Re: Eclipse

[Stuart Ballard]

Tom Tromey wrote:
> According to japi we aren't really missing much API at all.  I suspect
> some of the "errors" listed on the comparison page are actually things
> that changed in later JDK releases (it would be cool if japi knew how
> to do a 3 way comparison so we could filter these out).

Actually thanks to Michael Koch I know exactly how to implement that - I 
just haven't found the time to integrate it yet. His initial patch was a 
3-liner which provided the key insight into how to make it work, but 
there are some subtleties that need to be taken into consideration that 
I didn't think of to start with.

For example, against JDK1.0 the entire java.awt.peer package is listed 
as missing; that should naturally be skipped based on this filter, but 
then what do you show as the percentage for that package? 100% missing 
but no errors, 100% good even though "good" simply means one big error 
that was skipped, or (ideally) "100% skipped", where skipped is a new 
category that doesn't count against us *or* for us in the statistics.

Savannah has accepted the japitools project, btw, and I'm waiting for 
them to import my CVS history now. Then anyone who wants to help hacking 
on this can get access to the repository, and we can make a japitools 
mailing list etc.

I'm hoping that having the project open to other developers will allow 
me to delegate a few of the simpler tasks that need to be done. And 
maybe the more complicated tasks, too - I'm unlikely to be able to make 
any attempt to interpret the 1.5 bytecode features or incorporate them 
into japi files in the forseeable future, but I'd love if someone else 
could work on this. Any volunteers? :)

Stuart.

--
Stuart Ballard, Senior Web Developer
NetReach, Inc.
(215) 283-2300, ext. 126
http://www.netreach.com/