[Insight-developers] TIFF and LSM images failing in sanchome.megason-lab.hms : release

[Luis Ibanez]

Hi Kishore,

Thanks a lot for the report.

Now that we know this is reproducible, it is worth to track it
before the release of ITK 3.8


It is really unfortunate that this
only fails when compiling in release mode...  :-/



Looking at the list of TIFF tests, only some of them fail.

itkTIFFImageIOMultiPagesTest  Ok
itkTIFFImageIOTest            Ok
itkTIFFImageIOTest1           Ok
itkTIFFImageIOTest2           Failed
itkTIFFImageIOTest3           Failed
itkTIFFImageIOTest4           Ok
itkTIFFImageIOTest5           Failed
itkTIFFImageIOTest6           Failed
itkTIFFImageIOTest7           Ok
itkTIFFImageIOTest8           Ok
itkTIFFImageIOTest9           Ok
itkTIFFImageIOTest10          Failed
itkTIFFImageIOTest11          Ok
itkTIFFImageIOTest12          Ok


One way to start is to identify what is special about the five
tests that are failing.

Interestingly, all of them use the same executable, so the
differences are in the type of TIFF images that are feed
as input.


For example

itkTIFFImageIOTest2  reads RGBTestImageCCITTFax3.tif
itkTIFFImageIOTest3  reads RGBTestImageCCITTFax4.tif

So, the two Fax format fail...

then

itkTIFFImageIOTest5  reads RGBTestImageHuffmanRLE.tif
itkTIFFImageIOTest6  reads RGBTestImageJPEG.tif
itkTIFFImageIOTest10 reads RGBTestImageZIP.tif


Looking at the output images, all of them have grid-like artifacts,
which leads to suspects that the compression/decompression part of
process is the one failing.

Due to the way in which the test is setup, we don't know
if what is failing is the reading or the writing.

These test read an input TIFF image, save it again as TIFF,
and then compare the saved image to a TIFF baseline.

We are therefore adding duplicates of tests

       itkTIFFImageIOTest2
       itkTIFFImageIOTest3

numbered now as

       itkTIFFImageIOTest13
       itkTIFFImageIOTest14

They will read TIFF and Write MetaImage.

This will help us discriminate whether the problem is on
reading or during the writing of the images.


If you have a chance, please submit an experimental build
from your Fedora machine, or at least report your observations
of the output of these tests back to the mailing list.


     Thanks a lot for your help,


        Luis


--------------------------
Kishore Mosaliganti wrote:
> Hi Luis,
> 
> The tests you mention are failing in my machine as well: 
> http://www.cdash.org/CDash/viewTest.php?onlyfailed&buildid=131560 
> <http://www.cdash.org/CDash/viewTest.php?onlyfailed&buildid=131560>
> 
> The configuration of my machine is:
> Linux Fedora Core 9, gcc-4.3.x86_64, compiled in Release mode.
> 
> We face the same trouble on 3 PCs with similar configuration.
> 
> 
> Kishore
> 
> 
> Date: Wed, 23 Jul 2008 10:54:05 -0400
> From: Luis Ibanez <luis.ibanez at kitware.com <mailto:luis.ibanez at kitware.com>>
> Subject: [Insight-developers] TIFF and LSM images failing in
>        sanchome.megason-lab.hms : release
> To: 'Alexandre Gouaillard' <Alexandre_Gouaillard at hms.harvard.edu 
> <mailto:Alexandre_Gouaillard at hms.harvard.edu>>,
>        Insight Developers <Insight-developers at itk.org 
> <mailto:Insight-developers at itk.org>>
> Message-ID: <4887460D.70606 at kitware.com <mailto:4887460D.70606 at kitware.com>>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
> 
> 
> Hi Alex,
> 
> The release submission from
> 
> 
>             sanchome.megason-lab.hms
> 
> 
> has failing tests for TIFF and LSM images.
> 
> This happend in the release build and not in the debug build.
> 
> 
> Do you have any hints on why these tests are failing in this
> platform ?
> 
> We have submissions from other Linux 64 bits that seem to be
> working fine.
> 
> Can we suspect of a configuration problem in this machine ?
> 
> 
> 
>    Thanks for any help,
> 
> 
>        Luis