For international users and manual translators,
I found the trouble of building the manual when we run Ant with non English OS.
I know it's an OS independent problem because I'm using Japanese Windows and Linux.
The trouble is that we cannot build pages with European characters such as "umlaut"s even if the pages are written by UTF-8. This looked strange for me because I used Java, Ant and UTF-8 encoding. Why?
The trouble comes from Ant copy task with filter. The JBoss manual uses filters: @TODAY@ is embedded in the opening page of the manual, for example. The copy task with filtering="yes" will read/write files with default encoding, not UTF-8. That's why even UTF-8 xml files cannot converted into html correctly.
The bottom line is "Set LANG=C before build the manual".
This info is also important for manual translators who uses UTF-8 as xml encoding of the JBoss manual.