Re: [DotGNU]Bug - null objects and stack traces.

[Fergus Henderson]

On 06-Oct-2002, Gopal V <address@hidden> wrote:
> If memory serves me right, Rhys Weatherley wrote:
> > It is perfectly legitimate to call a method with a null "this"
> > object in CLI.  Object references are checked at the point they
> > are first dereferenced.  Dereferences occur when you access a
> > field or make a virtual method call (which needs to access the
> > vtable in the object).  Because "Push" is not virtual, the first
> > time that "s" is dereferenced is within the method body.  MS'es
> > CLR does the same thing.
> 
> "call" instruction does not throw a NullReferenceException , but
> "callvirt" does ... But "why ?" is not something I can answer.

Two reasons:

- Efficiency.  A typical implementation of "callvirt" will trap for
  null references anyway, whereas a typical implementation of "call" won't.

- Support for other languages.  The CLI should support languages whose
  procedure calls don't trap when the "this" pointer (or equivalent) is null.

-- 
Fergus Henderson <address@hidden>  |  "I have always known that the pursuit
The University of Melbourne         |  of excellence is a lethal habit"
WWW: <http://www.cs.mu.oz.au/~fjh>  |     -- the last words of T. S. Garp.