Since I opened the issue I copied the same archive to a workstation using the Sun JDK V1.5 Update 6 and recreated the same problem. I think this means that it is unlikely that the IBM JDK is causing the problem.

I have also modified the velocity-1.4.jar file in the .maven\repository\velocity\jars directory. I replaced org\apache\velocity\runtime\defaults\velocity.properties after making the following changes.

#----------------------------------------------------------------------------

1. T E M P L A T E E N C O D I N G
   #----------------------------------------------------------------------------

input.encoding=UTF-8
output.encoding=UTF-8

I recreated the problem after making this change. I like to think that the problem is in velocity, but this change did not affect the outcome.

Next I will move to Red Hat Linux and try the suggested work around there.

Lance

NOTE: Although it has no affect on this problem, I have discovered a defect in the test case I supplied. The properties files in src\i18nBundles have not been converted to the required ASCII encoding. I expected the translators to return ASCII encoded files, but they used some native format instead. As a result, the navigation items, the section, headers, and the subsection headers will appear wrong, even if the rest of the page is generated correctly.

I will attach an updated test case when I have converted the properties files correctly. I first have to find out what encoding the translators used (it is obviously not UTF-8) and fix them with native2ascii.

Lance

I found an old Red Hat Linux system where I could run the supplied test case. Precisely, it is Red Hat Enterprise Linux V4 update 3 for i386. I installed Maven 1.1-beta-2 with maven-xdoc-plugin-1.9.2 and the attached test case. I created a script that matches the actions in build_de.bat, build_en.bat, build_fr.bat, and build_zh_TW.bat.

Except for the unrelated problem caused by poison properties files in src\i18nBundles (see the problem report in a previous comment), the HTML was generated CORRECTLY.

NOTE: I did NOT have to modify the LANG or LC_CTYPE environment variables, as suggested in the xDoc plugin FAQ or in http://jira.codehaus.org/browse/MPXDOC-184 . By default, LANG was already set to LANG="en_US.UTF-8". I dumped the Java system properties and observed that file.encoding=UTF-8 by default.

So, that begs the question, "Why doesn't this work on a Windows XP system when you use -Dfile.encoding=UTF-8 to override the default file encoding?" Its a mystery.

This is an updated test case where the poison properties files have been converted to ASCII using the native2ascii utility. When you run the supplied batch files on a Windows system, the navigation pane, the section names, and the subsection names will appear correctly, however, special characters in the other content will be scrambled.

Lance