http://cr.yp.to/sigs/h0.html?

Following the description on http://cr.yp.to/sigs/h0.html, I wrote a
routine to expand a SHA-1 hash into a 1536-bit hash (the complete source
code is attached):

static void
expand_h (uint32_t h[48])
{
  int i;
  for (i = 5; i < 48; ++i)
    h[i] = h[i-5] + ((h[i-2] + s[i]) ^ ROT (h[i-1], 5));
}

When I use this routine to expand the 5 test cases following the
description, none of them fully coincide.  The first and third columns
match one value, the second and fourth columns match two values, and the
fifth column matches a whopping six values.

Are the test cases bogus, or did I misunderstand the hash expand
function?

--

-- 
Matthew Dempsky <mrd <at> alkemio.org>