On Thu, 2018-02-01 at 09:08 -0800, Darrick J. Wong wrote:
> On Thu, Feb 01, 2018 at 12:31:36PM +0100, Arnd Bergmann wrote:
> > On Thu, Feb 1, 2018 at 11:23 AM, Geert Uytterhoeven
> > <ge...@linux-m68k.org> wrote:
> > > With gcc-4.1.2:
> > >
> > > fs/xfs/scrub/agheader.c: In function ‘xfs_scrub_agfl’:
> > > fs/xfs/scrub/agheader.c:770: warning: missing braces around
> > > initializer
> > > fs/xfs/scrub/agheader.c:770: warning: (near initialization for
> > > ‘sai.oinfo’)
> > >
> > > The first member of struct xfs_scrub_agfl_info is no longer an integral
> > > type, but a struct. Add the missing curly braces to fix this.
> >
> > I suspect gcc-4.5 is affected as well, but not 4.6+
> >
> > > --- a/fs/xfs/scrub/agheader.c
> > > +++ b/fs/xfs/scrub/agheader.c
> > > @@ -767,7 +767,7 @@ int
> > > xfs_scrub_agfl(
> > > struct xfs_scrub_context *sc)
> > > {
> > > - struct xfs_scrub_agfl_info sai = { 0 };
> > > + struct xfs_scrub_agfl_info sai = { { 0 } };
> > > struct xfs_agf *agf;
> >
> > Looks ok to me, but
> >
> > struct xfs_scrub_agfl_info sai = { };
> >
> > might be slightly better in case the first member changes again.
>
> Frankly I'd rather see it changed to memset(&sai, 0, sizeof(sai)); and
> stop having to field all these gcc warnings that vary depending on
> compiler version...
trivia:
memset should also be preferred if the structure
could be copied to userspace as the "= {}" is not
guaranteed to zero any possible padding or member
alignment holes.
