non-standard to
me, but gdcm handles it on i386/amd64. I think it might have to do with
assumptions about endianess gdcm has to make when it can't figure things
out from the file, on i386/amd64 it guesses right, on big-endian
systems, it guesses wrong. At least that is my current guess, not more
than an hunch.
-Paul
Reply to: