[PATCH v2] tests/mknod.c: add workaround for ppc bug in travis

[Ákos Uzonyi]

On Tue, 25 Aug 2020 at 16:02, Dmitry V. Levin <ldv at altlinux.org> wrote:
> On Tue, Aug 25, 2020 at 03:39:24PM +0200, Ákos Uzonyi wrote:
> > On ppc travis builds syscall(mknod) sometimes does not fail with EEXISTS
> > as expected, but returns successfully. This commit uses sprintrc for
> > errno printing, which handles successful return values too.
> >
> > * tests/mknod.c (call_mknod): set errno to 0 before syscall.
> > (main): Use sprintrc for printing errno.
> > ---
> >  tests/mknod.c | 34 ++++++++++++++++++----------------
> >  1 file changed, 18 insertions(+), 16 deletions(-)
> >
> > diff --git a/tests/mknod.c b/tests/mknod.c
> > index 44c6f028..2570b523 100644
> > --- a/tests/mknod.c
> > +++ b/tests/mknod.c
> > @@ -10,6 +10,7 @@
> >
> >  #ifdef __NR_mknod
> >
> > +# include <errno.h>
> >  # include <stdio.h>
> >  # include <sys/stat.h>
> >  # include <sys/sysmacros.h>
> > @@ -20,6 +21,7 @@ static const char *sample;
> >  static long
> >  call_mknod(unsigned short mode, unsigned long dev)
> >  {
> > +     errno = 0;
> >       unsigned long lmode = (unsigned long) 0xffffffffffff0000ULL | mode;
> >       return syscall(__NR_mknod, sample, lmode, dev);
> >  }
>
> I don't see why you might need to change this part.
> Besides that, LGTM.

Oh yes, this is not needed anymore, as sprintc checks whether the
return value is -1.

Should I resend?