Prepare your environment for cross-platform C/C++ development with NetBeans, and put the C/C++ Pack to work creating a native library for Java applications


When NetBeans 5.5 was released in late 2006, it radically changed its own value proposition by offering first-class support for a language that doesn’t run inside a JVM. The NetBeans C/C++ pack provided to C/C++ programmers most features Java developers were already used to: advanced source editing with syntax highlighting and code completion, built-in CVS support, hyperlinks to navigate function declarations, a class hierarchy browser, an integrated debugger, and integration with the make tool.

This article focuses on how the C/C++ pack can help Java developers. Although I’m sure you all would like to code the whole world in pure Java, reality frequently challenges us to interface with native code, be it legacy systems, a device vendor SDK or a high-performance math library. Also, sometimes we need to use native code to improve the user experience, by means of tighter integration with the underlying operating system. Wouldn’t it be better to do all this from the same IDE we already use for Java development?

We’ll show how to leverage NetBeans and the C/C++ Pack to develop portable native libraries using C/C++, and how to integrate them with Java code in a way that eases deployment in multiple platforms.

NetBeans C/C++ Pack is more than just C/C++ coding support for the Java developer. It also suits many native code projects very well. The sidebar “Other open source C/C++ IDEs” compares the Pack with some popular open-source
IDEs for C/C++.

Installing NetBeans C/C++ Pack

Installing the C/C++ Pack per se will be a no-brainer for most users. No matter if you’ve installed the NetBeans IDE using the zip package or one of the native installers, you only need to run C/C++ Pack’s installer and point it to your NetBeans IDE installation directory. (Note that, although the C/C++ Pack is mostly Java code with just one tiny native library, there’s no multiplatform zip archive like the ones provided for the IDE.)

The installer itself will work the same for all supported platforms: Windows, Linux and Solaris. But configuring your environment for using C/C++ Pack may not be so easy. Just like the core NetBeans IDE needs a compatible JDK installation, the C/C++ Pack will require a C/C++ compiler and standard libraries and headers. So you need to install and configure these in advance.

To meet the Pack’s prerequisites, we’ll rely on the popular suite formed by the GNU C Compiler (GCC), GNU Binutils, GNU Make and GNU Debugger (GDB). This is the suite that received most of the QA effort of the C/C++ Pack developer team1, and it’s portable to Windows, Linux and Solaris environments.

Using the same compiler suite for all platforms greatly simplifies dealing with portable (and even non-portable) C/C++ code, as you won’t need to spend time fighting compiler directives, runtime library inconsistencies and language dialects. Besides, you’ll find that in most cases the GNU toolset competes head-to-head with other C compilers in both speed and optimization quality.


1 The only other compiler suite supported so far is the Sun Studio product for Solaris and Linux.


Installing the GNU toolset on Linux

Linux users should have no problem obtaining the GNU toolset for their preferred platform. Mine is Fedora Core 6, and as I installed a “development workstation” using Anaconda I already had everything ready for NetBeans C/C++ Pack. Users who didn’t install Linux development tools when configuring their systems should have no problem using either yum, up2date, yast or apt to install the GNU toolset.


Stay clear of CD-bootable mini-distros like Knoppix for real development work. Instead, install a full-featured distro in a native Linux partition in your main hard disk. The few additional gigabytes used will prove to be a small cost for all the hassle you’ll avoid.


Solaris users will also find it easy to install the GNU toolset; there are detailed instructions on the NetBeans Web site. But be warned: if you think you’d be better served by the native platform C compiler (Sun Studio), think again. This is because NetBeans C/C++ Pack’s debugger needs the GNU Debugger, and GDB has some issues running code generated by Sun compilers. So you can use Sun’s compiler to produce final code, but you’d better use the GNU toolchain for development.

Installing the GNU toolset on Windows

Windows users won’t be able to use native C/C++ compilers from Microsoft, Borland or Intel, and will have to stick with a Windows port of the GNU toolset. There are two options: Cygwin
and MinGW.

