Re: Crash invoke()'ing Method with "double" as return type

[Martin Egholm Nielsen <martin at egholm-nielsen dot dk>]

Hi,

> > Martin> Just ran into a strange problem on my target: When calling 
> > invoke() on
> > Martin> a java.lang.reflect.Method returning a double, my target 
> > crashes with
> > Martin> an "Illegal instruction" attempting to return:
> >
> > Martin> And it's not working on double in general that sucks - output 
> > line 3
> > Martin> is the output after invoking the same method in a non-reflective
> > Martin> manner.
> >
> > This I don't recall.
> >
> > Martin> This is on my ancient (hacked) GCJ 3.4.x, running on my PPC405EP
> > Martin> target... I know it's old - but does it ring a bell?
> >
> > The only problem I recall is that returning a double via libffi on
> > PPC64 could cause a buffer overflow.  

> Maybe it's a similar problem. I've tracked down the problem to occur 
> invoking "ffi_call_SYSV" in "sysv.S" (from ffi.c's "ffi_call"). And 
> that's where my trace stop.
>
> So, next step is to try with SVN trunk - if possible...

And that did not change anything - it still fails invoking "ffi_call_SYSV".
Maybe there is a similar buffer overflow on PPC32?
Can anybody else verify/try this on PPC32?
Maybe I'm just not doing it right...

BR,
 Martin Egholm - changing the return type to java.lang.Double for now...