LDP/LDP/howto/docbook/C++-dlopen/C++-dlopen.xml

783 lines
26 KiB
XML

<?xml version="1.0" encoding="utf-8"?>
<!DOCTYPE article PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN"
"http://docbook.org/xml/4.2/docbookx.dtd">
<article id="C---dlopen">
<articleinfo>
<title>C++ dlopen mini HOWTO</title>
<author>
<firstname>Aaron</firstname>
<surname>Isotton</surname>
<affiliation>
<address><email>aaron@isotton.com</email></address>
</affiliation>
</author>
<pubdate>2006-03-16</pubdate>
<revhistory id="revhistory">
<revision>
<revnumber>1.10</revnumber>
<date>2006-03-16</date>
<authorinitials>AI</authorinitials>
<revremark>Changed the license from the GFDL to the GPL. Fixed usage
of dlerror; thanks to Carmelo Piccione. Using a virtual destructor
in the example; thanks to Joerg Knobloch. Added Source Code
section. Minor fixes.</revremark>
</revision>
<revision>
<revnumber>1.03</revnumber>
<date>2003-08-12</date>
<authorinitials>AI</authorinitials>
<revremark>Added reference to the GLib Dynamic Module
Loader. Thanks to G. V. Sriraam for the pointer.</revremark>
</revision>
<revision>
<revnumber>1.02</revnumber>
<date>2002-12-08</date>
<authorinitials>AI</authorinitials>
<revremark>Added FAQ. Minor changes</revremark>
</revision>
<revision>
<revnumber>1.01</revnumber>
<date>2002-06-30</date>
<authorinitials>AI</authorinitials>
<revremark>Updated virtual destructor explanation. Minor changes.</revremark>
</revision>
<revision>
<revnumber>1.00</revnumber>
<date>2002-06-19</date>
<authorinitials>AI</authorinitials>
<revremark>Moved copyright and license section to the
beginning. Added terms section. Minor changes.</revremark>
</revision>
<revision>
<revnumber>0.97</revnumber>
<date>2002-06-19</date>
<authorinitials>JYG</authorinitials>
<revremark>Entered minor grammar and sentence level changes.</revremark>
</revision>
<revision>
<revnumber>0.96</revnumber>
<date>2002-06-12</date>
<authorinitials>AI</authorinitials>
<revremark>Added bibliography. Corrected explanation of extern
functions and variables.</revremark>
</revision>
<revision>
<revnumber>0.95</revnumber>
<date>2002-06-11</date>
<authorinitials>AI</authorinitials>
<revremark>Minor improvements.</revremark>
</revision>
</revhistory>
<abstract>
<para>How to dynamically load C++ functions and classes using
the <function>dlopen</function> API.</para>
</abstract>
</articleinfo>
<section id="intro">
<title>Introduction</title>
<para>
A question which frequently arises among Unix C++ programmers is
how to load C++ functions and classes dynamically using the
<function>dlopen</function> API.
</para>
<para>In fact, that is not always simple and needs some
explanation. That's what this mini HOWTO does.</para>
<para>An average understanding of the <systemitem>C</systemitem>
and <systemitem>C++</systemitem> programming language and of the
<function>dlopen</function> API is necessary to understand this
document.</para>
<para>This HOWTO's master location is <ulink
url="http://www.isotton.com/howtos/C++-dlopen-mini-HOWTO/"/>.</para>
<section id="copyright">
<title>Copyright and License</title>
<para>This document, <emphasis>C++ dlopen mini HOWTO</emphasis>, is
copyrighted (c) 2002-2006 by <emphasis>Aaron Isotton</emphasis>.
Permission is granted to copy, distribute and/or modify this
document under the terms of the GNU General Public License, Version
2, as published by the Free Software Foundation.</para>
</section>
<section id="disclaimer">
<title>Disclaimer</title>
<para>
No liability for the contents of this document can be
accepted. Use the concepts, examples and information at your
own risk. There may be errors and inaccuracies, that could be
damaging to your system. Proceed with caution, and although
this is highly unlikely, the author(s) do not take any
responsibility.
</para>
<para>
All copyrights are held by their by their respective owners,
unless specifically noted otherwise. Use of a term in this
document should not be regarded as affecting the validity of
any trademark or service mark. Naming of particular products
or brands should not be seen as endorsements.
