On Sat, Mar 27, 2021 at 10:51 AM Florian Weimer <fw@deneb.enyo.de> wrote:
>
> * H. J. Lu via Libc-alpha:
>
> > Run strnlen tests on strings without the null byte around the page
> > boundary.
> > ---
> >  string/test-strnlen.c | 26 ++++++++++++++++++++++++++
> >  1 file changed, 26 insertions(+)
> >
> > diff --git a/string/test-strnlen.c b/string/test-strnlen.c
> > index 61eb521dc1..a49d93afa2 100644
> > --- a/string/test-strnlen.c
> > +++ b/string/test-strnlen.c
> > @@ -196,6 +196,31 @@ do_page_tests (void)
> >      }
> >  }
> >
> > +/* Tests meant to unveil fail on implementation that does not access bytes
> > +   without the null character around the page boundary accordingly.  */
>
> I don't understand the comment.  Based on it, I would have expected a
> null byte at the page boundary, which is incorrectly ignored by the
> implementation.

It places strings ending at page boundary without the null byte.  If an
implementation goes beyond EXP_LEN, it will trigger the segfault.

> > +static void
> > +do_page_2_tests (void)
> > +{
> > +  size_t i, exp_len, offset;
> > +  size_t last_offset = page_size / sizeof (CHAR);
> > +
> > +  CHAR *s = (CHAR *) buf2;
> > +  memset (s, 65, page_size);
>
> I think this needs to use wmemset for the WIDE case, see
> wcsmbs/test-wcsnlen.c.

Fixed.

Here is the v2 patch.

-- 
H.J.