Eclipse Project Release Notes

Release 4.0.0
Last revised July 27, 2010

This software is OSI Certified Open Source Software.
OSI Certified is a certification mark of the Open Source Initiative. 

1. Target Operating Environments
2. Compatibility with Previous Releases
3. Known Issues
4. Running Eclipse
5. Upgrading a Workspace from a Previous Release
6. Interoperability with Previous Releases

1. Target Operating Environments

In order to remain current, each Eclipse Project release targets reasonably current operating environments.

Most of the Eclipse SDK is "pure" Java code and has no direct dependence on the underlying operating system. The chief dependence is therefore on the Java Platform itself. Portions are targeted to specific classes of operating environments, requiring their source code to only reference facilities available in particular class libraries (e.g. J2ME Foundation 1.1, J2SE 1.4, Java 5, etc).

In general, the 4.0 release of the Eclipse Project is developed on a mix of Java 5 and Java 6 VMs. As such, the Eclipse SDK as a whole is targeted at all modern, desktop Java VMs. Most functionality is available for use on Java 5 everywhere, and extended development capabilities are made available on the VMs that support them.

Appendix 1 contains a table that indicates the class library level required for each bundle.

There are many different implementations of the Java Platform running atop a variety of operating systems. We focus our testing on a handful of popular combinations of operating system and Java Platform; these are our reference platforms. Eclipse undoubtedly runs fine in many operating environments beyond the reference platforms we test. However, since we do not systematically test them we cannot vouch for them. Problems encountered when running Eclipse on a non-reference platform that cannot be recreated on any reference platform will be given lower priority than problems with running Eclipse on a reference platform.

Eclipse 4.0 is tested and validated on the following reference platforms (this list is updated over the course of the release cycle):

Operating System Version Hardware JRE Windowing System
Windows 7 x86 32-bit Sun Java 5 Update 22
IBM Java 5 SR11
x86 64-bit
Vista x86 32-bit Sun Java 5 Update 22
IBM Java 5 SR11
Oracle JRockit 27.6.5
x86 64-bit Sun Java 5 Update 22
IBM Java 5 SR11
XP x86 32-bit Sun Java 6 Update 17
Sun Java 5 Update 22
IBM Java 5 SR11
Oracle JRockit 27.6.5
x86 64-bit Sun Java 5 Update 22
IBM Java 5 SR11
Red Hat Enterprise Linux 5.0 x86 32-bit Sun Java 6 Update 17
Sun Java 5 Update 22
IBM Java 5 SR11
4.0 x86 64-bit Sun Java 5 Update 22
IBM Java 5 SR11
SUSE Linux Enterprise Server 11 x86 32-bit Sun Java 5 Update 22
IBM Java 5 SR11
x86 64-bit
Ubuntu Long Term Support 10.04 x86 32-bit Sun Java 5 Update 22
IBM Java 5 SR11
x86 64-bit
Apple Mac OS X 10.5 Universal 32-bit Apple Java 10.5 Update 2 Cocoa
Universal 64-bit

As stated above, we expect that Eclipse works fine on other current Java VM and OS versions but we cannot flag these as reference platforms without significant community support for testing them.

The Eclipse SDK is designed as the basis for internationalized products. The user interface elements provided by the Eclipse SDK components, including dialogs and error messages, are externalized. The English strings are provided as the default resource bundles.

Latin-1 and DBCS locales are supported by the Eclipse SDK on all reference platforms; BIDI locales are supported by the Eclipse SDK everywhere but on Motif.

The Eclipse SDK supports GB 18030 (level 1), the Chinese code page standard, on Windows XP and 2000, Linux/GTK and the Macintosh.

German and Japanese locales are tested.

2. Compatibility with Previous Releases

Compatibility of Release 4.0 with 3.6

Eclipse 4.0 is compatible with Eclipse 3.6 (and all earlier 3.x versions).

API Contract Compatibility: Eclipse SDK 4.0 is upwards contract-compatible with Eclipse SDK 3.6 except in those areas noted in the Eclipse 4.0 Plug-in Migration Guide . Programs that use affected APIs and extension points will need to be ported to Eclipse SDK 4.0 APIs. Downward contract compatibility is not supported. There is no guarantee that compliance with Eclipse SDK 4.0 APIs would ensure compliance with Eclipse SDK 3.6 APIs. Refer to Evolving Java-based APIs for a discussion of the kinds of API changes that maintain contract compatibility.

Binary (plug-in) Compatibility: Eclipse SDK 4.0 is upwards binary-compatible with Eclipse SDK 3.6 except in those areas noted in the Eclipse 4.0 Plug-in Migration Guide . Downward plug-in compatibility is not supported. Plug-ins for Eclipse SDK 4.0 will not be usable in Eclipse SDK 3.6. Refer to Evolving Java-based APIs for a discussion of the kinds of API changes that maintain binary compatibility.

Source Compatibility: Eclipse SDK 4.0 is upwards source-compatible with Eclipse SDK 3.6 except in the areas noted in the Eclipse 4.0 Plug-in Migration Guide . This means that source files written to use Eclipse SDK 3.6 APIs might successfully compile and run against Eclipse SDK 4.0 APIs, although this is not guaranteed. Downward source compatibility is not supported. If source files use new Eclipse SDK APIs, they will not be usable with an earlier version of the Eclipse SDK.

Workspace Compatibility: Eclipse SDK 4.0 is upwards workspace-compatible with earlier 3.x versions of the Eclipse SDK unless noted. This means that workspaces and projects created with Eclipse SDK 3.6 .. 3.0 can be successfully opened by Eclipse SDK 4.0 and upgraded to a 4.0 workspace. This includes both hidden metadata, which is localized to a particular workspace, as well as metadata files found within a workspace project (e.g., the .project file), which may propagate between workspaces via file copying or team repositories. Individual plug-ins developed for Eclipse SDK 4.0 should provide similar upwards compatibility for their hidden and visible workspace metadata created by earlier versions; 4.0 plug-in developers are responsible for ensuring that their plug-ins recognize metadata from earlier versions and process it appropriately. User interface session state may be discarded when a workspace is upgraded. Downward workspace compatibility is not supported. A workspace created (or opened) by a product based on Eclipse 4.0 will be unusable with a product based on an earlier version of Eclipse. Visible metadata files created (or overwritten) by Eclipse 4.0 will generally be unusable with earlier versions of Eclipse.

Non-compliant usage of API's: All non-API methods and classes, and certainly everything in a package with "internal" in its name, are considered implementation details which may vary between operating environment and are subject to change without notice. Client plug-ins that directly depend on anything other than what is specified in the Eclipse SDK API are inherently unsupportable and receive no guarantees about compatibility within a single release much less with earlier releases. Refer to How to Use the Eclipse API for information about how to write compliant plug-ins.

3. Known Issues

3.1 General problems
     3.1.1 Early Adopter Release Issues
     3.1.2 Startup
     3.1.3 GCJ
     3.1.4 64-bit Java HotSpot(TM) VM
3.2 Platform
     3.2.1 Core
     3.2.2 Ant
     3.2.3 User Assistance
     3.2.4 UI
     3.2.5 Text
     3.2.6 SWT
     3.2.7 Team and CVS
     3.2.8 Install/Update
     3.2.9 Debug
     3.2.10 Compare
3.3 Java development tools (JDT)
3.4 Plug-in Development Environment (PDE)

Note: Bug numbers refer to the Eclipse project bug database at

3.1 General problems

3.1.1 General - Early Adopter Issues

The Eclipse SDK 4.0 release is an Early Adopter Release. It is made available at this early stage to allow plug-in and application writers an opportunity to begin migrating their applications to use it prior to a wider scale commercial release. Due to the large amount of new technology in this release, there are some general problems listed here that early adopters and users should be aware of. See also the Missing Features wiki page for a high level overview of some of the incomplete areas of the release.