The C/C++ Pack’s docs at provide detailed instructions for using Cygwin, but I strongly advise you to use MinGW instead. The reason is that Cygwin relies on a Unix emulation layer, while MinGW uses native Windows DLLs for everything. Code compiled with Cygwin uses the standard GNU runtime library (glibc) on an emulation of Unix system calls, and semantics like mount points, pipes and path separators. But code compiled with MinGW will use standard Microsoft runtime libraries such as MSVCRT.DLL.

Cygwin has its uses, as many Linux and Unix software (specially open-source software) that has not yet been ported to Windows is easy to run under Cygwin without virtualization overhead. But I doubt you’d want to compromise stability and compatibility with the native platform when developing native libraries for use with Java applications. So MinGW is the way to go. The sidebar “Installing MinGW” provides detailed instructions.

Checking prerequisites

Whatever your platform of choice, you need access to the GNU toolset from your operating system command prompt. It may be necessary to configure the system PATH before using NetBeans C/C++ Pack. You can check that you have all prerequisites are available before proceeding by using the commands displayed in Figure 1. (Although this figure shows a Windows command prompt, you’ll be able to run the same commands from either the Linux or Solaris shells.) If you get software releases older than the ones shown, consider upgrading your GNU toolset.


Figure 1. Verifying that the GNU toolset is installed and configured correctly, and is using compatible releases.

When pure Java is not enough

Now that you have NetBeans C/C++ installed and its prerequisites configured, let’s present this article’s use case. You’re developing a desktop Java application with cryptographic features, which saves sensitive data such as key rings and private keys in a local file system folder. You want to be sure that only the user who’s running the application can read (and of course write) files to that folder.

The standard Java libraries provide methods in the class for checking if a file can be read or written by the current user, but these methods don’t check if other users can also read or write the same files. There are new methods in Java SE 6 that deal with file permissions, and work in progress under JSR 293; but if your application has to support Java 5 or 1.4, there’s no escaping from native code. So our application will use native system calls to verify local folder permissions during initialization, and refuse to start if it finds the folder is not secure.

Java doesn’t provide an easy way to declare external methods, like Free Pascal or Visual Basic, but it does of course provide the Java Native Interface, a standard and portable way to call native code from Java and vice versa. With the above use case in mind, we have to design an abstraction that hides platform details and the corresponding native code from the higher application layers. In the end, the apparent complexity of dealing with JNI may actually be an advantage, because it forces us to design the interface between Java and native code, instead of just going ahead and invoking operating system APIs directly.

The Java wrapper code

Let’s get our feet wet. Start NetBeans, create a Java Class Library Project, and name it “OSlib”. This project will contain all interfaces between our hypothetical application and the native operating system. Then create a new class named “FilePermissions”, with the code shown in Listing 1.

Listing 1. – Utility class with a native method.

package org.netbeans.nbmag3.util;


public class FilePermissions {

static {

public FilePermissions() {}

// Checks if a file or folder can only
// be read/written by the current user
public static native boolean isPrivate(String path);

The native keyword, you’ll remember, means that the method’s implementation will be provided by a native dynamic library. That library in our code is loaded by a static initializer in the class itself.

Following Test-Driven Development practices, I’ll create unit tests instead of creating a test application for the OS interface. Right click Test Packages in the Projects window and select New>Test for Existing Class to generate a skeleton for testing the native method. Then change this skeleton to make it look like Listing 2.