</para>
</section>
<section id="credits">
<title>Credits / Contributors</title>
<para>
In this document, I have the pleasure of acknowledging (in
alphabetic order):
</para>
<itemizedlist>
<listitem>
<para>Joy Y Goodreau <email>joyg (at) us.ibm.com</email> for
her editing.</para>
</listitem>
<listitem>
<para>D. Stimitis <email>stimitis (at) idcomm.com</email>
for pointing out a few issues with the formatting and the
name mangling, as well as pointing out a few subtleties of
<literal>extern "C"</literal>.</para>
<para>Many unnamed others pointing out errors or giving tips to
improve this howto. You know who you are!</para>
</listitem>
</itemizedlist>
</section>
<section id="feedback">
<title>Feedback</title>
<para>
Feedback is most certainly welcome for this document. Send
your additions, comments and criticisms to the following email
address: <email>aaron@isotton.com</email>.
</para>
</section>
<section>
<title>Terms Used in this Document</title>
<variablelist>
<varlistentry>
<term><function>dlopen</function> API</term>
<listitem>
<para>The <function>dlclose</function>,
<function>dlerror</function>,
<function>dlopen</function> and
<function>dlsym</function> functions as described in the
<literal>dlopen(3)</literal> man page.</para>
<para>Notice that we use
<quote><function>dlopen</function></quote> to refer to
the individual <function>dlopen</function>
<emphasis>function</emphasis>, and
<quote><function>dlopen</function> API</quote> to refer
to the <emphasis>entire API</emphasis>.</para>
</listitem>
</varlistentry>
</variablelist>
</section>
<!-- Translations - ->
<section id="translations">
<title>Translations</title>
<para>
Pointers to available translations are nice.
Also your translators tend to give very important inputs.
</para>
<itemizedlist>
<listitem>
<para>
<ulink url="http://tldp.org/">French Translation</ulink>
provided by Individual <email>someone (at) somewhere.fr</email>
</para>
</listitem>
<listitem>
<para>
<ulink url="http://tlpd.org/">German Translation</ulink>
provided by Individual <email>someone (at) somewhere.de</email>
</para>
</listitem>
</itemizedlist>
</section>
-->
</section>
<section id="theproblem">
<title>The Problem</title>
<para>At some time you might have to load a library (and use its
functions) at runtime; this happens most often when you are
writing some kind of plug-in or module architecture for your
program.</para>
<para>In the C language, loading a library is very simple (calling
<function>dlopen</function>, <function>dlsym</function> and
<function>dlclose</function> is enough), with C++ this is a bit
more complicated. The difficulties of loading a C++ library
dynamically are partially due to <link linkend="mangling">name
mangling</link>, and partially due to the fact that the
<function>dlopen</function> API was written with C in mind, thus
not offering a suitable way to load classes.</para>
<para>Before explaining how to load libraries in C++, let's better
analyze the problem by looking at name mangling in more
detail. I recommend you read the explanation of name mangling,
even if you're not interested in it because it will help you
understanding why problems occur and how to solve them.</para>
<section id="mangling">
<title>Name Mangling</title>
<para>In every C++ program (or library, or object file), all
non-static functions are represented in the binary file as
<emphasis>symbols</emphasis>. These symbols are special text
strings that uniquely identify a function in the program,
library, or object file.</para>
<para>In C, the symbol name is the same as the function name:
the symbol of <function>strcpy</function> will be
<computeroutput>strcpy</computeroutput>, and so on. This is
possible because in C no two non-static functions can have the
same name.</para>
<para>Because C++ allows overloading (different functions with
the same name but different arguments) and has many features C
does not &mdash; like classes, member functions, exception
specifications &mdash; it is not possible to simply use the
function name as the symbol name. To solve that, C++ uses
so-called <emphasis>name mangling</emphasis>, which transforms
the function name and all the necessary information (like the
number and size of the arguments) into some weird-looking
string which only the compiler knows about. The mangled name
of <function>foo</function> might look like
<computeroutput>foo@4%6^</computeroutput>, for example. Or it
might not even contain the word <quote>foo</quote>.</para>
<para> One of the problems with name mangling is that the C++
standard (currently <citation>ISO14882</citation>) does not
