Re: [sv-ec] Abstract classes and virtual methods

From: Gordon Vreugdenhil <gordonv_at_.....>
Date: Mon Jan 23 2006 - 06:47:00 PST
Steven Sharp wrote:
>>From: "Rich, Dave" <Dave_Rich@mentor.com>
> 
> 
>>Maybe we need to
>>say the presence of a default needs to match, but not the expression
>>itself.
> 
> 
> A little thought shows that this part is required.  The caller can
> leave out an argument only when there is a default.  How can the
> compiler determine whether it is legal for the caller to leave out
> an argument to a virtual method, when it doesn't know which
> implementation will be called until runtime, unless all of them
> must match?  Surely you aren't suggesting that this be a runtime
> check?


No, the presence of defaults must be (at latest) an elab time
check.  Note that it can be as late as elaboration, as can
the check as to whether a virtual function override is valid.

I've noticed a few places where the LRM uses "compile/compiler"
for checks (like this) that are really checks that must be done
"no later than elaboration" but certainly can't always be done at
initial analysis time (i.e. "compile").

Gord

> Steven Sharp
> sharp@cadence.com
> 

-- 
--------------------------------------------------------------------
Gordon Vreugdenhil                                503-685-0808
Model Technology (Mentor Graphics)                gordonv@model.com
Received on Mon Jan 23 06:47:05 2006

This archive was generated by hypermail 2.1.8 : Mon Jan 23 2006 - 06:47:59 PST