Re: Compiling ORBit is fun, isn't it?

[Martin Baulig <martin home-of-linux org>]

On Wed, 21 Oct 1998, Michael Hudson wrote:

> On 20 Oct 1998, Elliot Lee wrote:
> > On Tue, 20 Oct 1998 16:27:20 +0100 (BST), Michael Hudson <mwh21@cam.ac.uk>
> > wrote:
> > 
> > >1) My usual routine is to type
> > >./autogen.sh --disable-static && make clean && make
> > >This fails for a bundle of reasons.
> > >1.1) I get messages about not finding the static library for libIDL.la.
> > >OK, I can cope with that, take --disable-static out. Still, I *like*
> > >shared libraries (I'm running out of hard disk). Is there a good reason
> > >for this?
> > 
> > The ORBit/test/ programs need the static libs. 
> 
> So does orbit-idl, for some reason. I've commented the line in
> src/idl-compiler/Makefile.am that makes it link statically out. It seems
> to work just fine.
> 
> > If you submit a patch to
> > build test/ only when static libs are enabled, we can work it in.
> 
> OK, this is very simple minded, but it works. This includes the
> src/idl-compiler/Makefile.am patch mentioned above:

Short question:

why do the tests require static linking - can't we make them work with
shared libraries as well ?

Martin

> 
> -- start patch --
> Index: Makefile.am
> ===================================================================
> RCS file: /cvs/gnome/ORBit/Makefile.am,v
> retrieving revision 1.5
> diff -u -r1.5 Makefile.am
> --- Makefile.am	1998/09/21 05:54:35	1.5
> +++ Makefile.am	1998/10/21 16:55:56
> @@ -1,4 +1,4 @@
> -SUBDIRS=libIDL src test
> +SUBDIRS=libIDL src @TEST_SUBDIR@
>  
>  bin_SCRIPTS=orbit-config
>  
> Index: configure.in
> ===================================================================
> RCS file: /cvs/gnome/ORBit/configure.in,v
> retrieving revision 1.39
> diff -u -r1.39 configure.in
> --- configure.in	1998/10/18 17:58:42	1.39
> +++ configure.in	1998/10/21 16:55:56
> @@ -148,6 +148,12 @@
>  ], , AC_DEFINE(HAVE_LIMITED_WRITEV),AC_DEFINE(HAVE_LIMITED_WRITEV))
>  AC_SUBST(HAVE_LIMITED_WRITEV)
>  
> +if test "$enable_static" = "no"; then
> +	TEST_SUBDIR=
> +else
> +	TEST_SUBDIR=test
> +fi
> +AC_SUBST(TEST_SUBDIR)
>  AC_CONFIG_SUBDIRS(libIDL)
>  
>  AC_OUTPUT(Makefile
> Index: src/idl-compiler/Makefile.am
> ===================================================================
> RCS file: /cvs/gnome/ORBit/src/idl-compiler/Makefile.am,v
> retrieving revision 1.13
> diff -u -r1.13 Makefile.am
> --- Makefile.am	1998/09/10 22:06:06	1.13
> +++ Makefile.am	1998/10/21 16:55:57
> @@ -3,7 +3,7 @@
>  INCLUDES=-I$(top_builddir) -I$(top_builddir)/src \
>  	 -I$(top_srcdir) -I$(top_srcdir)/src @GLIB_CFLAGS@
>  orbit_idl_DEPENDENCIES = $(top_builddir)/libIDL/libIDL.la
> -orbit_idl_LDFLAGS = -static
> +#orbit_idl_LDFLAGS = -static
>  orbit_idl_LDADD = \
>  	$(top_builddir)/libIDL/libIDL.la \
>  	@GLIB_LIBS@ \
> -- end patch --
>  
> > >1.2) make clean is broken when the directory is clean anyway.
> > >Something in `test' depends on something that is built into `src'. This is
> > >fine if ORBit is built, but not if it isn't.
> > >Does that make sense?
> > >This would seem to be (after poring over make -d|less) because `Makefile'
> > >depends on $(BUILT_SOURCES), and BUILT_SOURCES contains `echo.h', which is
> > >built from `echo.idl' by $(top_builddir)/src/idl-compiler/orbit-idl. This
> > >will not exist if the directory was not built up.
> > 
> > automake problem, NMF... :)
> > 
> > -- Elliot
> 
> Yes, automake seems less magical than it did before I started fighting
> with this.
> 
> 
> Michael Hudson
> Jesus College
> Cambridge
> mwh21@cam.ac.uk
> 
> 
> 
> 
> -- 
>          To unsubscribe: mail gnome-list-request@gnome.org with 
>                        "unsubscribe" as the Subject.
> 

-----------------------------------------------------------------
   Martin Baulig - Angewandte Mathematik - Universitaet Trier
   martin@home-of-linux.org, http://www.home-of-linux.org/
------------------------------------------------------------------