Listing 2. Unit tests for FilePermissions native methods
package org.netbeans.nbmag3.util;

import java.util.Properties;
import junit.framework.*;

public class FilePermissionsTest extends TestCase {
Properties paths = null;

public FilePermissionsTest(String testName) {

protected void setUp() throws Exception {
paths = new Properties();

protected void tearDown() throws Exception {}

public void testIsPrivateOk() {
String fileName = paths.getProperty("FilePermissions.test.privateOk");
assertTrue("File does not exist", new File(fileName).exists());
boolean result = FilePermissions.isPrivate(fileName);
assertEquals(true, result);

public void testCanReadButNotWrite() {
boolean result = FilePermissions.isPrivate(
assertEquals(false, result);

public void testCanBeReadByOthers() {
boolean result = FilePermissions.isPrivate(
assertEquals(false, result);

public void testCanBeWrittenByOthers() {
boolean result = FilePermissions.isPrivate(
assertEquals(false, result);
# For testing under Linux / Unix
FilePermissions.test.privateOk = /home/fernando/privateOk
FilePermissions.test.readButNotWrite = /home/fernando/readButNotWrite
FilePermissions.test.readByOthers = /home/fernando/readByOthers
FilePermissions.test.writrenByOthers = /home/fernando/writtenByOthers

# For testing under Windows
#FilePermissions.test.privateOk = C:\\test\\privateOk
#FilePermissions.test.readButNotWrite = C:\\test\\readButNotWrite
#FilePermissions.test.readByOthers = C:\\test\\readByOthers
#FilePermissions.test.writrenByOthers = C:\\test\\writtenByOthers

The unit tests use a properties file (shown in the same listing) to get each test’s target filesystem path. This way, all file paths can be easily changed to comply with native-platform naming conventions, without needing to recompile the tests themselves. Also, don’t forget to create the target files and give them appropriate permissions.

If everything is fine so far, running the tests (by selecting the FilePermissionsTest class and pressing Shift+F6) should give the output shown in Figure 2. The UnsatisfiedLinkError exception is thrown because we haven’t yet provided the native method implementation.


Figure 2. Running JUnit tests for the unfinished native method.


Other open-source C/C++ IDEs

C and C++ are of course much older than Java, and are still the languages of choice for many high-profile open-source projects. Based on that, on could guess there would be many other strong cross-platform and open-source C/C++ IDEs. You’ll find that NetBeans C/C++ Pack may be the strongest one around, however. Let’s look at some C/C++ Pack’s competitors.


DevCPP is very popular among Windows developers. It’s lightweight, well supported, and, like NetBeans, relies on external make tools and C/C++ compilers. Additionally, it supports a wide variety of C/C++ compilers. Though DevCPP is written using Borland Delphi, an attempt to port it to Linux (using Kylix) failed. So DevCPP is not an option for cross-platform C/C++ development.


The Watcom C/C++ compiler is cross-platform but offers no Unix support; it targets Windows and OS/2. Though not very user-friendly, it comes with an integrated debugger and a help system. It was once the compiler of choice for high-performance C/C++ applications, with its enhanced code optimizer and support for all Intel processor variants. When Sybase bought Watcom, though, the C/C++ compilers and IDEs fell into obscurity. Later the tools were released as open-source software. Nowadays, it looks like the community project is going well, but there’s still no support for Unix and Linux systems. This makes OpenWatcom essentially a Windows-only IDE and not suitable for our purposes.


Anjuta is based on the complete GNU toolset for C/C++ development. In addition to the tools supported by C/C++ Pack, it supports the GNU Autotools, a set of scripts that simplifies generating Makefiles for multiple operating systems and compilers. It’s also focused on GNOME development, so it provides templates for GTK, Gnome and Glade applications.

While DevCPP and OpenWatcom are Windows-only, Anjuta and KDeveloper (see next) are Unix-only. Some users have reported success running both under Cygwin, but they are still far from providing robust support for compiling and debugging native Windows applications.

For Unix developers, Anjuta provides integrated access to man pages and GNOME documentation. Its integrated debugger, like C/C++ Pack, relies on GDB. The latest releases provide integration with Glade, the Gnome visual UI builder.


Everything said before about Anjuta applies to KDevelop, if you just replace GTK/Glade/GNOME with Qt/QtDesigner/KDE. Anjuta and KDevelop are strong C/C++ IDEs for open-source desktops, but they don’t cut it as cross-platform IDEs.

