> On Oct 4,  1:43pm, Andrew Cagney wrote:

> The comment should say r3/r4, not r3/r3.  Likewise here:

Changed.

> +      if (inval)
> +	{
> +	  /* This matches SVr4 PPC, it does not match gcc.  */
> +	  /* The value is padded out to 8 bytes and then loaded, as
> +	     two "words" into r3/r3.  */
> 
> In a comment prior to do_ppc_sysv_return_value(), please specify
> precisely which ABIs this function covers.  The function name implies
> that it's only SysV, but it looks to me like it's for SysV, Altivec,
> and e500.  I don't think it's worth making the function name more
> verbose, but I do think it's important to list the other ABIs that
> someone looking at this function needs to consider.  Otherwise, a lot
> of it doesn't make sense.

Done, they call themselves "suplements".


> I realize that this was lifted from code that existed elsewhere before,
> but after comparing this to the e500 ABI doc that I have, this doesn't
> look right.  The ABI says:
> 
>     Functions shall return values of 64-bit DSP types (__ev64_opaque__)
>     in r3.

I suspect a cut/paste tipo.

> 	  /* The e500 ABI places return values for the 64-bit DSP types
> 	     (__ev64_opaque__) in r3.  However, in GDB-speak, ev3 corresponds
> 	     to the entire r3 value for e500, whereas r3 only corresponds
> 	     to the lower 32-bits.  So place the 64-bit DSP type's value in
> 	     ev3.  */

I've added a comment.

How is the attached?

Andrew