Model corruption preventing Eclipse to start

In the 4.0 Early Adopter Release, there are circumstances where the model representing the user interface state can become corrupted. This can prevent startup from completing, or cause the user interface to start in a non-functioning state. In this situation, restarting with the -clearPersistedState command line argument will cause the user interface state to be discarded. The user interface will then start normally but will revert to the default perspective and layout.

Memory leaks

There are some known memory leaks in this release. Running the SDK for long periods will generally cause memory usage to continue increasing. Periodic restarts may be required to recover memory when the SDK is used extensively.

Widget disposed errors in the log during shutdown

There are circumstances where "widget is disposed" errors are printed to the log during shutdown. These are caused by bugs in user interface cleanup and generally don't affect subsequent sessions. See bugs 320854 and 319847 for details.

Changing theme does not update all widgets dynamically

When changing to a different them on the General > Appearance preference page, some changes to the appearance only take effect after a restart. See bug 321050 for details.

Failure to start when proxy passwords have been saved by earlier Eclipse version

Eclipse SDK 4.0 fails to start when proxy passwords have been saved in the global secure storage area by a previous version of Eclipse. The workaround is to clear or move the secure storage area prior to first launch of Eclipse SDK 4.0. See bug 320932 for further details on the bug and workaround.

3.1.2 General - Startup

Installation/Configuration issues that can cause Eclipse to fail start

Here are some common problems that can cause Eclipse not to start:

Invalid characters in install directory prevents Eclipse from starting

Eclipse will fail to launch if installed in a directory whose path contains certain invalid characters, including :%#<>"!. The workaround is to install Eclipse in a directory whose path does not contain invalid characters. (bugs 3109 and 17281)

Hanging during class loading when out of permanent generation memory

The Sun VM may hang indefinitely during class loading if it runs out of permanent generation memory. This will cause CPU usage to stay at 100% until the process is ended. See the section Running Eclipse for details on addressing this VM problem.

3.1.3 General - GCJ

GCJ is an effort by the GCC team to provide an open source Java compiler and runtime environment to interpret Java bytecode. Unfortunately, the GCJ runtime environment is not an environment that is often tested on by Eclipse developers.

The most common problems surrounding GCJ are:

The workspace's log file is a good place to check to identify whether GCJ is being used or not. Every Eclipse log session is prepended with information about the runtime environment that was used to run Eclipse. The log may include something like the following:

java.fullversion=GNU libgcj 4.2.1 (Debian 4.2.1-5)

If Eclipse does start, one can check which runtime environment is being used to run Eclipse by going to Help > About Eclipse SDK > Installation Details > Configuration. The About dialog itself can also provide other information, the build identifier can be of particular interest as it is tagged by some distributions. This allows the user to identify whether Eclipse was downloaded through the distribution's package management system or directly from the web site.

Eg: Build id: M20070212-1330 (Ubuntu version: 3.2.2-0ubuntu3)

The two most common workarounds are:

To download Eclipse, try one of the links below:

It is imperative that 64-bit builds are downloaded and used if a 64-bit Java runtime environment has been installed. Below are two sample tarball names of version 4.0.0 of the Eclipse SDK packaged for 32-bit and 64-bit processors.
eclipse-SDK-4.0-linux-gtk.tar.gz (32-bit)
eclipse-SDK-4.0-linux-gtk-x86_64.tar.gz (64-bit)

To run Eclipse with an alternate Java runtime environment, the path to the Java virtual machine's binary must be identified. With an Eclipse installation from the distribution, altering the $PATH variable to include the path to the alternate Java runtime environment is often not enough as the Eclipse that Linux distributions package often performs a scan internally to pick up GCJ by itself whilst ignoring what's on the $PATH. An example of the terminal's output is shown below:

searching for compatible vm...
testing /usr/lib/jvm/java-7-icedtea...not found
testing /usr/lib/jvm/java-gcj...found

Once the path to the virtual machine's binary has been identified, try running Eclipse with the following command:

./eclipse -vm /path/to/jre/bin/java

For an actual example, it might look something like the following:

./eclipse -vm /usr/lib/jvm/sun-java-6/bin/java
./eclipse -vm /opt/sun-jdk-

If this seems to solve the problem, it is likely that the problem really was related to the use of GCJ as the Java runtime for running Eclipse. The eclipse.ini file located within Eclipse's folder can be altered to automatically pass this argument to Eclipse at startup. An example of its content is presented below:


Note that every argument must be on its own line. More information about the eclipse.ini file can be found at

If problems persists after downloading an installation of Eclipse from and using a supported Java runtime environment (a list of which may be found above), you can seek further assistance through the newsgroups, the IRC channel, and/or bugzilla.

3.1.4 General - 64-bit Java HotSpot(TM) VM

There is a known issue with the Java HotSpot(TM) 1.6.0 VM compiler which causes eclipse to crash (see Sun bug, and Eclipse bug 214092). The crash usually occurs within a VM CompilerThread when attempting to compile the method org.eclipse.core.internal.dtree.DataTreeNode.forwardDeltaWith.

This problem has been addressed in Sun Java 6 update 11, so the simplest resolution is to obtain the latest JRE release for your platform. To work around the issue you can exclude the method org.eclipse.core.internal.dtree.DataTreeNode.forwardDeltaWith from being compiled with the following VM argument:


This VM argument can be placed in the eclipse.ini file after the -vmargs line like the following:


There have been reports of other classes that cause the compiler to crash. If all else fails you can disable the compiler with the VM arg "-Xint".

3.2 Platform

3.2.1 Platform - Core

Installing plug-ins by unzipping them into the plugins directory

New plug-ins can be installed into the system by unzipping them into the plugins directory. However this is not recommended, and the dropins directory should be used for this purpose instead. Note that unzipping a different version of a plug-in that is already installed will have no effect. To change the version of a plug-in installed in your system, you need to either perform an update, or install a feature patch.

XML files with UTF-8 byte order mark fail to have content type detected

Eclipse will fail to detect the proper content type for XML files that have a UTF-8 byte order mark if Crimson is the XML parser (as it is on Sun 1.4 JREs, but not on Sun 1.5 JREs). This problem will prevent actions normally available when files of the affected content types are selected from being presented to the user. The workaround is to ensure the default XML parser supports UTF-8 BOMs (such as Xerces does). (bug 67048)

No branding with old config.ini

If you have an old config.ini file and use it with a new Eclipse build, you may not get the correct product branding. This is because the id of the standard Eclipse product changed. Users in shared install scenarios may end up in this situation as previous builds of Eclipse automatically generated config.ini files in some cases. The work around is either to delete the local config.ini or update the eclipse.product line to read eclipse.product=org.eclipse.platform.ide.

Problems with classloaders in created threads

There is a known issue with trying to load classes from a newly-created thread using a class loader different from the plug-in class loader. The result will be a ClassNotFoundException. As a workaround, do the following:

  1. Create a thread in which to run your code.
  2. Send yourThread.setContextClassLoader(yourClassLoader); // you can find your classloader by grabbing a class it loaded (YourPluginClass.class.getClassLoader())
  3. Run your code in the newly created thread.

