[Mimedefang] MD not clearing up after itself

[David F. Skoll]

On Thu, 23 Jan 2003, Clayton, Nik [IT] wrote:

> MD doesn't always seem to clean up after itself in some cases when it
> exits.  On a busy mailserver we see mdefang-* directories in the spool
> directory that are over a day old.

I've noticed this occasionally.  I don't quite understand it; the cleanup
code is in mimedefang.c, so even if the multiplexor kills off the Perl
process, the cleanup should still happen.

> I've got a couple of untested theories about what might be causing them.
> One is that mimedefang process takes too long to do it's job, so the
> multiplexor kills it, so the stale directories are left hanging around.
> The other is that some messages might be sufficiently complex (either
> innocently, or maliciously) that they cause MD to blow out it's memory
> allocation, and segfault (or similar), and again, the cleanup code doesn't
> get the chance to run.

Again, even if that happens, the cleanup code is in the milter process,
so it shouldn't matter if the Perl filter dies (unless you use the "-k"
flag to keep failed directories around.)  I can only conclude that it's
a bug in libmilter, and that sometimes the cleanup callback isn't getting
called.

--
David.