Re: [boost] [valid_ptr] proposing a new type of pointer object that allows validity tracking without memory management and object ownership

Subject: Re: [boost] [valid_ptr] proposing a new type of pointer object that allows validity tracking without memory management and object ownership
From: Dan Walters (dan683_at_[hidden])
Date: 2011-03-17 07:56:31

Peter,

I understand now. That is a brilliant feature of the shared_ptr design
that allows an object to release itself while not being fully released
until its member function returns (when the shared_ptr on the stack
goes of scope).

However, with a normal pointer this would not be safe - it would be
very bad practice for an object that was not wrapped by a shared_ptr
to release itself - this would result in an invalid this pointer.

Locking the object before accessing is useful / essential when the
object may want to destroy itself or even just to avoid run time
errors when the programmer makes a mistake. However, I dont think it
is necessarily a bad thing to forgo locking when the object is not
meant to be able to release itself by design (in a ST environment).

Valid_ptr is no where near as smart as a shared_ptr/weak_ptr, it works
much more like a normal pointer, it takes a passive relationship to
the object it points to, resulting in a more passive design. I think
there are strong cases where the strengths of shared_ptr/weak_ptr are
just not required and something between a normal and smart pointer is
required.