If you set the context class loader for the current thread, you are competing with other users of the thread (all of Eclipse), so the results will be unpredictable. However, there should be no problem in practice provided you reset the context class loader back to its original value when your use in the current thread is complete. (bug 8907)

Deadlock creating executable extension in Plugin.startup

If Plugin.startup code is too complex and performs tasks such as creating an executable extension, a deadlock situation can be created. Only simple bookkeeping tasks should be performed in Plugin.startup code. (bug 5875)

Potential Problems Converting Plug-in Manifests

If your plug-in ships with a plug-in manifest and not an OSGi bundle manifest, is shipped as a JAR file, and contains a nested JAR file then there may be problems in the automatic generation of the bundle manifest file. The packages defined in the nested JAR may not be exported correctly in the Export-packages bundle manifest header. To work around this you should ship your plug-in with a bundle manifest. (bug 97689)

Location for Debug Options File on Mac OS

If you are running in debug mode on Mac OS, the default location for the .options file is inside the application bundle in the directory (like the eclipse.ini). (bug 88782)

Issues with JNI that use FindClass

There may be issues when using a JNI implementation that uses FindClass in a function where the JNIEnv pointer is not available, such as in a C callback (bug 125250). The reason is that FindClass, in this case, uses the application class loader to find the class. If the desired class is in the classpath used for the application classloader (e.g. defined by the VM argument -cp <classpath>), as it would typically be in a stand-alone application, there is no problem. However, under Eclipse, the application classloader does not have access to classes contained in plug-ins. Eclipse uses its own class loader to find classes contained in plug-ins.

The proper plug-in class loader is used by FindClass in JNI functions which are passed the JNIEnv pointer, but not when you have to use AttachCurrentThread to get the JNIEnv pointer. In this case the application classloader is used.

For example, the following will fail because AttachCurrentThread is used to get the JNIEnv pointer:

static JavaVM* jvm;  // Global variable

void myCallback(void) {
    JNIEnv* env;
    jvm->AttachCurrentThread((void**)&env, NULL);
    // Fails if some/class is not in the application classloader:
    jclass cls = env->FindClass("some/class");
    jmethodID methodID = env->GetMethodID(cls, "methodName",
      "(Ljava/lang/String;)V or whatever signature");
    env->CallVoidMethod(callback, methodID, ...);

A solution is to cache the method ID, for example:

static jmethodID mid;  // Global variable

JNIEXPORT jint JNICALL JNI_OnLoad(JavaVM *vm, void *reserved) {
  // Store the JavaVM pointer
    jvm = vm;

  // Find the class and store the method ID
  // Will use the class loader that loaded the JNI library
    jclass cls = env->FindClass(className"some/class");
    if(!cls) goto ERR;

    mid = env->GetMethodID(cls, "methodName",
      "(Ljava/lang/String;)V or whatever signature");
    if(!mid) goto ERR;

void myCallback(void) {
    JNIEnv* env;
    jvm->AttachCurrentThread((void**)&env, NULL);
    env->CallVoidMethod(callback, mid, ...);
     // Handle error ...

3.2.2 Platform - Ant

UTF-8 encoded buildfiles with Byte Order Mark

UTF-8 encoded buildfiles with byte order marks will fail to be parsed correctly depending on the XML parser being used for the build. Therefore a valid buildfile will fail to build with an error message similar to: "BUILD FAILED: C:\workspace\bom.xml:1: Document root element is missing.". To succeed in building with these files, ensure to include Xerces jars on the Ant runtime classpath so that the Xerces parser is used to parse the XML. As well the context menu for these files in the Navigator or Package Explorer will not have the run shortcuts for Ant builds. (bug 67048)

Custom Ant tasks and Ant types must be separate from plug-in library JARs

Including the class files for custom Ant tasks or Ant types in the regular code JAR for your plug-in causes problems. These class files must be provided in a separate JAR that is contributed to the org.eclipse.ant.core.antTasks or antTypes extension point (and not declared as a library in the plug-in's manifest). This ensures that the Ant tasks and types are loaded by the special Ant class loader and not by a plug-in classloader. (bug 34466).

Concurrent Ant builds not supported

Eclipse can run Ant in the same JVM as the rest of Eclipse. Several aspects of Ant and its use of global Java resources (such as System.out and System.err), make it unsafe to run more than one Ant build concurrently in the same JVM. (bug 24129).

Running certain Ant tasks cause memory leakage

Certain Ant tasks are known to leak memory. Please see the bug report for details, patches, and possible workarounds. (bug 24448)

Tasks that require input lock up workspace

As with using Ant from the command line, prompts for input from the console is not handled. This is not the same as making use of the <input> task, which works correctly within Eclipse. (bug 21748)

"version" property is always set when running Ant in the same VM as Eclipse

The Xalan libraries set system properties including a version property. These get set as properties within the Ant build and therefore the "version" property cannot be set within an Ant buildfile due to the immutable nature of Ant properties. This property will always be set to "2.4.1" for Ant builds in the same VM as Eclipse. (bug 45717)

XDoclet support from within Eclipse

Since there are differences when running Ant from the commandline and within Eclipse, some extra steps may be needed to have XDoclet support function correctly within Eclipse. Problems may occur creating XDoclet subtasks. The workarounds and full discussion can be found in bug report. (bug 37070)

Ant Editor code completion based on Ant 1.6.1

Code completion provided by the Ant editor does not respect the user-specified version of org.eclipse.ant.core plug-in or ANT_HOME. Code completion proposals are mostly based on Ant 1.6.1 with some updates to Ant 1.6.5 (bug 30886)

Eclipse can hang due to implementation of the Ant <property> task (Windows 9X only)

On Windows 9X, using:<property environment="env"/> will cause Eclipse to hang if the build occurs in the same VM as Eclipse. Running the build in a separate VM will hang the build but not Eclipse. (bug 44196)

Setting build loggers not supported when debugging Ant builds

When debugging Ant builds within Eclipse, setting -logger as a program argument will be ignored.

Renaming an External Tool builder set to run during auto-build will cause errors

If you rename an existing external tool builder that is configured to run during auto-builds, you will get the following error: Errors during build. Errors running builder "Integrated External Tool Builder" on project <PROJECT_NAME>. The builder launch configuration could not be found. The workaround is to first disable the builder for auto-builds and then rename the builder. (bug 118294)

Slow typing/saving of the Ant editor with imports that define numerous macrodefs

The Ant editor is slow on saving with buildfiles that have <import> declarations of buildfiles that have numerous <macrodef>s. See bugs 92640 and 125117 for possible workarounds

Failure to run Ant builds on non-Windows platforms if Eclipse installed in location with spaces in the path

Due to a bug in Ant 1.7.0, Ant builds will fail with an IllegalArgumentException if the Eclipse installation is in a location with spaces in the path. Embedded usage of Ant builds, such as plug-in export will also fail. See bug 187993 for possible workarounds

3.2.3 Platform - User Assistance

Welcome page not displayed properly (Linux/Unix)

The default Welcome implementation is HTML-based and requires a supported browser in order to work. If no supported browser can be found, Welcome falls back to its Forms-based implementation, which has a different (simpler) appearance. Consult the SWT FAQ for supported browsers and setting up your browser to work with eclipse.

