From: Tim Wilkinson (tim@transvirtual.com)
Date: Mon Nov 16 1998 - 18:11:12 EST
Godmar,
I'd be happy with you patches - I was just proposing the alternative and don't
really have a strong feeling for it (actually my idea might be unworkable
anyhow).
Tim
> >
> > Guys,
> >
> > We could use the 'exceptObj' in the Thread context (which really should
> > be a java.lang.Throwable and not a kaffe.util.Ptr since that's what it
> > points at - I should fix that) to hold any pending exception, just like
> > the JNI does. However I take the point that we couldn't use the
> > standard mechanism to construct the exception object - so the throwable
> > would have to be hand crafted in the C code, which would be okay
> > I suppose.
> >
>
> I still see two independent issues here:
> first, do we want to rely on thread-specific data instead of passing a
> pointer along. Second, should we store pending as throwables or not.
>
> As for the first question:
> I'm willing to go with the first solution (except that I know we will
> have to trick in the initialization, which I don't really like.)
> Also, if Kaffe ever goes to multiple instances, we'll most likely
> have to carry a pointer along anyway. Also, it seems to me that
> the functions we're talking about are not in the inner loop of any
> application: it's all loading, linking, jitting, etc.
>
> As for storing throwables (and handcrafting them) I'm somewhat skeptical.
> That doesn't seem to be a good trade-off to me: save one word per thread ---
> which sums up the savings --- and introduce another dependency between
> Java and VM code.
>
> Would you guys mind if I finish up what I'm working on and check that
> in; so you can look at it and we can decide whether make change one
> and/or change two later?
>
> - Godmar
-- Tim Wilkinson Tel: +1 510 704 1660 Transvirtual Technologies, Inc., Fax: +1 510 704 1893 Berkeley, CA, USA. Email: tim@transvirtual.com
This archive was generated by hypermail 2b29 : Sat Sep 23 2000 - 19:57:03 EDT