With no erors or warnings javadoc simply produces a dummy (no entries)
file for the index of all methods (one of the more useful things it
does). It looks like it does this in at least the following
cases.

1) If any class in any package has a static initialisation
   block. eg

   public class Buggy
   {
     static
     {
       int init = 0;
     }

     Buggy ()
     {}
   }

2) If any class uses silly variable names that start
   with underbars

   public class ConfuseJavadoc
   {
     public int _thisIsAllItTakes;
   }

This is a royal pain, since NOTHING comes out for the index
even if there is only one offending class buried away in a big
mass of packets (eg the whole project...). Also with no error
messages or hints as to why it failed to produce the index
its tough to track down what is going on.

I'd really like to have the method index come out. Maybe you
know of a way around the static init problem...How did you
folks get it to work with all your classes ????

Thanks John Cownie
###@###.###
---------
I agree, javadoc should handle these cases, but at least
the error messages should be much better and
should pinpoint at least the class if not the member
where it stumbled.

I'm not that familiar with all our source code. Do we
have both cases in our source code? We're not doing
anything special.

-Doug Kramer
-------
This is the strangest thing. I'm sure there must be some static
initialisations somewhere in all that code but it looks like
javadoc somehow handled it all ok. This makes me think I'm
missing something or don't have a magic flag thats needed.
Maybe you have a high priest of javadoc there who made the
right offerings and ritals to perform before running the beast.

If you have any clues please let me know. Not getting the index
out was just something that is bugging me. I have a nice scheme
that was to regenerate the docs each night and link into our
html cvs tracking audit log system (to give it a rather grand title.)

John Cownie