Community
Participate
Working Groups
F1 on XP (Chinese I believe) shortcut: "C:\Eclipse F1\eclipse\eclipse.exe" -vm d:/j2sdk1.4.0/jre/bin/javaw.exe -data c:\רגפ\ I can see the first startup "Completing the install please wait", never the second one
The % char in En_us machines also does not work (URL problem?). We should have a README or update the docs on invalid chars for the workspace location.
Also, bellow is what I get in the .log file. If we can somehow anticipate/validate incorrect chars for the platform location a more informative message should be logged instead of this stack trace. !SESSION ---------------------------------------------------------------------- !ENTRY org.eclipse.core.launcher 4 0 May 22, 2002 11:04:12.158 !MESSAGE Exception launching the Eclipse Platform: !STACK java.lang.IllegalArgumentException at sun.net.www.ParseUtil.decode(ParseUtil.java:100) at sun.net.www.protocol.file.FileURLConnection.connect (FileURLConnection.java:85) at sun.net.www.protocol.file.FileURLConnection.getInputStream (FileURLConnection.java:164) at java.net.URL.openStream(URL.java:825) at org.eclipse.core.launcher.Main.load(Main.java:869) at org.eclipse.core.launcher.Main.loadProperties(Main.java:840) at org.eclipse.core.launcher.Main.loadConfiguration(Main.java:797) at org.eclipse.core.launcher.Main.processConfiguration(Main.java:691) at org.eclipse.core.launcher.Main.run(Main.java:642) at org.eclipse.core.launcher.Main.main(Main.java:476)
Hum, interesting On the XP machine, in the browser, the char רגפ are supposed to be some indian or other alphabet char that the HTML didn't encode or that my NT cannot read ;-) Just FYI I looked at the HTML source and cannot see them encoded.. oh well, now you know
see also bug 7711.
It looks like neither IBM nor Sun JREs are able to correctly accept command line arguments containing Unicode characters on a Unicode-enabled Windows. The platform fails trying to find an existing configuration file or creating a new one in the workspace, because the Unicode chars specified in the command line are replaced by ?s (question marks), which are invalid characters for a file name on Windows. The .log file is not written for the same reason. The Windows API provides functions for retrieving command-line arguments containing Unicode characters, but it seems they are not being used in the Java VMs at all.
I don't think there is an easy way to overcome this limitation. Eclipse gets its options from the command line, and the JRE is not providing the right characters to Eclipse. But I think that when we are not able to write the log to the standard location (the workspace .metadata directory), we should try to write the error messages to an alternative location (maybe the temp dir) before exiting. And we should at least document this limitation. Any suggestions, DJ?
In order for a Java application to get the right characters from the command- line in XP, you must set the right language as the default language for non- unicode programs (Control Panel -> Regional and Language Options -> Advanced). This is probably the default for a localized version of Windows XP. I just opened another bug (bug 32026) to address the situation where the workspace location in the file system cannot be written for some reason and we fail to write to the log. Closing as works-for-me.