Eclipse CDT

C/C++ development support in Eclipse is almost as old as Eclipse IDE itself, but it has not matured as fast as the support for Java. Although currently labeled as release 4.0, Eclipse CDT doesn’t provide many features beyond those in NetBeans C/C++ Pack (which is younger).

Also like NetBeans, Eclipse CDT doesn’t integrate yet with visual development tools for Gnome, KDE or Windows. It has the advantage of supporting compilers other than the GNU compilers, but this won’t be a real plus if your goal is developing cross-platform C code.

Red Hat is developing GNU Autotools and RPM generation plug-ins which, when they are released as production level, may become Eclipse CDT’s real advantage over NetBeans C/C++ Pack (at least for Unix/Linux users). On the other hand, NetBeans is the development IDE for Open Solaris, so don’t expect it to fall short in enhancements for Unix developers.


The only flaw one would find in C/C++ Pack, comparing it to other open-source alternatives for C/C++ development, is the lack of operating-system and third-party library documentation support in the help system. That would be also its main drawback when compared to proprietary C/C++ IDEs. But if you evaluate alternatives for cross-platform C/C++ development, the strongest (and only) competitor for NetBeans is also its main competitor in the Java space, that is, Eclipse.


The native code project

Our unit tests are ready, but getting native code working alongside Java code is not trivial. We’ll mock the native method implementation so we can focus on how to build a native library that can be called by Java code. Start by creating a C/C++ Dynamic Library Project, as shown in Figure 3. Name the project “NativeOSlib” and clear the “Set as main project” checkbox.

Figure 3. Creating a C/C++ Project in NetBeans.

New C/C++ projects are created empty, except for a generated Makefile (see Figure 4), and are structured in virtual folders organized by file type – not by package names like Java projects. You’ll be pleased to know that NetBeans C/C++ Pack includes a Makefile editor (even though there’s still no support for running arbitrary Makefile targets as there is for Ant buildfiles).

Figure 4. The new C/C++ project in NetBeans’ Projects window.

Generating JNI Stubs

We’re ready to begin writing our C code. First remember that all JNI-compliant native methods should use the declaration generated by JDK’s javah tool. You could turn to the operating system command prompt to generate the C JNI stubs, but there’s a better solution. It’s the JNI Maker project, a plug-in module that adds a context menu for generating JNI header files from Java classes. Just get the nbm package from and install it using NetBeans’s Update Center. After restarting the IDE, you should see a new menu item as shown in Figure 5.

Figure 5. Generating a JNI stub using the JNI Maker plug-in module.

Before generating JNI stubs, make sure you’ve built the Java project. JNI Maker uses the distribution JARs.

Now select Generate JNI Stub from the FilePermissions class’s context menu. NetBeans shows a standard File Save dialog, where you can select a folder to save the generated FilePermissions.h header file. Move into the NativeOSlib project folder and create a new src folder (C/C++ Projects do not have a default file structure with separate source and test folders like Java projects do). Save the header file there. The output window will look like Figure 6 if the operation is successful.


Figure 6. Output from the Generate JNI Stub command.


JNI Maker Release 1.0 will only work correctly under Windows, but the generated code will compile and run fine on Unix/Linux. The project developers have been contacted about the module’s cross-platform issues and by the time you read this there should be a new release that will work on all platforms supported by NetBeans C/C++ Pack


Using the JNI Maker module has the same effect as running the following command from the operating system prompt, assuming the OSlib project folder is the current directory and NativeOSlib project folder is a sibling:

$ javah -classpath dist/OSlib.jar -jni -o ../NativeOSlib/src/FilePermissions.h



Installing MinGW

The MinGW project provides a native port of the GNU toolset for Windows platforms. Included in the base distribution are GNU C, C++, Objective-C, Ada, Java and Fortran compilers, plus an assembler and a linker; there’s also support for dynamic libraries and Windows resource files. Additional packages provide useful tools like Red Hat Source Navigator, Insight GUI debugger and a handful of Unix ports like the wget download manager.

