Working now (Was Re: HELP: Compiles one libsvn_* but not others.

On Wed, 19 Jun 2002, David Summers wrote:

> On Wed, 19 Jun 2002, Greg Stein wrote:
> > On Wed, Jun 19, 2002 at 01:09:32AM +0100, Philip Martin wrote:
> > > David Summers <david@summersoft.fay.ar.us> writes:
> > >
> > > > gcc -shared compose_delta.lo compose_editors.lo default_editor.lo diff.lo
> > > > diff_file.lo svndiff.lo text_delta.lo vdelta.lo xml_output.lo xml_parse.lo
> > > > -Wl,--rpath -Wl,/usr/local/apache2/lib -L/usr/local/apache2/lib -lcrypt
> > > > -ldl -L/usr/lib -lsvn_subr-1 -laprutil -lgdbm -ldb-4.0 -lexpat -lapr -lm
> > > > -lcrypt -lnsl -ldl -Wl,-soname -Wl,libsvn_delta-1.so.0 -o
> > > > .libs/libsvn_delta-1.so.0.0.0
> > > > /usr/bin/ld: cannot find -lsvn_subr-1
> > > > collect2: ld returned 1 exit status
> > > > libtool: install: error: relink `libsvn_delta-1.la' with the above command
> > > > before installing it
> > > > libtool: install: warning: remember to run `libtool --finish /usr/lib'
> > > > cd subversion/libsvn_ra_dav ; /bin/sh
> > > > /home/david/rpms/build/subversion-0.13.0/libtool --mode=install
> > > > /home/david/rpms/build/subversion-0.13.0/ac-helpers/install-sh -c
> > > > libsvn_ra_dav-1.la /home/david/rpms/tmp/subversion-0.13
> > > >
> > > > It sounds like the link commands are not getting the -L directory
> > > > correctly when the link happens so that it can't find the -lsvn_subr-1
> > > > can't locate the file correctly.
> >
> > You failed to note that the above problem happens at *install* time, rather
> > than during the build (as Philip was showing). *BIG* difference.
> >
> > Is the order of library installation correct? svn_subr needs to be installed
> > first, then delta, then others.
> >
> > What is the install prefix? I'm guessing /usr from the above output.
> >
> > Cheers,
> > -g
>
> Bingo. Sorry to not state that important distinction. Yes, this is
> happening at install time rather than build (compile) time. My RPM does a
> configure in the RPM setup phase, then a build (compile) in the build
> phase, and then a "make install" in the install phase.
>
> AFAICT, however, that is the normal build and install procedure. The only
> difference is I don't want to install directly in to the /usr/*
> directories immediately but install into the RPM "staging area" which is a
> directory mimicing the "real" directory structure in $RPM_BUILD_ROOT
> directory.
>
> How to best go about doing that? It used to work and then some change
> made it not work now and I'm getting around it by setting the -L
> directories during the compile.
>
> Actually I have the --prefix set for /usr in the build (compile) phase and
> $RPM_BUILD_ROOT/usr in the "install" phase.
>
> What *should* I be doing?
>
> Thanks!
>
>

Well, I'm not sure what I did wrong yesterday but now it is working
without any LDFLAGS at all in the configure.

Sorry for the false alarm. However it would be entertaining to figure out
what I was doing wrong yesterday that caused it not to work without that
stuff.

Don't spend any more time on it. I'm about to commit the new version
without all that -L stuff.

    Thanks for your input,
    - David

-- 
David Wayne Summers          "Linux: Because reboots are for upgrades!"
david_at_summersoft.fay.ar.us   PGP Key: http://summersoft.fay.ar.us/~david/pgp.txt
PGP Key fingerprint =  C0 E0 4F 50 DD A9 B6 2B  60 A1 31 7E D2 28 6D A8 
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org