I'm doing something wrong, but I just don't get it.

[Bruce Korb]

I am building both a library and a program that uses it.
When they are both installed, I do _NOT_ want it to be necessary for
people to have to fiddle with LD_LIBRARY_PATH in order to be able to
run the program.  I _ALSO_ do not want someone building a new package
to be forced to install the new library before it can be tested.  I keep trying
various variations and sometimes I can get it to work for me, but then other
folks have conflicts and when I fix their conflicts, then my build environment
is messed up.  I keep thinking, "OK.  I see the problem" and fix it only to
have someone eventually complain.

I guess it boils down to this:  I see the libtool argument "-rpath" referenced
all over the place, but *EXACTLY* which LDFLAG or LDADD or LDWHATEVER
is it where I put "-rpath" such that the build one is seen during the build and
the install one is used for installation?   If I omit it entirely, my test runs 
link
against the installed library.  I need a cookbook "do it like this", I guess,
because that automake manual is for automake and not libtool and the libtool
manual is for using libtool and not how to use libtoolized libraries in 
automake.
Even the Goat book doesn't make it clear.  *sigh*.  Thank you. - Bruce.