MinGW stands for “Minimalist GNU for Windows”. But it’s “minimalist” only when compared to the Cygwin environment. (Cygwin tries to emulate a full Unix shell, complete with bash scripting, user commands and a Unix-like view of the filesystem.)

In fact, MinGW is complete to the point of providing Win32 API header files, and many popular open-source applications like Firefox have their Windows releases compiled using it. (Recent Cygwin releases include many MinGW enhancements as a cross-compiling feature, showing how Windows development is “alien” to MinGW alternatives.)

If you check the project’s website, it looks like MinGW development has been stalled for quite a few years; the problem is that the site was automatically generated by a script that read the project’s SourceForge area, and developers simply got tired of catching up with’s design changes. However, MinGW is a very healthy project with active mailing lists and frequent file releases.

There is an installer for the base distribution named mingw-x.x.exe that downloads selected packages from SourceForge and installs them. The same installer can be used to update an existing MinGW installation.

Individual packages are downloaded to the same folder where the installer was started. This allows you to later copy the entire folder to another workstation and install MinGW there without the need of an Internet connection. Most extra packages provide their own installers or can simply be unpacked over an existing MinGW installation.

To satisfy C/C++ Pack’s prerequisites, you’ll need to download and install three MinGW packages: the base distribution itself, the GDB debugger, and the MSys distribution.

Installing MinGW

Download MinGW-5.1.3.exe (or newer) from the project’s current file releases at, then launch it to see a standard Windows installer.

On the third step of the wizard (the second screen in Figure S1) you only need to select “MinGW base tools” and optionally “g++ compiler”. Also, the Java Compiler may be interesting to play with, because of its ability to generate native machine code from Java sources and bytecodes, but it’s not supported by NetBeans yet. Interestingly, the g77 (Fortran) compiler will be officially supported very soon.

After downloading all selected packages, the installer will ask for the destination directory and unpack all packages there. It’s left to the user to configure environment variables so that MinGW tools can be used from the Windows command prompt.

Installing GDB

As we’ve seen, NetBeans C/C++ Pack needs GDB to be able to debug C programs. The MinGW distribution packages GDB as a stand-alone installer.

At the time of writing, the latest stable MinGW package for GDB was release 5.2.1, which won’t refresh the NetBeans debugger’s Local Variables window correctly. To solve this, download gdb-6.3-2.exe (or newer) from MinGW Snapshot Releases to a temporary folder and run it. Though you don’t need to install GDB over MinGW, your life will be easier if you do, as you won’t need to add another folder to your PATH system environment variable.

Installing MSys

The MinGW base distribution already includes a make tool named mingw32-make.exe, but NetBeans C/C++ Pack won’t be happy with it. MinGW’s make tool is patched to be more compatible with other native Windows C compilers, and NetBeans expects a Unix-style make tool. NetBeans generated Makefiles even expect to find standard Unix file utilities such as cp and rm.

The MinGW MSys package satisfies these dependencies. It is a “Minimal System” that provides a Unix-style shell and file utilities, and allows open-source projects based on GNU Autotools to be easily built using MinGW.

Download msys-1.0.10.exe or newer to a temporary folder and start it. At the final installation step, a batch script configures the integration between MSys and MinGW. You will still have to add the MSys programs folder to the system PATH (in my case, E:\MSys\1.0\bin), as you did for the MinGW base distribution.

That’s it. After running three installers and downloading about 23 MB, we are ready to develop C/C++ applications and libraries using the NetBeans IDE and C/C++ Pack on Windows.



Figure S1. Screens from MinGW’s base distribution installer.



(The command is broken to fit the column width, but it should be typed in a single line, of course.)

Now add the generated C header file to the NativeOSlib project. Right click Header Files inside the NativeOSlib project folder in NetBeans’ Projects window, and select Add Existing Item. Then browse to the file src/FilePermissions.h and open it. The code will look like Listing 3.

Listing 3. FilePermissions.h – JNI Stub for native methods in the FilePermissions class.

