Re: [PATCH 10/11] ELFv2 ABI: structure passing / return

["Ulrich Weigand" <uweigand at de dot ibm dot com>]

Joel Brobecker wrote:

> > +/* Walk down the type tree of TYPE counting consecutive base elements.
> > +   If *FIELD_TYPE is NULL, then set it to the first valid floating point
> > +   or vector type.  If a non-floating point or vector type is found, or
> > +   if a floating point or vector type that doesn't match a non-NULL
> > +   *FIELD_TYPE is found, then return -1, otherwise return the count in the
> > +   sub-tree.  */
> > +
> > +static LONGEST
> > +ppc64_aggregate_candidate (struct type *type,
> > +			   struct type **field_type)
> 
> I think we can reasonably change the return type to int, here?
> That would avoid the cast you're making at:
> 
> > +      struct type *field_type = NULL;
> > +      LONGEST field_count = ppc64_aggregate_candidate (type, &field_type);
> [...]
> > +	      if (n_elts)
> > +		*n_elts = (int) field_count;
> 
> The other option is to make everything use LONGEST for this count.
> That way, everything would be consistent and we'd be avoiding a cast
> as well.

So I've chosen those types deliberately :-)  In ppc64_aggregate_candidate
we find *any* aggregate as long as all members are of the same type.  It
is theoretically possible that the number of such members exceed 2G, e.g.
if the aggregate is a very large array.  Thus the obvious return type is
the type we use for array indices, i.e. LONGEST.

However, in ppc64_elfv2_abi_homogeneous_aggregate we only ever set n_elts
if the aggregate is one that is to be returned in registers, which means
that the number of members can never exceed 8.  Thus the obvious type of
n_elts is int.

Note that the cast is guaranteed to be safe, since the value is bounds-
checked before:

+      if (field_count > 0)
[...]
+         if (field_count * n_regs <= 8)
[...]
+               *n_elts = (int) field_count;

Given the above, it seems best to me to chose types as I did orignally ...

Bye,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU/Linux compilers and toolchain
  Ulrich.Weigand@de.ibm.com