Re: Bugs in Soot (fwd)

[Winghong Felix KWOK <wkwok@sable.mcgill.ca>]

> 
> Now,
> 
> For some reason when I do
> 
> java soot.Main -O -c -d lib --soot-classpath $SOOT_CLASSPATH --process-path
> lib
> 
> only the kaffe and com packages get sooted, leaving java and javax unsooted
> 
> 
> So I wrote my own directories recursive descent program.

That's a little strange. It should do everything as far as I can tell.

> 
> Now it seems Soot breaks java.lang.Object, since while sooting every file in
> turn, if I stop the process right before going through Object, kaffe works,
> but if I sootify Object, it then core
> dumps.
> 
> Also, if I sootify all the packages, but replace Object by one I had kept
> aside before sootifying, it works
> 
> Attached Object.j is the dejava of Object.class not sootified, Object.j.soot
> is the one sootified. Object.java is the source.
> 
> 
I ran into the same problem with java.lang.Object. For some reason, it
doesn't sootify correctly. I don't know why, but I think it's a bug in
Soot. Right now, just use the unsootified class file. We'll try to find
out why it doesn't work.

> Anybody know if the best (speed, size) byte code generated is through  javac
> then Soot?
> For example, jikes is fast but bad on code generation.
> 
Hmm, don't know that one. I'll let someone else answer that.

Felix