svn commit: r305091 - head/sys/sys

[Mateusz Guzik]

On Tue, Aug 30, 2016 at 03:35:39PM -0700, Conrad Meyer wrote:
> On Tue, Aug 30, 2016 at 2:48 PM, Mateusz Guzik <mjg at freebsd.org> wrote:
> > Author: mjg
> > Date: Tue Aug 30 21:48:10 2016
> > New Revision: 305091
> > URL: https://svnweb.freebsd.org/changeset/base/305091
> >
> > Log:
> >   fd: simplify fd testing in fget_locked by casting to u_int
> >
> > Modified:
> >   head/sys/sys/filedesc.h
> >
> > Modified: head/sys/sys/filedesc.h
> > ==============================================================================
> > --- head/sys/sys/filedesc.h     Tue Aug 30 21:43:57 2016        (r305090)
> > +++ head/sys/sys/filedesc.h     Tue Aug 30 21:48:10 2016        (r305091)
> > @@ -201,7 +201,7 @@ fget_locked(struct filedesc *fdp, int fd
> >
> >         FILEDESC_LOCK_ASSERT(fdp);
> >
> > -       if (fd < 0 || fd > fdp->fd_lastfile)
> > +       if ((u_int)fd > fdp->fd_lastfile)
> >                 return (NULL);
> >
> >         return (fdp->fd_ofiles[fd].fde_file);
> >
> 
> 
> I notice that fd_lastfile is an 'int'.  Won't this trigger warnings
> about the differing signedness of the two sides of the comparison?
> Should fd_lastfile just be u_int as well?  (If not (there is some
> valid negative value), this change may be invalid.)
> 

It is -1 just after inception and is supposed to grow immediately after.
That is, the table with -1 should never be accessible.

But now that you mention it I agree this is bad style. This
unnecessarily differs from the check in fget_unlocked so I'll just unify
it.

-- 
Mateusz Guzik <mjguzik gmail.com>