[R] minus I and minus L flags

[Dirk Eddelbuettel]

On 20 February 2005 at 11:12, Charles Geyer wrote:
| I have been RTFM/doc/www, but I'm still lost.
| 
| How does one tell R CMD INSTALL and R CMD check (both) that libraries
| are installed in non-usual places and so -I/APPS/include (or whatever)
| is necessary in CPPFLAGS and -L/APPS/lib (similarly) is necessary
| in LDFLAGS?
| 
| I know I can hardwire them in pkg/src/Makevars, but this requires hand
| editing of that file by each installer (yuck!)
| 
| rgentlem suggested some tricks with configure, but after a lot of grovelling
| in autoconf manuals and books, I still don't get it.
| 
| Surely this is a pre-solved problem (I hopefully assert).  Is there a standard
| solution, and if so where is the example I follow?

FWIW, I have followed other packages when I crafted this for RQuantLib.

i)  src/Makevars.in  is a stanza that gets filled by Autoconf:

    edd at chibud:~/src/debian/QuantLib/RQuantLib-0.1.11> cat src/Makevars.in
    PKG_CXXFLAGS=@pkg_cxxflags@
    PKG_LIBS=@pkg_libs@

ii) The actual work is done by configure at package built time. That is the
single best method.  You can snoop tricks from dozens of CRAN packages
that do the same:

I crafted a reasonably short file:

    edd at chibud:~/src/debian/QuantLib/RQuantLib-0.1.11> wc -l configure.in
    149 configure.in

that tests for both QuantLib and Boost headers and libraries. Personally, I
find going by working example the easiest, with the documentation (for
autoconf) at my side.

That said, I also benefited from tips by Kurt who wears a black belt
in autoconf magic.

    edd at chibud:~/src/debian/QuantLib/RQuantLib-0.1.11> head -8 configure.in
    # RQuantLib configure.in by Dirk Eddelbuettel <edd at debian.org>
    # Borrowed pieces from RPgSQL, GNU Gretl, GNU R and QuantLib
    #
    # Greatly simplified thanks to quantlib-config
    # Another contribution by Kurt Hornik gratefully acknowledged
    #
    # $Id: configure.in,v 1.1 2004/12/22 04:07:57 edd Exp $

The fact that this credits RPgSQL is probably a reflection of the point in
time when I first wrote this several years ago -- CRAN was considerably
smaller at the time.

Hope this helps. Dirk

-- 
Better to have an approximate answer to the right question than a precise 
answer to the wrong question.  --  John Tukey as quoted by John Chambers