On Wed, Apr 16, 2014 at 1:10 AM, Mark H Weaver <mhw@netris.org> wrote:

Can you try "make V=1" and show us the output that corresponds to
the above?

See attached makelog file. It looks like the libtool file itself has the ´hard code´uninitialised (libdir)´. This is probably not very useful, I'm sorry. I will try again, but I am working on a machine that is not net connected.

The relevant rule starts with "libguile-@GUILE_EFFECTIVE_VERSION@.la:"
in libguile/Makefile.in (line 2241 in guile-2.0.11). In the generated
libguile/Makefile, the relevant rule starts with "libguile-2.0.la:".

I see "-rpath $(libdir)" in the associated command, but it seems as if a
bare "(libdir)" is getting passed to libtool somehow.

What kind of system are you doing this on? What version of make?

See the makev attachment to this message. It's the output of uname, ldconfig and make -v

Can you search for "rl_get_keymap" in "config.log" and see what went
wrong here? It tries to compile a simple test program, and it would be
useful to see what error message(s) it printed.

See attatched readline_prob file. It looks like something to do with curses. Does the test program need ncurses?

You probably already know this, but on a GNU/Linux system it's important
to run "ldconfig" as root after installing libraries from source code.

I used to know that, but I'd forgotten it! I recall the bad old days when upgrading glibc was a major PITA, and everything was in such a state of flux that one ended up needing seven different versions of glibc, then it was necessary. But I haven´t done this for a long time ...

Thanks for your attention, and sorry for the reply latency.

Ian