define how names have to be mangled; thus every compiler
mangles names in its own way. Some compilers even change their
name mangling algorithm between different versions (notably
g++ 2.x and 3.x). Even if you worked out how your particular
compiler mangles names (and would thus be able to load
functions via <function>dlsym</function>), this would most
probably work with your compiler only, and might already be
broken with the next version.</para>
</section>
<section>
<title>Classes</title>
<para>Another problem with the <function>dlopen</function> API
is the fact that it only supports loading
<emphasis>functions</emphasis>. But in C++ a library often
exposes a class which you would like to use in your
program. Obviously, to use that class you need to create an
instance of it, but that cannot be easily done.</para>
</section>
</section>
<section id="thesolution">
<title>The Solution</title>
<section id="externC">
<title><literal>extern "C"</literal></title>
<para>C++ has a special keyword to declare a function with C
bindings: <literal>extern "C"</literal>. A function declared
as <literal>extern "C"</literal> uses the function name as
symbol name, just as a C function. For that reason, only
non-member functions can be declared as <literal>extern
"C"</literal>, and they cannot be overloaded.</para>
<para>Although there are severe limitations, <literal>extern
"C"</literal> functions are very useful because they can be
dynamically loaded using <function>dlopen</function> just like
a C function.</para>
<para>This does <emphasis>not</emphasis> mean that functions
qualified as <literal>extern "C"</literal> cannot contain C++
code. Such a function is a full-featured C++ function which
can use C++ features and take any type of argument.</para>
</section>
<section id="loadingfunctions">
<title>Loading Functions</title>
<para>In C++ functions are loaded just like in C, with
<function>dlsym</function>. The functions you want to load
must be qualified as <literal>extern "C"</literal> to avoid
the symbol name being mangled.</para>
<example>
<title>Loading a Function</title>
<para>main.cpp:</para>
<programlisting><![CDATA[#include <iostream>
#include <dlfcn.h>
int main() {
using std::cout;
using std::cerr;
cout << "C++ dlopen demo\n\n";
// open the library
cout << "Opening hello.so...\n";
void* handle = dlopen("./hello.so", RTLD_LAZY);
if (!handle) {
cerr << "Cannot open library: " << dlerror() << '\n';
return 1;
}
// load the symbol
cout << "Loading symbol hello...\n";
typedef void (*hello_t)();
// reset errors
dlerror();
hello_t hello = (hello_t) dlsym(handle, "hello");
const char *dlsym_error = dlerror();
if (dlsym_error) {
cerr << "Cannot load symbol 'hello': " << dlsym_error <<
'\n';
dlclose(handle);
return 1;
}
// use it to do the calculation
cout << "Calling hello...\n";
hello();
// close the library
cout << "Closing library...\n";
dlclose(handle);
}]]></programlisting>
<para>hello.cpp:</para>
<programlisting><![CDATA[#include <iostream>
extern "C" void hello() {
std::cout << "hello" << '\n';
}
]]></programlisting>
</example>
<para>The function <function>hello</function> is defined in
<filename>hello.cpp</filename>as <literal>extern
"C"</literal>; it is loaded in <filename>main.cpp</filename>
with the <function>dlsym</function> call. The function must be
qualified as <literal>extern "C"</literal> because otherwise
we wouldn't know its symbol name.</para>
<warning>
<para>There are two different forms of the
<literal>extern "C"</literal> declaration: <literal>extern
"C"</literal> as used above, and <literal>extern "C" {
&hellip; }</literal> with the declarations between the
braces. The first (inline) form is a declaration with extern
linkage and with C language linkage; the second only affects
language linkage. The following two declarations are thus
equivalent:
<informalexample>
<programlisting>extern "C" int foo;
extern "C" void bar();
</programlisting>
</informalexample>
and
<informalexample>
<programlisting>extern "C" {
extern int foo;
extern void bar();
}</programlisting>
</informalexample>
As there is no difference between an
<literal>extern</literal> and a
non-<literal>extern</literal> <emphasis>function</emphasis>
declaration, this is no problem as long as you are not
declaring any variables. If you declare
<emphasis>variables</emphasis>, keep in mind that
<informalexample>
<programlisting>extern "C" int foo;</programlisting>
</informalexample>
and
<informalexample>
<programlisting>extern "C" {
int foo;
}</programlisting>
</informalexample>
are <emphasis>not</emphasis> the same thing.</para>
<para>For further clarifications, refer to
