Hello Gianfranco,
On Thu, 2015-05-07 at 13:09 +0000, Gianfranco Costamagna wrote:
> (I guess you already have the patch :p
> http://itk.org/gitweb?p=ITK.git;a=commitdiff;h=7f54e864 )
As I said, currently I have no time to look into it. Anyway, we
shouldn't do a new upload until 4.7 has passed the NEW queue.
> >For itk3.20 it's a bit more complicated, because upstream doesn't
> >support it anymore, It would probably be better to file bugs against all
> >packages that still depend on ITK3.20 and ask to port them (or get new
> >versions into the archive), and when this is done we could get rid of
> >this old version altogether.
> the problem is the same, aka #780659...
> So many packages should drop non amd64/i386 builds...
"many" is actually very little:
libinsighttoolkit3.20 Reverse Depends:
python-vmtk + libvmtk1.0:
the new version 1.2 doesn't list ITK as build dependency on their
homepage
itksnap:
latest version in experimental is already ITK4
libgofigure0 + gofigure2:
unmaintained since 2011 ...
ginkgocadx:
currently doesn't build with ITK4 and VTK6.
libcamitk3:
seems to support itk4.
> (no problem for me, maybe itk4 might increase the compatibility archs?)
As of #780659 one of the problems seems to be GDCM on big endian.
> exactly I got them after applying the patch on ITK.
>
>
> ../../bin/libITKIO.so.3.20.1: undefined reference to `gdcm::SerieHelper::SetDirectory(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, bool)'
> ../../bin/libITKIO.so.3.20.1: undefined reference to `gdcm::SerieHelper::CreateUniqueSeriesIdentifier[abi:cxx11](gdcm::File*)'
> ../../bin/libITKIO.so.3.20.1: undefined reference to `gdcm::StringFilter::ToString[abi:cxx11](gdcm::Tag const&) const'
This is actually C++11 and not C11, but yes, I read elsewhere that C++11
may not be abi-compatible to C++98/03. The workaround here would be to
disable C++11, (i.e compile with -std=c++03).
> So I'm wondering if we should just remove it from the archive...
IMHO the proper approach is to initiate a transition phase (like it is
done with vtk and dcmtk). Given that these are very few packages it
should be possible to get this done until stretch goes into freeze.
> (or maybe a gdcm rebuild fixes the issue?)
Probably, but this would mean to wait until gcc-5 becomes the official
default.
Best,
Gert
Attachment:
signature.asc
Description: This is a digitally signed message part