[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Soot or SableVM bug?

Ondrej Lhotak wrote:
On Thu, Jul 08, 2004 at 12:20:56AM -0400, Etienne Gagnon wrote:

Hi there,

Does anybody have an idea what's the cause of

  java.lang.RuntimeException:
  Aborting: can't find classfile[C
  ...

in: hhttp://sablevm.org/bugs/10#c3

Thanks for your precious help, 


This is because some new javac compilers (and/or javac clones) generate
code with references to methods on arrays, when before, the reference
would always be to the method on java.lang.Object. Coffi used to not be able to
handle references to methods on arrays. Method resolution is supposed to
convert them to java.lang.Object, but resolution happens after Coffi.

See:
http://www.sable.mcgill.ca/sootbugs/incoming?id=85;page=2;user=guest
http://www.sable.mcgill.ca/listarchives/soot-list/msg01254.html
http://www.sable.mcgill.ca/listarchives/soot-list/msg01288.html

This bug was fixed in revision 1391. However, no Soot releases have been
made with the fix.

I notice you're using Soot 1.2.5, which is ancient.

I'll just pipe in and say that I'm using the Soot trunk all the time to transform and add attributes to all of GNU Classpath (including java.lang.Object! woohoo!), and I'm not aware of any bugs when reading them with SableVM. To build it you need the polyglot CVS and Jedd runtime (although I'm still using r1514 myself, which doesn't require Jedd, should probably update soon).

Cheers,
Chris