<citation>ISO14882</citation>, 7.5, with special attention
to paragraph 7, or to <citation>STR2000</citation>,
paragraph 9.2.4.</para>
<para>Before doing fancy things with extern variables, peruse
the documents listed in the <link linkend="seealso">see
also</link> section.</para>
</warning>
</section>
<section id="loadingclasses">
<title>Loading Classes</title>
<para>Loading classes is a bit more difficult because we need
an <emphasis>instance</emphasis> of a class, not just a
pointer to a function.</para>
<para>We cannot create the instance of the class using
<literal>new</literal> because the class is not defined in the
executable, and because (under some circumstances) we don't
even know its name.</para>
<para>The solution is achieved through polymorphism. We define a
base, <emphasis>interface</emphasis> class with virtual
members <emphasis>in the executable</emphasis>, and a derived,
<emphasis>implementation</emphasis> class <emphasis>in the
module</emphasis>. Generally the interface class is
abstract (a class is abstract if it has pure virtual
functions).</para>
<para>As dynamic loading of classes is generally used for
plug-ins &mdash; which must expose a clearly defined interface
&mdash; we would have had to define an interface and derived
implementation classes anyway.</para>
<para>Next, while still in the module, we define two additional helper
functions, known as <emphasis>class factory
functions</emphasis>. One of these functions creates an instance of
the class and returns a pointer to it. The other function takes a
pointer to a class created by the factory and destroys
it. These two functions are qualified as <literal>extern
"C"</literal>.</para>
<para>To use the class from the module, load the two factory
functions using <function>dlsym</function> just <link
linkend="loadingfunctions">as we loaded the hello
function</link>; then, we can create and destroy as many
instances as we wish.</para>
<example>
<title>Loading a Class</title>
<para>Here we use a generic <classname>polygon</classname>
class as interface and the derived class
<classname>triangle</classname> as implementation.</para>
<para>main.cpp:</para>
<programlisting><![CDATA[#include "polygon.hpp"
#include <iostream>
#include <dlfcn.h>
int main() {
using std::cout;
using std::cerr;
// load the triangle library
void* triangle = dlopen("./triangle.so", RTLD_LAZY);
if (!triangle) {
cerr << "Cannot load library: " << dlerror() << '\n';
return 1;
}
// reset errors
dlerror();
// load the symbols
create_t* create_triangle = (create_t*) dlsym(triangle, "create");
const char* dlsym_error = dlerror();
if (dlsym_error) {
cerr << "Cannot load symbol create: " << dlsym_error << '\n';
return 1;
}
destroy_t* destroy_triangle = (destroy_t*) dlsym(triangle, "destroy");
dlsym_error = dlerror();
if (dlsym_error) {
cerr << "Cannot load symbol destroy: " << dlsym_error << '\n';
return 1;
}
// create an instance of the class
polygon* poly = create_triangle();
// use the class
poly->set_side_length(7);
cout << "The area is: " << poly->area() << '\n';
// destroy the class
destroy_triangle(poly);
// unload the triangle library
dlclose(triangle);
}]]></programlisting>
<para>polygon.hpp:</para>
<programlisting><![CDATA[#ifndef POLYGON_HPP
#define POLYGON_HPP
class polygon {
protected:
double side_length_;
public:
polygon()
: side_length_(0) {}
virtual ~polygon() {}
void set_side_length(double side_length) {
side_length_ = side_length;
}
virtual double area() const = 0;
};
// the types of the class factories
typedef polygon* create_t();
typedef void destroy_t(polygon*);
#endif]]></programlisting>
<para>triangle.cpp:</para>
<programlisting><![CDATA[#include "polygon.hpp"
#include <cmath>
class triangle : public polygon {
public:
virtual double area() const {
return side_length_ * side_length_ * sqrt(3) / 2;
}
};
// the class factories
extern "C" polygon* create() {
return new triangle;
}
extern "C" void destroy(polygon* p) {
delete p;
}
]]></programlisting>
</example>
<para>There are a few things to note when loading classes:</para>
<itemizedlist>
<listitem>
<para>You must provide <emphasis>both</emphasis> a creation
and a destruction function; you must
<emphasis>not</emphasis> destroy the instances using
<literal>delete</literal> from inside the executable, but
always pass it back to the module. This is due to the fact
that in C++ the operators <literal>new</literal> and
<literal>delete</literal> may be overloaded; this would
cause a non-matching <literal>new</literal> and
<literal>delete</literal> to be called, which could cause
anything from nothing to memory leaks and segmentation