Help browser tool bar buttons do not work for some documents

The Help browser's Print, Synchronize, and Bookmark buttons do not work for pages that are not actually installed with the product. However, you can always use the print command in the browser's context menu to print the page you're reading. (bug 44216)

Help documents not displayed in a browser or very slow document loading (Windows only)

If your LAN settings are not properly configured for local host access, your Help browser might open to a blank page or display an HTTP error instead of a help page, or you may experience long delays when loading help documents. Your system administrator can configure your LAN settings so that help documents can be accessed from the local help server.
  1. In the Control Panel, open Internet Options, select the Connections tab and choose LAN Settings.
  2. If your host was configured to use DHCP for IP assignment, make sure that the "Automatically detect settings" check box is cleared.
  3. If you use a proxy server, ensure that the "Bypass proxy server for local addresses" is selected.
  4. In "Advanced" settings for proxies, add ";localhost" to the "Exceptions" if these addresses are not listed.
  5. If you are using an automatic configuration script for proxy settings, and are not sure that the script is correct, clear the "Use automatic configuration script" check box.

Working disconnected from the network (Windows only)

If you are experiencing problems when not connected to the network, you must install the loopback adapter from the Windows installation CD. (bug 831)

Using Internet Explorer in offline mode (Windows only)

If you have been using Internet Explorer in Offline mode, when you access the help system you will get a message indicating that the web page you requested is not available offline or a blank page will display. Click Connect or deselect "Work Offline" in the Internet Explorer "File" menu to return the system behavior to normal.

Help topics not highlighted in High Contrast mode (Windows only)

Windows High Contrast settings are not consistently picked up by Internet Explorer when they are set from the Accessibility Options utility as opposed to when they are set using the predefined schemes. On Windows XP, it is recommended to set High Contrast as follows: Right click the desktop, chose properties, select Windows Classic style from the Windows and buttons drop down on the Appearance tab, and choose your scheme (for example High Contrast Black) from Color Scheme drop down. (bug 28609)

3.2.4 Platform - UI

High contrast settings

Eclipse was tested for High Contrast using 1152 x 864 resolution in Windows XP High Contrast mode. You can select this mode by selecting Accessibility Options > Display > Use High Contrast from the Windows XP Control Panel menu.

Default text file encoding may be detected incorrectly (Windows XP/2000 only)

Note: the bug report associated with this problem has been fixed. If you run Eclipse with JDK 1.5 or greater you should not have to use the workaround stated below any longer. However, the problem still exists when running Eclipse with JDK 1.4.x or lower, so in this case the workaround is still required .

The "Text file encoding" value displayed in the Preferences dialog under "Editors" may be wrong on platforms running Windows XP (or 2000) when the user locale and system locale differ. 

Example of the manifestation of the bug: A Japanese user using Japanese Windows 2000 works in New York, United States. The user has selected English (United States) as the user locale. The "Text file encoding" value displayed by Eclipse is incorrect: "Cp1252" (English). It should display the system locale "MS932" (Japanese).

Workaround: The user can modify the user locale so that user locale and system locale are identical. In the example above, this means the user should set Japanese as the user locale. Then restart Eclipse. The "Text file encoding" value will then be correct: "MS932" (Japanese).

For Windows XP:

For Windows 2000:

(bug 20641)

Dirty state not tracked properly for OLE documents (Windows only)

The dirty state for an OLE document is not updated properly. This causes Eclipse to prompt to save the contents of the editor when the document is closed, even if the contents have already been saved. (bug 2564)

OLE document crashes can cause Eclipse to also crash (Windows only)

If an OLE document crashes, Eclipse can crash, or the workbench menus can become inconsistent.

2.1 Presentation based workspaces incorrectly get new Min/Max behavior

Workspaces that are currently using the Eclipse 2.1 Presentation will incorrectly 'inherit' the new min/max behavior when opened with 3.3.


  1. Go to the 'Preferences -> Appearance' page, change the current presentation to 'Default' and select apply
  2. Change it back to the 2.1 Presentation, select 'OK' and 'Yes' to the restart prompt
When the workbench re-opens the old min/max behavior will be restored.

Toolbars only containing contributed controls exhibit display errors on Mac/Linux

Currently there is no way on the Max or Linux platforms to define the height for controls contributed to toolbars, nor will those platforms respect the size returned by the control's computeSize method. If you encounter this issue there is currently no truly viable workaround. (bug 183003)

3.2.5 Platform - Text


3.2.6 Platform - SWT

Eclipse plug-in based on the SWT Browser throws exception

The SWT Browser widget uses a platform-specific web browser to render HTML. The org.eclipse.swt.SWTError exception ("No more handles") is thrown on platforms that don't meet the requirements for running the Browser widget. Supported platforms and prerequisites are listed on the SWT FAQ item "Which platforms support the SWT Browser?".

Opening File Dialog crashes eclipse (Vista only)

On Vista, launching eclipse using -vmargs -Xmx[any size] can crash eclipse when the FileDialog is opened. The workaround is to use the default heap size, i.e. do not use the -Xmx VM args. (bug 188317)

Crash while editing text (Windows XP with SP2 only)

Some users who have installed Service Pack 2 on Windows XP have experienced crashes while using editors in Eclipse. The workaround is to place a working version of Windows\System32\USP10.DLL in the Eclipse startup directory or uninstall Service Pack 2. (bug 56390)

Input Method broken (Motif only)

Some versions of RedHat Linux such as Fedora Core 3 and Enterprise Linux WS release 4 use a new technology called IIIM (Intranet/Internet Input Method Framework) to replace the old XIM (X input method). When running on these new systems, Eclipse will crash if you attempt to enter any DBCS character. The workaround is to use a XIM based input method such as chinput. This problem may be fixed in newer releases of RedHat. (bug 89722)

Eclipse does not start on Linux-Motif with Xinerama and a UTF-8 locale

The Linux-motif build of Eclipse does not launch properly when run on a computer with Xinerama (provides support for dual head monitors) and a UTF-8 locale. The workaround for this problem is to change the locale to a non-UTF-8 value, or to disable Xinerama. (bug 38843)

Eclipse hangs when pasting from an unresponsive application (GTK only)

If the application that is supplying the clipboard material is unresponsive, the paste operation hangs Eclipse for several minutes. This situation can be encountered when copying from an Eclipse target workbench, suspending the target workbench at a breakpoint and pasting into the hosting Eclipse workbench. (bug 44915)

Unable to drag data between applications in simplified Chinese locale (Motif only)

When configured for the simplified Chinese locale, it is not possible to drag data between applications running on the Motif window system. This is a known limitation of the Open Motif library. (bug 29777)

Crash when attempting to launch file browser (AIX Motif only)

There is a known AIX graphics bug affecting certain levels of AIX releases. Ensure that the AIX install includes the necessary service updates as described in the "Install notes/requirements for Eclipse on AIX" attachment to Eclipse bug report number 34524.

Available colors on 8-bit Linux (Linux only)

Typically, in Gnome Linux installs running with 8-bit visuals (i.e. 256 color mode), before the Eclipse application is started there are no free colors. This may mean that Eclipse is unable to allocate the default widget background color, causing it to display a white background. The functionality, however, is otherwise unaffected.

IME-related crash (Linux Motif only)

When using Linux Motif and GB18030 IME "chinput", Eclipse can crash if the IME client window is left open when the parent window is disposed. (bug 32045)

IME conversion problem (Solaris GTK only)

When typing Japanese text, the conversion to Kanji must be done one ideogram at a time. (bug 226636)

Key bindings can stop working on Debian (GTK+ only)

On some versions of Debian, Eclipse key bindings may stop working. In this context the only way to make the key bindings work again is to restart Eclipse.

