> On Sep 20,  8:36pm, Andrew Cagney wrote:
> 
> 
>> Note the way I implemented it - the above wrap a generic 
>> ppc64_sysv_abi_return_value function and that handles all cases.  I was 
> 
> 
> Nice.
> 
> 
>> finding that the more traditional technique (per my ppc post) was too 
>> error proned.  Converting it to this style caused several mysterious 
>> regressions to mysteriously disappear.
>> 
>> I'm thinking of proposing that this technique become the norm.
>> 
>> Anyway, ok to commit?
> 
> 
> Yes, okay.

I've committed it with one tweak.  The ..._return_value() method now 
returns "enum return_value_convention" instead of an int.  This is to 
better align its function signature with the new gdbarch_return_value 
method I've posted.

Andrew