Re: [boost] Re: auto_array_ptr

From: David Abrahams (abrahams_at_[hidden])
Date: 2001-04-06 11:00:52

std::auto_ptr is weird; it is modified when it is copied.

----- Original Message -----
From: "Stephen C. Gilardi" <squeegee_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Friday, April 06, 2001 11:44 AM
Subject: Re: [boost] Re: auto_array_ptr

> >  > The caveat about temporary smart pointer objects still applies. My
rule is
> >  > "never create an unnamed temporary smart pointer." The non-const
reference
> >  > idiom tries to enforce this rule... although I have another reason
for
> >  > preferring pass by non-const reference:
> >  >
> >>  void f(T arg);
> >  >
> >  > T t;
> >>
> >>  f(t);
> >  >
> >  > // now t is unchanged, right?
> >
> >Wrong, of course. Usually, the declaration of f isn't visible near the
call,
> >so you can't assume much about what's changed. But this case could be
seen
> >as an argument for using the unnamed temporary... then there's no object
> >hanging around with a confusing value.
>
> I don't see how t's value could be changed in the above example.
> Could you please explain?
>
> It seems to me that t was passed by value.  My understanding is that
> an unnamed temporary of type T is constructed using T's copy
> constructor and sent to the body of f.  This leaves t completely
> unmolested.
>
> If my understanding is incorrect, I'll be very happy to learn what
> really happens.