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

Neil,

 

Just so we are all in synch on the issue of overriding virtual
functions. This is what the 2003 ISO/IEC C++ standard says:

 

    If a virtual member function vf is declared in a class Base and in a
class Derived, derived directly or

    indirectly from Base, a member function vf with the same name and
same parameter list as Base::vf is

    declared, then Derived::vf is also virtual (whether or not it is so
declared) and it overrides Base::vf.

 

And, barring some other bug, this is the way in which Vera should
behave.

 

For the record, I believe that allowing an override of a virtual
function with a non-virtual is a poor design feature that violates the
assumptions of the base class and is likely to cause more problems. If
we do not enforce this behavior then I would be in favor of an error.

 

Arturo

 

-----Original Message-----
From: owner-sv-ec@eda.org [mailto:owner-sv-ec@eda.org] On Behalf Of Neil
Korpusik
Sent: Monday, February 13, 2006 9:26 AM
To: Gordon Vreugdenhil
Cc: Rich, Dave; sv-ec@eda.org
Subject: Re: [sv-ec] Abstract classes and virtual methods

 

Gordon Vreugdenhil wrote On 02/04/06 21:34,:

>>4. Optional 'virtual' keyword

>> 

>>   Page 2, first paragraph.

>> 

>>   "Once a method has been identified as virtual, it shall remain
virtual in

>>   any subclass that overrides it. The virtual keyword may be used in
later

>>   declarations, but is not required."

>> 

>>   This appears to be an enhancement. Not only that, I don't believe
that we

>>   want to add this capability. It should be up to the user to
determine if

>>   a method should be virtual or not.

>> 

>>Neil

> 

> 

> 

> Neil, do I understand that you want to require the user to

> have an explicit "virtual" in every derivative class?  

 

Yes, that is what I proposed.

 

> What

> if the user does not use the word "virtual" if the base has

> a virtual method?  Does that hide the inherited one or is that

> an error?

 

It would not be an error to leave off the word 'virtual' in a subclass.

It would override the virtual method in the base class just like it

would for any other virtual method. The difference would be that

this method would no longer be virtual for

any classes derrived from this subclass.

 

> 

> I don't see any particular problem with not requiring "virtual"

> for virtual overrides in derivative classes.

> 

 

It looks like we have a choice of two existing strategies for handling

this situation.

 

I checked the 1998 version of the ISO/IEC C++ standard to see what it
had

to say on this matter. It was quite clear that what Dave has proposed is

indeed the way that virtual member functions are handled in C++. See
section

10.3 Virtual functions.

 

I checked the OpenVera documentation but was unable to find sufficient
details

on this to definitively say which way it should work in Vera. I wrote a

small vera example and observe that it works in a manner similar to what
I

was suggesting in my post to the reflector.

 

That leaves us with two existing implemetations to coose from.
Orgininally

I was thinking that C++ and Vera treated virtual methods the same way. I
now

realize that they don't.

 

 

 

 

-- 

---------------------------------------------------------------------

Neil Korpusik                                     Tel: 408-720-4852

Senior Staff Engineer                             Fax: 408-720-4850

Frontend Technologies - ASICs & Processors (FTAP)

Sun Microsystems

email: neil.korpusik@sun.com

---------------------------------------------------------------------