[ITK-dev] HDF5 TransformIO broken with System HDF5

[Bradley Lowekamp]

Matt,

Here is my nightly build on RHEL6 (I should have said RHEL6 not RHEL5 below.) against the system HDF5:

http://open.cdash.org/viewNotes.php?buildid=3319862

It's passes OK. Looking at it closer, I see the following HDF5 file in the build directory:

./ITK-build/Testing/Temporary/UCharImage.hdf5
./ITK-build/Testing/Temporary/RGBImage.hdf5
./ITK-build/Testing/Temporary/FloatImage.hdf5
./ITK-build/Modules/IO/TransformHDF5/test/Transforms_double.hdf5
./ITK-build/Modules/IO/TransformHDF5/test/TransformsBad_double.hdf5

I didn't see those before, which made me skeptical of ITK functionality.

When SimpleITK is configured the same, HDF5 transform IO fails on the system. Not sure where the issue could be then.

Brad

On May 7, 2014, at 1:23 PM, Matt McCormick <matt.mccormick at kitware.com> wrote:

> Hi Brad,
> 
> It looks like our code coverage for the HDF TransformIO is pretty good
> [1], but I do not see a dashboard submission for those platforms that
> use a system HDF5.  I will set one up on faraway.kitware and try to
> reproduce the SimpleITK test failure.
> 
> Thanks,
> Matt
> 
> [1] http://lorensen.github.io/VTKCodeCoverage/ITK/IO/TransformHDF5/src/index.html
> 
> On Wed, May 7, 2014 at 8:59 AM, Bradley Lowekamp <blowekamp at mail.nih.gov> wrote:
>> Hello,
>> 
>> I wrote some SimpleITK tests to write out HDF5 transforms. They are failing when using ITK built with system HDF5. Specifically with RHEL5 and Ubuntu 12.04. I believe this is a critical issue as it's how ITK is being built for Linux distributions.
>> 
>> Looking through my ITK nightly build testing directory I didn't see any hdf5 files which leads me to believe this IO transform format may not be well tested.
>> 
>> I would greatly appreciate some assistance on this issue.
>> 
>> Here is my related SimpleITK issue.
>> https://issues.itk.org/jira/browse/SIMPLEITK-535
>> 
>> Thanks,
>> Brad
>>