It seems from my naive perspective that more information could be automatically
extracted from ant scripts. It may be that what is needed are some enhancements
to the ant tool itself to make identifying some of this information easier
(given that it clearly manages to parse its own files).

1. Libraries of jar could be compiled from <path> and <classpath> declarations
and <war>s <lib> sub-element. This won't include javadocs automatically, but at
least all the jars that the IDE will use will be those that the build will
actually use. If the IDE generated libraries the javadocs could be associated
with them later.

2. Java source paths could be identified against <javac>. Interestingly this
possibility could allow NetBeans to cope with the case where unit tests and
library code are intermingled if <include> and <exclude> were examined (although
I guess that could require significant internal changes).

3. Ant scripts can have optional behaviour based upon whether parameters are
set, or may expect particular parameters to be set outside of the script. Such
parameters will be used but never set in the script. Parsing could extract these
parameters and make the setting of them available in the project properties
(global options doesn't seem like the right place - they are project options
after all).