faults. The same is true if different standard libraries
are used to link the module and the executable.</para>
</listitem>
<listitem>
<para>The destructor of the interface class should be
virtual in any case. There <emphasis>might</emphasis> be
very rare cases where that would not be necessary, but it
is not worth the risk, because the additional overhead can
generally be ignored.</para>
<para>If your base class needs no destructor, define an
empty (and <literal>virtual</literal>) one anyway;
otherwise you <emphasis>will have problems</emphasis>
sooner or later; I can guarantee you that. You can read
more about this problem in the comp.lang.c++ FAQ at <ulink
url="http://www.parashift.com/c++-faq-lite/"/>, in
section 20.</para>
</listitem>
</itemizedlist>
</section>
</section>
<section id="source">
<title>Source Code</title>
<para>You can download all the source code presented in this howto as an
archive: <ulink url="examples.tar.gz"/>.</para>
</section>
<section id="faq">
<title>Frequently Asked Questions</title>
<qandaset>
<qandaentry>
<question>
<para>I'm using Windows and I can't find the
<filename>dlfcn.h</filename> header file! What's the problem?</para>
</question>
<answer>
<para>The problem is that Windows doesn't have the
<function>dlopen</function> API, and thus there is no
<filename>dlfcn.h</filename> header. There is a similar API
around the <function>LoadLibrary</function> function, and
most of what is written here applies to it, too. Please refer to the
<ulink url="http://msdn.microsoft.com/">Microsoft Developer Network
Website</ulink> for more information.</para>
</answer>
</qandaentry>
<qandaentry>
<question>
<para>Is there some kind of <function>dlopen</function>-compatible
wrapper for the Windows <function>LoadLibrary</function>
API?</para>
</question>
<answer>
<para>I don't know of any, and I don't think there'll ever be one
supporting all of <function>dlopen</function>'s options.</para>
<para>There are alternatives though: libtltdl (a part of libtool),
which wraps a variety of different dynamic loading APIs, among
others <function>dlopen</function> and
<function>LoadLibrary</function>. Another one is the <ulink
url="http://developer.gnome.org/doc/API/glib/glib-dynamic-loading-of-modules.html">Dynamic
Module Loading functionality of GLib</ulink>. You can use one
of these to ensure better possible cross-platform compatibility.
I've never used any of them, so I can't tell you how stable they
are and whether they really work.</para>
<para>You should also read section 4, <quote>Dynamically
Loaded (DL) Libraries</quote>, of the <ulink
url="http://www.dwheeler.com/program-library">Program Library
HOWTO</ulink> for more techniques to load libraries and
create classes independently of your platform.</para>
</answer>
</qandaentry>
</qandaset>
</section>
<section id="seealso">
<title>See Also</title>
<itemizedlist>
<listitem>
<para>The <function>dlopen(3)</function> man page. It explains
the purpose and the use of the <function>dlopen</function>
API.</para>
</listitem>
<listitem>
<para>The article <ulink
url="http://www.linuxjournal.com/article.php?sid=3687">
<citetitle>Dynamic Class Loading for C++ on
Linux</citetitle></ulink> by James Norton published on the
<ulink url="http://www.linuxjournal.com/">Linux
Journal</ulink>.</para>
</listitem>
<listitem>
<para>Your favorite C++ reference about <literal>extern
"C"</literal>, inheritance, virtual functions,
<literal>new</literal> and <literal>delete</literal>. I
recommend <citation>STR2000</citation>.</para>
</listitem>
<listitem>
<para><citation>ISO14882</citation></para>
</listitem>
<listitem>
<para>The <ulink
url="http://www.dwheeler.com/program-library">Program Library
HOWTO</ulink>, which tells you most things you'll ever need
about static, shared and dynamically loaded libraries and how
to create them. Highly recommended.</para>
</listitem>
<listitem>
<para>The <ulink
url="http://tldp.org/HOWTO/GCC-HOWTO/index.html">Linux GCC
HOWTO</ulink> to learn more about how to create libraries
with GCC.</para>
</listitem>
</itemizedlist>
</section>
<bibliography>
<bibliomixed>
<abbrev>ISO14482</abbrev> <title>ISO/IEC 14482-1998 &mdash; The
C++ Programming Language</title>. <releaseinfo>Available as
PDF and as printed book from <ulink
url="http://webstore.ansi.org/"/>.</releaseinfo>
</bibliomixed>
<bibliomixed>
<abbrev>STR2000</abbrev>
<author><surname>Stroustrup</surname> <firstname>Bjarne</firstname></author>
<title>The C++ Programming Language</title>, <edition>Special
Edition</edition>.
ISBN <isbn>0-201-70073-5</isbn>.
<publishername>Addison-Wesley</publishername>.
</bibliomixed>
</bibliography>
</article>