Re: japi note

[Stuart Ballard]

Tom Tromey wrote:
> I hacked around a bug in Japi (without really understanding it), and
> as a result removed some of the bogus results in the comparison page.
> Some still remain -- there seems to be a problem with computing the
> serialization uid for some classes (eg java.applet.Applet.AccessibleApplet).
>
> The nightly japi runs should improve a little starting tonight.
> There weren't very many errors, the results improve less than a point.

For a (possibly) larger improvement, would you consider building GNU 
JAXP as part of classpath.japi.gz? Since Classpath has apparently made 
the official decision that this is the recommended source for javax.xml 
for the moment (at least), it would give a fairer reflection of the 
state of Classpath if it were included.

I'd like to see a similar thing done for Tritonus as well; I don't know 
whether Classpath will be officially recommending that for javax.sound 
but it seems to be the leading free implementation. Kaffe, which bundles 
it, scores 100% on javax.sound where Classpath scores 0%.

I've regenerated my jdk japi files using your patch and I'm regenerating 
the results pages now. I don't know if the JDK was somehow immune from 
the effects of the bug, but it can't hurt. Still planning to do a new 
release with the patch.

Stuart.

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