Milter-greylist

> I fear you have a problem with your compiler. The zero'ing you did is
> three lines below this:
>         memset(sr, 0, sizeof(*sr));

Yes, you are right. And in fact the problem has come back after
setting nmatch to zero. The problem is the seg fault does not seem to
be very predictable. Before setting nmatch to zero, Milter greylist
would segfault within 15 min on the first regex match. After the fix,
it has run for many hours, hitting many regexes. That's why I figured
it must have been resolved. But this was probably just coincidence,
since 1) the fix shouldn't do anything as you pointed out and 2) the
seg fault has returned.

> try without any compiler optimization (and without your fix)?

I will do that.