Endianess problem with dcmseg_roundtrip test

[alciregi]

Hello.
I'm trying to create an RPM package using the Fedora build infrastructure.
While dcmtk 3.6.6 builds successfully and all the tests pass on various architectures (i.e. x86_64, aarch64, ppc64le, etc.) I encountered a problem with s390x.
The dcmseg_roundtrip and dcmect_roundtrip tests fail, because (forgive me, but I'm not a developer):

Code: Select all

Dump produced: 
# Dicom-Data-Set
# Used TransferSyntax: Big Endian Explicit
...
Dump expected: 
# Dicom-Data-Set
# Used TransferSyntax: Little Endian Explicit
...

Here you can find the full log: https://kojipkgs.fedoraproject.org//wor ... /build.log of the build process.

Thank you if you would provide any hint.

[J. Riesmeier]

Could you please test the latest development version (from the public git repository)? I guess this issue has already been fixed with this commit: http://git.dcmtk.org/?p=dcmtk.git;a=com ... c5b1b32175

[alciregi]

Thank you for the answer.
I applied the patch. I'm still able to compile for the various architectures; for x360s the dcmseg_roundtrip test now passes, but dcmect_roundtrip still fails.

Code: Select all

261/263 Test #263: dcmect_roundtrip .............................................***Failed    0.05 sec
FAILED test 'dcmect_roundtrip' at /builddir/build/BUILD/dcmtk-3.6.6/dcmect/tests/t_roundtrip.cc:622: frame[pix] == numInstance + 1
FAILED test 'dcmect_roundtrip' at /builddir/build/BUILD/dcmtk-3.6.6/dcmect/tests/t_roundtrip.cc:622: frame[pix] == numInstance + 1
FAILED test 'dcmect_roundtrip' at /builddir/build/BUILD/dcmtk-3.6.6/dcmect/tests/t_roundtrip.cc:622: frame[pix] == numInstance + 1
FAILED test 'dcmect_roundtrip' at /builddir/build/BUILD/dcmtk-3.6.6/dcmect/tests/t_roundtrip.cc:622: frame[pix] == numInstance + 1
FAILED test 'dcmect_roundtrip' at /builddir/build/BUILD/dcmtk-3.6.6/dcmect/tests/t_roundtrip.cc:622: frame[pix] == numInstance + 1
FAILED test 'dcmect_roundtrip' at /builddir/build/BUILD/dcmtk-3.6.6/dcmect/tests/t_roundtrip.cc:622: frame[pix] == numInstance + 1
FAILED test 'dcmect_roundtrip' at /builddir/build/BUILD/dcmtk-3.6.6/dcmect/tests/t_roundtrip.cc:622: frame[pix] == numInstance + 1
FAILED test 'dcmect_roundtrip' at /builddir/build/BUILD/dcmtk-3.6.6/dcmect/tests/t_roundtrip.cc:622: frame[pix] == numInstance + 1

Thanks.

[alciregi]

So, debugging a little, it seems that

Code: Select all

frame[pix] == numInstance + 1

is

Code: Select all

256 == 1

and

Code: Select all

512 == 2

[alciregi]

Indeed, if I use

Code: Select all

            OFCHECK(__bswap_16(frame[pix]) == numInstance + 1);

on line 622, it works.

[Michael Onken]

Hi,

I'll look into it on Monday. Thank you for debugging.

Best regards,
Michael

[Michael Onken]

Hi alciregi,

I added the swap in a corresponding way in DCMTK -- if it goes through the internal testing tonight, I'll publish the fix to master (and let you know).

Best,
Michael

[Michael Onken]

Hi, the related commits are now public on DCMTK master. It would be great if you could try this out in the next days.

Thank you,
Michael

[alciregi]

Hi, the related commits are now public on DCMTK master. It would be great if you could try this out in the next days.

Thank you very much.
I will try ASAP.

[alciregi]

Yes.
Using this patch http://git.dcmtk.org/?p=dcmtk.git;a=com ... a348922bc6 now the test passes successfully.
Thanks!