> Il Mercoledì 2 Luglio 2014 13:15, Gert Wollny <gw.foss...@gmail.com> > The failing command seems to be > > /usr/bin/gccxml -fxml-start=_cable_ > -fxml=/«PKGBUILDDIR»/BUILD/Wrapping/Modules/ITKCommon/vcl_complex.xml > --gccxml-gcc-options /«PKGBUILDDIR»/BUILD/Wrapping/Modules/ITKCommon/gcc_xml.inc > -DCSWIG -DCABLE_CONFIGURATION -DITK_MANUAL_INSTANTIATION > /«PKGBUILDDIR»/BUILD/Wrapping/Modules/ITKCommon/vcl_complex.cxx > /usr/bin/cmake -E cmake_progress_report /«PKGBUILDDIR»/BUILD/CMakeFiles > > Seems like something is wrong with the generated > Wrapping/Modules/ITKCommon/gcc_xml.inc >
> is anybody making any progress with this? > I found that with DITK_WRAP_PYTHON:BOOL=OFF > but this seems not so helpful, unless we want to remove the python support, > right? Indeed. > I'm really lost about any possibility to fix this bug. I had another take on it, and it seems that it is not actually a mis-configuration on the part of gccxml, but ITK seems to expects that SSE is enabled when emmintrin.h is available. If emmintrin.h is available then it is included. In gccxml for gcc-4.7/4.8 the replacement emmintrin.h does #ifndef __SSE2__ # error "SSE2 instruction set not enabled" #else ... #endif but in gcc-4.9 some gcc pragmas are used that are probably not understood by gccxml. Still, I do not understand how the compile went through with earlier versions of gcc, considering that the default build flags completely disable SSE on Debian-i386 (since it still supports i486 processors) and the #error directive should have been hit. The relevant include file tests that sets ITK_HAVE_EMMINTRIN_H comes from Modules/Core/Common/CMake/CheckSupportForSSERounding.cmake and it's set in BUILD/Modules/Core/Common/itkConfigure.h > The package is quite big, and studying it requires too much effort for my time > > I hope somebody else will find a solution for this bug, Well, I've added Matt from Kitware to the CC, maybe he can shed some light on what might be going on here. Best Gert -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
