forcing tail/sibling call optimization

[Robert Dewar]

<<So, any comments?  If I were to provide a clean patch that implements
this feature, together with documentation giving a clear specification,
would you be happy to have this change included in future versions of
GNU C?
>>

How about two steps here. First the clear specification, then the decision
as to whether this extension makes sense (I have trouble seeing what this
clear specification would look like). I think it is very important that
extensions to GNU be very clearly defined.