Re: Nightly Japi comparisons

[Stuart Ballard <sballard at netreach dot com>]

Tom Tromey wrote:
> > > > > > "Anthony" == Anthony Green <green at redhat dot com> writes:
>
>
> Anthony> I'm equally puzzled by the Class.getPackage() comment.  It
> Anthony> claims that libgcj's implementation throws a
> Anthony> NoSuchMethodException, but I see no such exception being
> Anthony> thrown.
>
> This one I don't know about.  Maybe this is a japi bug?  I've CC'd
> Stuart, in case he knows.

The following line (it's a single line but my mailer will probably 
wordwrap it) appears in the libgcj.japi.gz file that I'm downloading 
from you guys.

+java.lang,Class!getPackage() Pcinu 
Ljava/lang/Package;*java.lang.NoSuchMethodException

That means the problem, whatever it is, is happening either during the 
Japize process or during compilation. If any of you know about bytecode 
formats, you could try verifying that the .class file produced by the 
compiler doesn't include this exception. If it does, that's a compiler 
bug, of course (luckily, the compiler is presumably gcj, so this is 
going to the right list... ;) ).

If it doesn't, then we're looking at either a bizarre algorithmic bug in 
japize itself (possible, I guess, but I find it hard to imagine what bug 
could produce that symptom) or a bug in the Jode code that Japize uses 
to read bytecode.

Depending on how much time or inclination you have to look into this, 
I'd suggest adding debugging statements in the constructor of 
JodeMethod.java and/or the japizeClass method of Japize.java to verify 
whether the bad information is coming from Jode itself or being 
introduced by Japize later.

If it's coming from Jode, I'd suggest mailing the suspect .class file to 
the Jode maintainer, Jochen Hoenicke (sp?) whose contact information can 
be found at http://jode.sourceforge.net/, along with a brief description 
of the problem. Be sure to mention that japitools uses a (very lightly 
patched) Jode 1.1.x and not the current development version.

If you don't have the time or inclination to look into it yourselves, 
I'll try to take a look at it sometime but my time availability is 
pretty limited at this point. If you could at least rule out the 
possibility of a compiler bug, that would be much appreciated, because 
I'm really not sure how I'd do that exactly (but someone on the GCJ list 
must, surely! ;) ).

Stuart.

--
Stuart Ballard, Programmer
NetReach - Internet Solutions
(215) 283-2300, ext. 126
http://www.netreach.com/