Viktor Dukhovni:
> -#define HAVE_NEXTHOP_STATE(state) (STR((state)->iterator->request_nexthop)
> != 0)
> +#define HAVE_NEXTHOP_STATE(state)
> (STR((state)->iterator->request_nexthop)[0] != 0)
That is clearly a bug, considering the semantics of the *_NEXTHOP_STATE()
primitives:
FREE_NEXTHOP_STATE() set iterator->request_nexthop to empty
SET_NEXTHOP_STATE() set iterator->request_nexthop to non-empty
HAVE_NEXTHOP_STATE() should test if iterator->request_nexthop is non-empty
This bug caused plaintext connections for different destinations
to be cached under the same lookup key.
Thanks to Viktor for finding the cause, to Sahil for discovering
that there was a problem, and to myself for structuring Postfix
code such that I made this mistake in only one place.
Wietse