Re: [RFA] JDWP single step callback

[Keith Seitz <keiths at redhat dot com>]

Tom Tromey wrote:

> How about making a patch to add a new news entry to the gcj web 

I'll work on that.

> Keith> +      fprintf (stderr, "libgcj: internal error: could not find class for method while single stepping -- continuing\n");
>
> I'm not super happy about doing this but I see we already have some in
> jvmti.cc.  And it isn't immediately clear what would be better.

Yeah, there's not a whole lot we can do. This really shouldn't happen, 
though. I guess I could change these to assertions, as I did in a few 
other places. Better?

> Keith> +  jobject instance = NULL; // FIXME
>
> What's up with this?  Just put a short explanation (a couple words,
> say) in the FIXME and I think this patch is ok.  Thanks.

That's an oversight on my part. I forgot to add the code for that. [dig 
dig] Oddly, I remembered it as part of the breakpoint callback... In any 
case it's the same code:

  JvAssert (thread->frame.frame_type == frame_interpreter);
  _Jv_InterpFrame *iframe
    = reinterpret_cast<_Jv_InterpFrame *> (thread->interp_frame);
  jobject instance = iframe->get_this_ptr ();


Let me know about switch the fprintfs to assertions.

Keith