The problem is that a focus issue exists in GTK+ 2.6.7 and earlier, for which SWT has a workaround. This workaround is incompatible with the GTK+ 2.6.7 fix, so a GTK+ version check is done at runtime to determine whether the workaround should be used or not. However, Debian backported the GTK+ focus fix into their libgtk+2.0 (2.6.4-2) package, so the SWT workaround and GTK+ fix are both incorrectly applied in this context.

To work around this problem, either get the Debian unstable version of GTK+, compile your own GTK+, or hack SWT's Shell.gtk_realize(int) and change the version that it checks. See SWT bug 107013 and GTK+ bug 109246 for more information.

Typing in an editor crashes with IBM 1.5 VM (Linux GTK PPC only)

When running on the IBM Java 5.0 VM, Eclipse crashes while the user is typing in an editor. If using this VM you must disable the JIT with the -Xnojit vm argument to avoid the crashes (see bug 116730). The command line for launching Eclipse with this vm should be:
eclipse -vmargs -Dosgi.locking=none -Xnojit

Eclipse won't start (Linux GTK PPC only)

Eclipse fails to create a lock file with reason "No locks available". To launch eclipse you must disable file locking using the osgi.locking property. For example, you could launch eclipse as follows:
eclipse -vmargs -Dosgi.locking=none

SWT for Carbon cannot be used with Mac OS X JRE version 1.6 (Mac OS X 10.5 only)

On Leopard (Mac OS X 10.5), JDK version 1.6 assumes that pointers have a size of 64 bits, but SWT's Carbon port only uses 32-bit pointers, so SWT and Eclipse cannot be used on Leopard with JDK version 1.6. The workaround is to use the Cocoa version of SWT or an earlier supported version of the Mac OS X JDK.

Snow Leopard (Mac OS X 10.6) has both a 32-bit and 64-bit version of JDK 1.6, so this restriction does not apply on that version of the operating system.

Strings may be truncated or incorrectly wrapped on RHEL5 (Linux GTK only)

Strings on wrapping Controls may not appear correctly in some locales on RHEL5 as a result of a bug in Pango version 1.14.x. This problem can be fixed by upgrading the installed Pango library to a version that is newer than 1.14.x. (bug 231951)

Block Selection functionality provided by StyledText is not BIDI aware

When the orientation of characters under the left and right edges of the block selection rectangle are not the same, the actual selection ranges (in memory) differ from the visual representation of the selection. (bug 277929)

Older versions of some Windows screen readers no longer work with Eclipse

JAWS versions 8 and 9 and Window-Eyes version 6 no longer work well with Eclipse and other SWT applications. Window-Eyes 6 will cause Eclipse to crash, and JAWS 8 and 9 can cause SWT applications to crash. This happens because IAccessible2 support was added to SWT for Eclipse 4.0, but these older screen reader versions contain partial implementations of IAccessible2 that do not follow the current IAccessible2 specification.

The workaround for these cases is to specify Java property org.eclipse.swt.accessibility.UseIA2 with value false, which will instruct SWT to not attempt to use IA2 interfaces. An easy way to set this property is to specify VM argument -Dorg.eclipse.swt.accessibility.UseIA2=false when launching Eclipse or your SWT application. (bug 313182)

3.2.7 Platform - Team - CVS

The following are known problems with the CVS repository provider only, and do not apply to other repository providers. Additional information on how to use CVS from Eclipse can be found in the Eclipse CVS FAQ.

CVS server compatibility

The CVS plug-in parses messages returned from the CVS server. If the format of these messages is not as expected, some of the plug-in's functionality may be missing. The CVS plug-in is compatible with all stable 1.11.X builds of the CVS server, and should be compatible with future releases in that stream unless text message formats change (the last tested server was 1.11.22). As for the 1.12.X feature releases of CVS, the Eclipse CVS client has been tested with builds up to 1.12.13. However, future releases could easily break the Eclipse CVS client. Basic functionality, such as Checkout, Commit, and Update, should always work, but there may be problems with more advanced commands such as Synchronizing and Browsing the repository.

SSH2 proxy settings lost upgrading to 3.3

CVS now uses the Platform proxy settings. As a result, any CVS proxy settings will be lost and must be re-entered on the General>Network Connections preference page.

Connection cannot be found after initially missing

If a connection initially fails due to a network problem, the connection may continue to fail even when the network problem is fixed. In order to establish the connection you must exit and restart Eclipse. (bug 9295)

"Received broken pipe signal" error from server

Eclipse sometimes performs multiple commands within a single connection to the server. This may cause problems with CVS servers that are running server scripts in response to certain commands. (bugs 23575 and 23581)

"Terminated with fatal signal 10" error from server

There is a bug in the CVS server related to some compression levels. If you get this error, changing the compression level on the CVS preference page may help. (bug 15724)

"Unknown response" error using ext connection method

There are a few situations that can result in an "Unknown response" error messages when using the ext connection method. One situation involves using an external communications client (e.g. rsh or ssh) that adds CRs to the communications channel (bug 21180). Another involves Eclipse not properly reading the stderr output of the external communications tool. (bug 11633)

A disabled CVS capability may not be auto-enabled in existing workspaces

New in 3.0 is the ability to disable capabilities and the CVS support in Eclipse can be disabled. However, for backwards compatibility the CVS capability is auto-enabled in existing workspaces that already contain CVS projects. The auto-enabling function may not run if the team support plugin is not loaded at startup. (bug 66977)

Builder output files may appear as changed

When folders containing build output are shared they may get improperly marked as dirty when build output is generated.

Enabling GNOME proxy support

GNOME applications can make use of proxy settings defined in this environment. If set, Eclipse will use it prior to proxy settings declared using env variables. This feature is disabled by default, to enable it launch Eclipse with "" switch. That is,


3.2.8 Platform - Install/Update

Manually installing features and plug-ins on a FAT file system (Windows only)

When features and plug-ins are manually installed on top of an Eclipse-based product install located on a FAT file system that has already been run at least once, the product must be explicitly restarted with -clean. That is,

eclipse.exe -clean

Connecting to untrusted sites using https

You cannot install or update software from a site using https whose certificate is not chained to a trusted root certificate in your local certificate store. This typically means the server is using a self-signed certificate, or a certificate authenticated by an unknown third party.

Extension location is lost if the install path changes

A previously configured extension location may be temporarily removed if the install is moved or mounted under a different path. This only happens when the link file that configures the extension location uses a relative path that points to a directory under the Eclipse install. On a second startup using the same install path, the extension location is added again (bug 95403).

Feature patches can only be installed from Eclipse 3.4-based update sites

Feature patches can only be installed from update sites designed for Eclipse 3.4 or greater. Specifically, the update site must have the necessary metadata for Equinox p2 (a content.xml or content.jar file). This data can be generated by building an update site using Eclipse 3.4 or newer, or running the p2 metadata generator on an update site built using earlier versions of the Eclipse SDK. See the help topic Generating p2 metadata for more details on running the p2 metadata generator (bug 244483).

3.2.9 Platform - Debug

None. (Known problems with the Java debugger appear below in the JDT section.)

3.2.10 Platform - Compare


3.3 Java development tools (JDT)

Multiple regions formatting in a given source snippet

In version 3.4, the new API method org.eclipse.jdt.core.formatter.CodeFormatter.format(int, String, IRegion[], int, String) has been added to allow the formatting of several regions in a source snippet with a single pass.
Even if specified, this method does not currently accept comments of the following kinds: This will be fixed in a future release (bug 233967).

