[Soot-list] tracking down class-resolver problems

[Chris Pickett]

Will Benton wrote:
> Thanks for all your help, Chris.

It's fine, really... after all, many of us benefit from figuring this 
stuff out.

>> Sorry, what did you have to delete?
> 
>  From chart, I had to delete com.lowagie.servlets.ITextOutputFilter.  It 
> appears to not be referenced anywhere else in the application and it was 
> the only reference to the SAXmyHandler class.
> 
>>> Yes, the problem I've had with xalan (and fop) is that I can't build 
>>> and analyze them against GNU Classpath (which I want to do because I 
>>> will be running them in Jikes).  They won't build against any of the 
>>> GNU Classpath libraries I have installed (because they are JDK 1.5 
>>> libraries), and if I build them against the Sun JDK, they will 
>>> contain references to classes that aren't in GNU Classpath, so they 
>>> won't analyze.  If anyone has a solution for this (or a suggestion 
>>> for a suitably ancient version of Classpath), I'd appreciate hearing 
>>> about it.
>>
>> Try -source 1.4
> 
> I did try this, both in the ant properties and on the command line.  I 
> am under the impression that -source only affects syntactic issues (e.g. 
> assert, enum, generics, etc.) and that -target only affects the 
> generated bytecodes.

I think so too.

> Neither -source nor -target solves the problem I'm having -- namely, 
> that the org.w3c.dom.Node interface has some methods in Java 1.5 that it 
> does not in Java 1.4, and classes depended upon by fop and xalan 
> subclass the "old" version of this interface.  Since these classes don't 
> implement all of the methods in the "new" version of the interface, it 
> looks like I'm stuck.
> 
> I suppose I could a very old version of Classpath for these two 
> benchmarks, or hack the Node interface in the current release, but 
> neither is a particularly appealing solution.

Don't bother with old CP releases.  If it's just interface and not 
implementation issues, it sounds quite fixable.  You could hack the 
interface or the subclasses or both.

Chris