Date: Wed, 27 Oct 2021 08:11:13 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Gleb Smirnoff <glebius@freebsd.org> Cc: mjg@freebsd.org, jhb@freebsd.org, jeff@freebsd.org, arch@freebsd.org Subject: Re: rwlock(9) and mutex(9) definitions Message-ID: <YXjfcclPCfbhYYCQ@kib.kiev.ua> In-Reply-To: <YXjadvxmJRoXwWUj@FreeBSD.org> References: <YXiw1afVlQyEhQyc@FreeBSD.org> <YXizhiRnByvyisXe@kib.kiev.ua> <YXi0PM6babQKFulv@kib.kiev.ua> <YXjS16MQZiKm4E/r@FreeBSD.org> <YXjYCj2zh0cjL/%2Bq@kib.kiev.ua> <YXjadvxmJRoXwWUj@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Oct 26, 2021 at 09:49:58PM -0700, Gleb Smirnoff wrote:
> On Wed, Oct 27, 2021 at 07:39:38AM +0300, Konstantin Belousov wrote:
> K> > K> Both rw_rlock and rw_wlock are in tail context. You cannot _return_ void.
> K> >
> K> > You actually can return void to hint compiler for a tail call optimization.
> K> > It is not a wrong syntax.
> K> >
> K> > Other code that is working with true void functions (e.g. with WITNESS) and
> K> > doesn't work with "do {} while" is:
> K> >
> K> > void
> K> > something(bool clue)
> K> > {
> K> > return (clue ? rw_rlock(lock) : rw_wlock(lock));
> K> > }
> K> >
> K> > This is explicitly allowed in 6.5.15 of the C11 standard.
> K> >
> K> > Of course all this code can be written in some other way, so constraint
> K> > of KPI not being true functions can be worked around, but I believe better
> K> > it be fixed.
> K>
> K> Hum. I know about ternary operator allowing void-typed expressions on both
> K> sides of ':'. What I replied to is the following, from C17
> K>
> K> 6.8.6.4 The return statement
> K> Constraints
> K> 1 A return statement with an expression shall not appear in a function
> K> whose return type is void. A return statement without an expression
> K> shall only appear in a function whose return type is void.
> K>
> K> So syntax above is explicitly prohibited by the standard. I was quite
> K> surprised that both gcc and clang silently accept this, unless -pedantic
> K> is specified. There is no mention of this extension in gcc manual.
> K> Compiler explorer demonstrates that compilers like msvc do warn about
> K> the construct, and some even error out (tendra):
> K> https://godbolt.org/z/xqcPssTcY
> K>
> K> Another part of the confusion, perhaps, is that
> K> return <void expression>;
> K> is explicitly allowed by the C++ standard (I looked at 2020 version),
> K> unlike C.
>
> Hmm, so I mistakenly transferred my knowledge about the tail call
> optimization hint from C++ to C.
>
> Okay, let's put return aside. This would compile with true
> functions (e.g. WITNESS), otherwise not:
>
> void
> something(bool clue)
> {
> clue ? rw_rlock(lock) : rw_wlock(lock);
> }
>
> And this is correct code per 6.5.15.
So why cannot you write it as
...
if (clue)
rw_rlock(lock);
else
rw_wlock(lock);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?YXjfcclPCfbhYYCQ>