[R] shared libraries on FreeBSD 4.*

[Kurt Hornik]

>>>>> loren mccarter writes:

> Thanks for your help. As was suggested by Prof Ripley, the problem
> appears to be related to the environment variables for FLIB, LIBS and
> LD_LIBRARY_PATH. I was able to solve the problem with RODBC by setting
> the LIBS variable in the current shell immediately before the install
> (LIBS was already set correctly in $RHOME/etc/Makeconf but RODBC must
> not have been reading it). Below is a log that might be useful for
> understanding the issue. It uses a modified version of the
> RODBC/configure script that simply prints to screen the relevant
> environment variables before AC_SEARCH_LIBS is invoked. I have yet to
> figure out what's happening with RSPython, but that is probably more
> approporate for omegahat-help if Prof Ripley's suggestion does not
> work. Thanks!

> Loren

> /* First showing that build does not work despite correct settings in  
>  * LD_LIBRARY_PATH  
>  * Note that LIBS='' despite correct settings in $RHOME/etc/Makeconf
>  */

Hmm ... why did you think that LIBS from R_HOME/etc/Makeconf was being
used?  (This is the value that was obtained from the R configure
process, and I think it is a leftover from the past.  Older versions of
R would add $LIBS when linking shlib bundles for add-on packages.)

The recommended is documented in the INSTALL help page:

     Some package sources contain a `configure' script that can be
     passed arguments or variables via the option `--configure-args'
     and `--configure-vars', respectively, if necessary.  The latter is
     useful in particular if libraries or header files needed for the
     package are in non-system directories.  In this case, one can use
     the configure variables `LIBS' and `CPPFLAGS' to specify these
     locations (and set these via `--configure-vars'), see section
     `Configuration variables' in ``R Installation and Adminstration''
     for more information.  One can also bypass the configure mechanism
     using the option `--no-configure'.

but of course you can also set LIBS and export before calling INSTALL.
The question is whether

	LIBS=... R CMD INSTALL /path/to/package

should work (in which case INSTALL should always export LIBS and
CPPFLAGS)?

It is not clear whether one should use the information about libs and
headers in non-standard locations which was used when building R also
when building add-on packages.  What we could maybe do was to provide
some autoconf macros for accessing such information, and we definitely
need to do this for BLAS (which has been on my TODO list for some
time).

Best,
-k
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._