Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Building shared libraries in subdirectories

Tags:

c

r

makefile

I am trying to build an R package that uses some C code. I have a C library that is compiled into an executable, that can be called from the command line. There is a Makefile associated with it.

I am trying to grok the information here, where it says

If you want to create and then link to a library, say using code in a subdirectory, use something like

 .PHONY: all mylibs

 all: $(SHLIB)
 $(SHLIB): mylibs

 mylibs:
         (cd subdir; make) 

Be careful to create all the necessary dependencies, as there is a no guarantee that the dependencies of all will be run in a particular order (and some of the CRAN build machines use multiple CPUs and parallel makes).

If I create a new subdirectory of the src folder in my package, called someLibrary, with the code and Makefile unchanged, and in turn, in the original Makevars file for my package I add the above code unchanged, then I will be able to build that shared library to be exported using useDynLib?


Edit 1:

Following information here, I changed the Makefile to create a shared library by adding

CFLAG = -fPIC -g -O3 
LDFLAGS= -shared

However, this leads to the problem that the .so file is not exported directly to the libs directory of the package. If I hard code the path into the target, then the file is sent to the libs directory of the package (this is all by the way of calls to R CMD INSTALL myPackage).


Edit 2:

Lastly, I would like to know how to make calls to the shared library, given that it has a main() method that I could call from the command line executable.

What is the procedure to expose this to the R NAMESPACE, so that it can be called via .Call?

PS. Please let me know if I should make the last bit a separate question.

like image 377
tchakravarty Avatar asked Jun 16 '13 12:06

tchakravarty


People also ask

What is the difference between a directory and a subdirectory?

Files are organized by storing related files in the same directory. In a hierarchical file system (that is, one in which files and directories are organized in a manner that resembles a tree), a directory contained inside another directory is called a subdirectory.

Is a subdirectory a folder?

A subdirectory is a type of website hierarchy under a root domain that uses folders to organize content on a website. A subdirectory is the same as a subfolder and the names can be used interchangeably.

How many subdirectories can a directory have in an OS?

Answer. What is the maximum number of subdirectories that a single directory might contain? Subdirectories might be limited by the number of available inodes and maxdirsize setting. There is a limit of 99,998 directories per sub-directory.

What are the subdirectories?

In a computer file system, a subdirectory is a directory that is contained another directory, called a parent directory. A parent directory may have multiple subdirectories.


Video Answer


2 Answers

The original question author asked, in the comments on the question, for an example of using Automake, Libtool and LDADD to link a program compiled in one directory with a shared library compiled in a second directory. This is a complete, standalone, fully worked example of how to compile a library and program in separate directories of the same source tree using GNU Autotools.

Directory structure

We need to set up a directory structure as follows:

├ A/
│ ├ Makefile.am
│ ├ helloworld.c
│ └ helloworld.h
├ B/
│ ├ Makefile.am
│ └ foo.c
├ configure.ac
└ Makefile.am

The shared library will be compiled in directory A/, and the program that uses it in directory B/.

Writing the source code

There are three source files.

A/helloworld.c is the source code of the library. It exports one procedure, say_hello(), which prints the message "Hello world!" to standard output.

#include <stdio.h>
#include "helloworld.h"

void
say_hello (void)
{
  printf ("Hello world!\n");
}

A/helloworld.h is the header file that contains the declaration of the say_hello() function. It only has one line:

void say_hello (void);

Finally, B/foo.c is the source code of the program that uses the shared library. It includes the library's header file, and calls say_hello().

#include <helloworld.h>

int
main (int argc, char **argv)
{
  say_hello ();
  return 0;
}

Compiling the library

We will use Automake and Libtool to compile the shared library. Both of these tools are very powerful, and actually remarkably well-documented. The manuals (Automake, Libtool) are definitely worth reading.

The A/Makefile.am file is used by automake to control the compilation of the library.

# We're going to compile one libtool library, installed to ${libdir},
# and named libhelloworld.
lib_LTLIBRARIES = libhelloworld.la

# List the source files used by libhelloworld.
libhelloworld_la_SOURCES = helloworld.c

# We install a single header file to ${includedir}
include_HEADERS = helloworld.h

Compiling the program

The B/Makefile.am file controls compilation of the library. We need to use the LDADD variable to tell automake to link against the library we compiled earlier.

# Compile one program, called foo, and installed to ${bindir}, with a single C
# source file.
bin_PROGRAMS = foo
foo_SOURCES = foo.c

# Link against our uninstalled copy of libhelloworld.
LDADD = $(top_builddir)/A/libhelloworld.la

# Make sure we can find the uninstalled header file.
AM_CPPFLAGS = -I$(top_srcdir)/A

Controlling the build

Finally, we need a top level Makefile.am to tell Automake how to build the project, and a configure.ac file to tell Autoconf how to find the required tools.

The top-level Makefile.am is fairly straightforward:

# Compile two subdirectories.  We need to compile A/ first so the shared library is
# available to link against.
SUBDIRS = A B

# libtool requires some M4 scripts to be added to the source tree.  Make sure that
# Autoconf knows where to find them.
ACLOCAL_AMFLAGS = -I m4

Finally, the configure.ac file tells Autoconf how to create the configure script.

AC_INIT([libhelloworld], 1, [email protected])

# This is used to help configure check whether the source code is actually present, and
# that it isn't being run from some random directory.
AC_CONFIG_SRCDIR([A/helloworld.c])

# Put M4 macros in the m4/ subdirectory.
AC_CONFIG_MACRO_DIR([m4])

# We're using automake, but we want to turn off complaints about missing README files
# etc., so we need the "foreign" option.
AM_INIT_AUTOMAKE([foreign])

# We need a C compiler
AC_PROG_CC

# Find the tools etc. needed by libtool
AC_PROG_LIBTOOL

# configure needs to generate three Makefiles.
AC_CONFIG_FILES([A/Makefile
                 B/Makefile
                 Makefile])
AC_OUTPUT

Testing it out

Run:

$ autoreconf -i
$ ./configure
$ make
$ B/foo

You should see the required output: "Hello world!"

like image 51
Peter T.B. Brett Avatar answered Oct 12 '22 13:10

Peter T.B. Brett


callable.c

#include <stdio.h>

int main(int argc, char **argv) {
    printf("Hello World\n");
    return 0;
}

To get a .so file from the C file callable.c use

R CMD SHLIB callable.c

So now we have callable.so

R normally requires that all C arguments be pointers, but if your main method ignores argc, then we can get around this. So to call main from callable.so from R we can write a method like so.

main <- function() {
    dyn.load("callable.so")
    out <- .C("main", argc=0, argv="")
}

Calling that main function will run the C main function.

like image 28
lthreed Avatar answered Oct 12 '22 13:10

lthreed