Bug ID: JDK-4979996 Should default to headless java when DISPLAY is null on unix

Type: Enhancement
Component: client-libs
Sub-Component: java.awt
Affected Version: 1.4.2,5.0

Priority: P2
Status: Resolved
Resolution: Fixed
OS: linux,solaris_10
CPU: x86,sparc

Submitted: 2004-01-18
Updated: 2004-03-13
Resolved: 2004-03-13

Other
5.0 b43Fixed

Currently, in order to use headless java, the headless flag must be 
specified when launching the VM:  java -Djava.awt.headless=true  

Judging by traffic on the java forums, and from bugs we have received, many 
people find this confusing.  It has recently been suggested that we should 
default to headless java on unix and linux systems when the DISPLAY 
variable is null.  

Note: we should not default to headless when the DISPLAY is invalid.  

Also, this heuristic should not override user settings: if the DISPLAY is null, 
but headless java has been disabled at launch (-Djava.awt.headless=false) 
we should not choose headless.

CONVERTED DATA BugTraq+ Release Management Values COMMIT TO FIX: tiger-beta2 FIXED IN: tiger-beta2 INTEGRATED IN: tiger-b43 tiger-beta2

24-08-2004

SUGGESTED FIX We should specify a new property, perhaps: Boolean sun.awt.assumeHeadless which we would set to true when the DISPLAY is null and no direction has been given to the VM about whether to use headless or not. In this case, we would also set java.awt.headless to true. We would use sun.awt.assumeHeadless rather than java.awt.assumeHeadless since this would not be a publically supported property: it would be for our internal use only. Therefore we should not need to check whether a user has set it. I assume these properties would be available as environment variables, too? ###@###.### 2004-01-17 Joe Kowalski brought up the case in which the DISPLAY variable is set but has no value. We should consider this case as well. I think the right thing may be to treat it the same as an invalid value: we would not assume headless. ###@###.### 2004-01-19 Name: atR10251 Date: 04/09/2004 ------- GraphicsEnvironment.java ------- *** /tmp/sccs.7WaGX2 Tue Feb 10 19:40:52 2004 --- GraphicsEnvironment.java Tue Feb 10 19:40:16 2004 *************** *** 40,49 **** --- 40,54 ---- * The headless state of the Toolkit and GraphicsEnvironment */ private static Boolean headless; /** + * The headless state assumed by default + */ + private static Boolean defaultHeadless; + + /** * This is an abstract class and cannot be instantiated directly. * Instances must be obtained from a suitable factory or query method. */ protected GraphicsEnvironment() { } *************** *** 91,118 **** * a display, keyboard, and mouse; <code>false</code> * otherwise * @see java.awt.HeadlessException * @since 1.4 */ ! public static boolean isHeadless() { return getHeadlessProperty(); } /** * @return the value of the property "java.awt.headless" * @since 1.4 */ private static boolean getHeadlessProperty() { if (headless == null) { ! String nm = (String)java.security.AccessController.doPrivileged( ! new sun.security.action.GetPropertyAction( ! "java.awt.headless", "false")); ! if (nm.equals("true")) { ! headless = Boolean.TRUE; ! } else { ! headless = Boolean.FALSE; ! } } return headless.booleanValue(); } /** --- 96,149 ---- * a display, keyboard, and mouse; <code>false</code> * otherwise * @see java.awt.HeadlessException * @since 1.4 */ ! public static boolean isHeadless() { return getHeadlessProperty(); } /** + * @return warning message if headless state is assumed by default; + * null otherwise + * @since 1.5 + */ + static String getHeadlessMessage() { + if (headless == null) { + getHeadlessProperty(); // initialize the values + } + return defaultHeadless == null ? null : + "\nNo X11 DISPLAY variable was set, " + + "but this program performed an operation which requires it."; + } + + /** * @return the value of the property "java.awt.headless" * @since 1.4 */ private static boolean getHeadlessProperty() { if (headless == null) { ! java.security.AccessController.doPrivileged( ! new java.security.PrivilegedAction() { ! public Object run() { ! String nm = System.getProperty("java.awt.headless"); ! ! if (nm == null) { ! String osName = System.getProperty("os.name"); ! headless = defaultHeadless = ! ("Linux".equals(osName) || "SunOS".equals(osName)) && ! (System.getenv("DISPLAY") == null) ? ! Boolean.TRUE : Boolean.FALSE; ! } else if (nm.equals("true")) { ! headless = Boolean.TRUE; ! } else { ! headless = Boolean.FALSE; ! } ! return null; ! } ! } ! ); } return headless.booleanValue(); } /** ------- HeadlessException.java ------- *** /tmp/sccs.O.ai52 Tue Feb 10 19:40:53 2004 --- HeadlessException.java Tue Feb 10 17:53:40 2004 *************** *** 18,23 **** --- 18,35 ---- public class HeadlessException extends UnsupportedOperationException { public HeadlessException() {} public HeadlessException(String msg) { super(msg); } + public String getMessage() { + String superMessage = super.getMessage(); + String headlessMessage = GraphicsEnvironment.getHeadlessMessage(); + + if (superMessage == null) { + return headlessMessage; + } else if (headlessMessage == null) { + return superMessage; + } else { + return superMessage + headlessMessage; + } + } } ###@###.### ======================================================================

24-08-2004

EVALUATION Should address this for tiger-beta2. ###@###.### 2004-01-17 Note that there has been some discussion of printing an error message to stderr (or perhaps logging some kind of notification), or adding an explanatory message to the HeadlessException if we are assuming headless java because of a null DISPLAY. I'm not sure if we need to consider localizing such messages. ###@###.### 2004-01-17

17-01-2004

Duplicate :	JDK-5032424 - 'new StyleSheet()' fails for headless applications
Relates :	JDK-4710324 - rmic uses AWT ?!