Re: variable initializer circularity needs to be defined

At first reading, this looks good. I'm forwarding this to the Working 
Group for discussion, and I'll let you know when there is something to 
report.

Jonathan

Per Bothner wrote:

> Per Bothner wrote:
> 
>> Presumably it wouldn't be difficult to define circularity
>> in terms of dependencies between variables and functions. 
> 
> 
> Jonathan Robie wrote:
> 
>> Do you have a definition you would propose?
> 
> 
> Would something like the following work?
> 
> A variable declaration V1 directly depends on a variable declaration
> V2 if V1 has an initializing expression that contains a variable 
> reference that is bound to V2.
> 
> (This begs the question of the meaning of "bound", but presumably
> that is defined in the formal semantics.  If we disallow variable
> redefinitions, we can define "bound" in terms of matching QNames.)
> 
> A function declaration F1 directly depends on a variable V2
> if the function body of F1 countains a variable reference that is
> bound to V2.
> 
> A function declaration F1 directly depends on a function declaration F2
> if the function body of F1 contains a call bound to F2.
> 
> A variable declation V1 directly depends on a function declaration F2
> if V1 has an initializing expression that contains a call bound to F2.
> 
> A (function or variable) declaration D1 depends on declaration D2
> if one of:
> (a) D1 directly depends on D2; or
> (b) there is some declaration Dx such that D1 depends on Dx and Dx
> depends on D2.
> 
> It is a circularity error if a variable depends on itself.