Searching for constant field references

Search does not find references to constant fields inside binaries because the Java Language Specification mandates that constant field values be inlined in the class file's byte codes, leaving no trace of a field reference. (bug 12044)

Cut, copy, paste not working for linked resources in views showing Java elements

The cut, copy, and paste actions do not work for linked files and folders appearing in views that show Java elements, including the Package Explorer. The workaround is to use these actions from the Navigator view instead. (bug 34568)

Java working sets not working correctly for elements from JRE system library container

Applying a working set consisting entirely of elements from the JRE System library container as a filter to the packages view might result in an empty Package Explorer. (bug 30442)

Breakpoints unreliable running Sun 1.6.0_14

Developers debugging applications on Sun's 1.6.0_14 virtual machine should be aware that breakpoints are unreliable (i.e. do not always suspend execution). The problem occurs on Windows and Linux platforms. This is an issue with the VM and not with Eclipse. The workaround is to use the -XX:+UseParallelGC VM option. (bug 279137).

Side effects of Step into Selection and Run to Line

The actions "Step into Selection" and "Run to Line" optimistically set breakpoints on the line the user has chosen to step into or run to. However, the debugger can not determine if or when execution will ever reach the chosen line. The breakpoints set by the underlying implementation are not visible to the user and can cause execution to suspend unexpectedly at a later time, when the associated line is actually executed. (bug 51507)

Some refactoring script operations fail with Sun 6.0 JREs

Creating and applying refactoring scripts sometimes fails with Sun 6.0 JREs due to a bug in the XML parser that is shipped with those VMs. A workaround is to use a J2SE 5.0 VM or an IBM 6.0 VM. Another workaround is to replace the XML parser using the Java Endorsed Standards Override Mechanism:

  1. Get original versions of xml-apis.jar, xercesImpl.jar, xalan.jar, and serializer.jar from Apache, e.g. from here.
  2. Unpack the archive and copy the 4 JARs into <path-to-your-JavaSE6.0-install>\jre\lib\endorsed\.
(see Sun bug, Eclipse bug 262820)

Suspend on uncaught exception overrides exception breakpoint location filters

Exception breakpoints can be configured with location filters (inclusive and exclusive). When an unchecked exception is configured to not suspend execution in a specific class, execution will still suspend when the user preference to suspend on uncaught exceptions is on. (bug 66770)

Running Java programs with non-Latin-1 characters in package or class names

You get a java.lang.NoClassDefFoundError when running Java programs with non-Latin characters in the package or class names. The workaround is to package the class files as a JAR file and run the program out of the JAR and not from the file system directly. (bug 4181)

Cannot run or debug class in a project with GB18030 characters in project name

Most class libraries do not properly support the creation of a system process (via java.lang.Runtime.exec(...)) when the specified command line contains GB18030 characters. This limitation means the debugger cannot launch applications when the command line it generates contains GB18030 characters. (bug 32206)

Cannot detect installed JRE with GB18030 characters in path name

Automatic JRE detection fails when the JRE is stored in a directory containing GB18030 characters in its name. (bug 33844)

Cannot generate Javadoc for packages with GB18030 characters in the name

Most class libraries do not properly support the creation of a system process (via java.lang.Runtime.exec(...)) when the specified command line contains GB18030 characters. Since Javadoc is created using the Javadoc executable provided with the JDK, generating Javadoc fails if the package or class name contains GB18030 characters. (bug 32215)

Unable to debug stack overflows

If a debug session suspends on a java.lang.StackOverflowError exception (due to an exception breakpoint), the debugger may not be able to retrieve any debug information from the target JVM. As well, the debugger may not be able to reliably interact with the target JVM past this point. (bug 19217)

Evaluation limitation

The debugger uses threads in the target JVM to perform evaluations (both explicit evaluations that the user requests, and implicit evaluations such as toString() invocations in the Variables view). The Java Debug Interface (JDI) requires that the thread in which an evaluation is performed be suspended by a user event (that is, a breakpoint or step request). Evaluations cannot be performed on threads suspended by the suspend action. As well, when a breakpoint is configured to suspend the JVM rather than just the individual thread, the threads which did not encounter the breakpoint are not in a valid state to perform an evaluation. When an evaluation is attempted in a thread that is not in a valid state to perform an evaluation, an error message will appear to the effect of "Thread must be suspended by step or breakpoint to perform method invocation". (bug 34440)

Missing debug attributes

The debugger requires that class files be compiled with debug attributes if it is to be able to display line numbers and local variables. Quite often, class libraries (for example, "rt.jar") are compiled without complete debug attributes, and thus local variables and method arguments for those classes are not visible in the debugger.

Using Hot Code Replace

Hot code replace is supported on JDK 1.4.x VMs, and IBM J9 VMs. The debugger will attempt to replace all class files that change in the workspace as the user edits and builds source code. However, hot code replace is limited to changes that a particular virtual machine implementation supports. For example, changes within existing methods may be supported, but the addition or removal of members may not be.

Note that hot code replace and stepping on JDK 1.4.0 VMs was unreliable. The underlying VM problems were fixed in JDK 1.4.1, and later.


Setting a breakpoint inside a scrapbook page is not supported.

When a snippet is run in the scrapbook which directly or indirectly calls System.exit(int), the evaluation cannot be completed, and will result in a stack trace for a com.sun.jdi.VMDisconnectedException being displayed in the scrapbook editor.

Terminating a scrapbook page while it is performing an evaluation results in a com.sun.jdi.VMDisconnectedException being displayed in the scrapbook editor.

Debugging over slow connections

A global Java debug preference specifies the debugger timeout, which is the maximum amount of time the debugger waits for a response from the target VM after making a request of that VM. Slow connections may require that this value be increased. The timeout value can be edited from the Java > Debug preference page. Changing the timeout value only affects subsequently launched VM, not VMs that are already running.

Updating of inspected values

When inspecting the result of an evaluated expression in the debugger, it is important to note that the result displayed is the result of that expression at the time it was evaluated. For example, when inspecting a simple integer counter (primitive data type), the value displayed in the Expressions view is the value when the expression was evaluated. As the counter is changed in the running program, the inspected result will not change (since the view is not displaying the value bound to a variable - it is displaying the value of an expression, and the value of a primitive data type cannot change). However, if an expression results in an object, fields of that object will be updated in the inspector as they change in the running program (since the value bound to fields in an object can change).

Stepping over native methods that perform I/O

When the debugger steps over native methods that perform I/O to System.out or System.err, the output may not appear immediately unless the native performs a flush on the output buffer.

VM and process termination running on IBM 1.3 JVM on Linux (Linux only)

Terminating a launch, debug target, or system process associated with a debug target running on the IBM 1.3 JVM on the Linux platform does not work when the associated debug target has a suspended thread. To remove such debug targets from the debug UI, select Terminate and Remove from the debug view's pop-up menu (or use the shortcut "delete" key). Associated system processes in the OS may not be properly cleaned up. If a debug target has no suspended threads, termination works properly. (bug 1631)

Memory View (Linux only)

The feature to automatically load segments of memory while scrolling in the Memory view does not work on Linux. Instead the user must use the "Next Page" and "Previous Page" actions to manually load memory segments (bug 74559)

Java 6 and MacOS

Apple JavaSE-1.6 VMs only execute on 64-bit architectures but JDT will detect 1.6 VMs installed on 32-bit architectures when a new workspace is started or when the user presses the "Search..." button on the Installed JREs preference page. Error messages will appear in the log each time JDT attempts to determine which execution environments a 1.6 VM is compatible with. JDT can be configured to ignore 1.6 JREs by removing them from the Installed JREs preference page. (bug 262542)

Java Annotation Processing

Some methods in the processing API are unimplemented when compiling within the IDE, and will throw UnsupportedOperationException.

