Re: [Bug-gsl] array, vector indices out-of-bounds in "linalg/bidiag.c"

[Andrew W. Steiner]

In case it helps, see
http://www.gnu.org/software/gsl/design/gsl-design.html#SEC31
for why the loop is constructed that way.

On Sun, Apr 20, 2008 at 3:58 AM, Julian Seward <address@hidden> wrote:
>
>  >       for (j = N; j > 0 && j--;)
>
>  This is an extremely strange for-loop header, and I wonder if
>  it is what the author(s) really intended.  I _think_ it might
>  be equivalent to
>
>    for (j = N; j > 0 && j != 0; /*no step action*/) {
>       j--;
>       /* now the rest of the loop body */
>    }
>    j--; /* because the condition is consulted 1 more time than the
>            loop body runs */
>
>  so there isn't necessarily an overrun at j = N.
>
>  I wonder if it would not be cleaner to use a standard idiom:
>
>    for (j = N-1; j >= 0; j--)
>
>  IMO even a C language lawyer would have a hard time figuring out
>  what the exact behaviour is here, which doesn't bode well for
>  end-user understanding of the code.
>
>  You might want to try Valgrind's Memcheck tool to see if there
>  are in fact any overruns happening.
>
>
>
>  J
>
>
>  _______________________________________________
>  Bug-gsl mailing list
>  address@hidden
>  http://lists.gnu.org/mailman/listinfo/bug-gsl
>