On Sun, Jul 29, 2018 at 12:45:23AM +0300, Konstantin Belousov wrote: > On Sat, Jul 28, 2018 at 01:59:16PM -0700, Conrad Meyer wrote: > > On Sat, Jul 28, 2018 at 1:32 PM, Konstantin Belousov > > <kostik...@gmail.com> wrote: > > > Our existing practice is to not provide an advise in the man pages > > > for the userspace code which would harm the portability or the standard > > > conformance. > > > > Perhaps, but you do not have to look beyond this manual page to find > > that such a practice, if it is even nominally applied, is not followed > > consistently. The remainder of the document spells the terminating > > value as plain "NULL." FreeBSD manual pages consistently refer to > > NULL as a pointer value. > > > > The Linux manual page spells the terminating value (char *)NULL and > > explicitly documents why. That is perhaps both pedantically correct > > and clear. I am amenable to such a change. > I do not see much difference between (char *)0 and (char *)NULL. Both > are better there, because both are correct. > > If you prefer (char *)NULL, please use it. > > > > > > I am not so sure about CheriBSD, for instance. > > > > I observe simply that CheriBSD is not FreeBSD and that CHERI is an > > exotic architecture. > We should not give an advise which breaks other architectures, even > if not included into the mainline repo.
CheriBSD would be fine because FreeBSD uses (void *)0 for NULL. That said, C++ requires that NULL be a bare 0 or nullptr so (char *)NULL is likely the most portable. -- Brooks
signature.asc
Description: PGP signature
