Re: Guile 1.5.6 beta available for testing.

[Steve Tell]

On 11 Mar 2002, Marius Vollmer wrote:

> Steve Tell <address@hidden> writes:
> 
> > I guess this may be more of a libtool problem, but is there any hope of
> > helping libtool do the right thing here?
> 
> What happens when you include /usr/guile-1.5.6/lib in LD_LIBRARY_PATH?

I was definitely setting LD_LIBRARY_PATH when running, and I think I had
set it for "gmake install" too.
 
> Other than that, I have no clue.  Where does the -L/usr/lib thing come
> from?  Is it because libtool accidentally links libguilereadline
> against /usr/lib/libguile instead of against
> /usr/guile-1.5.6/lib/libguile?

As near as I can tell by wading through the libtool script and sh -x
output, libtool adds "-L/usr/lib" because libguile.la says that libguile
needs /usr/lib/libltdl.la, and /usr/lib/libltdl.a says that libltdl has
its further dependendencies specified as "-ldl" (instead of as an absolute
path to a libdl.la?)

I found two workarounds that involve hacking the libtool script.  
Both get libtool to link libguile into libguilereadline as
"/usr/guile-1.5.6/lib/libguile.so" instead of "-lguile". Even though
-L/usr/lib remains on the gcc command line, this does the right thing.


One is based on that recent posting in the libtool mailing list, and goes
like this:

*** libtool.orig        Mon Mar 11 00:22:33 2002
--- libtool     Mon Mar 11 01:21:54 2002
***************
*** 2180,2187 ****
              add="-l$name"
            else
              # We cannot seem to hardcode it, guess we'll fake it.
!             add_dir="-L$libdir"
!             add="-l$name"
            fi
  
            if test $linkmode = prog; then
--- 2180,2188 ----
              add="-l$name"
            else
              # We cannot seem to hardcode it, guess we'll fake it.
! #           add_dir="-L$libdir"
! #           add="-l$name"
!             add="$libdir/$linklib"
            fi
  
            if test $linkmode = prog; then

(this hack would probably better be done in ltmain.sh)

The other is changing a libtool configuration variable to read
"hardcode_direct=yes".  It has the same effect on that code above, but I'm
unclear on how it is used elsewhere. I'm temporarily joining the libtool
list to see if any of the experts can shed some light.  It definitely
looks like libtool is doing the wrong thing, or at least a less correct
thing.

Steve



 
--
Steve Tell  address@hidden