[exim] Little surprise with headers from DATA ACL and exisca…

This is probably just a "that's the way it works" heads-up, but
I thought it might be worth putting on record.

We've just moved to a new(er) exim setup, with spamassassin now driven
via exiscan-acl.

In the old setup, we had some stanzas in the DATA ACL which computed
warning headers based on certain header tests that we were doing
there; and then we spam-rated those extra headers in spamassassin.

Now it's looking as if spamassassin isn't getting shown any headers
that have been added by the DATA ACL. We've deliberately placed the
first call of any spamassassin function /after/ the point where the
headers would be generated, and the headers /can/ be seen in the
resulting mail, located /before/ the spamassassin report; but
presumably the mechanism is such that they aren't included in what is
sent to spamassassin, since we're not seeing any spam-rating points
from them.

I guess this is "just the way things are"?

The only quick-fix that I can think of at the moment is to calculate,
in the ACL, the spam-rating points that these headers /had/ been
intended to generate, and to add this into the spamassassin result
when comparing it to the rejection threshold. Simple enough, but a
bit messy (this spam-bonus score won't feature in the $spam_bar and
$spam_report).

A more fundamental solution would be to add a custom eval: function to
spamassassin, as a functional replacement for our existing DATA ACL
stanzas; but I'm not keen on messing with that, to be honest.

None of this is a show-stopper, just a useful supplement (in our
particular context) to the mainline antispam mechanisms which we use.