[kaffe] verifier and threads

[Dalibor Topic]

Hallo Helmer,

--- Helmer Krämer <hkraemer@freenet.de> wrote:
> On Tue, 21 Jan 2003 11:13:57 -0700
> Patrick Tullmann <tullmann@cs.utah.edu> wrote:
> 
> > Helmer wrote:
> > > seems like I tracked this one down. The problem
> was that
> > > Class.forName () catched the VerifyError and
> threw a
> > > ClassNotFoundException instead. I've attached a
> pretty
> > > simple patch that should fix this.
> > 
> > While I admit that's weird to replace one
> exception with the other I'm
> > quite certain it was done for a reason.  Since
> there aren't any
> > comments of note, I'm guessing that the code you
> removed intentionally
> > changed an unchecked Error into a checked
> Exception, so that users of
> > the method would have a chance to deal with it.
> 
> Yep, I don't think it was changed without reason
> either, but
> 
> - the api spec states that forName throws a
> LinkageError and
>   VerifyError is a subclass of it
> 
> - sun's jdk throws VerifyErrors
> 
> - if you simply convert a VerifyError into a
> ClassNotFoundException
>   you can't tell whether the class was not found or
> whether the
>   class was found but contains invalid byte code (or
> whatever)
> 
> - how would you handle a VerifyError? 

Yeah, and additionally there is no discussion of it on
the mailing list, as I'd assume in the case it was
done to get an application to run, and the comment is
very unclear. Beside that, your patch doesn't make any
regression test fail, so due to the lack of evidence
to the contrary, I've checked it in. ;)

Whatever the reason was to catch the verifier error,
it doesn't seem to be present anymore.

best regards,
dalibor topic


__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com