Re: [DebianGIS] HDF4 changes
On Wed, Mar 11, 2009 at 06:00:17PM +0100, Francesco P. Lovergine wrote:
> Hi folks
> As some of you could already know by looking into the d-gis repository,
> HDF4 4.2r4 edition packaging is in progress, since quite a long time.
> Historically, Debian provided for years a C/Fortran mixed shared lib
> for HDF4. This was different in respect with upstream, who provided
> at the time only not-shared versions of the two libs. The aim of
> that was providing packaged program a single binary.
> This is now changed upstream: 4.2 is distributed with shared lib support
> for C functions. The Fortran API is now still available, but only in
> static mode. The whole library is libtoolized with a new appropriate
> SONAME and version 0:0:0. A prospective new Fortran-only shlib will
> be with all probability distributed as a unified libhdf4_fortran.so.
> This is planned AFAIK.
> So what? Well, I see no reasons to still diverge from upstream, because
> that is not expected by the casual developer and it is not justifiable.
> We will do exactly the same, because any developers would expect so.
> This implies that we are going to DROP FORTRAN SUPPORT (both shared
> and static) and eventually reintroduce
> it only WHEN the expected Fortran shlib edition will be available by
> upstream, with a well defined interface. People that need to
> link the Fotran HDF4 library will have to do that themselves.
> This is not of great interest, because
> - they are very few :)
> - a static library can be easily embedded into their source package
> - that is what they should have yet done for years now.
Some good news
Guess what? I built working shared C/Fortran libraries by changing
the automake configuration in straightforward manner. I tried
linking and running with the usual test units. I'm quite
confident they should not give problems to Fotran folks but
some feedbacks about that would be nice. I'll upload this
new edition in experimental, as well. I talked with upstreams
about Fotran shlibs and they consider that a low priority
feature, so I'm inclined to maintain in main the mixed
shlibs using the upstream SONAME if none had problems with
Francesco P. Lovergine