Fwd: [Pnet-developers] Porting work remaining question

[Peter Colson]

Begin forwarded message:

> More interestingly though:
>
> > > > > > > > stacktop(llx)= 80090dd2e8
> > > > > > > > &(stacktop[-CVM_WORDS_PER_LONG]) = 88090dd2e0
> > ..
> > > I'm missing something here - where are you getting
> > > 0x7FFFFFFF8 from?
> >
> > 88090dd2e0 > 80090dd2e8 (I'm assuming that you have
> > this in the printfs)
> >
> > fprintf(stderr, ">>>>> stacktop(llx) = %llx",
> > stacktop);
> > fprintf(stderr, ">>>>>
> > &(stacktop[-CVM_WORDS_PER_LONG]) = %llx",
> > &(stacktop[-CVM_WORDS_PER_LONG]));
> >
> > So then why are these two printf's showing so much
> > difference in value (my calc showed 0x7FFFFFFF8).
>
> Indeed why?
>
> I've gone back to this and can see (as you did):
>
> stacktop = 0x80090dd2e8
> &(stacktop[-CVM_WORDS_PER_LONG]) = 88090dd2e0
>
> However,
>
> stacktop-CVM_WORDS_PER_LONG = 80090dd2e0
>
> which seems to make more sense. Now, are:
>
> &(stacktop[-CVM_WORDS_PER_LONG])
> stacktop-CVM_WORDS_PER_LONG
>
> equivalent? At first glance I would have thought yes...which
> makes the differences quite odd and it's in the second '8' of
> '88090...'. I'll look into it more...

OK. I've found some strangeness:

If stacktop = 80040dd2e8

Then:

&(stacktop[-1]) = 80040dd2e0
stacktop-1 = 80040dd2e0

Which is the same. But using expanded CVM_WORDS_PER_LONG:

&(stacktop[-((sizeof(ILInt64) + sizeof(CVMWord) - 1) / 
sizeof(CVMWord))] = 88040dd2e0
stacktop-((sizeof(ILInt64) + sizeof(CVMWord) - 1) / sizeof(CVMWord)) = 
80040dd2e0

Which re-introduces the difference seen above.

This seems to be either precedence (which I doubt as everything
including final result is bracketed prior to negation) or something
in the evaluation of 8+8-1/8? I'll keep looking, but if there's anything
obvious, let me know.


Regards,
Peter Colson.