Marking select(2) as restrict
Warner Losh
imp at bsdimp.com
Thu Feb 22 14:43:13 UTC 2018
On Thu, Feb 22, 2018 at 3:56 AM, Konstantin Belousov <kostikbel at gmail.com>
wrote:
> On Thu, Feb 22, 2018 at 11:27:52AM +0100, Tijl Coosemans wrote:
> > On Wed, 21 Feb 2018 12:44:00 +0200 Konstantin Belousov <
> kostikbel at gmail.com> wrote:
> > > On Tue, Feb 20, 2018 at 10:14:05PM -0800, Eitan Adler wrote:
> > >> On 20 February 2018 at 21:19, Warner Losh <imp at bsdimp.com> wrote:
> > >>> Once upon a time, this would break a lot of code. Perhaps times have
> > >>> changed.
> > >>
> > >> I've seen very little code that this would break though some of it
> > >> certainly exists.
> > > You certainly seen very little code, but the question was about the
> > > existed code.
> >
> > FWIW, it seems that glibc uses restrict since 2000 so there's unlikely to
> > be much fallout:
> > https://sourceware.org/git/?p=glibc.git;a=blob;f=misc/sys/select.h
> > https://sourceware.org/git/?p=glibc.git;a=commitdiff;h=
> 98cbe360d947b59e7a5eda068581f4cfeb4b99b3
>
> Clearly, nobody knowns. At least, glibc is used with gcc compilation, not
> with clang.
>
> Consider the recently changed devd code:
> select(n + 1, &fd, &fd, &fd);
> There, compiler can see that restrict is applied to arguments which are
> given same values. Since this leads to the self-contradicting statement
> fd != fd
> which cannot be true, compliler in its optimizing wisdom can assume that
> the code is never executing and remove it. I do not know whether clang
> actually makes such transformation, but it does not sound unfeasible
> looking at its other advances.
>
If the compilers affirmatively fails when this happens, then the exp run
will catch this stuff. If it doesn't, or just gives a warning, it likely
will not.
Absent that, nobody can say with certainty this change won't break
anything.
Warner
More information about the freebsd-hackers
mailing list