/* DO NOT EDIT THIS FILE - it is machine generated */
#include <jni.h>
/* Header for class org_netbeans_nbmag3_util_FilePermissions */

#ifndef _Included_org_netbeans_nbmag3_util_FilePermissions
#define _Included_org_netbeans_nbmag3_util_FilePermissions

#ifdef __cplusplus
extern "C" {
 * Class:     org_netbeans_nbmag3_util_FilePermissions
 * Method:    isPrivate
 * Signature: (Ljava/io/File;)Z
JNIEXPORT jboolean JNICALL Java_org_netbeans_nbmag3_util_FilePermissions_isPrivate
  (JNIEnv *, jclass, jobject);

#ifdef __cplusplus

Mocking native code

Due to space constraints, we won’t show you the final C code for the FilePermissions.isPrivate() native method, but the sources available for download will provide working implementations for both Windows and Unix (Posix) systems.

To create the C implementation file, right click Source Files and select New>Empty C File, then type “FilePermissions.c” as the file name and “src” as the folder name. A new node named FilePermissions.c should be created under Source Files.

Copy the C stub function declaration from FilePermissions.h to FilePermissions.c and change it to include the header file. Also add parameter names. The code should look like Listing 4. (Listing 3 highlights the declaration you have to copy, and Listing 4 highlights the changes after copying.)

Listing 4. FilePermissions.h – JNI mock implementation for the FilePermissions native methods.

#include "FilePermissions.h"

JNIEXPORT jboolean JNICALL Java_org_netbeans_nbmag3_util_FilePermissions_isPrivate
(JNIEnv *env, jclass clazz, jstring path) {
return JNI_TRUE;

At this point, Unix and Linux users should be ready to build the native code and run unit tests again2. But Windows users first have to change a few project properties to make MinGW generate Windows-compatible JNI DLLs. The sidebar “JNI and MinGW” details these


2 At least if you use JDK packages compatible with your distro package manager, like the IBM and BEA JDKs provided by RHEL and SuSE Enterprise, or the RPM Packages from If not, you’ll have to add your JDK include folder to the GNU C compiler include directory. The configurations will be similar to the ones presented in the “JNI and MinGW” sidebar, but you won’t need to change either the linker output file name or additional compiler options.

Right click the NativeOSlib project and select Clean and Build Project. If there are no errors, you should see make’s output as in Figure 7.

Figure 7. Building the NativeOSlib project under Linux.

Running unit tests again

You need to set the OSlib project’s java.library.path system property before running it, or you’ll still get UnsatisfiedLinkError exceptions. Open the project’s Properties dialog, select the Run category and change VM Options to specify the full path to the NativeOSlib project’s platform-specific native-library folder, which is inside the dist folder (see Figure 8). In Linux, this will be PROJECT_HOME/dist/Debug/GNU-Linux-x86; in Windows, PROJECT_HOME\dist\Debug\GNU-Windows.

Figure 8. Configuring the java.library.path property so unit tests can find the native code library on Linux.

Now run the unit tests again. The result should be as shown in Figure 9. Since the mock native code always returns true, some tests pass even if you have not created target test folders or forgotten to setup their access permissions. Anyway, the first test should fail because it takes an extra step to check if the target file path
actually exists.

Figure 9. Running unit tests using a mock native implementation.

JNI and MinGW

Unix and Windows native C/C++ compilers use different conventions for mangling function names1*, exporting global symbols from libraries and setting up stack frames. JNI on Windows uses Microsoft conventions for Windows DLLs, while GCC uses its own conventions for dynamic libraries. This means that if you simply try to compile and link a dynamic library, MinGW will stick to its Unix origins and produce a DLL that is incompatible with native Windows C/C++ compilers. The JVM won’t be able to get native method implementations from that library and will generate more UnsatisfiedLinkExceptions.

