Verily did Peter Seebach write:

> It's reasonably visible in low-memory systems.  I think someone explained
> that the rule was something like "if a process which is partially paged
> out exits and the parent process is also at least partially paged out,
> the child process's pages are not reclaimed".  It was arcane.

> Our system used to crash or need a reboot about every 3 days, with 8 megs
> of memory, and comfortably goes for 2 weeks, with 16.  Symptom: swap
> space runs out.

Hmm, I don't see a "leak", but more of a "liberal use of virtual mem"
on one of my systems.  i386, mailhost, dns server, router, 8M real mem,
64M swap, 32M /tmp (mfs, only 3k in use), not much in user-processes
running (just shells and a couple of irc processes).

Swap on this machine is currently idling at 71%, with an uptime of
37 days (longer than ftp.netbsd.org, I might add.  Hi Bill :).  I'd guess
that that if something was "leaking", then it's quite slow, and I couldn't
begin to put a finger on which app/area it is.

-- Rob
----------------------------------------
Internet: windsor@pobox.com      Life: Rob@Sunnyvale.California.USA.Earth
"Da Web": http://pobox.com/~windsor/

"Ain't much distance 'tween a pat on the back and a kick in the pants."
    -- David Lee Roth (1990), `The Dogtown Shuffle'