[Debian-med-packaging] Bug#672015: Build issue when building with gcc-4.7

[Andreas Tille]

Hi,

may I bring up this topic again.  From a Debian perspective we might be
on the safe side to ignore your new upstream "release" and just fix the
build issue in the current package.  However, in case you might be
interested in having the latest version in the next stable Debian
release it would help if you answer the question below.

Kind regards

        Andreas.

On Thu, May 10, 2012 at 11:59:38AM +0200, Andreas Tille wrote:
> Hi,
> 
> when building libhmsbeagle using gcc version 4.7 you get a build error
> which can be easily fixed (see patch in Debian Med SVN[1])
> 
> Before I upload a package featuring this patch I would like to clarify
> something which remains not clear to me.  The first thing is that I'm a
> bit unsure what version you might consider as "stable for Debian
> release".  The homepage[2] is featuring version 1.0.  However, our watch
> file (containing the metainformation from where to download the source
> which needs to point to tagged SVN releases because you are not
> providing a downloadable source tarball) points to SVN tags[3] and finds
> version 1.1.  So which one is the one you want Debian to distribute?
> BTW, questions like this will not occure if you would provide versioned
> a source tarball at the homepage directly.
> 
> The second issue is about the problem which was discussed in the
> beginning of this year[4] about SSE support.  I have noticed that you
> seem to have dealt with this but at least if I build not really
> successfully.  I can not build on my amd64 machine if I do not at least
> add this patch:
> 
> 
> --- libhmsbeagle-1.1.orig/configure.ac
> +++ libhmsbeagle-1.1/configure.ac
> @@ -211,6 +211,9 @@
>  if test  "$enable_sse" = yes; then
>         SSE_CFLAGS+="-DENABLE_SSE"
>      AM_CXXFLAGS="$AM_CXXFLAGS -msse2"
> +    AM_CONDITIONAL(HAVE_SSE2,true)
> +else
> +    AM_CONDITIONAL(HAVE_SSE2,false)
>  fi
> 
>  # ------------------------------------------------------------------------------
> 
> 
> Otherwise automake is claimin missing AM_CONDITIONAL statement for
> HAVE_SSE2.  Moreover I think the way you changed
> libhmsbeagle/CPU/Makefile.am is not sufficiently preventing that the
> build system tries to build libhmsbeagle-cpu-sse.la on non-Intel
> architectures.  The first line says:
> 
> lib_LTLIBRARIES=libhmsbeagle-cpu.la libhmsbeagle-cpu-sse.la
> 
> and thus libhmsbeagle-cpu-sse.la is an unconditional target which is not
> handled by the HAVE_SSE2 conditional.  It might be that the HAVE_SSE2
> condition has a completely different purpose than the patch from
> Debian[5] which was using HAVE_SSE and thus I would like to make sure
> I have understood things properly first before I upload.
> 
> Kind regards and thanks for providing libhmsbeagle as Free Software
> 
>       Andreas.
> 
> [1] http://anonscm.debian.org/viewvc/debian-med/trunk/packages/libhmsbeagle/trunk/debian/patches/gcc-4.7.patch?view=markup
> [2] http://code.google.com/p/beagle-lib/
> [3] http://beagle-lib.googlecode.com/svn/tags/
> [4] http://bugs.debian.org/656755
> [5] http://anonscm.debian.org/viewvc/debian-med/trunk/packages/libhmsbeagle/trunk/debian/patches/disable_cpu_sse_plugin.patch?revision=9433&view=markup
> 
> -- 
> http://fam-tille.de

-- 
http://fam-tille.de