Re: nullptr vs 0

From: Andrzej Krzemienski (akrzemi1_at_[hidden])
Date: 2022-05-25 05:34:18

niedz., 22 maj 2022 o 23:59 Andrey Semashev via Boost <boost_at_[hidden]>
napisał(a):

> On 5/22/22 18:35, Peter Dimov wrote:
> >> I did however not know that GCC doesn't warn on NULL. Interesting
> >> compromise, even though I don't quite see the point of it.
> >>
> >> Clang does warn on NULL, however.
> >
> > I do see the point of it now. Under GCC the warning is pragmatically
> > intended to catch actual mistakes such as mistyping
> >
> >     while( *p != 0 )
> >
> > as
> >
> >     while( p != 0 )
> >
> > so the warning only triggers on the actual 0, and is enabled under
> > C++03 as well.
> >
> > Clang, and the various static analysis tools such as the MS analyzer,
> > enforce a stylistic preference for nullptr, so they warn on both 0
> > and NULL, but Clang doesn't warn in C++03 mode.
> >
> > Either way, I'm not looking forward to adding a bunch of lines to
> > suppress the various warnings, instead of just telling the PR author
> > to use BOOST_NULLPTR instead of nullptr.
>
> We already have NULL as the portable macro, so I would still prefer it
> to a Boost invention.
>
> Again, that some compilers warn on NULL is a compiler bug and should be
> treated as such, IMO. Compiler vendors should keep their stylistic
> preferences to themselves, especially when they are impossible to follow
> (in C++03 case). The use of NULL as a null pointer is perfectly valid,
> in the sense of both formal C++ compliance and code readability.
>

While I agree with the above argument, I still support the idea of adding
BOOST_NULLPTR to Boost.Config.

My reasoning is that the macro is meant to be used by library implementers;
and the world of library implementers is different from the world of
application developers, who consume the libraries.
As a regular application developer, I know which version of C++ I am using,
and I know if I should use NULL or nullptr. I also know which compiler
warning to turn off.

In contrast, Boost is known to have served as a "platform" that screens the
users from compiler differences and compiler bugs. To say "file a bug
against the compiler and have Boost just use NULL" goes against this
tradition.
As a maintainer of a library that supports C++98 compilers with bugs, I
feel compelled to use something like BOOST_NULLPTR. In fact, I have been
hit by this very issue:
https://github.com/boostorg/optional/commit/76ff82d1912901f188599acea09cdcdb232353bd

As an application developer, I would never use BOOST_NULLPTR.

Regards,
&rzej;

>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost
>