The solution is to add a few command-line options when compiling C/C++ sources: ‑D_JNI_IMPLEMENTATION -Wl,--kill-at. Open the C/C++ Dynamic Library Project properties and expand C/C++>Command Line, then type these options in the Additional Options text field (see Figure S1).

You also need to add your JDK include folders (JAVA_HOME\include and JAVA_HOME\include\win32) to the project properties. Open
C/C++>GNU C Compiler>General and change the Include Directories field as shown in Figure S2.

You need one last change in the C/C++ Dynamic Library Project properties so you can generate a JNI-compatible DLL. By default, NetBeans chooses a library name that corresponds to Cygwin conventions, but we need to use native Windows conventions. So you need to enter the Linker>General category and remove the “cyg” prefix from the Output field (Figure S3).


1“Mangling” is the process used for generating public C++ function names in object files. It’s needed because the C language doesn’t support function overloading, and, to keep backward compatibility, C++ compilers generate a function name that encodes parameter types.


Figure S1. MinGW compiler options for generating JNI-compatible DLLs

Figure S2. Configuring JDK include folders for MinGW

Figure S3. Changing the output file name for compliance with Windows DLL naming conventions


Managing platform-specific compiler settings

NetBeans C/C++ Pack puts object files in the build and dist folders, inside subdirectories named after the target platform, for example GNU-Linux-x86 or GNU-Windows. But it won’t save different compiler options for each target, forcing you to have a different project for each platform if there’s a need for platform-specific compiler settings.

You can solve this using NetBeans C/C++ Pack’s multiple configurations feature. Open NativeOSlib’s project properties and notice the Configuration combo box on the top of the window (Figure 10). The default configurations are meant to save different compiler settings for Debug and Release builds, like keeping symbol information for Debug builds and optimizing code for Release builds. So if you want platform-specific configurations, you may need to create Release and Debug variants for each platform.

Figure 10. Combo box for changing compiler configurations for a C/C++ project.

The Manage Configurations button to the side of the combo box lets you create new configurations either from scratch or as a copy of an existing configuration (see Figure 11). You’ll notice I renamed the generated Debug configuration to Debug-Linux and copied it to a new configuration named Debug-Windows. Doing this lets you change the Windows configuration to include all options needed by MinGW for generating JNI-compatible DLLs, while keeping the default settings for the Linux configuration.

Figure 11. Creating, renaming or copying configurations.

NetBeans-generated Makefiles provide many extension points (like the Ant buildfiles generated by the IDE), and they can be used outside the IDE. For example, for building the Debug-Windows configuration you’d type the following command at the operating system prompt:

make CONF=Debug-Windows

Thus, you could have Continuous Integration servers for many platforms, all being fed by the same CVS or Subversion source tree. And thanks to GNU C cross-compiler features it would be possible to have a “compile farm” that generates native binaries for multiple platforms, without the need for multiple OS installations. For example, a Linux server could generate both Windows and Solaris
SPARC binaries.


NetBeans C/C++ Pack provides a rich environment for developing C and C++ applications and libraries. It’s useful for Java developers that need to interface with native code and, of course, for developing fully-native applications. Compiler configuration may pose some challenges for Windows developers if they never tried GNU compilers before, but the effort will certainly pay off because of the increased portability of both code and Makefiles.

  NetBeans C/C++ Pack home page

  MKO-JNI Stub Maker
  MKO JNI Stub Maker, a NetBeans plug-in for generating JNI headers
  Java Native Interface specification and related tips
  MinGW, the native GNU toolset for Windows
  SourceForge file releases for MinGW
  Cygwin, a GNU toolset for Windows.



Fernando Lozano
( is an independent consultant and has worked with information systems since 1991. He’s the Community Leader of the Linux Community at, webmaster for the Free Software Foundation and counselor to the Linux Professional Institute. Lozano helps many open-source projects and teaches at undergraduate and postgraduate college courses. He’s also a technical writer and book author, as well as Contributing Editor at Java Magazine (Brazil) and freelance writer for other leading IT publications.


Not logged in. Log in, Register