On Sun, 20 Apr 1997, Allen Gwinn wrote:
> Has anyone tackled this one yet, before I jump in? Also, I'm not really
> familiar with the differences between ndbm and gdbm enough to know if
> they're 1-for-1 compatible. If they are, I would suggest a "-lgdbm" in
> place of "-lndbm" in the Linux stuff.
There are problems with gdbm if it is compiled in such a way as to do
its own locking. This then conflicts with Exim's attempts to lock the
file. I have been sent some patches to overcome this, and am hoping to
do some work in the general dbm area reasonably soon. (There's a new
version of Berkeley DB out which also does its own locking...)
On Sun, 20 Apr 1997, Philip Blundell wrote:
> If you're prepared to live without squashing or logging of options, you
> can just change the `#if !HAVE_IPV6' immediately above the offending
> code to an `#if 0'.
That is certainly true, and since most versions of Linux, I'm told, have
IP source-routing turned off in the kernel anyway, you don't actually lose
anything.
On Sun, 20 Apr 1997, Allen Gwinn wrote:
> Not being that familiar with Exim (first shot at installing it), what do I
> give up, specifically? What is the impact?
Not very long ago code was added to Exim to detect "IP options", and in
particular, "IP source routing", which is a potential security threat
(see kill_ip_options and related). Many sites lock out IP source routing
in their routers, and I understand that Linux by default locks it out in
the kernel, but it was felt useful to have a check in Exim as well.
Unfortunately, Linux handles the detection of IP options totally
differently to every other OS that Exim runs on. What is worse, it
changed the name of a structure from "options" to "ip_options" between
releases. Exim 1.62 is supposed to detect which one to use, but things
may go wrong.
On Mon, 21 Apr 1997, Dr. Rich Artym wrote:
> I'm running the latest RedHat 4.1 with the nearly-latest gcc 2.7.2.1 and
> the almost-latest production kernel 2.0.29, and Exim 1.62 compiled without
> a single hitch (just the LOCAL blip, no hitch). I'm happy. :-)
Phew! I'm glad I got it right for at least *one* version of Linux...
--
Philip Hazel University Computing Service,
ph10@??? New Museums Site, Cambridge CB2 3QG,
P.Hazel@??? England. Phone: +44 1223 334714
