closure api return value types problem

[Andrew Haley]

On 07/04/10 20:10, Anthony Green wrote:
> On 04/07/2010 02:44 PM, Sergei Trofimovich wrote:
>> On Wed, 07 Apr 2010 08:48:17 +0100
>> Andrew Haley <aph@redhat.com> wrote:
>>
>>   
>>>>> I think this is just because of the common (universal?) C ABI
>>>>> convention that integer typed return values of rank less than int are
>>>>> returned as a whole word, suitably sign- or zero-extended.  It need
>>>>> not have been done that way.
>>>>>         
>>>> The thing is:
>>>> * it's _very_ unintuitive (and not only for me) as that closure callback
>>>>    (with args and resp parameters) is meant to be convention agnostic
>>>>       
>>> Yes.  I take your point.
>>>     
>> How about changing it in libffi-4.x.y with slight API change to break
>> wrong assumptions?
> 
> I'm open to API changes in 4.x if they make sense.

There's nothing slight about this API change, though.  I think it
would break all clients of the closure interface, and wouldn't it mean
changing all the back-ends too?  Perhaps it would make sense to add a
new API with the desired behaviour.  Sometimes, even though a decision
may have been wrong, it makes more sense to keep it than to
retroactively make a change.

Andrew.