Java 6 annotation processors are supported in the batch compiler and in the IDE. By design, Java 6 processors are only executed during a build, not while editing. (bug 188558)

Java 5 annotation processors are supported in the IDE only. Java 5 processors can be executed while editing, as well as during a build. Slow annotation processors can cause a slowdown of the editing experience. If this occurs, you may wish to turn off Enable processing in editor on the Java Compiler > Annotation Processing properties page of your Java project.

Java indexing encounters problems when a folder is used both as a source and a class folder

Java indexing encounters problems when a folder is used both as a source folder in a project and as a class folder in another project. Hence, when this peculiar setup is used, the Java Search might miss matches located in such a folder. To avoid this kind of problem, it is strongly advised to use different folders for sources and binary classes. (bug 309903)

3.4 Plug-in Development Environment (PDE)

Feature manifest editor does not preserve all comments

When a non-source page of the feature manifest editor is used, PDE will convert changes back into XML by regenerating the file. Although the overall content and most of the comments are preserved, some comments may be lost. (bug 59502)

PDE will not unzip source zips of some plug-ins

In the plug-in import wizard, when you choose to import plug-ins as "projects with source folders", PDE will not unzip the source for the org.apache.ant, org.eclipse.core.runtime.compatibility.registry, org.eclipse.osgi.util and This is because the source ZIPs contains code that will not compile when unzipped as it requires additional JARs that are not part of the SDK. To avoid the creation of plug-in projects that won't compile, PDE will import these plug-ins as binary and attach source, so you would still be able to read the source, you just won't be able to modify it. Also, PDE will not unzip the source for the org.eclipse.swt plug-ins. In this case, it is because, when shipped, the swt code is spread across a plug-in and a fragment, and when unzipped, it will require circular dependencies between the plug-in and fragment projects. These circular dependencies are at minimum marked as warnings by the JDT compiler and may result in unpredictable build behavior. Therefore, PDE always imports org.eclipse.swt as binary with source attached. (bug 66314)

Emacs key bindings do not work in manifest editor fields

Non-default key bindings currently do not work in fields on non-source pages of the PDE manifest editors. (bug 19482)

Plug-in import wizard does not allow plug-ins of different versions to be imported

The Eclipse platform allows two plug-ins with the same ID but different versions to coexist if the only thing they contribute is run-time libraries. However, PDE cannot handle these plug-ins because it creates project names using plug-in Ids during binary project import. (bug 18500)

Export of plug-in may silently drop classes

When exporting a plug-in using the plug-in, feature or product wizards, some classes might be dropped from the resulting archive if their fully qualified name is too long. This typical path limitation can be worked around by creating the jar of the problematic plug-in by using the Jar export wizard. (bug 97150)

Compilation errors when exporting projects not stored outside of the workspace

When exporting multiple plug-ins and one is stored outside of the workspace, compile errors occurs on export. To work around the problem, you can either export the plug-ins one by one, or change their location. (bug 98579)

Headless build needs to be run from a fully qualified path

When running a headless build using the scripts provided by pde build, the properties builder and buildDirectory must refer to a fully qualified path. (bug 139554)

Target Platform only sees installed plug-ins

With the new p2 provisioning system in 3.4, PDE introduced a preference to control how target platforms are built. By default, this preference is on if your target equals your host, otherwise it's off. When this preference is enabled, PDE attempts to read a target platform's configuration and build the target platform based in the target's list of installed plug-ins. If a configuration can't be found (a or platform.xml file), PDE will simply manually scan the target directory and populate the target platform's list of plug-ins. (bug 226037 and bug 225148)

Delta pack is not seen by PDE when installed

If you're using the delta pack, the target platform preference for building a target based on the target's installed plug-ins must be checked off. This is because a target's runtime configuration only contains plug-ins specific to the platform it's running on. (bug 230146)

The org.osgi.util.tracker package is exported at wrong version

The Equinox OSGI Framework (org.eclipse.osgi) exports the org.osgi.util.tracker package at the incorrect version of 1.4.2. The correct org.osgi.util.tracker package version for the OSGi Release 4 Version 4.2 specification is version 1.4.0. Bundles that wish to run on other vendor frameworks should import the org.osgi.util.tracker package at version 1.4.0. By default PDE will suggest the version 1.4.2 to be used when a bundle imports the org.osgi.util.tracker package (using Import-Package manifest header). Developers can manually change the import to use 1.4.0 version of the package to be able to run on both Equinox and other vendor frameworks. (bug 279622)

Importing plug-ins as source misses resources

When importing plug-ins as source (from associated binary plug-ins and source bundles), resources such as property files and images will not be imported into the workspace. The workaround is to retrieve projects from their associated CVS repository or import binary plug-ins with attached source (if the ability to edit the imported plug-ins is not required). (bug 280259)

4. Running Eclipse

After installing the Eclipse SDK in a directory, you can start the Workbench by running the Eclipse executable included with the release (you also need a Java SE 5 JRE, not included with the Eclipse SDK). On Windows, the executable file is called eclipse.exe, and is located in the eclipse sub-directory of the install. If installed at c:\eclipse-SDK-4.0-win32, the executable is c:\eclipse-SDK-4.0-win32\eclipse\eclipse.exe. Note: Set-up on most other operating environments is analogous. Special instructions for Mac OS X are listed below.

Allocating enough memory and solving OutOfMemoryErrors

By default, Eclipse will allocate up to 256 megabytes of Java heap memory. This should be ample for all typical development tasks. However, depending on the JRE that you are running, the number of additional plug-ins you are using, and the number of files you will be working with, you could conceivably have to increase this amount. Eclipse allows you to pass arguments directly to the Java VM using the -vmargs command line argument, which must follow all other Eclipse specific arguments. Thus, to increase the available heap memory, you would typically use:

eclipse -vmargs -Xmx<memory size>

with the <memory size> value set to greater than "256M" (256 megabytes -- the default).

When using a Sun VM, you may also need to increase the size of the permanent generation memory. The default maximum is 64 megabytes, but more may be needed depending on your plug-in configuration and use. When the VM runs out of permanent generation memory, it may crash or hang during class loading. This failure is less common when using Sun JRE version 1.5.0_07 or greater. The maximum permanent generation size is increased using the -XX:MaxPermSize=<memory size> argument:

eclipse -vmargs -XX:MaxPermSize=<memory size>

This argument may not be available for all VM versions and platforms; consult your VM documentation for more details.

Note that setting memory sizes to be larger than the amount of available physical memory on your machine will cause Java to "thrash" as it copies objects back and forth to virtual memory, which will severely degrade your performance.

Selecting a workspace

When the Workbench is launched, the first thing you see is a dialog that allows you to select where the workspace will be located. The workspace is the directory where your work will be stored. If you do not specify otherwise, Eclipse creates the workspace in your user directory. This workspace directory is used as the default content area for your projects as well as for holding any required metadata. For shared or multi-workspace installs you must explicitly specify the location for your workspace using the dialog (or via the "-data" command line argument).

