Re: ITK debian packages
On Fri, Jan 20, 2006 at 05:51:05PM -0500, Steve M. Robbins wrote:
> However, the binaries in question are those of gcc-xml which you agree
> should be packaged separately anyway. So the question now is more
> like: can cableswig be built to use an external gcc-xml?
I'm pleased to report that it can be.
I uploaded gccxml package yesterday. It's new and currently awaiting approval,
but you can fetch a copy from http://people.debian.org/~smr/gccxml/
Then there are two changes that need to be done to cableswig to use
the external gccxml package.
1. Don't build GCC and GCC_XML:
diff -u -b -B -r1.20 CMakeLists.txt
--- CMakeLists.txt 7 Jun 2005 13:34:18 -0000 1.20
+++ CMakeLists.txt 22 Jan 2006 17:14:28 -0000
@@ -80,7 +80,7 @@
# Tell GCC-XML where to install.
SET(GCCXML_INSTALL_ROOT ${CableSwig_INSTALL_ROOT}/lib/CableSwig)
-SUBDIRS(SWIG Cable GCC GCC_XML Executables)
+SUBDIRS(SWIG Cable Executables)
2. Use "/usr/bin/gccxml" in the CableSwigConfig.cmake. I tried to do
this by fiddling with debian/CMakeCache.txt.debian or setting on the
cmake command line, but failed. So here's the crude way to do this:
diff -u -b -B -r1.2 CableSwigConfig.cmake.in
--- CableSwigConfig.cmake.in 16 Jun 2004 14:39:03 -0000 1.2
+++ CableSwigConfig.cmake.in 22 Jan 2006 17:19:24 -0000
@@ -4,5 +4,5 @@
SET(CableSwig_cableidx_EXE "@CableSwig_cableidx_EXE_CONFIG@")
SET(CableSwig_cswig_EXE "@CableSwig_cswig_EXE_CONFIG@")
-SET(CableSwig_gccxml_EXE "@CableSwig_gccxml_EXE_CONFIG@")
+SET(CableSwig_gccxml_EXE "/usr/bin/gccxml")
SET(CableSwig_DEFAULT_LIB "@CableSwig_DEFAULT_LIB_CONFIG@")
I then built the insighttoolkit sources with python wrappers using
this cableswig. The build is complete, but I have not tested it.
One refinement you can make is to completely omit GCC and GCC_XML
from the source.
Also I noticed that cableswig has never been released and you have
to pull it from CVS. You packaged it with the ITK release number
(2.4); is that what the ITK folks recommend? Or should we use a
CVS version scheme, e.g. 0+cvs20060122?
Regards,
-Steve
Reply to: