Re: [boost] 'function' idea

From: Paul Mensonides (pmenso57_at_[hidden])
Date: 2002-11-10 01:44:57

----- Original Message -----
From: "Aleksey Gurtovoy" <agurtovoy_at_[hidden]>

> Paul Mensonides wrote:
>
> > What do you mean by "tweaked"?
>
> With the cleaned up "type deduction success/error?/failure garbage" :)

By cleaned up, I mean any template function when the declaration is
instantiated yields a semantically invalid function type, in *any* way, it
should remove the function from the overload set.  Right now, there is no
such thing as an third "error" alternative.  The standard only mentions type
deduction failure or success.  There is *no way* that type deduction can
succeed with a semantically invalid declaration.  Likewise, there is no way
it can fail according to the "list" of ways type deduction can fail.  The
only other possibility is to have a compile-time error.  If this is the
result, the mere presence of a template function with a certain name can
completely break the overload set of that name entirely.

If it was changed to make type deduction fail if it yields a semantically
invalid declaration, it would be even easier to implement that and many
other things.

Paul Mensonides