[Boost-users] Re: ~intrusive_ptr bug?

From: David Abrahams (dave_at_[hidden])
Date: 2005-05-21 22:37:16

Jean-François Brouillet <verec_at_[hidden]> writes:

>>>>> What is the name you would choose for the "addRef" parameter?
>>>>> "managed" ? Or maybe, simply "addRef" after all, since you
>>>>> would _assume_ by virtue of the symmetric behaviour of a  
> destructor
>>>>> w.r.t a constructor, that whatever the constructor does, the
>>>>> destructor un-does it, and whatever the constructor does NOT do,
>>>>> the destructor does NOT un-do.
>>>>
>>>> Bad assumption with smart pointers.  Ctors don't allocate
>>>> resources, yet dtors deallocate them.
>>>
>>> Which might only indicate that they are not so "smart" after
>>> all... ;-)
>>
>> Clever quips aside, it indicates that your design principle doesn't
>> generalize.
>
> I fail to see the "generality" in the current design. 

I wasn't talking about generality of design.  I was talking about
whether the design principle that ctors and dtors must be opposites
holds up in general.

> An example where it actually makes sense to not "addRef" on ctor but
> to "remRef" on dtor would be most welcome.

I have no opinion about that issue.

>> If you're arguing that the design of intrusive_ptr should be
>> changed,
>
> I don't have such pretensions. I was only suggesting to *rename* the
> parameter ``addRef'' into something less confusing because of the
> asymmetric behavior.

Naming is part of the design.

> Initially, I thought it was a *bug*, hence the title of this thread.
> I was told that it is not so, and that what I call a bug is actually
> a design feature. That's fine with me. Except for the confusing
> name, first, and the inability anyone, so far, has had to
> demonstrate a useful example where this asymmetric behavior makes
> sense.
>
> I'm all hears if you can convince me that the only road-block
> between my goals and the boost xxxx_ptr family is just my lack of
> understanding.

I don't have such pretensions ;-)

>> Whatever specific change you're arguing for may have merit, but not
>> on the basis of ctor/dtor symmetry.
>
> That's an *opinion*, probably as valid as mine, which is the
> opposite, regardless of which has the most "mind-share".
>
> My stance is that symmetry is at the very heart of the C++
> object model. "destruct" is the precise opposite to "construct".

In most cases, that's true.

> You can break it on whatever grounds you find suitable, but you
> can't expect people to have an easy time a) not being surprised, and
> b) understanding what was the overriding reason.

Fair enough.

P.S. Your use of quotation marks comes off as rather needlessly
sarcastic and disparaging of those you're corresponding with.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com