Skip to content

Problem with dyn.load()

2 messages · Francisco J Molina, Brian Ripley

Original
Francisco J Molina
# 
I tried to dynamically load a c++ function. Everything seems to be OK
when dyn.load() is  executed but then I get this error


Error in .C("integral", sumInMean, sumInVar, boundError,
numberSampleClasses,  : 
        C/Fortran function name not in load table


The code of the function is something like this:

#include <cmath>

extern "C" {
#include "/usr/local/include/gsl/gsl_statistics_double.h"
#include "/usr/local/include/gsl/gsl_randist.h"
#include "/usr/local/include/gsl/gsl_roots.h"
#include "/usr/local/include/gsl/gsl_sf_erf.h"
#include "/usr/local/include/gsl/gsl_math.h"
#include "/usr/local/include/gsl/gsl_integration.h"

void integral( arguments ){
...
return;
}
}

I compile with:

g++ -c integral.C -fPIC -I/usr/local/include/gsl -o integral.o; R CMD
SHLIB integral.o

What could go wrong? The function modify its arguments ( the arguments
are  pointers )

I put my function with the C header directives in extern{}.

gsl is a library written in C, so I put the gsl's header directives also
inside extern{}.

I placed in my compiling directory a Makevars file with this content:

PKG_LIBS = -lm -lgsl -lgslcblas

and my R code looks like:

dyn.load('/usr/local/lib/libgslcblas.so',F,)
dyn.load('/usr/local/lib/libgsl.so',F,)
dyn.load('/home/f/p/integral.so')

Any hint?
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Brian Ripley
#
On Mon, 15 Apr 2002, Francisco J Molina wrote:

            

      
      
      
    

        
Why are you doing this?  Only the last should be necessary. The
dynamic module integral.so will be linked against any dynamic libraries
on which it depends by the dlopen call underlying dyn.load (at least
on common implementations of dlopen).

The first thing to do is to check what symbols integral.so exports (with nm
-g) and what it depends on (with R CMD ldd integral.so since the R script
sets LD_LIBRARY_PATH).

You haven't told us your OS or compiler.  An experiment with what you gave
(and no content to the integral function) worked for me on Solaris 2.7 and
gcc 3.0.4 with just the final dyn.load.  (Not a complete test, of course,
since gsl was not really being used.)