Specifying the Java virtual machine

Here is a typical Eclipse command line: 

eclipse -vm c:\jdk5\jre\bin\javaw

Tip: It's generally a good idea to explicitly specify which Java VM to use when running Eclipse. This is achieved with the "-vm" command line argument as illustrated above. If you don't use "-vm", Eclipse will look on the O/S path. When you install other Java-based products, they may change your path and could result in a different Java VM being used when you next launch Eclipse.

To create a Windows shortcut to an installed Eclipse:

  1. Navigate to eclipse.exe in Windows Explorer and use Create Shortcut on the content menu.
  2. Select the shortcut and edit its Properties. In the Target: field append the command line arguments.

Opening this shortcut launches Eclipse. (You can drag the shortcut to the Windows Desktop if you want to keep it in easy reach.)

Mac OS X

On Mac OS X, you start Eclipse by double clicking the Eclipse application. If you need to pass arguments to Eclipse, you'll have to edit the eclipse.ini file inside the Eclipse application bundle: select the Eclipse application bundle icon while holding down the Control Key. This will present you with a popup menu. Select "Show Package Contents" in the popup menu. Locate eclipse.ini file in the Contents/MacOS sub-folder and open it with your favorite text editor to edit the command line options.

On MacOS X you can only launch a UI program more than once if you have separate copies of the program on disk. The reason for this behavior is that every UI application on Mac can open multiple documents, so typically there is no need to open a program twice. Since Eclipse cannot open more than one workspace, this means you have to make a copy of the Eclipse install if you want to open more then one workspace at the same time (bug 139319).

If you need to launch Eclipse from the command line, you can use the symbolic link "eclipse" in the top-level eclipse folder. It refers to the eclipse executable inside the application bundle and takes the same arguments as "eclipse.exe" on other platforms.

On Mac OS X 10.4 and later, you may notice a slow down when working with significant numbers of resources if you allow Spotlight to index your workspace. To prevent this, start System Preferences, select the Spotlight icon, then the Privacy tab, then click the Add button ("+") and find your workspace directory in the dialog that appears.

Shared Install

The startup speed of a shared install can be improved if proper cache information is stored in the shared install area. To achieve this, after unzipping Eclipse distribution, run Eclipse once with the "-initialize" option from an account that has a write access to the install directory.

5. Upgrading Workspace from a Previous Release

Users who don't use "-data"

If you weren't previously using "-data" to specify your workspace, follow these steps to upgrade:

  1. Find the workspace directory used by your old version of Eclipse. Typically this is located inside the directory in which Eclipse was installed in a sub-directory called "workspace". If you are using a shortcut or script to launch Eclipse, then it will be under the current working directory of that shortcut or script in a sub-directory called "workspace". For Windows users, this is specified by the "Start in:" argument in your shortcut properties.
  2. Copy this workspace directory to a new, empty location outside of any Eclipse install directory.
  3. Install the new version of Eclipse in a new location, separate from any old version of Eclipse.
  4. If you had installed additional features and plug-ins into your old Eclipse, you should re-install them in the new Eclipse.
  5. Start this new version of Eclipse and select this location using the workspace chooser dialog at startup (or use "-data" command line argument to pre-select the workspace location).

Users who do use "-data"

If you were previously using the "-data" argument to start Eclipse, your upgrade path is much easier:

  1. Optionally copy your workspace directory to a new, empty location outside of any Eclipse install directory as a backup.
  2. Install the new version of Eclipse in a new location, separate from any old versions of Eclipse.
  3. If you had installed additional features and plug-ins into your old Eclipse, you should re-install them in the new Eclipse.
  4. Start this new version of Eclipse and select this location using the workspace chooser dialog at startup (or use "-data" command line argument to pre-select the workspace location).

Note: Copying your workspace is recommended because, after you've upgraded your workspace, you won't be able to use it again with an older version of Eclipse. If you ever want to go "back in time" to an earlier release, you will need that backup.

Users who use User Libraries or classpath containers that contain JARs referencing other libraries via Class-Path in the MANIFEST.MF

If you want the referenced JAR files to be included in the classpath, you can do one of the following:

6. Interoperability with Previous Releases

6.1 Interoperability of Release 4.0 with previous releases

Sharing projects between heterogeneous Eclipse 4.0 and 3.6

Special care is required when a project in a team repository is being loaded and operated on by developers using Eclipse-based products based on different feature or plug-in versions. The general problem is that the existence, contents, and interpretation of metadata files in the workspaces may be specific to a particular feature or plug-in version, and differ between versions. The workspace compatibility guarantees only cover cases where all developers upgrade their Eclipse workspaces in lock step. In those cases there should be no problem with shared metadata. However, when some developers are working in Eclipse 4.0 while others are working in Eclipse 3.6, there are no such guarantees. This section provides advice for what to do and not to do. It addresses the specific issues with the Eclipse SDK.

The typical failure mode is noticed by the 4.0 user. 4.0 metadata is lost when a 3.6 user saves changes and then commits the updated metadata files to the repository. Here's how things typically go awry:

Here are some things to watch out for when sharing projects between Eclipse 4.0 and earlier 3.x releases:

Using Eclipse 4.0 to develop plug-ins that work in Eclipse 3.6

It is also possible (and reasonable) to use Eclipse 4.0 to develop a plug-in intended to work in Eclipse 3.6 or earlier. Use the Plug-in Development > Target Platform preference page to locate non-workspace plug-ins in an Eclipse 3.6 install. This ensures that the code for your plug-in is being compiled and tested against Eclipse 3.6 APIs, extension points, and plug-ins. (The above list of concerns do not apply since they affect the layout and interpretation of files in the plug-in project but none affect the actual deployed form of the plug-in.)

Sun, Solaris, Java and all Java-based trademarks are trademarks of Oracle Corporation. in the United States, other countries, or both.

IBM is a trademark of International Business Machines Corporation in the United States, other countries, or both.

Microsoft, Windows, Windows NT, Vista, and the Windows logo are trademarks of Microsoft Corporation in the United States, other countries, or both.

Apple and Mac OS are trademarks of Apple Computer, Inc., registered in the U.S. and other countries.

QNX, Neutrino, and Photon are trademarks or registered trademarks of QNX Software Systems Ltd.

Other company, product, and service names may be trademarks or service marks of others.

(c) Copyright IBM Corp. and others 2009, 2010

Appendix 1: Execution Environment by Bundle

In the table below, the "4.0 minimum execution environment" column indicates the minimum Java class library requirements of each bundle for the 4.0 release, where the value is one of:

Entry Meaning
J2ME Foundation 1.0 - indicates that the bundle can only be run on Foundation 1.0 or greater. Note that with the exception of some MicroEdition IO classes, Foundation 1.0 is a subset of J2SE 1.3.
J2ME Foundation 1.1 - indicates that the bundle can only be run on Foundation 1.1 or greater. Note that with the exception of some MicroEdition IO classes, Foundation 1.1 is a subset of J2SE 1.4.
J2SE 1.3 - indicates that the bundle can only be run on JSE 1.3 or greater.
J2SE 1.4 - indicates that the bundle can only be run on JSE 1.4 or greater.
Java SE 5 - indicates that the bundle can only be run on Java SE 5 or greater.
Java SE 6 - indicates that the bundle can only be run on Java SE 6 or greater.
n/a Unknown at the time of this revision.

Table of minimum execution environments by bundle. (See also the Equinox Project plan for the execution environment requirements of bundles contributed via that project.)


not specified
not specified
not specified
not specified
not specified
not specified
not specified
not specified
not specified
not specified
not specified
not specified
not specified